Picoammeter Design

[dreaquil]

Hi,

Does anyone know of any reference material which I could use to design a decent picoammeter? I know of the Low Level Measurement Handbook and a few other articles but I want to know as much as I can because it's such a noise prone design.

[Gyro]

I don't know about decent but this design using the LMC662 might be a useful starting point...

http://www.vk2zay.net/article/251

Edit: Apart from that, the Keithley Low Level Measurement Handbook that you've already found is the best reference I know of.

[Cerebus]

I'll second the Keithley Low Level Handbook.

I've found digging out older service manuals for instruments, the ones that have proper schematics, to be very helpful as examples of good design. You could start with the Keithley picoammeter that Dave repaired a way back. The service manual is linked from the episode page.

[unitedatoms]

I watched a documentary about madam Curie measuring the discharge rate of pF capacitor (leak caused by radioactivity of Uranium ore) using electrometer with rotating mirror and beam of light. The electrometer was working as a nullmeter where opposite side was connected to piezocrystal, generating an accurate DC by pulling end of crystal with weight.

So for picoammeter level accurate comparative measurements one needs a piezo DC source, scale, set of weights, electrometer, capacitor with shield, clocks and earth's gravity.

This design can give so many ideas for calibration of picoammeter.

[Gyro]

In terms of practical advice....

- Put the picoammeter, or at least your sensing head containing the input amp and feedback resistor as close as possible to what you're measuring (ie, make it small).

- Don't use any cables (in line with keeping it close). Even the low noise ones are piezoelectric / triboelectric / leaky.

- Shield everything in a metal enclosure.

- Air wire everything that's high impedance - lift the IC input pin if using a PCB. Air is the best insulator and isn't piezoelectric / triboelectic.

- Use a decent input connector, BNCs are available in high voltage variants that have an extended ribbed ptfe rear insulator. Use it as one of the supports for the internal air wiring. Don't use a mating BNC plug - insert a suitably sized solder pin on wire into the center contact (you could use the guts of a BNC plug)

- Wash everything in IPA and allow to air dry, then never touch it with fingers.

The close proximity and screening ought to help keep noise under control.

[dreaquil]

Thanks for the replies guys, some very useful information. I'm going to look into all the suggestions and I'll get back to you I've got a prototype on the way but I'm looking at improvements for rev 2

[Gyro]

Good luck, I'm intending to build my own

[rvalente]

The HP 4140B is a pA meter and precision voltage source, I'd see how they did and replicate...

Here is the service manual: http://www.keysight.com/upload/cmc_upload/All/04140-90021.pdf
Datasheet: http://www.keysight.com/upload/cmc_upload/All/59528837.pdf

[Skimask]

I watched a documentary about madam Curie measuring the discharge rate of pF capacitor (leak caused by radioactivity of Uranium ore) using electrometer with rotating mirror and beam of light. The electrometer was working as a nullmeter where opposite side was connected to piezocrystal, generating an accurate DC by pulling end of crystal with weight.

Was that the one on PBS recently, like in the past week or so?

[unitedatoms]

I watched a documentary about madam Curie measuring the discharge rate of pF capacitor (leak caused by radioactivity of Uranium ore) using electrometer with rotating mirror and beam of light. The electrometer was working as a nullmeter where opposite side was connected to piezocrystal, generating an accurate DC by pulling end of crystal with weight.

Was that the one on PBS recently, like in the past week or so?

Yes, it was on PBS yesterday. Lots of nice stories about discovering chemical elements. I was impressed what a person can do in a lab with few glass bottles, wires, pencil and paper. Madam Curie measured Radium atomic mass to 3.5 digit accurate with basically bare hands.

So considering that capacitor was about 0.1pF, voltage may be in range of 100V, time constant in ballpark of 100 sec. The resistance could be 10^15 Ohm. So currents were around hundred femtoAmperes. May be it is possible to estimate the electrons count per second, like divide current in Amperes by elecron's charge in Cuolombs and estimate the decay rate of ore. But I am too lazy. May be thousands-to-millions of electrons per second.

Edit: OK wolframAlpha answer to "electrons per second in 100 femtoamperes" is one electron per 1.6 usec.

[fcb]

The Signal Path did a great teardown/repair of a Keithley 220 - you'll get some great hints from this video (teflon standoffs etc...). And you get to see Shahriar wearing a natty little hat.

https://youtu.be/pMYK5qoQvYo

[plesa]

Look on how Keithley did it in 6485 https://www.eevblog.com/forum/testgear/keithley-64856487-teardown/
They used LMC662, LMC6081.
Or consider buying or check at least layout on LMP7721 dev kit.
In other thread was mentioned some discount code on TI development boards,maybe it can be used for this kit as well.

[Gyro]

What strikes me, looking at the teardowns, is the tremendous advantage you have in making a one-off home unit - air wiring, isolated battery power etc, that just wouldn't be feasible in a production unit (not to mention losing that dust generating cooling fan!)

The LMC662 in dip pkg looks a very good bet for avoiding a PCB and associated leakage, also availability and low cost. The LMP7721 might spec a little better (well for input protection, not necessarily i/p current) but would be a pain to mount and more importantly clean to keep package surface leakage low. Input protection on the LMC662 shouldn't be an issue because you can put a large value series resistor directly on its input (I think the i/p protection diodes can still manage 5mA... versus 10mA for the LMP7721).

A keep it simple approach eg. Using an existing external voltage source should help a lot too.

[Kleinstein]

A TIA can cover a rather large current range. So ranges could be in 1:100 steps or similar, unless very high resolution is needed.
So it might be easier to build a complete separate input stage just for the lowest range - this way saving on the switches, with all the trouble with extra leakage. Also consider battery power, as the very low bias OPs usually don't need mich current.

The higher ranges are far less critical.

[dom0]

What strikes me, looking at the teardowns, is the tremendous advantage you have in making a one-off home unit - air wiring, isolated battery power etc, that just wouldn't be feasible in a production unit

Keithley 640 :-)

[plesa]

Or use decent DMM with known input impedance  (10M) on lowest DCV range. There was not mentioned sensitivity level required :-)
100pA will be 1mV.
Or check uCurrent from Dave.

[Gyro]

Keithley 640 :-)

Well I think the vibrating capacitor might be a bit challenging for home construction 

Or use decent DMM with known input impedance  (10M) on lowest DCV range. There was not mentioned sensitivity level required :-)
100pA will be 1mV.
Or check uCurrent from Dave.

I think a Picoammeter worthy of the name ought to be able to resolve at least single digit pA, should be easily possible with the single digit fA input current of an LMC662. That would open it up to serious insulation evaluation, home made ionisation chambers etc. (the fun stuff  )

[plesa]

Oki, so for your application will be enough to build not whole picoammeter with multiple ranges, but rather TIA. So it is much more simplier ;-)

[Gyro]

I think it's still practical to do multiple ranges by using the LMC662 in standard inverting mode using a single high value feedback resistor switched from different proportions of the op-amp output. That way the range switching is 'low' impedance and the only high impedance node is one end of the feedback resistor/input connector/LMC662 -ve input. (ok, I will include a high value resistor between this node and the LMC662 input pin for input protection.

This style:

http://www.tradeofic.com/uploadfile/ic-circuit/200972431748582.gif

... but voltage output to an external DVM.

[dreaquil]

I got my first prototype back today and tested it out. Worked pretty well for a rough estimate .. used it to calculate the insulation resistance of a cable at 2G and 500G. It's in the range and the readings are repeatable. Now I managed to get my hands on a 6485 Picoammeter for calibration and reference I'll tell you how it goes.

Just so you know, I didn't air solder. I did a surface mount construction with a guard ring surrounding the input. Removed the top solder mask to avoid static build up and used a 1% 1G resistor with an LM6462 (150fA input bias). I know I can improve this with an LM662. My application wasn't precision measurement but rather ball park measurement so it works perfectly so far however I would like to get it to a precision stage.

EDIT: The capacitor is there to limit the bandwidth due to the Johnson equation.

[Gyro]

Wow, that was quick! 

Just one thing - I'd be tempted to put a resistor (10-100k?) directly in series with the input of the op-amp (pin 6), after the feedback parts to give its protection diodes a better chance at fighting off ESD strikes.

[Kleinstein]

Its better to have the resistor directly at the input: this way it also prevents oscillation in case of a highly capacitive source, like a long cable.
At 100 M in the Feedback,  a 100 K resistor can give a maximum of about 5 mV drop, before the OP goes into saturation. So no real need to worry about this.

[dreaquil]

Yeah that would probably be a smart move. I'm curious to how it stacks up against the picoammeter. I have a number of improvements for rev 2.

[Gyro]

I've finally got around to putting my own Picoammeter together. It uses the LMC662 as previously discussed.It seems to work pretty well and readings seem to be stable and repeatable. Maximum current is around +/-4.5nA, that will drop to +/-2.5nA at battery end of life voltage, assuming that I risk leaving it in there that long. Supply current is just under 1mA, even adding the most frugal of power LEDs would double this, so I haven't bothered.

Accuracy seems to be fine - within the limits of my test capability anyway. Tested with my only 10G ohm resistor (also 2%) it reads 1.022nA at 10V, close enough. Input voltage offset is less than 60uV using a randomly chosen LMC662.

EDIT: Ignore the 60uV, that was due to me not having nulled the Vos properly the other day. 



I've included a protection resistor within the feedback loop which should be good to +/-500V on the input, that's with 0.5mA through the LMC662 protection diodes (spec limit 5mA) so in practice is limited by the breakdown voltage of the resistor (I used a 1M large bodied carbon film).

I've also included a null adjustment pot for the op-amp offset voltage (range approx 1.35mV).  I've successfully nulled it to 1uV on my bench DVM but in practice it's ridiculously sensitive to external influences at that (fA) level. The LMC662 Vos also drifts over the supply voltage range by a few 10s of uV, so in practice it's fine for zeroing at the 100fA level (see pic). Vos TC is only 2uV/'C so not an issue.

I've included rudimentary output protection, the two 1k resistors are to ensure stability against capacitive loads (oscillation shows up as a few mV of offset).



Construction is my favorite double sided Manhattan style, using single sided copperclad. The input circuit air-wired and parts cleaned with IPA (it took several hours to stabilize afterwards - mainly the polystyrene cap I think).

[Vgkid]

Nice job  .

[krivx]

Looks good, but the point where the Zeners meet looks very close to the +Battery node. Maybe move that.

[SeanB]

Hopefully you did not wash the ps cap with IPA, as it will have degraded it rather badly. Styrene dissolves in alcohol, so hopefully it was only exposed to the vapour and the long stabilisation diffused it back out without degrading the capacitor.

Nice work, and you can still see the milling marks on the box lid, and the extra sleeve on the Duracell is a nice extra touch. I would replace the zeners with reverse biased bipolar transistor BE junctions, less sensitive to light and lower leakage. 6-8V breakdown as well.

[Gyro]

Thanks for the feedback folks 

Glad you like to battery sleeve - I wanted to provide a bit of extra insulation as the battery is 'floating' +and - from the case, then found that the plastic blister pack makes an excellent sleeve if you cut it a bit and fold it in on itself. Nice and springy, it holds the battery nicely in place without additional padding. One to remember before throwing the packaging away.

The zeners (which are clearly a rather embarrassing afterthought!) are actually well clear of the board, up on the ends of the 4mm sockets. I think the sockets look shorter because I put some bits of silicone rubber tube around them (bodge battery buffers). I should really have put them on the board before the 1k resistors to give better protection, but I didn't want to desolder the input and pull the board out again. Good idea about the reverse biased b-e junctions, but I checked the zeners and there's no significant leakage at max. 4.5V (for a 2k source impedance anyway). Just thought, I am now actually able to test transistor b-e junction leakage  I've already found that 1N4148s do actually leak about 3nA at 10V (spec 25nA max @20V) I expected much lower for some reason.

Interesting point on the polystyrene capacitors - I did some googling beforehand and there seemed to be a consensus  that polystyrene is only mildly soluble in IPA after sustained immersion testing. After I finished the unit a couple of days ago I tested another capacitor after a quick swab with IPA... After a quick air dry, leakage hit the end stop at 4.5nA and was still the same an hour or so later (though I'd killed it). Next morning it was down to very sub pA again. No physical signs of dissolving or crazing. I can only think that the IPA draws some moisture into the surface, which takes a long time to evaporate again. It certainly explains the stabilization time anyway!

[Marco]

Would a plain high voltage ceramic capacitor suffice too?

[SeanB]

1N4148 leakage is a worst case, typically at 125C, but it suffers rather badly in being sensitive to light, IR ( it is after all a silicon photodiode in a glass package) and to temperature and stress of the leads. Plastic packages TO92 devices should be a lot lower, but the SMD versions are not as good, just from the smaller package.

nice that you connected the PS cap correctly, with the outer foil to ground. If you get a larger value ( I have some 1uF polystyrene caps around somewhere, which are pretty big, even at a 50V rating device) you will see just how badly they perform with contamination, though they are very good and stable otherwise, only PTFE and vacuum is better. Ceramic, even NPO, is terrible in comparison, it drifts too much in value with both time, applied voltage and has a lot of dielectric absorption in comparison, even a NPO type. Only advantage is that it is very cheap and will handle temperature over 70C.

[krivx]

Do you have part/series numbers for the cap, connector and 1G resistor for anyone wanting to take a crack at this?

[Gyro]

@Marco

Would a plain high voltage ceramic capacitor suffice too?

Somewhere between maybe and no. I just tried a fresh from bag 220pF 1kV Murata ceramic. It reads around 300fA at 10V, surprisingly low  but an un-cleaned polystyrene 330pF from the parts box comes out at around 100fA or less (not sure how I'm going to clean them now!). A nice clean 2n2 @ Y1 (5kV test) ceramic reads over 6pA at 10V. So a 330pF 5kV part might get near the ballpark but probably still not as good as a clean low voltage Polystyrene.

@SeanB

I shielded the 1N4148s from light as much as possible but not in dark condition. I did try sliding a piece of heatshrink over one and the leakage shot up to over 1nA  Not the world's greatest insulator then! Haha, yes the end marking of the PS survived its wash. As I said, I'm worried about how to clean them now. The highest value I have is 100nF, so 1uF must be big!

[Vgkid]

What about using a jfet as the protection diode, especially the to-18 canned variants.

[SeanB]

Fist sized, and they are pretty heavy for a part with such thin leads. You also discovered heatshrink is conductive ( surprise surprise carbon black is a conductor), and to clean the ps caps you wash with deionised water and dry with cool air. Solder in using a heat shunt as well, there were only a few series of potted Murata polystyrene capacitor that would survive board cleaning at all, or reflow soldering. Du Pont stopped making the polystyrene film used in them a few years ago, which was the end of them being made new by most manufacturers, only the cheapest off brands are using Chinese film.

[Gyro]

Do you have part/series numbers for the cap, connector and 1G resistor for anyone wanting to take a crack at this?

I've had the glass resistor for years (stripped from an old ion gauge meter together with a 100M and 10G). I think it's a Welwyn 3800 series. RS components list several 1G resistors in non-glass packages:

http://uk.rs-online.com/web/c/?searchTerm=1G+ohm&sra=oss&r=t

 296-0667 looks a good compromise between size and cost (and not having to buy a bag of 5). It's 1% 100ppm/'C too. They also have 160V polystyrene capacitors (and of course LMC662s).

The extended insulation BNC came from my parts box too, but a clean standard one should be just as good.

[Gyro]

What about using a jfet as the protection diode, especially the to-18 canned variants.

They certainly work from the leakage point of view, Datron (among others) use them for that purpose, they wouldn't provide the Zener action for use in the output though (to allow normal output swing). The internal input protection diodes in the LMC662 are really good, they're bootstrapped to achieve the low input bias current while still being good for 5mA (the LMP7721 uses the same trick and achieves 10mA protection).

Fist sized, and they are pretty heavy for a part with such thin leads. You also discovered heatshrink is conductive ( surprise surprise carbon black is a conductor), and to clean the ps caps you wash with deionised water and dry with cool air. Solder in using a heat shunt as well, there were only a few series of potted Murata polystyrene capacitor that would survive board cleaning at all, or reflow soldering. Du Pont stopped making the polystyrene film used in them a few years ago, which was the end of them being made new by most manufacturers, only the cheapest off brands are using Chinese film.

Haha they must be. The old Plessey ones I have include a thicker wire fused into each end with the fine wire brought out and soldered externally. Probably even worse for transferring soldering heat though. Ah, I hadn't thought about the carbon black! Something to be wary of if covering high voltage connections, which it very well might be used for. It was only in very light contact with the 4148 leads too. Yes I knew DuPont had stopped producing the film - I was surprised when I just checked RS for kirivx that they still have them in stock.

I'm not sure how well just deionised water would work against finger grease. I must do some sacrificial tests with kitchen detergent (...or just not touch them in the first place  )

I must find the article Bob Pease wrote on low leakage diodes (maybe in one of his books). I think he highlighted 1N914s being a problem because they were gold bonded. They used to be used more or less interchangeably with 1N4148s in most applications.

[Gyro]

Do you have part/series numbers for the cap, connector and 1G resistor for anyone wanting to take a crack at this?

Ah, RS do have the extended insulation BNCs (for a price!) if you really want to go the extra mile.

http://uk.rs-online.com/web/p/bnc-connectors/2127400/

As I said, a standard BNC socket should be fine though. I haven't tried Farnell etc for pricing. You could try ebay for the resistor and socket too.

EDIT: Remember, the shortest leakage path is between the pins on the LMC662 dip package, nothing to be done about that apart from getting it really clean.

By the way, I made the input 'plug' from the insulator of a standard BNC plug with a barb pin pressed in as center pin and a loop test point soldered to the top. It keeps the input connector clean (long leakage path) when handling and saves pushing wires up the center of the socket.

[edavid]

I must find the article Bob Pease wrote on low leakage diodes (maybe in one of his books). I think he highlighted 1N914s being a problem because they were gold bonded. They used to be used more or less interchangeably with 1N4148s in most applications.

1N914s and 1N4148s use the same gold-doped die, so both have fairly high leakage.

I'd suggest 1N3595 as a good cheap low leakage diode.

Too bad PN4117s are out of production, but you can still get them.

[Alex Nikitin]

I must find the article Bob Pease wrote on low leakage diodes (maybe in one of his books). I think he highlighted 1N914s being a problem because they were gold bonded. They used to be used more or less interchangeably with 1N4148s in most applications.

1N914s and 1N4148s use the same gold-doped die, so both have fairly high leakage.

I'd suggest 1N3595 as a good cheap low leakage diode.

Too bad PN4117s are out of production, but you can still get them.

A good low leakage diode for this kind of circuit is FJH1100 from Fairchild Semi (RS #7729256) however it is very expensive (almost £14 each, inclusive of VAT) and light sensitive (not a problem inside a box though). On the other hand it is good down to some fA at room temperature near 0V for the input protection to ground.

Cheers

Alex

P.S. - I plan to measure leakage currents on a number of various "low leakage" diodes sometime soon, with both forward and reverse bias voltages applied. There are some potentially good protection diodes available - ESD9R from ON Semi for example.

[Gyro]

LEDs are supposed to be good too due to their high bandgap voltage. Of course that equates to a high forward voltage too and of course they are also light sensitive! I haven't tested one yet though.

[Alex Nikitin]

LEDs are supposed to be good too due to their high bandgap voltage. Of course that equates to a high forward voltage too and of course they are also light sensitive! I haven't tested one yet though.

I'll try few LEDs as well, no problem. I've just assembled a tri-axial cable for my Keithley 617, so should be able to measure leakages reliably down to about 10fA. Need to build a test box though.

Cheers

Alex

[BravoV]

Thanks for sharing  , subscribed.

Made me want to build one, problem is finding that Giga ohm resistor. 

[Alex Nikitin]

Thanks for sharing  , subscribed.

Made me want to build one, problem is finding that Giga ohm resistor. 

Well, you can build one without a resistor. There is a second way to measure low currents - with an integrator. Replace the resistor with a capacitor and a reset switch, add a comparator and a timer. 10pF capacitor would create a slope of -1V/s with +10pA input current.

Cheers

Alex

[Gyro]

I'll try few LEDs as well, no problem. I've just assembled a tri-axial cable for my Keithley 617, so should be able to measure leakages reliably down to about 10fA. Need to build a test box though.

Hope you have better luck than I just did (probably wasn't dark enough). I got it from Bob Pease's 'Troubleshooting Analog Circuits'. Apparently the c-b junctions of most 2N3707s are good too, as are 2N4117A jfets.

P.S. I need to build a test box too, AC pickup is a killer!

[Alex Nikitin]

I'll try few LEDs as well, no problem. I've just assembled a tri-axial cable for my Keithley 617, so should be able to measure leakages reliably down to about 10fA. Need to build a test box though.

Hope you have better luck than I just did (probably wasn't dark enough). I got it from Bob Pease's 'Troubleshooting Analog Circuits'. Apparently the c-b junctions of most 2N3707s are good too, as are 2N4117A jfets.

P.S. I need to build a test box too, AC pickup is a killer!

I design electrometers (and other precision analogue stuff) and have to deal with femtoamp current levels everyday at work  . For obvious reasons I won't disclose the details of these circuits but will be happy to help with an advice where I can.

Cheers

Alex

[Gyro]

Haha, a stopwatch and a voltmeter, my reactions are getting too slow for that one. Another alternative is to use two back to back semiconductor junctions as the feedback resistor and make a logarithmic scale meter, that would take some experimentation and calibration though!

I don't know where you are based BravoV. Your local equivalent of RS? I'm sure DIgikey must stock them too. There are lots on ebay from Hong Kong (watch the physical size though), glass ones from Bulgaria / Ukraine / Russian Federation too.

[Gyro]

I design electrometers (and other precision analogue stuff) and have to deal with femtoamp current levels everyday at work  . For obvious reasons I won't disclose the details of these circuits but will be happy to help with an advice where I can.

Interesting, thanks Alex. I don't know about advice - I'd just throw in the gems that we haven't thought of yet!     One that tripped me was when I tried shorting the input to null it. Of course when you do that with a transconductance amplifier it overrides the feedback resistor and turns the opamp into a comparator, so the optput hits one of the rails. Then I referred to the Keithley book and realized that an ammeter should always be checked open circuit (but screened).  I suppose you could also argue that feeding into a virtual earth, any thermal emf should create a very high current so the output should hit the rail anyway.

[Alex Nikitin]

Then I referred to the Keithley book and realized that an ammeter should always be checked open circuit (but screened).  I suppose you could also argue that feeding into a virtual earth, any thermal emf should create a very high current so the output should hit the rail anyway.

In your design you can null the offset by connecting the input to the output (-) with a bit of wire. That should remove the input bias component. After that you remove the wire and cap the input. Any output voltage now is the result of the input bias current. Otherwise you may actually introduce some DC offset on the input, compensating for the bias current drop on the 1G resistor! If your LMC662 clean and undamaged, you should see very little difference on the output after removing the wire (and the circuit settles down).

Cheers

Alex

[Gyro]

Ah, I hadn't thought of that, thanks. My DVM is high impedance (not 10M) so the 1k output resistor shouldn't introduce any error (unless my zeners are leakier than I thought).

EDIT: Yes that worked very well (didn't initially, then I realized I hadn't got it turned on!). I was able to null the opamp cleanly to 0uV with the link attached. With the link removed, things obviously get rather noisier but if I take the average offset it looks less than 5uV (5fA bias current through the 1G) and even the peak readings still comes out less than 10fA. I must have got the package reasonably clean then.    Actually, leaving it a bit longer, the noise seems rather more symmetrical around zero.

[Alex Nikitin]

Ah, I hadn't thought of that, thanks. My DVM is high impedance (not 10M) so the 1k output resistor shouldn't introduce any error (unless my zeners are leakier than I thought).

EDIT: Yes that worked very well (didn't initially, then I realized I hadn't got it turned on!). I was able to null the opamp cleanly to 0uV with the link attached. With the link removed, things obviously get rather noisier but if I take the average offset it looks less than 5uV (5fA bias current through the 1G) and even the peak readings still comes out less than 10fA. I must have got the package reasonably clean then.    Actually, leaving it a bit longer, the noise seems rather more symmetrical around zero.

Yes, a good result. With 1G resistor and ~0.5Hz bandwidth you have due to 330pF cap, the noise of the resistor is about 3uV RMS at room temperature so the equivalent current noise is about 3fA RMS. You are near the limit of detection for 1G resistor, well done!

Cheers

Alex

[dreaquil]

Glad to see this thread has come back! I'm in the process of setting up my one. Before I fried one of the opamps (waiting for that to arrive) I measured 0pA input offset at around 350fA with an LMC6462 ( I stuffed up the design by not having the power rails large enoguh for the LMC662). I'll get back to you when I have more results.

For anyone looking for a good feedback resistor I'm using an Ohmic 10G MOX-750231008FE.

[Gyro]

Hi dreaquil, nice to see you back. I was wondering how you were getting on with yours.  I seem to have hijacked your thread a bit in the meantime, sorry. I'll be interested to see how low you can get with your guarded PCB...and the LMC6462. The air wired approach works very well but isn't as duplicatable (new word ) as a PCB.

Yes, a good result. With 1G resistor and ~0.5Hz bandwidth you have due to 330pF cap, the noise of the resistor is about 3uV RMS at room temperature so the equivalent current noise is about 3fA RMS. You are near the limit of detection for 1G resistor, well done!

Many thanks Alex, that's really good to know. I've left it a few hours and I'm hard put now to see much of an offset underneath the noise. It needed a bit of copper tape over the BNC to get there in the end.

Chris

[splin]

Funnily enough whilst reading this thread I just received an email from Analog devices announcing the ADA4530-1 electrometer amp with impressive specs -

  Ib  <1fA typical,  20fA max (-40C to 85C). (Production test 20fA max at 25C)
  Voffset 9uV typical, 70uVmax (0C to 125C)
  Voffset drift .13uV/K typical .5uV/K max.
  Input resistance 100Tohms
  Supply .9mA, 4.5V to 16V


Its a bit noisy at 4uVp-p 0.1 to 10Hz and pricey at $15, 100+ (plus distributer margins).

It also has a built in guard amplifier and esd protection diodes up to 10mA. Don't know when they will be available though.

http://www.analog.com/en/products/amplifiers/operational-amplifiers/ada4530-1.html#product-overview

[EDIT] Added input resistance and supply current
[EDIT] Corrected part number

[Gyro]

That's an impressive looking device! Actually all the application information (also the app note) is as interesting as the device itself. Lots of detail.

Edit: The eval board details too.

The input bias current at 85'C is stunning - the LMP7721 is 900fA at the same temperature. It's a neat trick bringing out the protection diode guard buffer, the 662 and 7721 presumably have one, but only internal. It's surprisingly noisy as you say, the LMP7721 is much quieter whilst having a higher GBP.

Time to start saving the pennies... and for an ultrasonic cleaner that's safe to use with IPA (see app note). I though that was a no-no due to cavitation and flammable fogs.

Edit: Damn, the eval board is $250 and doesn't have anywhere for a protection resistor (inside the feedback loop anyway).

[plesa]

Nice find of LMP alternative.
Ultrasonic is good for dirty stuff, you can clean components by vapor or only wash them and dry them with hot air gun.
So I will rather buy the LMP or the AD dev kit. instead of ultrasonic
Some components like ceramic caps I needs to bake for several days to stabilize their leakage current.
BTW there are other cleaning methods but these are more industrial like cleaning with plasma - this will decompose the fingerprints, flux and oil residues.

[exmadscientist]

Just a few notes on cleaning from an ex-scientist:

When I worked in an clean-room environment, there were three pieces to our cleaning strategy. It's rather "low-tech" and uncontrolled -- this wasn't exactly a semiconductor fab -- but you'd be amazed at how well you can do with some basic stuff:

• Ethanol was the backbone of our cleaning regimen for stuff that just had to be "regular clean". It is significantly more aggressive than isopropanol and goes after grease and organics very effectively.
• We used ultrapure (DI) water for all critical cleaning and rinsing. It's quite aggressive chemically, particularly in going after ionic anything. (I've seen it rust 316 stainless steel, which some poor, misinformed "experts" say cannot ever happen! Granted, it was slow corrosion and in extreme circumstances, but it's possible.)
• Our only detergent was Alconox laboratory powder detergent. It's available on Amazon (among other places) and can clean just about anything off of anything. It's gentle yet powerful. Two warnings: don't ever do laundry with it (it pulls dyes out of fabrics), and don't use it in automatic washers (it foams). There's a low-foaming version called Alcojet, but in my experience Alcojet is significantly less effective and should only be used when Alconox can't be.
• After cleaning with Alconox, we rinsed critical items with ultrapure water until the rinsate conductivity fell to zero on a small handheld conductivity meter. That was usually less than a few minutes even for larger items.
• Ultrasonic cleaners with an Alconox bath were our first choice for cleaning anything small enough and robust enough to go through them. This is how we cleaned a lot of things that had to be seriously clean.
• Adding citric acid to the Alconox bath will clean metal surfaces, particularly copper ones, to just about any degree of cleanliness you could need.

[Alex Nikitin]

Funnily enough whilst reading this thread I just received an email from Analog devices announcing the ADA4530-1 electrometer amp with impressive specs -

  Ib  <1fA typical,  20fA max (-40C to 85C). (Production test 20fA max at 25C)
  Voffset 9uV typical, 70uVmax (0C to 125C)
  Voffset drift .13uV/K typical .5uV/K max.
  Input resistance 100Tohms
  Supply .9mA, 4.5V to 16V


Its a bit noisy at 4uVp-p 0.1 to 10Hz and pricey at $15, 100+ (plus distributer margins).

It also has a built in guard amplifier and esd protection diodes up to 10mA. Don't know when they will be available though.

http://www.analog.com/en/products/amplifiers/operational-amplifiers/ada4530-1.html#product-overview

[EDIT] Added input resistance and supply current
[EDIT] Corrected part number

A very nice chip (speaking from personal experience  ).

Cheers

Alex

[splin]

Funnily enough whilst reading this thread I just received an email from Analog devices announcing the ADA4530-1 electrometer amp with impressive specs -

  Ib  <1fA typical,  20fA max (-40C to 85C). (Production test 20fA max at 25C)
  Voffset 9uV typical, 70uVmax (0C to 125C)
  Voffset drift .13uV/K typical .5uV/K max.
  Input resistance 100Tohms
  Supply .9mA, 4.5V to 16V


Its a bit noisy at 4uVp-p 0.1 to 10Hz and pricey at $15, 100+ (plus distributer margins).

It also has a built in guard amplifier and esd protection diodes up to 10mA. Don't know when they will be available though.

http://www.analog.com/en/products/amplifiers/operational-amplifiers/ada4530-1.html#product-overview

[EDIT] Added input resistance and supply current
[EDIT] Corrected part number

A very nice chip (speaking from personal experience  ).

Cheers

Alex

I was pretty sure you would know all about them 

Presumably you've got engineering samples - any idea when the general public will be able to buy them?

[Gyro]

I was pretty sure you would know all about them 

+1 

Just been reading a bit more... 3 HOURS ultrasonic cleaning with IPA for No-Clean Flux 

[Alex Nikitin]

I was pretty sure you would know all about them 

Presumably you've got engineering samples - any idea when the general public will be able to buy them?

I do have some engineering samples, however I can not give you more information, sorry*.

Cheers

Alex

* P.S. - I will have an opportunity to ask about it next week. If I'll get permission to say more, I will  .

[Gyro]

@exmadscientist,

Did you ever come across a cleaner called "Micro" by International Products Corp. I was given a sample bottle of it years ago (no idea why now) and still got it. Google shows up Micro-90, might be the same stuff because it was probably the '90s when it was given to me! 

It claimed to be very thorough but gentle (apart from removing the oxide layer from Aluminium that is!) and was suitable for standard wash, ultrasonic etc.

[exmadscientist]

@exmadscientist,

Did you ever come across a cleaner called "Micro" by International Products Corp. I was given a sample bottle of it years ago (no idea why now) and still got it. Google shows up Micro-90, might be the same stuff because it was probably the '90s when it was given to me! 

It claimed to be very thorough but gentle (apart from removing the oxide layer from Aluminium that is!) and was suitable for standard wash, ultrasonic etc.

Micro-90 rings a faint bell somewhere in the back of my mind, though that wasn't one of the cleaners we used. It's definitely a good laboratory-grade detergent and probably quite suitable for this type of cleaning. Alconox Inc. claims it's equivalent to their Liquinox product, which sounds about right to me. (It's also probably still fine even if it is from the '90s -- I think one of our Alconox cartons was at least that old, and we still trusted the stuff for critical cleaning.) I wouldn't hesitate to try it.

[dreaquil]

Wow! That ADA4530 is a beast. I'll let you know what levels I get down to. However I have a reed relay on the input also because of the specs of my application. Although it shouldn't introduce too much leakage as it has a minimum insulation resistance of 10^12ohm.

If anyone gets any info on when the ADA4530-1 is coming out please post here to let us know! It sounds like a really great opamp.

[BravoV]

Thanks for sharing  , subscribed.

Made me want to build one, problem is finding that Giga ohm resistor. 

Well, you can build one without a resistor. There is a second way to measure low currents - with an integrator. Replace the resistor with a capacitor and a reset switch, add a comparator and a timer. 10pF capacitor would create a slope of -1V/s with +10pA input current.

Cheers

Alex

So viewing the measurement result thru scope ? Mind elaborate more on the detail circuit, I'm just hobbyist here.

[dreaquil]

@Gyro im getting about 1.4pA of leakage with all connectors attached , reed relay and my test voltage applied (40.96V) since I'm using it as a high impedance tester. On the other hand when I turn off the test voltage and disconnect the input from anything using a seperate reed relay I'm getting leakage on the order of 50fA. The LMC6462 has a typical input bias of 150fA which implies that theres probably some leakage thats cancelling out the bias current

[Gyro]

Micro-90 rings a faint bell somewhere in the back of my mind, though that wasn't one of the cleaners we used. It's definitely a good laboratory-grade detergent and probably quite suitable for this type of cleaning. Alconox Inc. claims it's equivalent to their Liquinox product, which sounds about right to me. (It's also probably still fine even if it is from the '90s -- I think one of our Alconox cartons was at least that old, and we still trusted the stuff for critical cleaning.) I wouldn't hesitate to try it.

Thank you. I'll give it a try on a sacrificial board first. According to the comparison chart on the website it handles Flux, Fingerprints, lubricants, mineral deposits, oxides, scale, rust, Blood etc. Unfortunately it's no good on Proteins, Eggs, Butter, Fruit Stains and Urine. As Blood IS covered I think I can probably live with those shortcomings. 

[Gyro]

@Gyro im getting about 1.4pA of leakage with all connectors attached , reed relay and my test voltage applied (40.96V) since I'm using it as a high impedance tester. On the other hand when I turn off the test voltage and disconnect the input from anything using a seperate reed relay I'm getting leakage on the order of 50fA. The LMC6462 has a typical input bias of 150fA which implies that theres probably some leakage thats cancelling out the bias current

That sounds pretty good for that opamp (at least the opamp isn't dominating the overall leakage), I know you said you had supply voltage issues with the LMC662, the  LMP7721 is spec'd down to 1.8V, but of course it's a single, not dual. Alex's method worked well for discriminating leakage from bias current.

Relay choice is an interesting one, I've seen old Solartron arrays where the coils and poles were actually spaced alongside the reed capsules both for leakage and thermal emf. I don't know if you can still get unencapsulated reed relays where you could maybe squeeze a thin wrap of ptfe tape between the capsule and coil. Hopefully it's getting you within your target anyway!

P.S. Any chance of a picture? 

[dreaquil]

It was actually the fact that the input common mode range is V+-2.3V Which I overlooked! The input common mode on the LMC6462 can take inputs all the way up to the supply voltage. This is a problem because I shifted the non inverting input to +4.096V so that it can swing negatively. Ill get you a picture when I open it up again because its in production line for testing (this is my thesis - a piece of manufacturing test equipment.)

[Gyro]

Ah yes, I understand now. I went back and had another look at your comments and schematic in reply #19. It looks like it's holding up well against your expectations then 

[pmcouto]

Wow! That ADA4530 is a beast. I'll let you know what levels I get down to. However I have a reed relay on the input also because of the specs of my application. Although it shouldn't introduce too much leakage as it has a minimum insulation resistance of 10^12ohm.

If anyone gets any info on when the ADA4530-1 is coming out please post here to let us know! It sounds like a really great opamp.

According to the information on AD site, it seems that the ADA4530-1 is now available:

Product Lifecycle
This product has been released to the market. The data sheet contains all final specifications and operating conditions. For new designs, ADI recommends utilization of these products.

[dreaquil]

Hmm not on digikey or mouser yet though.

[Gyro]

I do have some engineering samples, however I can not give you more information, sorry*.

Cheers

Alex

* P.S. - I will have an opportunity to ask about it next week. If I'll get permission to say more, I will  .

Hi Alex,

I'm not sure if you've had your meeting yet, but I have one that you might want to throw into the mix. If you're making a device with 'ultimate' specs like that, then why stuff it into a bog standard SO8 package? They seem to be making it harder for people trying to use it, with PCB layout and leakage concerns. Reading their datasheet and app note where it starts talking about advanced PCB materials (possibly laminated into a standard FR4 stack) it sounds a lot more expensive for the customer than their added cost in using a better package. Surely they could at least put it into an an S014 or SO16 with part populated leadframe? It's been done before for creepage on offline switchers etc. Is it just lack of imagination? 

Thanks,

Chris

P.S. It can't be a pin compatibility issue because I don't think anything else has the same pinout, the LMP7721 is different.

[Alex Nikitin]

I do have some engineering samples, however I can not give you more information, sorry*.

Cheers

Alex

* P.S. - I will have an opportunity to ask about it next week. If I'll get permission to say more, I will  .

Hi Alex,

I'm not sure if you've had your meeting yet, but I have one that you might want to throw into the mix. If you're making a device with 'ultimate' specs like that, then why stuff it into a bog standard SO8 package? They seem to be making it harder for people trying to use it, with PCB layout and leakage concerns. Reading their datasheet and app note where it starts talking about advanced PCB materials (possibly laminated into a standard FR4 stack) it sounds a lot more expensive for the customer than their added cost in using a better package. Surely they could at least put it into an an S014 or SO16 with part populated leadframe? It's been done before for creepage on offline switchers etc. Is it just lack of imagination? 

Thanks,

Chris

Hi Chris,

The meeting is on Friday. I personally see no big problem with the package if the results are as good as the datasheet shows (and my own experience confirms). I have achieved good results even with the SO-8 version of the LMC662 with it's standard pin-out, if a good board material is used, and with the ADA4530-1 clearances are much better due to the non-standard pin-out.

Cheers

Alex

[Gyro]

Hi Alex,

Thanks. I'm sure you're right in terms of a perfectly executed and cleaned design (which I'm certain yours are) It does raise the question of how good it could have been, ie. how many of the remaining artifacts are due to package / unavoidable PCB spacings vs die, also improved width of guard tracks, more space for via fence in terms of PCB bulk leakage. It would also help with the 'odd spec of dust' during the operating life of the equipment. The guard pins must help a lot though, an improvement on the two unconnected pins in the same position on the 7721 (I know they said they had redesigned pinout to improve performance).  Just a though anyway.

Thanks,

Chris

P.S. A useful side effect would be to give me more space to air-wire it! 

[krivx]

P.S. A useful side effect would be to give me more space to air-wire it! 

If someone manages a PCB on specialty materials this could be a good candidate for a group-buy.

[Gyro]

The Eval board would be an excellent candidate aside from the eye-watering $250 price.

[Alex Nikitin]

It does raise the question of how good it could have been, ie. how many of the remaining artifacts are due to package / unavoidable PCB spacings vs die, also improved width of guard tracks, more space for via fence in terms of PCB bulk leakage.

Well, the graphs in the datasheet show the input leakage below 0.1fA for <60% RH, that is about 600 electrons per second. For my application it is actually useful to have a small SMD package with this kind of performance.

Cheers

Alex

[plesa]

I'm thinking about getting one of these eval boards to make an electrometer, which is one of the really good ways to measure very high value resistors accurately.  Probably, I will use SHV connectors on the "business end" of the device.

Why are you considering SHV and not some triaxial connectors with guard?

[Alex Nikitin]

The Eval board would be an excellent candidate aside from the eye-watering $250 price.

If you look at the docs for the eval board, you will see that this is a special PCB with teflon board material laminated to both sides of an FR4 board.  *THAT* has got to be rather expensive, and there are proportionally not many board houses that can even do that-- so the price reflects that.

Add to that the rather expensive chip, and all the other parts, plus labor, *very* special cleaning schedule, etc. -- and I think you can see that the price of the eval board is actually a bargain!

I'm thinking about getting one of these eval boards to make an electrometer, which is one of the really good ways to measure very high value resistors accurately.  Probably, I will use SHV connectors on the "business end" of the device.

The evaluation board already uses a very good (and very expensive) triaxial connector.

Cheers

Alex

[Gyro]

1000V through 1G ohm (or 100V through 100M) is 1uA. I don't think you are going to need an electrometer for that, it would register on a 50uA moving coil meter.  (Edit:) or more practically, pretty much any opamp, bipolar or Fet.

I think you are a few orders of magnitude out on your measurement voltage and target resistance values. I'm using a 1G ohm as feedback on my Picoammeter and getting 1mV / pA!

P.S. Maybe you should suggest a group buy to them Alex! Glad I had the luxury of using air though.

[plesa]

For the far end of the high-value resistor [not the input to the electrometer amp-- the connector it has is OK already].  If you are applying 1000V to the far end of the high-value resistor [say, 1G-ohm], then a good, safe, HV connector is needed.  On the low-voltage end, the triaxial connector is just fine.  Probably I would follow the output with a gain stage, so I can get 10V out [using a 10M-ohm feedback resistor] with 1000V into the high-value resistor.  [100V is I'm measuring a 100M-ohm resistor].

Oh, I get it:)

[Gyro]

Sure, I understand, but I still think you're making things difficult (and expensive) for yourself. 1T ohms is highest you're likely to get in practice, at 1000V that is 1nA (or 100pA at 100V). A low bias opamp (such as a $1 LMC662 at ~ 3fA) is easily going to give you much more accuracy than you need. It's going to be very difficult to maintain a 1T ohm standard resistor at any sort of precision. For lower resistance values, the opamp bias becomes even more insignificant.

[dreaquil]

Those triax cables are exorbitantly expensive. 120eur for a 1m cable or so. Couldn't do without it or it would render the keithley 6487 useless.

[plesa]

Those triax cables are exorbitantly expensive. 120eur for a 1m cable or so. Couldn't do without it or it would render the keithley 6487 useless.

It depends if you really needs low noise with graphite coated insolation.
Quite good quality cable (including triaxial) are made by Habia.
Belden 9220 yellow triaxial is not low noise but it is also relativelly cheap (Pomona use it for their triax-banana adaptor)
http://www.belden.com/techdatas/english/9222.pdf
For connectors check ebay or other surplus, you can find them quite cheap.

BTW with Keithley 6487 you will receive low noise triax-banana cable.

[dreaquil]

Those triax cables are exorbitantly expensive. 120eur for a 1m cable or so. Couldn't do without it or it would render the keithley 6487 useless.

It depends if you really needs low noise with graphite coated insolation.
Quite good quality cable (including triaxial) are made by Habia.
Belden 9220 yellow triaxial is not low noise but it is also relativelly cheap (Pomona use it for their triax-banana adaptor)
http://www.belden.com/techdatas/english/9222.pdf
For connectors check ebay or other surplus, you can find them quite cheap.

BTW with Keithley 6487 you will receive low noise triax-banana cable.

Unfortunately the cables were missing with the 6487. The pomona one was specifically the one I bought from Mouser. 122Eur excluding shipping or VAT -.- http://eu.mouser.com/ProductDetail/Pomona-Electronics/5342/?qs=sGAEpiMZZMv8kklI404QlRE%2fFt1BM97a

[dreaquil]

Does anyone know of any board connectors i can use which are known to be low leakage which I can use to run some single strange PE wires to my front panel (where my connectors will be). The connectors will make life easier in dismantling the board from the connectors for debugging and troubleshooting.

EDIT: I meant to write PE not PTFE.

[splin]

Does anyone know of any board connectors i can use which are known to be low leakage which I can use to run some single strange PTFE wires to my front panel (where my connectors will be). The connectors will make life easier in dismantling the board from the connectors for debugging and troubleshooting.

How strange exactly?   

How about using PTFE stand-offs with 2mm bullet connectors or similar soldered on top?

[Gyro]

What about Oxley Snaplox test point and connector....

http://www.oxleygroup.com/images/newproductspdf/126.pdf

[dreaquil]

I meant to write PE insulated wires not PTFE. Those Oxley Snaplocks look ideal, but I'm not sure ill be able to order them because I can't order from farnell and my needs are very low volume (20 max?). I came across these though http://www.digikey.com/product-search/en?keywords=182596000
They look pretty good with a 5mm pitch for higher insulation resistance and it looks like they have an air gap built in.

[edavid]

Does anyone know of any board connectors i can use which are known to be low leakage which I can use to run some single strange PE wires to my front panel (where my connectors will be). The connectors will make life easier in dismantling the board from the connectors for debugging and troubleshooting.

If they are single wires, how can the connector type have an effect on the leakage?

[Gyro]

What about using a single berg pin type female connector?  The sort they use for those cheap jumper leads to plug onto Arduino headers. As edavid says, if it's a single wire, there's no leakage unless something touches the connector or the wire. The only issue I can see is persuading a single berg pin to stay vertical while you solder it!

eg. http://www.ebay.co.uk/itm/Jumper-Leads-Cables-Pin-type-Female-to-female-20cm-length-40-wires-/301112648485?hash=item461bb66325:g:Ki4AAOSwA4dWI7qW

[Gyro]

Probably one for Alex...

I've been playing around with spark gap input protection for ESD (at some point my 1M series resistor is going to flash-over). I've tested a few telecom type spark gap tubes, the ceramic ones are better than the glass ones (probably less humidity prone) but they come in at 5pA+ after cleaning. I have a 1kV one that came from an old Solartron meter, but it's still single digit pA. I tried a small neon bulb, but higher too (anything glass is bad). My best option at the moment seems to be to make a small air-gap with a wire from the ground tag. Is it normal to include spark gap input protection on picoammeters?

Edit: Leakage tests were carried out at 10V

[Marco]

at some point my 1M series resistor is going to flash-over

Use 10 10kV (VR37) 100k resistors then ... your feed through will be a good enough spark gap at that point.

[Gyro]

Thanks Marco, unfortunately not an option due to size constraints (see photo in reply #23). I used the biggest body size I could (2W carbon film). The input connector is also an extended insulation one for low surface leakage.

I'm probably being over-cautious as resistor and op-amp input protection diodes are good for at least 500V (resistor voltage spec), probably double that in practice. I've also managed to bore out the center of a BNC connector shell to clear my input insulator and provide a total ground shield when not in use. I may implement an air gap to break down at around 1kV.

Edit: ... It looks as if anything with a package is going to compromise my (currently single digit fA) resolution.

[SeanB]

Tried soldering in a shorting wire and then snipping it with some clippers to leave 2 sharp points almost touching. 1mm gap will arc over at around 1kV, and a pair of sharp points will make the initiation easier.

[Alex Nikitin]

Is it normal to include spark gap input protection on picoammeters?

No, usually a couple of low leakage diodes to ground on the input after a large resistor is enough.

Cheers

Alex

[Gyro]

Many thanks guys.

Sean, that's the sort of thing I was thinking, I have the solder tag of the ground terminal as a fairly rigid mounting so it should be reasonably easy to maintain a small gap.

I was briefly concerned about corona / ionization current leakage across the gap until I remembered that the input is a virtual ground node, the only time it will see voltage is if the test item is shorted (must remember to include a current limit resistor in case the gap triggers).

I'm probably going overboard in that case Alex, it is a bit belt and braces but I'm not sure if I've been particularly lucky with this LM662 sample and don't want to subtly degrade it with multiple strikes.

Chris

[plesa]

For basic input protection as Alex pointed use only low leakage diodes (1N3595  or DPA1 or low leakage  JFET) and if you are going to connect voltage above 200V create additional box with resistor in series.

[Gyro]

The low leakage diodes are already taken care of - they are the bootstrapped protection diodes within in the LMC662 itself, they're capable of sustaining 5mA max. With the 1M series resistor I'm only asking them to handle 500uA at 500V input, I can't see the resistor actually breaking down until >1kV. You can see my schematic and photos in reply #23. The 1M protection resistor is within the feedback loop to avoid parasitic effects on input impedance (it's a transconductance amp). Note that the 1G ohm feedback resistor is also rated at 500V. I think I'm covered pretty well - the spark gap is mainly precautionary, I just need it to break down before the resistors in case of a major zap (which I ought to be careful enough to avoid!)

[Gyro]

Zero leakage spark gap added... and yes it is more stable than it looks! 

The gap is similar to the 1-2kV sawn spark gap protector from my junk box (came from a CRT base PCB).

[3roomlab]

i have a noob question, after seeing this thread and this http://www.vk2zay.net/article/251
yates' version drives a LCD readout, i was wondering if this a good way to buffer the LMC662 to drive something else like a cheap LED readout sprawled all over ebay?

[dom0]

If you want to filter incoming noise from the meter inputs I'd say a T-filter network is a better choice. Assuming, of course, there is such noise, and that it affects the amplifier.

[3roomlab]

im not sure about noise, i just happen to have those standalone 5 digit LED volt display. so i am thinking of the unit as stand alone w/o requiring a DMM to read it. the input impedance is few xxxk so i dont think i can connect direct to the LMC

[dreaquil]

You should be able to hook it up directly. The LMC662 is specified for 2k and 600ohm loads. It also has a minimum output current of 14mA for the AI version so it should be just fine.

[Gyro]

The advantage of using a DMM is that you get range switching - you can sense to just over +/-4nA (+/-4V out) using the 1G feedback resistor and with a fresh 9V battery. With a 5 digit display however that isn't as much of an issue.

The LM662 should drive that sort of load without any issues. Remember though that the output is inverting (-ve for positive input) and it is biased mid way between the battery supply rails (mid rail bias is provided by the other half of the 662) so you need to make sure that the DVM module can handle that if you're planning to use the same supply rails as the 662.

[Tartan5]

Hi everyone,

I am looking to make a board that includes multiple (50 or more) TIA amplifiers.  I'd like to be able to resolve 0.1 nA DC.  How well would the LMC662 work from two linear regulators (noise wise) vs a battery?

Thanks!

[Gyro]

Hi Tartan5,

It should work quite well. I found that with battery operation the input offset voltage does shift over the battery voltage range (mind you I'm talking a 2:1 range there), which regulated supplies would prevent.

Yes, you will need to be careful of noise, but the supply current will be low so you can easily apply good filtering. Also be careful to avoid ground loops.

[bdivi]

Gyro,

What was the reason to use split supply instead of single ?

LMC662 datasheet specifies that the input should work to ground and the output is rail to rail.

Thanks,

[Gyro]

Hi,
The split rail is so that I don't have to worry about input polarity. It will do +/- 4.5nA full scale on a fresh battery (+/- 2.5nA by change time). On a picoammeter with BNC input connector it's not a simple as swapping the input leads, the outer is effectively ground.

[bdivi]

I understand - thank you.

Regarding the split supply in your schematics. Is it the virtual ground buffer with resistor in the feedback and capacitor to ground ? The way it is now it is some sort of an integrator and does not make sense to me.

[Gyro]

They are there (together with the resistor on the output) for stability, the rail splitter half of the LMC662 has a tendency to oscillate without them, it is basically a voltage follower at DC though.

[zlymex]

I made two similar ones before, battery powered, split supply and air wiring at input pin of the opamp. I use 100G and 1T as feedback resistors to achieve lower current noise. Also, I use a piece of twisted PTFE insulated wire pair as the feedback capacitor.
I tried different opamps and different resistors, they are not all good.

[mmagin]

zlymex: What is this cute MX-6.5 you have there?  Homebrew or commercial?

[zlymex]

zlymex: What is this cute MX-6.5 you have there?  Homebrew or commercial?

Homebrewed hand held 6 and half digits.
http://bbs.38hot.net/thread-4594-1-1.html

[dannyf]

I have done quite a few experiments with a Pico ammeter built around cmos opamps, in topology almost identical what what's shown earlier. If you do a search you will see those posts. Mine was inspired by a keithley meter.

The biggest lessons I learned is that the use of a t-network, often seen in a mature and academic papers, is a crime in building those meters.

I ended up constructing a Na meter around a tl0x4 chip, as an adapter to a digital multimeter with relative measurement capabilities.

I think I tried quite a few of lose cmos opamps mentioned here with fa-level bias current.

[dannyf]

"The LMC662 in dip pkg looks a very good bet for avoiding a PCB and associated leakage,"

Unless you have access to super duper PCB with looooow leakage, point to point wiring is the only way to go. I used some aerospace grade wires in my built.

The most critical component in this type of build is that resistor. Very difficult to get a good resistor into the T or Gohm territory.

[mmagin]

zlymex: What is this cute MX-6.5 you have there?  Homebrew or commercial?

Homebrewed hand held 6 and half digits.
http://bbs.38hot.net/thread-4594-1-1.html

Cool.
May I ask what voltage reference and ADC you used?

[zlymex]

Cool.
May I ask what voltage reference and ADC you used?

Vref: ADR441
ADC: LTC2440

[bdivi]

I built the Gyro picometer myself and I am very impressed with the performance.

Once nulled and quiet it stays close to 0.000 mV and only moves +/- 3 uV which is 3 fA 

100mV with 100MOhm resistor measures 1000.2 pA - again pretty nice for 1% 1GOhm feedback resistor.

Very happy with this project

[David Hess]

Back when I was doing this, we graded LMC6081s for low input current and I do no remember finding any which were above 1 picoamp.  Using the DIP package, we bent the inverting input straight out and air wired it.

There was a Burr-Brown application note about a picoamp current inverter which could have been used as the basis of a picoamp input transimpedance amplifier; notably it used low leakage diodes in place of the high value feedback resistors.  In his book, Bob Pease described a similar circuit for making picoamp measurements; in his case, the logarithmic output was a feature.

Good 10 megohm input digital voltmeters can achieve 10 picoamp resolution or better.

[Tartan5]

I made two similar ones before, battery powered, split supply and air wiring at input pin of the opamp. I use 100G and 1T as feedback resistors to achieve lower current noise. Also, I use a piece of twisted PTFE insulated wire pair as the feedback capacitor.
I tried different opamps and different resistors, they are not all good.

Hi Zlymex,

Could you list what feedback resistors (mfg pn) you were testing?  I am trying to find a good source of low noise high resistance resistors to use for the feedback resistor.  I've heard that the thick film resistors tend to have quite a bit of inherent noise...

All:  Overall, what style of resistor seems to perform best for this application?

Thanks!
Tartan5

[zlymex]

Hi Zlymex,

Could you list what feedback resistors (mfg pn) you were testing?  I am trying to find a good source of low noise high resistance resistors to use for the feedback resistor.  I've heard that the thick film resistors tend to have quite a bit of inherent noise...

All:  Overall, what style of resistor seems to perform best for this application?

Thanks!
Tartan5

Top two in the first photo are glass-vacuum type, the one I tested are average. The quality may depend on manufactures a lot.
Next two(red) are film resistors made in China, good quality, I bought twice, they are my main resistors for week current measurement
The dark brown one is made in Japan, perform good in high voltage but very poor at low voltage as exhibit very large DA(Dielectric Absorption) effect
Bottom two are very bad, too much voltage co-efficient

I also tested some Ohmite resistors such as mini-mox(mox1125-23) and slim-mox(102E and 204E), good for this purpose.

Photo only show my 100G resistors, I got other decade values(100M to 1T) but looks similar in appearance.

Edit: Even the performance from the same batch is different, this especially true for high resistors such as 100G and 1T. For value 10G or below, it is not critical.

[bdivi]

The Ohmite 104E I used is also very good - 1GOhm, 1% tollerance (happened to be better than 0.2%), TCR 25ppm/C, 0.25% voltage coefficient to full voltage rating and 0.25% humidity variation.

http://www.newark.com/ohmite/sm104031007fe/thick-film-resistor-1gohm-1-5w/dp/66K6874?ost=SM104031007FE&selectedCategoryId=&categoryNameResp=All%2BCategories&iscrfnonsku=false

[Tartan5]

Thanks Guys.  I'll get a few different ones on order.  I plan on laying out a few different op amp / resistor configurations on a test PCB.  I am going to try some surface mount feedback resistors with guarding as well.....

Thanks!
Tartan5

[Tartan5]

Quick question..... wouldn't the thermal noise of the 1G or 10G resistor swamp out the current reading?  I know I can LPF the output, but what if I wanted to sample at 10 hz or so?

Thanks!

[Tartan5]

1G @ 10hz = 13fA rms resistor noise, approximately? That would be not including current and 1/f noise, correct?

Thanks

[Alex Nikitin]

1G @ 10hz = 13fA rms resistor noise, approximately? That would be not including current and 1/f noise, correct?

Thanks

If you sample at 10Hz the bandwidth should be about 5Hz and RMS thermal noise current of the resistor would be about 9fA at room temperature. To reduce it you need to use a larger resistance. A 100G resistor would give you ~ 0.9fA RMS noise. Amplifier noise components are added to that figure, though good electrometer-grade opamps add very little.

Cheers

Alex

[dannyf]

"wouldn't the thermal noise of the 1G or 10G resistor swamp out the current reading? "

Current noise vsm voltage noise.

The academic advise of using a T networks consisting of loww value resistors here is absolutely the wrong approach. The bigger resistor value, the better.

[pigeon]

I made two similar ones before, battery powered, split supply and air wiring at input pin of the opamp. I use 100G and 1T as feedback resistors to achieve lower current noise. Also, I use a piece of twisted PTFE insulated wire pair as the feedback capacitor.
I tried different opamps and different resistors, they are not all good.

Is the top op-amp circuit a constant current generator?

[Gyro]

Welcome pigeon,

No it's a rail splitter (basically a voltage follower with the input driven from the supplies by two equal value resistors). It sets the ground reference of the circuit to be 50% of the supply so that the opamp can swing positive and negative. This allows it to respond to both positive and negative input currents.

[bdivi]

Welcome pigeon,

No it's a rail splitter (basically a voltage follower with the input driven from the supplies by two equal value resistors). It sets the ground reference of the circuit to be 50% of the supply so that the opamp can swing positive and negative. This allows it to respond to both positive and negative input currents.

Only the resistor and capacitor positions must be swapped - resistor as a feedback and capacitor as a filter to ground.

[Cerebus]

Welcome pigeon,

No it's a rail splitter (basically a voltage follower with the input driven from the supplies by two equal value resistors). It sets the ground reference of the circuit to be 50% of the supply so that the opamp can swing positive and negative. This allows it to respond to both positive and negative input currents.

Only the resistor and capacitor positions must be swapped - resistor as a feedback and capacitor as a filter to ground.

Eh? It's setting the DC ground so why would you want to only AC couple it to ground? If I remember the circuit correctly, the capacitor you're talking about is there to cut the op amp bandwidth. Or have I misunderstood - you've not exactly made it clear what you're referring to?

[bdivi]

That is what I mean.

[Dave]

No, just no.

[iromero]

That is what I mean.

Look carefully, the 100 ohm and 1M resistors are actually in series going to the inverting input of the amplifier, forming the DC negative feedback, it's a bit confusing because both go to "ground", but ground is the output of that amplifier.

[bdivi]

And of course the simpler follower

[Gyro]

Ah, now I understand the confusion, pigeon quoted zlymex's post which also contained a (different) schematic.

As Cerebus and iromero point out there needs to be a DC feedback path. The output of the opamp is driving the 'ground' via the 100R to buffer it from capacitance involved. The opamp is sensing the 'ground' voltage via the 1M to establish the DC return path of the voltage follower. The 10n capacitor provides local HF feedback around the opamp for stability.

I found these component values to be needed for stability in my implementation. The LMC662 was 'twitchy' in driving the ground (low frequency instability). I suspect that it is because 'ground' in my case was the copper groundplane of the board. At that point I didn't want to significantly change the layout so I went with stabilisation values that worked solidly.

The circuit is however a simple voltage follower at DC, with the output at half-rail with fine trim to cancel the opamp input offset voltages. The 1M resistor also matches the one on the inverting input of the 'measuring' opamp which hopefully also minimises bias current offsets (this does seem to work as output offset at the uV level is maintained pretty well over normal ambient variations).

P.S. @bdivi, I hadn't noticed your post with the implementation of the "Gyro picometer" (undeserved fame at last! ). It looks a neat implementation and it's good to see that it performs so well.

[Cerebus]

The LMC662 was 'twitchy' in driving the ground (low frequency instability). I suspect that it is because 'ground' in my case was the copper groundplane of the board.

I'd guess that 'ground' was one plate of a 10-20 pF capacitor depending on the size of your board.

I think Gyro's mistake was missing that 'ground' was just another wire/net/node from the electrons' point of view. I find it helpful in schematics for this kind of setup to use a symbol* for an equipotential point that isn't the classic earth or chassis symbol and separately show a direct connection to ground from that symbol to reinforce the point that there is nothing magical about the 'ground' connection, that it's just another node.

* Can't be bothered to generate and upload a picture. I usually use the inverted hollow triangle with a single letter designator (A and D being the most common).

[Gyro]

I think Gyro's mistake was missing that 'ground' was just another wire/net/node from the electrons' point of view.

Hmm, maybe, maybe not. In this case that node is 'ground', It is connected to the board copper plane, the BNC connector outer and the metal case itself. In this case the ground symbol is rather appropriate. It's the battery terminals that are being 'positioned' by the opamp to span each side 'ground'. There shouldn't be any assumption that either of the battery terminals is a ground reference.

EDIT: On reflection a 'chassis' symbol would be more appropriate but I don't think that would be any less confusing than the ground symbol. It is still the node that will be grounded externally (via the ground terminal) during use.

Edit 2: Just looking back at the schematic, I did actually use the chassis symbol. 

[splin]

1G @ 10hz = 13fA rms resistor noise, approximately? That would be not including current and 1/f noise, correct?

Thanks

If you sample at 10Hz the bandwidth should be about 5Hz and RMS thermal noise current of the resistor would be about 9fA at room temperature. To reduce it you need to use a larger resistance. A 100G resistor would give you ~ 0.9fA RMS noise. Amplifier noise components are added to that figure, though good electrometer-grade opamps add very little.

Cheers

Alex

The noise bandwidth will be determined by the anti-alias filter; without one it will be the bandwidth of the amplifier and the analogue bandwidth of the ADC - which for a 1MSPS convertor (not uncommon on a micro-controller) could be 10, 20 or even 30MHz+. (All the noise above the Nyquist rate will fold down into the passband). By sampling much faster than 10Hz, a simpler anti-alias filter can be used with a digital low pass filter removing noise above 5Hz.

[Vtile]

Can someone explain for a newb why series connection of lower value resistors are nogo?? I Something related to noise?? Dielectric absorption is too high because of the "end cap" of leads and there is a series of voltage sources because of DA??

I happen to have a pair of OPA111AMs from B&B (salvaged board). Are they any good for this kind of job? Electronics aren't my area of specialty so I stuggle with the datasheets, but to me it seems that the low as possible input bias current is what you are after and what matters here and since that current is for OPA111AM a pA range (sub) it will be nogo, since it should be on femtoAmp range? Or is it just matter of complexity of balancing the input that goes beyond DIY.

[David Hess]

The OPA111 is a great choice down into the picoamp range but its performance will be limited by its change in bias current with temperature which doubles roughly every 10C.

[Gyro]

Can someone explain for a newb why series connection of lower value resistors are nogo?? I Something related to noise?? Dielectric absorption is too high because of the "end cap" of leads and there is a series of voltage sources because of DA??

I can't immediately see where the series connection being 'nogo' reference comes from. Series connection is quite often used in high voltage dividers, it spreads a voltage drop across multiple packages, increasing breakdown voltage and it can actually reduce shunt capacitance caused by the packages (both effects really coming from you effectively building a physically longer resistor). Of course the devil is in the detail - how you construct the chain, stray leakages if you construct on a PCB etc. I don't think dielectric absorbtion comes into the equation though.

[Vtile]

Can someone explain for a newb why series connection of lower value resistors are nogo?? I Something related to noise?? Dielectric absorption is too high because of the "end cap" of leads and there is a series of voltage sources because of DA??

I can't immediately see where the series connection being 'nogo' reference comes from. Series connection is quite often used in high voltage dividers, it spreads a voltage drop across multiple packages, increasing breakdown voltage and it can actually reduce shunt capacitance caused by the packages (both effects really coming from you effectively building a physically longer resistor). Of course the devil is in the detail - how you construct the chain, stray leakages if you construct on a PCB etc. I don't think dielectric absorbtion comes into the equation though.

I think I misunderstood somepoint of this thread, now I rereaded it through I don't know where I got such idea, but the subject goes over my (currently)head. Maybe I mimick this design and try to salvage those ($50 how on earth they are so expensive) OPA111s from that board and test if I can get to under nA range (without anykind of true accuracy). 

I made a quick calculations and it seems that Gohm range of resistor the price is pretty constant 35 to 55 euros per 10GOhm so the series idea wouldn't be a such economically practical solution anyway even with 1000 SMDs 

For mechanical construction firts thing that came to my mind were PTFE poles and zigzag air network of resistors between them.

[Cerebus]

The OPA111 is a great choice down into the picoamp range but its performance will be limited by its change in bias current with temperature which doubles roughly every 10C.

That's going to be an issue with anything with a JFET input. The only way to avoid that is to go with MOSFETs and down in the pA ranges the lower noise of a JFET is going to be a winner every time compared to MOSFETs, plus there's a tendency for MOSFET inputs to have a higher 1/f noise corner.

For instance, OPA111 (JFET, 800 fA typical bias) 8 nV/sqrt(Hz) @ 1 kHz, LMC662 (MOSFET, 2fA typical bias!) 22 nv/sqrt(Hz) @ 1 kHz. The LMC662 data sheet doesn't tabulate any noise figures other than @ 1kHz and the graphs show a very poorly defined 1/f corner such that noise is still falling at 10 kHz, the OPA111 has a well defined, if high, noise corner at around 1kHz (The more modern OPA627 has a much more respectable 100 Hz 1/f noise corner).

[mmagin]

I made a quick calculations and it seems that Gohm range of resistor the price is pretty constant 35 to 55 euros per 10GOhm so the series idea wouldn't be a such economically practical solution anyway even with 1000 SMDs 

Wow, that's expensive.  At least here in the US, mouser has 10G maxi-mox for under $6:
http://www.mouser.com/ProductDetail/Ohmite/MOX1125231008FE/?qs=sGAEpiMZZMtlubZbdhIBIC2FpSEBOOMsPx9T7amSmNA%3d

[Gyro]

You might want to try ebay (unless there is a major shipping charge hike for Finland):

http://www.ebay.co.uk/sch/General-Purpose-Resistors/181912/i.html?_from=R40&_nkw=10G+ohm

(or similar search strings)

[orolo]

I've been toying with the idea of building this circuit for years-- even have the components stored somewhere. One improvement I was thinking about was changing the mongo resistor for a tee network. Would it be advisable in this case?

[David Hess]

The OPA111 is a great choice down into the picoamp range but its performance will be limited by its change in bias current with temperature which doubles roughly every 10C.

That's going to be an issue with anything with a JFET input. The only way to avoid that is to go with MOSFETs and down in the pA ranges the lower noise of a JFET is going to be a winner every time compared to MOSFETs, plus there's a tendency for MOSFET inputs to have a higher 1/f noise corner.

For instance, OPA111 (JFET, 800 fA typical bias) 8 nV/sqrt(Hz) @ 1 kHz, LMC662 (MOSFET, 2fA typical bias!) 22 nv/sqrt(Hz) @ 1 kHz. The LMC662 data sheet doesn't tabulate any noise figures other than @ 1kHz and the graphs show a very poorly defined 1/f corner such that noise is still falling at 10 kHz, the OPA111 has a well defined, if high, noise corner at around 1kHz (The more modern OPA627 has a much more respectable 100 Hz 1/f noise corner).

That is exactly why I have used graded LMC6081s (practically the same as the LMC662 but with better precision) in the past when I needed lower input bias current than a JFET input although there *are* some JFET input operational amplifiers like the AD549 with a bias current significantly below that of the OPA111.

At high temperatures, bipolar input operational amplifiers become competitive.

[Vtile]

So as newb with a straight thinking, without much considerations of realities.
A deep frozen MOSFET input is the best? (in reality moisture and temp control would be a nightmare)
Is the time driven through those classic discrete multistage darlington and FET-BJT hybrids etc. amplifier designs, when measuring nothingness.

For my calculation of the cost of gigaohm resistors I used the easiest source for me (as non-professional, non institutional, non commercial, non-corporate buyer) it is farnel.uk (since my local electron shed no more takes custom orders to other as used to), so it is obviously not universally applicable. I'm going through my buckets of old stuff I hoarded as a kid, but I think I'm lucky if I find anything over 1GOhm and the condition of such parts is random.

[Kleinstein]

The T type network adds extra resistor noise. The than lower resistor has a higher current noise. As the noise voltage of those ultra low bias OPs is not so low, there is very little advantage to have a high signal level at the output. A simple (low noise is easy at low impedance) post amplification can do the same.

There is the option to use a charge amplifier instead of a classical TIA instead. This saves the high value resistor, but needs a way to reset the capacitor, without adding too much leakage. It can work very well at the low end, but is not easy either. One variation is using a photo-diode to compensate the current. Low leakage diodes might be easier to get and more stable than a 100 GOhms resistors.

Cooling MOSFETs is often tricky, they tend so to show quite some shift in offset / operating point and usually don't work so well at really low temperatures (e.g < -50C). If you really need to cool, JFETs may be more attractive - they are said to work reasonable to low temperature. Doubling the current every 5K also works when going down.

Humidity can be a night mare with low temperatures. I have not seen cooled amplifiers for pA range currents, except when the whole system is in vacuum (e.g. in vacuum tunneling microscope).

 Often a temperature like 10 K above room temperature is a good compromise: not too much semiconductor leakage and yet reduced relative humidity (< 50%) to get low surface leakage.

[orolo]

The T type network adds extra resistor noise. The than lower resistor has a higher current noise. As the noise voltage of those ultra low bias OPs is not so low, there is very little advantage to have a high signal level at the output. A simple (low noise is easy at low impedance) post amplification can do the same.

Thanks. I didn't think about the noise problem. I also noticed that the input offset voltage of the LMC662 will cause a large voltage offset at the output with the ground referenced T network. That can be calibrated away but is a major nuisance.

The offset problem reappears if the 10G resistor is sacrified for, say, a 10MEG one (10uV/pA) and then a precision low noise amplifier x1000 is added. The offset voltage of an LMC662 is 1mV typ, so a typical 1V offset with 1.3mV/ºC drift should be expected.

[Kleinstein]

The T-network usually only makes sense for a little divider, like maybe 1:10 or less. And it has its limitations.

As a rule of thumb it is a good idea if the current measured causes a voltage drop of at least 30 mV over the feedback resistor. If the current is from a source with shot noise (like a photo-diode) at this point the shot noise (charge quantization and counting) is about as large the resistor intrinsic noise. So a 10 M resistor is good for the nA range.

[David Hess]

The Analog Devices ADA4530-1 datasheet has a great discussion about noise in transimpedance amplifiers which use very high value feedback resistors including this gem:

Shot noise calculations are appropriate only for some legacy JFET-based electrometer amplifiers, where only a single junction is connected to the amplifier input pins. Modern high impedance amplifiers have several semiconductor junctions connected to the amplifier input pins. The most significant of these junctions are the ESD diode structures. The input bias currents are equal to the sum of these diode currents. The diode currents are designed to cancel each other, but the shot noise currents are uncorrelated and cannot cancel, which, in turn, makes calculating the shot noise from the input bias current impossible.

[EmmanuelFaure]

The T type network adds extra resistor noise. The than lower resistor has a higher current noise. As the noise voltage of those ultra low bias OPs is not so low, there is very little advantage to have a high signal level at the output. A simple (low noise is easy at low impedance) post amplification can do the same.

+1

With a transimpedance amplifier, the sensivity is a function of R, and the noise (dominated by R) is a function of sqrt(R). In consequence the signal-to-noise ratio is a function of sqrt(R). The bigger R, the better.

[razberik]

Here is my implementation of PAM. It is nearly the same that Gyro did. I use 1Gohm resistor Ohmite SM108031007FE. Styrene cap is 47pF. I didnt have anything else. New from sealed bag, wasnt wash with IPA. I used gloves while manipulating with sensitive parts. Resistors, SMA PTFE connectors were washed.
Didnt have time to evaluate performance, but it seems to be possible to zero.

[Kalvin]

Not sure if this has already been posted, but here's a nice article by Bob Pease about the transimpedance amplifiers ie. current-to-voltage converters and how to optimize the performance by a few external components:

http://electronicdesign.com/analog/whats-all-transimpedance-amplifier-stuff-anyhow-part-1

[razberik]

I have played around with my PAM little bit. It is possible to zero it.
I have built source box which contain 1.5V battery, 100G resistor and PTFE BNC connector. When I connect my PAM and this source with 30cm PTFE cable, I receive about 15.8pA. The noise is about 10-20fA. Is this a real number ? I really have to admit that I wasnt 100% precious while handling sensitive parts, but I tried my best and did handle with tweezers only and wash the parts with IPA.

I have 1G and 47pF, which gave me BW = 3.37Hz.
What way Gyro found out the 330pF (0.5Hz BW) ?
How do I determine the bandwidth I need ?

[Alex Nikitin]

I have played around with my PAM little bit. It is possible to zero it.
I have built source box which contain 1.5V battery, 100G resistor and PTFE BNC connector. When I connect my PAM and this source with 30cm PTFE cable, I receive about 15.8pA. The noise is about 10-20fA. Is this a real number ? I really have to admit that I wasnt 100% precious while handling sensitive parts, but I tried my best and did handle with tweezers only and wash the parts with IPA.

I have 1G and 47pF, which gave me BW = 3.37Hz.
What way Gyro found out the 330pF (0.5Hz BW) ?
How do I determine the bandwidth I need ?

For 1G resistor and 3.37Hz BW the noise (at 25C) should be around 7.5fA RMS or roughly ~50fA p-p. I don't know how you are measuring these "10-20fA" but it is at least in a correct order of magnitude.

Cheers

Alex

[Gyro]

What way Gyro found out the 330pF (0.5Hz BW) ?
How do I determine the bandwidth I need ?

Nice recycling. 

I picked 330pF as 0.5Hz seemed a reasonable round number without going for an overly large Polystyrene cap. It was also the one in my collection with the longest length of fused polystyrene at each end of the winding. Not very scientific I know. I'm not sure if there is a 'right' bandwidth as it is application dependent. 0.5Hz obviously yields a lower noise figure, as below (from earller in the thread):

I was able to null the opamp cleanly to 0uV with the link attached. With the link removed, things obviously get rather noisier but if I take the average offset it looks less than 5uV (5fA bias current through the 1G) and even the peak readings still comes out less than 10fA. I must have got the package reasonably clean then.    Actually, leaving it a bit longer, the noise seems rather more symmetrical around zero.

Yes, a good result. With 1G resistor and ~0.5Hz bandwidth you have due to 330pF cap, the noise of the resistor is about 3uV RMS at room temperature so the equivalent current noise is about 3fA RMS. You are near the limit of detection for 1G resistor, ...

[fcb]

What a great thread.  Nice designs and great execution.

[(*steve*)]

I have what I think is a silly question.  But I'm not sure why it is silly.

One of the schematics includes something like this to generate both the split rail and the bias to the other op amp.

Imagine my first attachment here.

My concern is twofold:

1) first, the adjustment range appears to be only +/- 0.5mV even though the datasheet specifies that the LMC662 can have up to 5mV input offset. (is this somehow related to using the other op-amp in the same package, assuming they're well matched, and the same error will be present in the split ground rail, so this only nulls out the difference?)

2) won't the voltage across the 2 15R resistors vary significantly as the battery ages, causing the difference between the ground voltage and the bias voltage to change?  Or is this so dependent on other factors that you need to set up the bias frequently anyway?

My thought was to use a diode drop in place of the two 15R resistors.  This should remain relatively constant as the battery voltage changes (I'm guessing at about 0.4V at around 20 to 40uA)

Imagine the second image here

Would this result in more noise?  Would capacitors across the 300k resistors help?

(this is the first time I've uploaded images, so I'm not sure if or where they'll appear)

[Gyro]

Ah yes, that was mine I think.

Not a silly question, undoubtedly the network could probably be improved with a bit more effort.

Regarding the limited trim range, no it doesn't cover the full worst case offset range of the LMC622, however in practice it was fine for the 'typical' offset of the sample I used (I had measure its offset at the start anyway). If neccessary the component values could be tweaked for greater span but I was shooting for maximum null resolution on the pot. If I was going for volume production I would have expanded the range but would probably screen the LMC662s too, as the ones at the offset limits might possibly be less than ideal in other parameters too (input current etc.). Yes, there probably is an element of offset tracking between the two opamps in the package too (not tested).

In practice the offset nulling tracks the opamp offset pretty well as the supply voltage declines. I haven't found any need adjust offset trim. My unit is still on it's original battery and without any intermediate trimming, is still zero to within the 0.1mV digit on the 200mV range of a normal handheld DMM. There don't appear to be any noticable temperature drift effects at this resolution over normal ambient.

I did consider adding a 5V micropower LDO, which would remove pretty much any battery voltage related offset drift, but I haven't got around to it and given the stability over the last two years, probably won't bother now (maybe a lucky opamp sample).

Looking at your revised circuit....

- The diode is another active component with it's own TC, hard to know how things would track or not. Also the diode would be operating at a very low current (less than approx 30uA on a full battery), where its forward voltage and stability might not be that predictable, sample to sample (hard to say off hand).

- Yes, the input to the opamp will  be seeing a much higher source impedance ~165k vs 7.5R (once the closest decoupling capacitors are taken into account), again hard to predict what effect this might, or might not, have without prototyping it. Maybe just a little bit noisier, extra capacitors would probably be unnecessary and might add stray leakage paths. The input reistance of the LMC662 is so high that it probably wouldn't make any difference in practice.

- More components, not an issue if it's 'better'.


As I mentioned above, adding supply regulation is probably the most effective way of removing any offset trim issues, at the expense of slightly less battery life and maximum current measurement.

Why not have a play with offset compensation networks for yourself - LMC662s (or another generic cmos opamp) are really cheap and you don't need the 1G feedback resistor if you are only interested in offset compensation experiments.

[razberik]

Here it goes, another version of Gyro's picoammeter.

Difference is use of ADA4530-1 and feedback resistor is 100Gigaohm RX1M-1009FE. Compensation is 10pF styrene cap. Theoretically 159mHz bandwidth.
Unfortunately I cannot disclose more details since it is a part of some test/development jigs in my work.
I attach picture of it.

I made some noise measurement. I have a jig which generates low current using 1.5V battery + 1TeraOhm resistor. See simplified schematic (omitted capacitors and ... everything).
When I made the picoammeter I made 1hour measurement. Then I went on vacation for 1 week and when I got back I ran once again confirmation measurement.
There is a theoretical nominal current which is calculated from real battery voltage 1.58703V and 943GOhm measured resistor (claimed to be 1T).
Current is calculated with real transimpedance of 99.08Gigaohm. Output voltage is read by 34401A in fast 6 digit mode and logged into Excel.

[Gyro]

Nice implementation and stability. 

[approxime]

Hi all,

Does 8V2 mean two oppositely oriented Zener diodes? If so, is it possible to use 1N3595 instead as somebody has proposed?
Thank you!

[magic]

Looks like output overvoltage protection. It needs to be a pair of zeners connected as drawn or a biderectional transil/TVS.

[mark03]

Does 8V2 mean two oppositely oriented Zener diodes? If so, is it possible to use 1N3595 instead as somebody has proposed?

No, it just means "8.2V" i.e. a zener diode.  The reason there are two of them is that the diagram shows two.  This is a common engineering notation where the letter takes the place of the decimal point.  It is commonly used for labeling supply voltages, e.g. 3V3 means 3.3 volts, and for small resistors, e.g. 5R6 means a 5.6 ohm resistor.  I assume this originated because CAD software had trouble with punctuation in symbol names?

[Gyro]

I just put the zeners there (yes, two 8.2V ones) as a bit of output overload protection. It seemed like a good idea at the time, but they are almost certainly unnecessary in practice, as long as you aren't going to do something silly with a bench PSU or large ESD hit.

[Marco]

Shot noise calculations are appropriate only for some legacy JFET-based electrometer amplifiers, where only a single junction is connected to the amplifier input pins. Modern high impedance amplifiers have several semiconductor junctions connected to the amplifier input pins. The most significant of these junctions are the ESD diode structures. The input bias currents are equal to the sum of these diode currents. The diode currents are designed to cancel each other, but the shot noise currents are uncorrelated and cannot cancel, which, in turn, makes calculating the shot noise from the input bias current impossible.

I don't believe the bias currents are matched, it makes little sense. Lets say you reverse bias both the ESD diodes at 10 mV+offset-error to keep them both reverse biased regardless of offset error (as long as it's smaller than 10 mV). The bias currents then aren't matched, they are simply both as small as they can be with a simple follower setup.

You could think up more complex ways of balancing them by creating a femptoampere meter in your picoampere meter, but I don't think it's realistic. AFAICS it's better to just keep the bias currents as small as possible and forget about matching them.

[ckocagil]

In Gyro's design, why is the non-inverting input of the bottom op-amp connected directly to the trimpot and not the ground?

Isn't the op-amp on the top is just a buffer for the ground point?

Edit: Nevermind, I think I see it now. The op-amp establishes the middle-of-supply ground point. The trimpot allows, well, trimming the non-inverting input to compensate the Vos.

[Gyro]

Edit: Nevermind, I think I see it now. The op-amp establishes the middle-of-supply ground point. The trimpot allows, well, trimming the non-inverting input to compensate the Vos.

Yes, that's right. 

[Cerebus]

Shot noise calculations are appropriate only for some legacy JFET-based electrometer amplifiers, where only a single junction is connected to the amplifier input pins. Modern high impedance amplifiers have several semiconductor junctions connected to the amplifier input pins. The most significant of these junctions are the ESD diode structures. The input bias currents are equal to the sum of these diode currents. The diode currents are designed to cancel each other, but the shot noise currents are uncorrelated and cannot cancel, which, in turn, makes calculating the shot noise from the input bias current impossible.

I don't believe the bias currents are matched, it makes little sense. Lets say you reverse bias both the ESD diodes at 10 mV+offset-error to keep them both reverse biased regardless of offset error (as long as it's smaller than 10 mV). The bias currents then aren't matched, they are simply both as small as they can be with a simple follower setup.

You could think up more complex ways of balancing them by creating a femptoampere meter in your picoampere meter, but I don't think it's realistic. AFAICS it's better to just keep the bias currents as small as possible and forget about matching them.

You do realise that you're quoting text from an Analog data sheet (for the ADA4530-1) and misattributing it to David?

The design being discussed therein is specified for typical input bias currents of < 1 fA (max ±20 fA @ -40ºC < T_a < +85ºC) and similar input offset currents. Having managed to hit that loose easy-to meet specification, I suspect that the chip designer's statements about the various bias currents might just about be correct. 

[Marco]

Yes? So? Doesn't allow them to escape physics.

If you try to balance the currents by simply matching the bias voltages you will always arrive at the conclusion that it's best to minimize the bias voltages period ... in which case matching goes out of the window. You could increase the bias voltages to get a better relative match, but the absolute error stays the same, so it gets you nothing.

I just don't see a cheap way to create matched bias currents in the ESD diodes, so I doubt. Do you see one?

PS. AFAIK the bias current for a low leakage diode at 10 mV reverse bias is in the fA range ... so really you need to do nothing more than have a <10 mV offset buffer and create +10/-10 mV guard voltages for the ESD diodes. Once the input escapes the buffer, additional diodes can carry the ESD to the rails. No matching needed. It's the most straightforward architecture.

[magic]

Rumor has it that LMC662 bootstraps ESD diodes. Send one to Zeptobars if you want to know

That being said, some cancellation appears to occur. I vaguely remember a tendency of some LMC660 circuit to stabilize at a particular voltage when one of the input pins was left floating. I presume it was the point where all leakages cancel out.

You may try this experiment: disconnect IN+, short IN- to OUT, pre-charge IN+ to either GND, VCC or VCC/2 and see how it drifts over time. Even the magnitude of the currents (or their difference, at least) could be estimated knowing the approximate input capacitance.

[Cerebus]

Yes? So? Doesn't allow them to escape physics.

If you try to balance the currents by simply matching the bias voltages you will always arrive at the conclusion that it's best to minimize the bias voltages period ... in which case matching goes out of the window. You could increase the bias voltages to get a better relative match, but the absolute error stays the same, so it gets you nothing.

I just don't see a cheap way to create matched bias currents in the ESD diodes, so I doubt. Do you see one?

PS. AFAIK the bias current for a low leakage diode at 10 mV reverse bias is in the fA range ... so really you need to do nothing more than have a <10 mV offset buffer and create +10/-10 mV guard voltages for the ESD diodes. Once the input escapes the buffer, additional diodes can carry the ESD to the rails. No matching needed. It's the most straightforward architecture.

Sigh. Sorry, I was trying to stop you wandering off into the weeds by (1) saying a working, verifiable commercial design was wrong and can't work like that, (2) going off at a complete tangent because the original you 'kind of quoted' was about shot noise and the basis for calculation thereof, (3) discussing a quote of a data sheet as if it was David's original post. Feel free to grab the wrong end of any stick you like...

[Marco]

saying a working, verifiable commercial design was wrong and can't work like that

Their explanation of how it works is likely simplified to the point of uselessness ... it is likely just bootstrapped.

A more realistic view does allow you to approximate worst case shot noise ... and funnily enough it's just the worst case leakage through a single junction. Because worst case leakage and worst case shot shot noise will have all the bias current going through one ESD diode and not cancelling out.

PS. well I guess that depends how you account for the buffer, regardless their explanation is completely misleading. There is almost certainly not some relatively large current through the ESD diodes cancelling out.

[fcb]

So the ESD structures/diodes on the inputs are held at a guard voltage derived from the input? Presume there are another set of ESD structures to protect the guard drivers?

[Marco]

There's a couple of options, none of which produce large bias currents which cancel out. They produce bias currents caused by offset voltage, which ideally cancel out a bit ... but if not, then not. Cancelling is incidental, keeping them small to begin with the important part.

PS. I think I was overcomplicating things when saying you had to keep the diodes reverse biased, might not matter if they become positive biased a bit.

[NeverDie]

Is this helpful?  TI very recently posted the gerber files for their LMP7721 eval board.  You can find the zip file here: http://e2e.ti.com/support/amplifiers/f/14/t/895571?LMP7721MAEVALMF-NOPB-Gerber-files

[ckocagil]

I built one of these too based on Gyro's design. I tried to stay faithful to the builds here since this is my first experience with a low current instrument (I haven't owned or used one prior to this). Even copied Gyro's box cover

Cap and resistor: 330pF PS cap, 1G 1% resistor (both from Taobao)
Input socket: high voltage BNC socket from Taobao
Cleaning: IPA and brush

Post-cleaning the output started to drift wildly, but it settled after about 10 hours.

Results: nulled to within 10fA, noise is +-10fA within 60 seconds. I'm happy with the result, but I'd still like to know if anyone has a suggestion to achieve a lower noise.

[Gyro]

Looks good, (I'm flattered  ).

I'm not sure how much of the thread you read, but the easiest way to zero adjust the opamp Vos is to connect the input to the -Ve output. Anything remaining when you remove the link is then bias current.

The Johnson (thermal) noise from the 1G resistor is approx. 3uV RMS (fA reading) for the unit's 0.5Hz bandwidth, so around 10fA Pk-Pk. There's no getting around the laws of Physics. It sounds as if you may be a little bit higher than that (20fA P-P?).  I had to screen the end of the BNC connector with a piece of copper tape* to reach the lowest noise (it is very sensitive), lid screwed on too of course.

The only way to reduce noise further would be to reduce the bandwidth (increase the value of the 330pF cap), but of course that would increase the settling time too.


P.S. * Or just hold a coin over the end.

[magic]

The only way to reduce noise further would be to reduce the bandwidth (increase the value of the 330pF cap), but of course that would increase the settling time too.

It has already been discussed here that you can increase SNR by increasing the feedback resistor.

[Gyro]

Johnson noise increases with resistance and I think it is the dominant noise source here. Unless you change the capacitor value too, to restore the same 0.5Hz bandwidth, you are not measuring like for like.

The other, more practical issue though, is that it would increase gain, reducing the maximum current reading possible with the 9V battery, and increase the contribution of the Opamp bias current, which cannot be sensibly nulled out. I took 1G as a sensible compromise (other compromises are available).


Edit: Actually, yes. Thermal noise with resistance isn't a linear relationship, 10G and adjusted for 0.5Hz BW would be quieter. It didn't meet my practical requirements above though.. No, it would still be noisier, just not 10 times as noisy.

[magic]

it would increase gain
it would still be noisier, just not 10 times as noisy.

There you have it, higher SNR. Alex Nikitin or somebody mentioned it long ago.

[SilverSolder]

Switchable gain, e.g. with a relay or even a panel switch?

[Kleinstein]

The current noise from the Johnson noise of the resistor goes down like square root of 1 /R when the resistor is increased. So more resistance helps, but only a little, at the price of a reduced range. With 10 G the Johnson noise would be about 1/3 but the range reduced to 1/10.

There can also be other sources of noise (e.g. the OPs noise - CMOS OPs can have quite some low frequency noise). 10 µV_pp is about the size expected the low frequency range. The effect of voltage noise is reduced linear with a higher resistor.
Another point can be surface charges and dielectric absorption with  isolating material inside the case. This is more like some drifty slowly relaxing background.

Range switching is tricky as the switches can have leakage. So it is possible but not that easy as it looks. Just changng the resistor would also change the time constant, so one would also need a second capacitor.

[mark03]

What if you got rid of the resistor and used a capacitor instead, i.e. an integrator, with some periodic reset mechanism and a sample/hold/display at the output?

[Kleinstein]

Integrate and reset is possible, but it has it's own limitations, e.g. when averaging over more than a discharge period.
It usually also takes some extra HW behind to do the reset and calculate the current from rate of change or at least send the data to a computer. The simple TIA can directly connect to a simple DMM or even analog meter.

edit:
The capacitor charge method gets interesting when the resistor noise gets much larger than the OPs voltage noise. So more as an alternative to resistors larger than some 100 G, not so much for the larger ranges lime 1 nA FS, where the OPs noise is often larger than Johnson noise.

[NeverDie]

In case anyone here is interested, on another thread it seems we've established that Gyro's design can accurately measure picoamps even at single digit picoamp levels:  https://www.eevblog.com/forum/beginners/static-control-requirements-for-picoamp-measurements-using-ucurrent-gold/msg3101836/#msg3101836

Congrats to Gyro for a great design!   

[arivalagan13]

Can anybody tell me how these PAM's can be tested? I mean can we have a simple current source in the picoampere range to test this picoammeter? I want to test a10^9 gain transimpedance amplifier. I want to design a simple current source to test the amplifier. Any suggestions?

Regards
ArM

[Gyro]

Hi,

I've attached the text of the PM that you sent me as it includes more detail than your post, it will hopefully help more people to make suggestions...

I got to know about you from the eevblog picoammeter design thread and thought.
I'm designing a precision Transimpedance Amplifier (TIA) to amplify a current wiggling between few picoamperes to around 100 picoamperes maximum. I built the amplifier with LTC6268 500MHz version.

The problem is when I tried to test the TIA.
I want a picoampere current source to test the TIA in the above said range. Could you please give me a suggestion or an outline on how do I do that in designing a picoampere current source.

Any advice would be much appreciated.

NeverDie's thread in the beginner's section (linked in his last post above) gives a good narrative on how to generate low source currents at DC.

The real reason for quoting your PM text is that you mention that you used the "LTC6268 500MHz version". This makes me think that you are planning some high frequency application. If so, it is going to be much harder to generate suitable low test currents needed for testing due to the influence of stray capacitances.

The LTC6268 seems to have very similar input current DC characteristics to the LMC662 (it bootstraps its input protection diodes too), which seems to be quite remarkable considering its frequency performance. Being able to utilise them at significant frequencies is a potentially a different matter (I notice that their datasheet typical application circuit uses a 20k feedback resistor for a 65MHz TIA, which is a long way from your 10^9 gain).

Maybe you can give us some more detail of your intended application and operating frequency?


P.S. The application information in the LTC6268 datasheet makes interesting reading, particularly with regard to controlling (and making use of) stray capacitances.

[arivalagan13]

Sure. I want to amplify a signal coming out of a Faraday plate detector (part of ion mobility spectrometer). The application does not demand high frequency response. It's speed is mentioned in terms of Full width at half maximum (FWHM) at 0.5 milliseconds (worst case)..this corresponds to an approximate bandwidth of few kilokertz. But, I'm given to design the amplifier for 50kHz at 7 microsecond rise time.

The signal amplitude is between few picoampere to around 100 picoampere. That's about the specs. I hope the information provided is sufficient.

Coming to the TIA design, yes, 10^9 gain is too much of gain. So, I'm going for a two stage design. The first stage amplifier has transimpedance gain of around 100 megohm (will iterate with 250 megohm as well) and the second stage gain is 10 (or 4).

Hope this information helps.

Regards
ArM

[Gyro]

This sounds like an interesting application, I'm afraid only a few fairly random thoughts come to mind at the moment:

- The LTC6268 does not have offset voltage adjustment pins, so there is no way to zero out the Vos contribution. You can still differentiate the relevant  contributions of input bias and Vos however, by the same method of shunting the feedback resistor, measuring the resulting output (which give you the Vos if the opamp at unity gain). Then checking the difference between this and the output with the feedback resistor alone and the input open (but screened) will give you opamp input bias current contribution. Just accept the Vos contribution as a fixed (0.4uV/'C) offset. In India I suspect that the ambient temperature contribution to input bias current may be more of an issue in terms of accuracy.

- Calibration of the TIA at DC is relatively simple, as NeverDie demonstrated. This is really a confirmation though - for practical purposes, the accuracy of the TIA is determined by the accuracy of the value of the feedback resistor.

- AC performance is heavily dependent on the physical implementation of the LTC6268 circuit. As I mentioned, stray capacitances are absolutely critical, and the 'Application Information' section of the Datasheet goes into this in significant detail. LT also have an eval board, the DC2414A, which covers three different PCB layouts.... https://www.digikey.com/catalog/en/partgroup/ltc6268-and-ltc6268-10-demonstration-board/66825 This uses a photodiode as a high frequency current source example. The DC2414A board manual contains even more detailed application information. These documents will probably give you the most specific and relevant information source.

- A 100M feedback resistor still sounds very optimistic given your frequency goals and the significant effects of parasitic capacitance at 7uS rise time.

- Maybe some form of low capacitance C or RC discharge source would give you what you need in terms of AC current calibration source.

[magic]

There is a similar thread in "repair" where somebody is trying to replace a broken 100MΩ amplifier with 30kHz bandwidth. Clearly, it can be done. That being said, I calculated that <0.05pF feedback capacitance was required to get there.

I think calibration at AC would be doable (connect to an AC voltage source through suitable resistor) but one needs to mind parasitic capacitance of the resistor, parasitic capacitance of the TIA and frequency-dependent effective input resistance of the TIA.

[Gyro]

Ah, I missed that one....  https://www.eevblog.com/forum/repair/repairing-a-current-to-voltage-preamplifier/

[RawCode]

Hello

I don't know if i'm going out of topic.

I'm trying to modify this desing making it multi-range, using a push-pull buffer inside the feedback to achieve higher current capability when reading milliAmps.
When i try to use the 1GOhm range, so using the circuit almost as it was design(except for the current buffer) i get a nasty saturated 50Hz output.
do you have any ideas why it happens? I shoud say that i air-wired only the part in the red area in the picture. Can this be the issue?

[Kleinstein]

Good shielding is definitively needed to keep the mains hum out. The switches will add more leakage current. So one would minimize the switching to a minimum, especially with the 1 G resistor active. Not all switches are suitable.

The buffer is not working well unless rare zero threshold fets are used. This can be a problem for the stability as it slows down the buffer at some current.

Even with a buffer the higher current range is not practical, as there would be too much heat at the resistor with more than some 10 mA. Even 1 mA may start to become tricky and requite a reduced range. The other point is that the TIA input may need some "isolation", not to have to much capacitance at the input. This would requite some series resistance, so the useful range in usually limited up to maybe some 100 µA - this would be a current range that does not need the extra buffer, at least not a very strong one.  The 1 mA range would be more like a completely separated TIA.

[Cerebus]

Even with a buffer the higher current range is not practical, as there would be too much heat at the resistor with more than some 10 mA. Even 1 mA may start to become tricky and requite a reduced range. The other point is that the TIA input may need some "isolation", not to have to much capacitance at the input. This would requite some series resistance, so the useful range in usually limited up to maybe some 100 µA - this would be a current range that does not need the extra buffer, at least not a very strong one.  The 1 mA range would be more like a completely separated TIA.

My "rule of thumb" for precision circuits is to keep resistor dissipation down to 10 mW or below. Obviously, like all rules of thumb, this doesn't obviate the need for proper analysis in cases that deserve it, where one will take into account tempco, thermal resistance, etc. etc. As we don't know the later for RawCode's intended components we can't do a proper analysis, but perhaps using rule of thumb would yield some insight.

Assumption, the high current buffer is only really intended to come into play in conjunction with the 1 ohm resistor. Dissipation at 10mA will be 1 * 0.01² mA = 100 uW. So, as long as my rule of thumb holds, we should be golden. Let's see if it does. Assumptions 200ppm resistor tempco (i.e. cheap, non precision resistor), 1/8W resistor (50ºC temp rise for 0.125W, so R_th to air = 400ºC/W). So, 100 uW * 400ºC/W = 0.04ºC self heating => 8 ppm (0.0008%) shift from self heating.

Conclusion, worst case effect of self-heating from a 10mA current on the 1 range is 8 ppm - negligible for a circuit of the likely precision class we'd assign for this class of picoammeter. A 100mA current would generate 10 mW of dissipation, 4ºC self heating and 800ppm (0.08%) shift (Same assumptions as above,  1/8W and 200 ppm/ºC. Obviously could be made better by picking a suitable precision resistor).

[Cerebus]

Hello

I don't know if i'm going out of topic.

I'm trying to modify this desing making it multi-range, using a push-pull buffer inside the feedback to achieve higher current capability when reading milliAmps.
When i try to use the 1GOhm range, so using the circuit almost as it was design(except for the current buffer) i get a nasty saturated 50Hz output.
do you have any ideas why it happens? I shoud say that i air-wired only the part in the red area in the picture. Can this be the issue?

You need to extend your air wiring and shielding to all the high impedance areas. That includes all of the switches that are currently outside your 'red' zone.

I would have designed it with the switches on the low impedance side (i.e. between the buffer output and the range resistors) and kept my highest impedance resistor in circuit at all times and just compensated for it always being in parallel by either different component values or by correcting after the ADC.

Also, in passing, if any of your switches are mechanical (including relays) there is a real risk of not pushing enough current through to overcome 'dry' contacts. If you're using mechanical switches or relays make sure that they are rated for 'dry' switching.

[Gyro]

Hello

I don't know if i'm going out of topic.

I'm trying to modify this desing making it multi-range, using a push-pull buffer inside the feedback to achieve higher current capability when reading milliAmps.
When i try to use the 1GOhm range, so using the circuit almost as it was design(except for the current buffer) i get a nasty saturated 50Hz output.
do you have any ideas why it happens? I shoud say that i air-wired only the part in the red area in the picture. Can this be the issue?

Not off topic as such (although by the time you get to 1R feedback resistors you're a long way from pA. )

Instinct says that you are going to need very different (incompatible?) switching mechanisms to tackle dry switching and stray leakage currents on the 1G range and the nasty contact resistance dependency of the 1R range. Do you actually need to cover this entire range on a single input? Maybe you could split them.

You might be better to bump or cross refer to (I guess I've done that now) to your https://www.eevblog.com/forum/projects/poor-linearity-of-a-opamp-push-pull-buffered/ regarding the issues around the push-pull MOSFET follower, and your https://www.eevblog.com/forum/beginners/line-filtering-in-transimpedance-amplifier/ regarding the 50Hz issues and filter attempts so that other members have some context of what you have already done and the existing replies.

[RawCode]

Thank you for all the replies, very very kind of you

The 50Hz interference behaviour is quite strange: depending on the load, there is some kind of a voltage/current "threshold". If the voltage/current exceeds this "threshold" the output signal is fine, almost without any 50Hz interference. If the voltage/current on the load is below that variable threshold though, in the output there is a 50Hz square wave(probably a saturated 50Hz sine). I mean, it's not always present, just when the current/voltage on the load is very small(200/500mV or so across the load). Could this behaviour be an issue related to the MOS current buffer? If yes, a BJT/LT1010 current buffer can solve the issue?

Yes this circuit should have this large current sensitivity, because i don't know which kind of device will be tested. To be honest I was taking inspiration from this circuit and the Keithley 220, which seems that can handle a large current range.
Watching at the schematics (i don't know if i'm allowed to share them, but you can find them in the 220's manual) it seems that the highest resistor is always connected, as Cerebus suggested, but how can i be sure about the value of the resulting parallel resistor? i mean, yes there is the formula, but how is it reliable considering both resistor tollerances? probably the best way is measuring the actual value of the resistors, but how can i do it without a specialized and very sensitive ohmmeter(considering that there is a 1GOhm resistor)?
Also, there are practically no capacitor in parallel with the sense resistors in the Keithley's schematic. Would just one capacitor (330pF) suffice for the entire range of resistors or are necessary different capacitor for each resistor?

Yes, I'm using mechanical switching(these relays, which should have at least 1.5TOhm insulation resistance: DIP12-1A72-12L). I'm assuming that you are referring to dry switching to the phenomenon where the relay's contacts get oxidised. Am I right? If not, what do you mean by dry switching?

Also, in passing, if any of your switches are mechanical (including relays) there is a real risk of not pushing enough current through to overcome 'dry' contacts.

Do you suggest the use of some kind of bootstrapping for the relays' coil, like a capacitor in parallel to a the series resistor with the coil to get a momentary current kick?

Instinct says that you are going to need very different (incompatible?) switching mechanisms to tackle dry switching and stray leakage currents on the 1G range and the nasty contact resistance dependency of the 1R range. Do

In the Keithley's schematic there is something strange, like a hybrid relay-NJFET switching mechanism, but i don't quite understand how it works. What about JFET's parassitic capacitances, leakage current? How just one NJFET can allow bidirectional current flow? Sure they are bidirectional unlike BJTs, but should it be necessary a PJFET in parallel to all of those NJFET to handle negative signals? The 220's circuit is quite complex, especially the circuit related to U319, and i don't quite understand what is going on to be honest, so please sorry if i'm saying something stupid: my electronics knowledge is still quite limited 

[Cerebus]

Watching at the schematics (i don't know if i'm allowed to share them, but you can find them in the 220's manual) it seems that the highest resistor is always connected, as Cerebus suggested, but how can i be sure about the value of the resulting parallel resistor? i mean, yes there is the formula, but how is it reliable considering both resistor tollerances? probably the best way is measuring the actual value of the resistors, but how can i do it without a specialized and very sensitive ohmmeter(considering that there is a 1GOhm resistor)?

You just have to work it out the standard way, and account for tolerances in the calculation. In real life tolerances tend to add in your favour (i.e. tend to cancel because statistics, but one can't rely on that).

Also, there are practically no capacitor in parallel with the sense resistors in the Keithley's schematic. Would just one capacitor (330pF) suffice for the entire range of resistors or are necessary different capacitor for each resistor?

The RC product (i.e. time constant) needs to be the same for each range. Again, in practice, this can often means that stray capacitance in combination with high value resistors like 1G can result in too big a time constant without a compensating capacitor at all. It gets messy. Look at the Keithley 220 schematics and try to figure out how they did the compensation.

Yes, I'm using mechanical switching(these relays, which should have at least 1.5TOhm insulation resistance: DIP12-1A72-12L). I'm assuming that you are referring to dry switching to the phenomenon where the relay's contacts get oxidised. Am I right? If not, what do you mean by dry switching?

Also, in passing, if any of your switches are mechanical (including relays) there is a real risk of not pushing enough current through to overcome 'dry' contacts.

Do you suggest the use of some kind of bootstrapping for the relays' coil, like a capacitor in parallel to a the series resistor with the coil to get a momentary current kick?

No, nothing to do with the coil currents and all to do with the formation of a high resistance layer on the relay contacts. Pretty much all relays are problematic when you're trying to switch picoamps with them - leakage currents, dry contacts and so on. The old way of getting around the dry contact problem was to use mercury wetted relay contacts but those are a thing of the past. Leakage currents generally calls for expensive relays that take this stuff into account (e.g. Coto) and provide you with contacts for guard rings etc.

[GerryR]

I have read through this thread and thought the information presented here was worth the "bump."  Enjoy the journey!

[Terry Bites]

Some more hints


https://www.analog.com/media/en/technical-documentation/user-guides/ADA4530-1R-EBZ_UG-865.pdf
https://www.analog.com/media/en/reference-design-documentation/reference-designs/CN0407.pdf

[GerryR]

Another op-amp candidate for this project, oldy but a goody:
(20fA bias current) 

[magic]

I found the cheap TS27L2 to be surprisingly decent.
I don't remember the exact number, but my sample's input leakage was something like either 10fA or 20fA, not far behind LMC662.

[Gyro]

Another op-amp candidate for this project, oldy but a goody:
(20fA bias current) 

Not a bad choice.

Unfortunately, when you get to the fA level, you get into the fuzzy and dangerous area of 'Typical' specs at room temperature rather than limit values, taking into account the upper temperature spec - hence the industrial temp LMC6482 shows 4pA max while the commercial temperature spec variant of the LMC662 shows 2pA (and 4pA for the industrial spec part).

It is generally believed (I don't think it has been confirmed) that the LMC662 achieves its particularly low (2fA Typ) input current by bootstrapping its input ESD protection diodes from internal guard supplies. I don't know if the LMC6482 [EDIT: or the TS27L2] have the same or not. It would be nice to have some die pictures to know for sure. [EDIT: Alex Nikitin would probably has the best placed on this stuff as he designs Picoammeters commercially and gets some inside IC info].

As I say Typical / headline figures vs guaranteed limits are a dangerous area, at least for volume design. For one-off projects you can take few more liberties with individual samples.

As the two parts are pin compatible it would be interesting to directly compare if you have some in stock. Not quite as quick and simple as simple socket swapping due to the need to air-solder the inverting input pin (if you chose to go that way) and subsequent cleaning and drying time.

[GerryR]

I've had the 6482's "in stock" for 20+ years.  When going through my stock data sheets, I came across the 6482's and it was like Christmas!  I'm waiting for the 662's to get here, along with the 1 G resistor and then I will have at it.  I did intend on comparing the two.
Cheers,
GerryR

[magic]

It is generally believed (I don't think it has been confirmed) that the LMC662 achieves its particularly low (2fA Typ) input current by bootstrapping its input ESD protection diodes from internal guard supplies. I don't know if the LMC6482 [EDIT: or the TS27L2] have the same or not. It would be nice to have some die pictures to know for sure. [EDIT: Alex Nikitin would probably has the best placed on this stuff as he designs Picoammeters commercially and gets some inside IC info].

LMC6001 is on zeptobars. It is believed to be one half of LMC662 or something very similar to that and indeed its die contains transistors for two channels, but only the left one is wired up.

Protection structures are near the input pads (left edge) with connections to GND and VCC and they look quite small but I will not pretend to understand how they work. Other than that, there are only polysilicon (green) traces going from there straight to the input stage (16 circular transistors in the center). These are better visible in the right channel, where metal doesn't obscure them.

There is a paper about LMC660 written by Monticelli which I linked in Noopy's opamp thread recently, but IIRC there was nothing about input protection there.

As the two parts are pin compatible it would be interesting to directly compare if you have some in stock. Not quite as quick and simple as simple socket swapping due to the need to air-solder the inverting input pin (if you chose to go that way) and subsequent cleaning and drying time.

One could bend the important pin sideways and attach some sort of single-pin socket to it. That's how I did it, with the rest of the circuit literally running on a breadboard and the whole experiment enclosed in a grounded metal box.

I verified input leakage by hanging an appropriately bent wire on the leads of the air-suspended feedback resistor to short it out.

[David Hess]

How just one NJFET can allow bidirectional current flow? Sure they are bidirectional unlike BJTs, but should it be necessary a PJFET in parallel to all of those NJFET to handle negative signals?

Only one polarity of FET, whether a JFET or MOSFET, is required to switch a signal.  CMOS multiplexers use both in parallel so that their signal range includes the entire supply voltage range.

Bipolar transistors actually are bidirectional but were seldom used that way because of various deficiencies.  They used to make symmetrical bipolar transistors for chopping applications but they were quickly replaced with JFETs and then MOSFETs when they became available.

Below is an example from the Tektronix 7T11 sampling sweep using 2N3904s for Q400 and Q402.  I have no idea why Tektronix did not use a JFET since they obviously had them available.  I have to assume that the bipolar transistors, 2N3904s in this case, actually performed better.

[David Hess]

LMC6001 is on zeptobars. It is believed to be one half of LMC662 or something very similar to that and indeed its die contains transistors for two channels, but only the left one is wired up.

The LMC6001 is an LMC6081 tested for 25 femtoamp input current.

I used a lot of LMC6081s for low drift high temperature integrators and measured many of them at 2 femtoamps.

[Gyro]

LMC6001 is on zeptobars. It is believed to be one half of LMC662 or something very similar to that and indeed its die contains transistors for two channels, but only the left one is wired up.

Protection structures are near the input pads (left edge) with connections to GND and VCC and they look quite small but I will not pretend to understand how they work. Other than that, there are only polysilicon (green) traces going from there straight to the input stage (16 circular transistors in the center). These are better visible in the right channel, where metal doesn't obscure them.

Thanks, I hadn't seen that. No, I wouldn't pretend to understand the fine detail without Noopy's closeups and interpretation either.

Another interesting comparison might be the LMP7721, which (similar to the LMC662) claims 3fA typ input current (20fA max @25'C), and from the datasheet/app note, definitely has internal guarding. I can't find it now, but I'm sure I saw a reference to its input protection being able to withstand 10mA continuous rather than the usual 5mA too.

It still seems to be a strange anomaly that the low cost LMC662 has a claimed 2fA typ input current, similar to the LMP7721, when others, such as the LMC6001 / LMC6081 and LMC68482 are all in the 20-25fA range (ok the LMC6001 is the tested limit rather than a typical). The LMC662 does seem to get used in commercial picoammeter input stages though, as indicated in Alex Nikitin's earlier posts, and observed in several equipment teardowns on the forum. It would be rather amusing if it all came down to the datasheet author having accidentally missed out a zero.

[magic]

The original spec for LMC660/662 was 40fA. It's in the Monticelli paper and the 1989 NS linear databook. So those other parts may have been an improvement at the time of their introduction, or at least on paper.

I don't know why the spec has been revised - due to an actual improvement, an improvement in their ability to measure it, or a typo. Somewhere there is a video of Bob Pease talking about their adventures building test fixtures capable of measuring Ib of those opamps, maybe it was even posted in this thread?

[Kleinstein]

In the sub pA range there can be a big difference between typical an maximum specs. Not too many DS show both typical an max values.  E.g. The LMC6462AM gives 150 fA typical and 200 pA_max. There can also be differences if there are more manufacturers and the letters at the end.

[Cerebus]

Somewhere there is a video of Bob Pease talking about their adventures building test fixtures capable of measuring Ib of those opamps, maybe it was even posted in this thread?

Probably, because I know that I've posted a link the forum more than once. But there's no harm with doing it again:

[Gyro]

Ah, yes, that was the one which mentioned the potential pitfalls of too much Teflon and cosmic rays. It ought to be pinned somewhere.

I wonder if the LMC662 headline figure partly came from that video [EDIT: No of course it can't, the datasheet pre-dates the video  ], it mentions that 99+% of devices came in below 5fA. Being an earlier device, they may have made a bigger deal of such a low typical figure (yes there's a big difference between typical and limit values, I mentioned earlier that it was a danger zone for volume production), than the LMC6001. Either that, or they got really lucky with that particular die layout.

The LMP7721 seems to be the only device where they have made the effort to optimise the pinout specifically for low leakage. Although it is only available in SO8, it has two dedicated n/c pins between the inputs and other pins, that can be included in the guarding scheme. The price seems to have come down too (at least looking at RS in the UK) it's about a third of the price of an LMC6001 (probably a result of the additional part selection step in the latter).

As an aside, I don't know how much benefit there is to be gained from using a dual device in practice. In my proto adaptor, having the rail splitter and input amp in the same package does seem to have benefited from offset voltage tracking between the two with supply and temperature, as the offset null seems to remain remarkably stable. Maybe just luck with a particular sample, tracking isn't something that appears in commodity opamp datasheets.

[Kleinstein]

The ADA4530 also has a special pinpout. It has great data but also a steep price.

[bsw_m]

ADA4530-1 input bias current measurement with 0.1Hz bandwith:
Perhaps later, I will measure the input currents for the LMC662 and LMP7721 in the same way.

[bsw_m]

ADA4530-1 input bias current measurement with 0.01Hz bandwith:

[magic]

That's nice.

On the other hand, LMP7721 is supposed to have less voltage noise than the rest of those opamps. It probably doesn't matter in this application.
The datasheet shows some dependence of leakage on common mode input voltage.

edit
Attached input protection and leakage paragraph from Monticelli.

[Cerebus]

ADA4530-1 input bias current measurement with 0.1Hz bandwith:
Perhaps later, I will measure the input currents for the LMC662 and LMP7721 in the same way.

Please forgive me for being that guy, but that chart appears to show typical readings in the region of 5x10^-17 A (50 attoamps), taken in 0.1 second samples. That's around 31 individual electrons measured in each sample. What instrument are you using that is capable of delivering valid readings under those circumstances?

[SilverSolder]

ADA4530-1 input bias current measurement with 0.1Hz bandwith:
Perhaps later, I will measure the input currents for the LMC662 and LMP7721 in the same way.

Please forgive me for being that guy, but that chart appears to show typical readings in the region of 5x10^-17 A (50 attoamps), taken in 0.1 second samples. That's around 31 individual electrons measured in each sample. What instrument are you using that is capable of delivering valid readings under those circumstances?

An abacus?   

[bsw_m]

What instrument are you using that is capable of delivering valid readings under those circumstances?

Old V7-45 electrometer, designed in MNIPI.
Edit:
Please note 0.1second per sample - this is ADC speed. Measuring interval (bandwith) is 0.1Hz or 0.01Hz.
Regarding the reliability of measurements, I'll just say that the performance verification procedure provides for checking the 20aA point, while the maximum permissible error is ± 8aA

[bsw_m]

An abacus?   

We can say that the abacus for individual electrons

[bsw_m]

As for ADA4530-1. This is an excellent opamp, suitable for making a device that will be able to measure current up to 1E-16A, albeit with some nuances.

[SilverSolder]

An abacus?   

We can say that the abacus for individual electrons

I just read your other thread about that instrument.  Those are some pretty cool toys you've got there! 

[David Hess]

The LMP7721 seems to be the only device where they have made the effort to optimise the pinout specifically for low leakage. Although it is only available in SO8, it has two dedicated n/c pins between the inputs and other pins, that can be included in the guarding scheme. The price seems to have come down too (at least looking at RS in the UK) it's about a third of the price of an LMC6001 (probably a result of the additional part selection step in the latter).

They would not have much choice in a surface mount package.  With the DIP LMC6081, there is plenty of space to place guard traces, or the input pin can be bent up to horizontal and air wired.

In the past they had a 10-lead TO-99 through-hole package for low input bias current parts.

[MegaVolt]

1G 1% resistor (from Taobao)

Can you share the link?

[MegaVolt]

It is possible instead of a large resistor to use a T-shaped divider of 3 resistors. That will allow to apply smaller resistors and reduce the noise level.

[Kleinstein]

It is possible instead of a large resistor to use a T-shaped divider of 3 resistors. That will allow to apply smaller resistors and reduce the noise level.

One can use the T network to replace a large resistor. However as a TIA / pA meter this comes with a higher noise level. For a TIA the input referred current noise goes down with a larger resistor. The larger resistor will have more voltage noise, going up with the square root of the resistance. However the current noise is voltage noise divided by the resistor and thus proportional to 1 over the square root of the resistance.

For a classical TIA with a resistor feedback, the noise of the resistor can be a major contribution and the very high level resistors (> 100Mohms) often have comparatively poor stability. For the very low current range there is the alternative to use a charge amplifier instead and than look at the rate of ouput voltage change. Good capacitors in the 100 pF range are better available than good resistors in the Gohms range. Another, a bit unusual way is to use a photodiode as feedback element - this gives some extra shot noise, but can still be a viable alternative.

[magic]

The short answer is that noise of R2 decreases, but it is now amplified by the ratio of R2/R1.
It's same thing as using a lower resistor and adding a noninverting gain stage.
If you do the math, you see you cannot win.

[MegaVolt]

If you do the math, you see you cannot win.

Good We do not win noise. But can we win in the stability and accuracy of resistors?

[Kleinstein]

The T circuit can have more stable / accurate resistors, as it can use lower value ones.  Above some 10 M it is rare to find really stable ones (e.g. wire wound).

[magic]

The T network also amplifies the opamp's flicker noise while preserving the original signal gain.
You throw away SNR and there will be a point where no amount of filtering will recover it.

By all means go and calculate the details or build a prototype if you want.

[Marco]

So would the lowest leakage opamp you can realistically make without a fab be a composite opamp with a 2n7002 differential pair? (the class 0 ESD ones) Ignoring the terrible offset drift for a moment, oven can take care of that.

[Cerebus]

You'd probably have better luck with a different discrete unprotected IGFET.

First off find something (if you can, choices in discrete small IGFETs are pretty slim nowadays) with a hermetic case - a glass seal will perform better leakage wise than epoxy. Secondly, find something with the substrate brought out to a terminal; then you can mess around with bootstrapping the substrate independently of the source (within limits obviously, you still need to keep any parasitic junctions turned off) and might shave a few percent of leakage off that way. Again, good luck with finding a 4 lead IGFET in anything other than NOS nowadays.

[magic]

Or try your luck with depletion mode MOSFETs, as they can be made to work at zero Vgs.
There is still the drain, but it comes from the bottom of the die and from the opposite side of the package (in SOT23 etc).

Or not from the bottom? Are those things vertical or lateral, actually?

[Kleinstein]

Small unprotected MOSFETs are difficult to get. Essentially all the modern one have a zener for protection and this gives quite some leakage.
Chances are some of the low leakage CMOS OP are easier to get than low leakage MOSFETs.

The other alternative is a small JFET, especially MMBF4117 with small drain-source voltage (e.g. some 2-3 V). This could be similar to the input section of the Keithley 617.

[Cerebus]

My own experience would seem to suggest that the x4117 family are in at best a few hundred fA territory rather than 10s or single fA territory. I think one is probably stuck with integrated CMOS solutions when aiming for absolute minimum leakage/bias current.

[Kleinstein]

The x4117 would not beat the dedicated low bias CMOS chips. In pA meter design the OPs bias is only one hurdle to take. In many designs the feedback resistor and PCB/relay leakage can be the larger problem.  So a low bias OP is onely a part of the solution and with relatively afordable CMOS OPs (e.g. LMC662 or LMC6482), maybe even more like the simpler part.

[Marco]

You'd probably have better luck with a different discrete unprotected IGFET.

Yeah, they seem to have completely ditched all the old parts. Nexperia has an application note from a mere 3 years ago where they still said BSS138P is unprotected ... yet it's AEC-101 certified now.

Linear Systems at least still selling new unprotected small signal MOSFETs, 5$ for a single in a nice TO72 isn't that bad. Just pray the distributor was careful parcelling them out.

[bsw_m]

At the moment I am thinking about the concept of a picoammeter on a vibrating reed capacitor.
Why do I need it.
Some time ago, I purchased about 50 pieces of resonant vibrating reed capacitors DRK-2 "New old stock". These capacitors were developed in specialized design engineering bureau (later became an institute MNIPI) of the Minsk Molotov plant by Shuklin A.S.  in the early 60s. The capacitors I purchased were released in 1987. These capacitors were purchased for study and experimentation. I'm wondering what results can be achieved today using these old parts in the front end of a transimpedance amplifier.
At the moment, the stability of the resonant frequency has been assessed. The leakage of the input insulator was also measured. The leakage current of the input insulator was about 2.5E-16A at a test voltage of 100V.

The resonant frequency measurement is given in the attached graph.
I do not like the stability of the resonant frequency, so I will modify these capacitors to get better stability.

What frustrates me, unfortunately, if I publish the full documentation for the resulting picoammeter, no one will be able to repeat it, due to the fact that unobtanium components  will be used in the input circuits. 

If you are interested, then as experiments and new data are obtained, the results can be published. If there is no interest, then I probably will not litter the forum.

[SilverSolder]

At the moment I am thinking about the concept of a picoammeter on a vibrating reed capacitor.
Why do I need it.
Some time ago, I purchased about 50 pieces of resonant vibrating reed capacitors DRK-2 "New old stock". These capacitors were developed by Shuklin A.S. in the early 60s. The capacitors I purchased were released in 1987. These capacitors were purchased for study and experimentation. I'm wondering what results can be achieved today using these old parts in the front end of a transimpedance amplifier.
At the moment, the stability of the resonant frequency has been assessed. The leakage of the input insulator was also measured. The leakage current of the input insulator was about 2.5E-16A at a test voltage of 100V.

The resonant frequency measurement is given in the attached graph.
I do not like the stability of the resonant frequency, so I will modify these capacitors to get better stability.

What frustrates me, unfortunately, if I publish the full documentation for the resulting picoammeter, no one will be able to repeat it, due to the fact that unobtanium components  will be used in the input circuits. 

If you are interested, then as experiments and new data are obtained, the results can be published. If there is no interest, then I probably will not litter the forum.

The project sounds very interesting, it doesn't matter if it is a one-off...  I'd certainly be interested in reading (reeding?) about it!

[ZhuraYuk]

I think when more details and technical ideas will be revealed, the more people will get interested in the project.
Also to make the implementation repeatable it is better to make it more universal when you can easily replace DRK with TI amp and thus get -16 lower measurement range instead of -18. Someone might even recreate DRK itself or MEMS DRK will appear on market soon.

[bsw_m]

I assembled a prototype amplifier on a slightly modified DRK-2.
The amplifier's own noises were tested, not entirely satisfied with the result. When checking noise, the amplifier worked in buffer mode (gain = 1)
But new additional knowledge was obtained on the operation of the DRK-2. Now I'm thinking about the possibility of creating a DIY vibrating reed capacitor.

[Kleinstein]

Te noise (or background) level looks quite high. It looks like part of this may be some AC background and not all noise. It may be worth looking at the FTT of the data.
The final use would likely be mainly for the very low frequency part ( e.g. << 10 Hz) and not so much the faster part. So the actual noise performance may not be that bad.

[Atomillo]

Hello all!

I've recently read this gem of a thread and started pondering and trying to do some simple noise calculations myself. While doing so, some doubts have appeared to me and I would be grateful if someone more knowledgeable could lend me a hand (I'm not creating a new thread because I believe these questions might be relevant to the original design and posterior discussion, but if not please let me know).

The first question is this: isn't the Johnson noise of the 1 Meg resistor connected directly at the input? Doesn't the low noise we achieve with a high feedback resistor get ruined by this? I assume not, since people with a lot of experience in this area have not mentioned it (in the various noise calculations done by Alex Nikitin it isn't taken into account for example), but I don't see how this is so. Is something cancelling the noise of this resistor?

The second question is related to T-networks. I considered the possibility of using them along with a capacitor, to simulate pF and sub-pF capacitors with larger, more accurate ones. But my pre eliminary (and probably erroneous) calculations show that the johnson noise current of the resistors used in the divider would translate directly to input refered current noise. Is this correct?

Many thanks all for this great thread and any help is very much appreciated!

[Kleinstein]

You are right about the T network: it gives additional gain for the ouput, just like a very high resistor, but the noise is the same as with the smaller actual resistor and thus more noise than a true large resistor. With capacitors the noise is not a problem though.

The protection resistor in series to the OP-amp does contribut, just like noise of the OP-amp. However this is usually not much compared to the noise from the FB resistor. The protection resistor becomes relevant when the FB resistor is comparable or smaller. Another case is when the DUT is relatively low resistance.

[Atomillo]

The protection resistor in series to the OP-amp does contribut, just like noise of the OP-amp. However this is usually not much compared to the noise from the FB resistor. The protection resistor becomes relevant when the FB resistor is comparable or smaller. Another case is when the DUT is relatively low resistance.

This is what I don't get. Shouldn't it be the other way (i.e the protection resistor much larger than the feedback resistor)? After all, the Johnson current of the protection resistor is proportional to 1/sqrt(R) too.
Just as a numerical example, with the 1Meg resistor at T=300K, we have a noise of 0.11pA/rt(Hz). Isn't this noise added directly to the input of the circuit? How is it that in the given calculations previously on this thread only the current noise of the 1G feedback resistor is taken into account? Does the op-amp cancel this noise in any way I'm not seeing?

You are right about the T network: it gives additional gain for the ouput, just like a very high resistor, but the noise is the same as with the smaller actual resistor and thus more noise than a true large resistor. With capacitors the noise is not a problem though

I was refering to the noise of the resistors that make the divider at the output of the op-amp. I arrived to the conclusion that it didn't make sense to remove the high value resistor for a capacitor to get rid of the noise if the noise of the lower-valued resistor in the divider was much worse (since their current noise appeared to translate directly to the input).

Many thanks for your help!
Cai

[Kleinstein]

The protection resistor has one side at the op-amps input and this a high impedance in series. So there is voltage noise, but no extra current noise.

If one uses the alternative version with the protection directly in series with the input, the protection resistor is in series with the signal source / DUT. So there would be high noise if the DUT is low impedance, but in this case the whole TIA amplifier concept does not work - there is plenty of current noise from the source.

[Atomillo]

The protection resistor has one side at the op-amps input and this a high impedance in series. So there is voltage noise, but no extra current noise.

If one uses the alternative version with the protection directly in series with the input, the protection resistor is in series with the signal source / DUT. So there would be high noise if the DUT is low impedance, but in this case the whole TIA amplifier concept does not work - there is plenty of current noise from the source.

Oh I get it, there's no return for the current due to the high impedance of the op amp input so we only have to consider the noise voltage!

Thanks it makes sense now!

[Atomillo]

I've been thinking about this and I must confess I don't fully understand the reasoning regarding the "high impedance" node. After all, the input impedance of the op-amp can never be infinite.
Let's say we connect a resistor from the inverting input of the op-amp to ground (Rtest in the attached schematic). Now the impedance at this node is no longer theoretically infinite. From which value would we be able to neglect the current noise of Rp? And why?

[Atomillo]

Sorry for the doble post but I think I figured it out! Turns out actually drawing the schematic instead of just thinking about it helps a great deal to see this stuff.

Rtest and Rp forms a voltage divider and this sets the voltage at the non-inverting node of the op-amp (Vx = Rtest/Rtest+Rp * en). Since no current flows into the op-amp, the actual input refered current is then Vx/Rtest. As Rtest grows, Vx diminishes. Eventually, we can consider it is zero and thus no contribution to input current.

Also, as Vx approaches zero, we also find that the voltage noise at the input is that of the resistor, also matching what we were expecting.

[magic]

AFAIK you can assume that Johnson noise is an AC current source appearing in parallel with the resistor, resulting from thermal motion of electrons.
If the current has no other path to complete the loop, it flows through the resistor and becomes an AC voltage in series with the resistor.
If there are other circuits in parallel with the resistor, it splits proportionally to their admittance.

[aleruggeri87]

A notice that might be of interest to readers of this thread:

Today I sorted the available opamps on the TI website by their input bias current, and the one with the lowest was no longer the "well-known" LMP7721, but the OPA928:
https://www.ti.com/lit/ds/symlink/opa928.pdf

Still in preview, it has a pinout "à la ADA4530-1", specs seems to be in line or better.
Looking forward to read comments and reviews from the pros 

Best,
Alessandro.

[David Hess]

Today I sorted the available opamps on the TI website by their input bias current, and the one with the lowest was no longer the "well-known" LMP7721, but the OPA928:
https://www.ti.com/lit/ds/symlink/opa928.pdf

I am confused about the absolute maximum ratings on page 4.  Is it a 16 volt part or a 40 volt part?

[SilverSolder]

Today I sorted the available opamps on the TI website by their input bias current, and the one with the lowest was no longer the "well-known" LMP7721, but the OPA928:
https://www.ti.com/lit/ds/symlink/opa928.pdf

I am confused about the absolute maximum ratings on page 4.  Is it a 16 volt part or a 40 volt part?

Recommended 16V, with absolute max 40V.   Maybe some specification or other becomes untenable over 16V,  but the device still survives and kind of works up to 40V?

[Alex Nikitin]

Today I sorted the available opamps on the TI website by their input bias current, and the one with the lowest was no longer the "well-known" LMP7721, but the OPA928:
https://www.ti.com/lit/ds/symlink/opa928.pdf

I am confused about the absolute maximum ratings on page 4.  Is it a 16 volt part or a 40 volt part?

Recommended 16V, with absolute max 40V.   Maybe some specification or other becomes untenable over 16V,  but the device still survives and kind of works up to 40V?

No, in the text it is clearly stated 16V max, 40V in the maximum ratings is an obvious error.

Cheers

Alex

[David Hess]

I know from experience that TI's was not kidding about the 18 volt absolute maximum rating of their old LinCMOS stuff.  At some point I got into the habit of adding 16 volt TVS diodes across the supply to protect against spikes.

[iMo]

Guys, let me kindly ask experts here - I've found in my junkbox an LMC662 SMD still soldered into a pcb and a "500000 Megaohm VICTOREEN" resistor in glass, my bet it means 500GOhm
While considering every hint in this thread - it is even feasible to try with it?
Like an desoldered SMD 662, cleaned up in IPA, its input bent off the pcb, that resistor with a couple of pF gimmic capacitor (made of enameled wire, for example) in parallel, all built on teflon standoffs (or dead bug in air).
Asking before spending too much time with elaborating an apparent nonsense

PS: below a scrap device with that resistor on it - the resistor is wired at the input of the HDIG1030 transistor, perhaps an ion detector or something like that..

PPS: I've been trying to read the pcb and it seems it is a TIA amplifier with the resistor in the feedback, the white block looks like 100J capacitor in parallel with the 500G resistor, and together with the uA776 it creates an amplifier with the socketed HDIG1030 at the input, basically the same as the PAM here.. Hmmm, I have to reverse engineer it..

PPS: HDIG1030 - PMOS ENH (3N164 eq. ??)

[iMo]

FYI - here is the schematics of the probe I quickly draw (it had something with probing ions in a chamber many decades back, afaik). I think they adjusted the offset from outside via the OFFSET input..
Interesting the RC constant of the FB (500G || 100pF, if I read the capacitor value properly).

Nope: the capacitor is Vitramon 10pF/500V, Porcelain

[magic]

It should work, I suppose. With such high resistance you may find you don't even need a capacitor for stability, unless signal source capacitance is really high.

This thing probably has a fraction of pF in itself, enough to slow the TIA down to a few Hz.

[zrq]

I have two of such Victoreen glass sealed resistors salvaged from old vacuum instruments, both of them are pretty inaccurate (up to -10% off) today and exhibit quite significant voltage coefficient. For TIA feedback with limited dynamic range it maybe fine, through.

[David Hess]

A t-network can replace the high valued resistor in a transimpedance amplifier at the cost of increased noise, offset, and drift.

[Gyro]

Personally, I think it would be a shame to strip that thing for its feedback resistor. I know it would be easy to remove but that assembly is nicely made, complete with a low leakage relay etc. I don't know what the gate leakage spec is for the HDIG1030, if intact, but it must be very low to make a 500G resistor worthwhile (btw, a quick forum search show that this mosfet is used in the Keithley 155 Null detector / Microvoltmeter). The other problem is that even the 2fA typical leakage of the LMC662 is probably going to cause a significant offset with a 500G feedback resistor. This might take some a fair amout of nulling out. You're going to end up with a fairly long time constant when you take leakage capacitances into acount [Edit: maybe not], which might be inconvenient for measurements - not to mention inconveniently high input sensitivity.

Your choice of course, but 1G resistors aren't hard to purchase these days, so don't pose a particular problem that would require a salvaged part. Likewise an LMC662, it is probably better to start by IPA cleaning an already clean fresh (preferably DIP) package than one that has been desoldered a board.

[iMo]

Before I remove the resistor from that probe - it could be the probe still works, thinking how to wire it such it shows some signs of life.. Unless the input fet is dead already (50/50) it may show something, there is nothing special there except the fet. It is definitely not clean enough anymore so I would not expect miracles.. 
The sim below shows it may work with a 10t 10k pot and a bias resistor.
With the relay ON I set zero (with the pot) at the output and with relay OFF I should get then +/-10V out with +/-20pA in.. Will try tomorrow.. 

[iMo]

I found a doc from Hughes Aerospace on some tests they did for NASA in 1973/74 - the HDIG1030 chip shot there.. (long before Noopy..)

[Gyro]

If it's NASA, that maximum Gate Leakage spec is probably over an extended temperature range. I would expect the typical gate leakage at ambient temperature to be a few orders of magnitude lower.

P.S. Looking at the Keithley 155 threads I mentioned, they used selected parts. I wouldn't be surprised if the part you have there is selected for low leakage too.

[iMo]

Yep, the transistor sits in a socket, thus they hopefully made a preselection, only the gate is wired to the resistor in free air. I've seen somewhere it got 20fA typically, but I cannot find the info anymore.
Btw - here is the Hughes' report for NASA..

https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=&ved=2ahUKEwi3h4TEqcj_AhVtxQIHHcCJCyAQFnoECA8QAQ&url=https%3A%2F%2Fntrs.nasa.gov%2Fapi%2Fcitations%2F19750002239%2Fdownloads%2F19750002239.pdf&usg=AOvVaw1uB3o3vRhWZqYLJ27LeBRi&opi=89978449

[iMo]

OK, I've invested some effort and wired the unknown probe as below..
As usual the most troubles made me the reed relay.. I almost resigned with my attempt to run this 40+ years old device, until I recognized the relay's contacts are switched "on" (shorted) when the coil is not powered. Sure, this was the great idea - as thanks the input shorted with the output (actually via 39k to opamp's output) helped the HDIG1030 survive rough handling and the transistor still WORKS!

I added inside the module an 0.47u capacitor (tantalum drop) across the opamp's power and a diode at the relay coil.

With the relay not powered (the contacts shorted) I can set zero (50k 10t) at the opamp's output (OUT1), the zero then fluctuates slowly some 500uV up and down in time. The voltage at the HDIG1030's source (IN-) is +2.855V with the 13k source resistor and 15V Vcc, when zeroed.

With the relay powered (the contacts disconnected) the OUT1 voltage settles from initial jump after about 20secs and fluctuates then some 80mV peak-peak from +300mV to +380mV in "quiet periods" (from 1.5m distance, while watching it I am interacting with the electrons at the input - the wave functions collapse, when not watching it could be much less  ), and much more based how you move yourself around. While messing in vicinity of the input it sometimes jumps into volts level, thus it seems the transistor works.

I have to investigate the role of the R3=39.2k, on the pcb there is written an old remark "R75=39.19" and the board is populated with the "39.2k" one.. Any idea?? 

PS: ok, the shorted part of the FB is outside the 39k, thus the 39k resistor creates the +300mV offset, it seems.

PPS: Correction - the capacitor is 10pF/500V Porcelain Vitramon

[Gyro]

Looks very promising. That n/c relay has clearly saved the day (and the gate). 

From experience with my Picoammeter (1G feedback resistor), you won't be able to get a true stable zero unless the outer case is fitted and the input needle is fully screened (not shorted), say, with some foil formed as a 'bulb' and held with an elastic band around the end of the case. It would be interesting to see how stable the zero is under those conditions.

Regarding the 39k2 resistor, I think that is probably a fine gain trim to compensate the actual, as opposed to marked, value of the 500G resistor - 39k2 is close enough to 39k19 (maybe its measured value is 39k19). It will be affected by (any) loading on the offset pin anyway.

[mawyatt]

Wow, a uA776

Haven't seen those used since 70s when we used them for a number of special applications.

Best,

[iMo]

..
Regarding the 39k2 resistor, I think that is probably a fine gain trim to compensate the actual, as opposed to marked, value of the 500G resistor - 39k2 is close enough to 39k19 (maybe its measured value is 39k19). It will be affected by (any) loading on the offset pin anyway.

My feeling is the in/out marked "OFFSET" in my schematics is an input for compensating the zero, like with a DAC. The value 39k was somehow important for them to get that default offset (after setting the "zero" with the voltage at IN+ input) into a certain DAC range.
The 500G resistor's and fet/opamp's TC may then fit into the "DAC offset range", like from 0.0 to 1.0V DAC's output (after the initial coarse setting by the voltage at IN+).
The actual "500G value" spread and long term drift could be by many orders of magnitude higher than the 39k value, my guess.. (ie with TC=10ppm/C, what is an absolutely unrealistic value, the 500G will change by 5Mohm with dT=1C).

[magic]

You will see nothing but noise if this thing isn't shielded from static fields. Get a strip of sheet metal and wrap it around the probe, or try to find a matching pipe that would fit around it. It all should be grounded and I'm pretty sure that the metal caps at the ends are grounded, so just connect with them.

If the capacitor is really 100pF, the time constant is ridiculously long, almost a minute.

[iMo]

Sure, this is just a smoke test, it must be perfectly shielded (and it was originally, afaik it was a part of a particle detector put into a metallic tubing, and that all was inside a metallic tank filled with some noble gases). Therefore they limited the number of parts on the pcb (ie none decoupling capacitors). The connector's pins go through glass. The capacitor is marked 100J (100p) for 500V in ceramic or glass, definitely something special. On the first glance the responses are much faster than the 500G||100p would allow, that is because the rather high contamination, imho (a lot of people who do not read Metrology section were touching it)

[Kleinstein]

The capacitor marked with 100J could be 10 pF as 10 * 10^0 pF. That would make more sense speed wise.

[iMo]

I've found an info on ebay 
10pF 500v 5% Vitramon Porcelain Axial Capacitor CY13C100J MIL-SPEC HiRel Ceramic  ($15) ..
Vitramon now part of Vishay

[iMo]

I put the probe only into a small metallic cookie box (A5 format) and grounded the box to GND at the test pcb (outside the box). Set the zero with the pot to -0.1mV. After opening the short after aprox 50-60seconds it settled at aprox +4mV DC, with noise aprox 2.5mVpp.
Thus those +300mV offset disapeared (??). While observing the stuff from a 3m distance the noise was still the same, there was a small DC move up by 2mV during first aprox 5minutes. Still sensitive to touching the box, however it jumps by aprox 50+mV.
I will continue later on with some better measurements, but have to organize the experiment, most notably grounding/shielding, there is an o-ring on the probe which may isolate the probe's body while laying in the box (the front aluminum part with by teflon isolated needle is floating). Promising results, indeed.

[iMo]

I arranged the measurement somehow with better grounding inside the box, I put the outside pcb into the box as well. Outside left are the voltage sources (723 based floating voltages). I added 330nF foil at the OUT1 (thus 1k/330nF low pass at the output). Still a little bit sensitive and picking noises from outside via the wires, though (touching the box causes 3-4mV peak).
This setup is a monster, I would say

Below logged a short measurement from yesterday - with 34401A, 100NPLC (4.012 secs sampling period). The DC did not move, moreover I actually measured the Vcc's 723 voltage reg stability (PS: and noise as well), thus the DC drift has little sense in this measurement, imho.

The lowest noise I saw was 20uV stddev (running over 100 samples), the average in this monster setup in peaceful periods around 30uV, afaik. I tried to calculate the noise in pA/fA but as a novice I got weird results.. 

So the experts may help here - the input free floating in the box, 500G||10pF, 4.012sec sampling period, and say 20uV RMS best case at the output.

[magic]

As a sanity check, you could try with shorted relay to measure noise of the electronics alone. At least their voltage noise, because gate current noise of the source follower (if there is meaningful leakage) will not be accounted for in this configuration.

Not sure how much shielding you have applied around the high impedance electronics (the "probe"). It really should be enclosed in a grounded can. A larger box may not be enough, any wire entering the box which carries voltage noise or is high impedance (has a potential to pick up noise from outside the box) may inject noise into the probe circuitry.

The large bursts of noise seem to be something external being picked up.

The source follower adds that 2.855V offset and so your IN+ is not GND. Any noise on IN+ will directly appear on the output; hopefully your filtering sorts this. This noise would of course also be included in the "shorted relay" test.

[iMo]

Yep, this setup is a nonsense.. The filtering at the IN+ is not sufficient, there is 4k7/330n low pass only. As I wrote above I made the 723 voltage regulator noise measurement actually.. This TIA architecture would require a low noise reference voltage at IN+, the better cleaning, shielding, the Vcc/Vss inside the box. I may try to power it from my 20V battery, with a rail splitter (the reed relay consuming the most current), but my motivation is low.
Btw shorting the relay contacts moved the DC up by some 700uV, and the noise so far is the same (around 35-40uV stddev so far).
Not sure whether it has sense to mess with this monster further on.. Best would be to desolder the 500G resistor and the 10pF capacitor and the rest trashed..

[magic]

Btw shorting the relay contacts moved the DC up by some 700uV, and the noise so far is the same (around 35-40uV stddev so far).

Bad news: the electronics are noisy for some reason.
Good news: it's easier to hunt noise sources in low impedance circuitry like it's now

The only difference the relay makes is shorting the 500GΩ, so 700μV shift implies that there was 1.4fA flowing through it and that's your input leakage current. Not too bad.

edit
It has occurred to me that the probe contains a bipolar opamp, so there is IN+ current noise which may or may not be a problem. Either calculate what's the resulting voltage noise when it flows through your 4k7||330n impedance or throw a larger cap there or short IN+ to ground and AC couple the output for a quick test.

It has also occurred to me that this level of noise is perhaps acceptable, considering the ridiculous sensitivity of this thing? It's 500μV per 1fA, if my math is right.

[Kleinstein]

The noise level is actually quite low for the TIA when taken into account the sensitivity. A large resistor naturally has some noise and also the FETs are not super low in noise. So there is no need for super low noise for the offset / bias at +in.

The fast variable parts still look a bit strange, like interference or maybe something oscillating. Some amplifier don't like capacitive loading (e.g. shilded cables or a DMM). So it may be be a good idea to have some 100-500 ohm series resistance at the amplifiers output.

[Gyro]

...
Not sure whether it has sense to mess with this monster further on.. Best would be to desolder the 500G resistor and the 10pF capacitor and the rest trashed..

I still think that would be a crying shame. That thing is insanely sensitive and the 1.4fA input leakage current is very respectable, it might be even better after cleaning if it has been lying around in a junk box and you've been touching the insulatoror other parts with your fingers. The noise level doesn't sound that bad (for DC measurement use). I'm not clear on the source, but it might be possible to substitute a more recent, lower noise opamp. The physical construction is good and would take some significant effort to reproduce.

Before pulling the resistor, you need to consider what purpose you want to use your picoammeter for. 0.5mV output per fA is going to limit you to a very few applications - it would be useless for measuring the leakage of most components, reverse leakage of semiconductors, where the output would immediately peg to the supply rail.

Having built a 1mV/pA picoammeter (reply #23), I have found this perfectly adequate for all my leakage measurement needs (even evaluating insulating sleeving). I would still find a 1G, or maybe 10G resistor if you're feeling the need for a little more sensitivity (and noise).

As for the capacitor, I hadn't realized that Porcelain capacitors were actually that low leakage, but it is something that could be easily matched with an axial Polystyrene capacitor with a lower value feedback resistor - you would want a higher value anyway (330pF in my case).

If you have low motivation for this unit at the moment, I would wrap it carefully and store it in dry conditions. Either that or pass it on to another member. I certainly wouldn't trash any of it, the mosfet for one, has value for other instrument repairs as previously mentioned.

Just my tuppence worth anyway.

[magic]

I wonder if leakage current could be reduced by biasing the drain to a lower voltage, maybe -3V or thereabouts to be opposite to source voltage. Like all bias cancellation tricks it will probably increase noise current, but that is maybe still lower than other noise sources. The feedback resistor alone has 0.18fA/rtHz Johnson noise density.

You could try different bias currents on the FET.


To hunt for noise sources, short things out. Short IN+ to ground to eliminate opamp current noise from the equation. Close the relay, short the 39k2 resistor (WTF is even its purpose anyway?) and then short the FET gate to source to eliminate the FET and see voltage noise of the opamp alone. Always keep in mind that 1mV is 2fA - this puts things in perspective.

Ideally, you want noise to be dominated by the resistor (and maybe leakage currents flowing through it). Which is to say, overall noise should decrease when the relay is closed.


Thermal drift of the FET's uncancelled Vgs may become another annoying factor. It may be possible to compensate for it by manipulating IN+ bias; decent thermal tracking between components is necessary, of course. Drift may also become zero at one particular drain current, possibly impractically large.

[iMo]

The last attempt: here is the schematics of the testing setup after some changes - the IN+ filter is now 33k/470u and I split the source resistor as well (6k8/5k6/220u). I set the zero to some +4mV two hours back and now I read +17mV out at this moment and waiting till it stabilizes somehow (because of the large leaky aluminum caps). Just now I see 35uV stddev (shorted contacts). I will report later on with some graphs.

@Kleinstein: I have the 1k output resistor at the opamp's output from the very beginning .

[David Hess]

I wonder if leakage current could be reduced by biasing the drain to a lower voltage, maybe -3V or thereabouts to be opposite to source voltage. Like all bias cancellation tricks it will probably increase noise current, but that is maybe still lower than other noise sources. The feedback resistor alone has 0.18fA/rtHz Johnson noise density.

Bootstrapping the drain to the source is commonly done to control and reduce JFET gate current.  (1)  The JFET and signal source are likely so noisy that any increase in noise from the bootstrap circuit will be insignificant.

I do not know if it would improve the gate leakage of a MOSFET; I have not tried it.

(1) JFET gate leakage increases with drain voltage through impact ionization?  Pease mentioned it in one of his articles after he and a friend rediscovered this phenomena.

[magic]

IN+ filter is now 33k/470u and I split the source resistor as well (6k8/5k6/220u). I set the zero to some +4mV two hours back and now I read +17mV out at this moment and waiting till it stabilizes somehow (because of the large leaky aluminum caps)

I suggest methodically locating noise sources rather than random component swaps

Bootstrapping the drain to the source is commonly done to control and reduce JFET gate current.  (1)  The JFET and signal source are likely so noisy that any increase in noise from the bootstrap circuit will be insignificant.

I do not know if it would improve the gate leakage of a MOSFET; I have not tried it.

I generally expect all things to have non-zero, and usually positive, admittance. The logic with MOSFET is of course that the gate and its dielectric covers the whole length of the channel, so if one end is positive and leaks positive current (as we found) taking the other end negative will hopefully make it leak negative current and balance things out.

I don't know what's the drain breakdown rating of this FET.

[iMo]

Btw the 776's supply current according to the Fairchild's DS is set to some 40uA with the 4M7 resistor, provided I identified the resistor properly (I saw around 5M).

[iMo]

.. I don't know what's the drain breakdown rating of this FET.

Here are some params..
I set the Ids to 1mA, it could be it is a wrong setting, provided the 776 runs micropower..

[ch_scr]

If it was meant to be in high vacuum, the parts will have a lot harder time cooling down, with no air convection.
Maybe originally the FET power dissipation was set similar to the 776, or in the ballpark at least?

[iMo]

This is the latest 4hours run, shorted relay contacts, as per latest schematics above.
STDDEV aver ~35uV, min ~20uV.
Drift some 200-250uV up.

[iMo]

I put the probe into a plastic tube covered with Al foil (the needle is not touching the foil nor the plastic), the pickup is much lower, the noise seems still the same on the first glance, I will do some measurements later on during night.
In the meantime I tried with opening/closing the relay (btw the relay's coil is the major contributor to the pickup) - see below. The DC with shorted contacts moved a little bit up and down as I put the stuff on the top of my HP meter, then I replaced it while still sampling data (as the meter creates heat, obviously)..
I also changed the fet's source resistor to 68k.
The best case difference between opened/closed is aprox 1.5mV in this measurement (I saw below 1mV already) after 20-30minutes.
Experts here may judge on other params from that charging shape..
PS: the time axis is in HH:MM

[zrq]

The minutes scale slow recovery after releasing the relay (like switching off Zero Check on Keithley electrometers) looks like dielectric absorption.
Maybe from the feedback capacitor or the resistor, charged by the thermal emf of the relay. Wrong, that would be much smaller.

[David Hess]

Bob Pease mentioned using relays which have electrostatic shielding between the coil and contacts, and driving them with the minimum change in voltage needed for operation, to minimize coupling from the drive signal to the switched signal.  Eventually he used a mechanical switch with a plastic rod for actuation.

I would have tried using a reed switch with an electrostatically shielded coil however I assume it would perform poorly because this does not seem to be common.

[iMo]

The minutes scale slow recovery after releasing the relay (like switching off Zero Check on Keithley electrometers) looks like dielectric absorption.
Maybe from the feedback capacitor or the resistor, charged by the thermal emf of the relay. Wrong, that would be much smaller.

While shopping I was thinking on that long recovery - I think it comes from the capacitor created by the needle and the outer AL foil. The needle is floating in air (it goes through the teflon isolation), then there is a plastic cap about 1cm over the needle (such the needle does not hit the AL foil), then air, then the plastic pipe and finally the grounded AL foil. That creates the capacitor.
The first sharp fast peak comes from the 10pF FB porcelain capacitor, imho, the slow recovery part from that "pipe capacitor", I would say. I will try to simulate that..

PS: the front part of the probe made of AL with teflon is floating, not grounded, perhaps it carries a charge too..

[Kleinstein]

The input see a relatively low impedance. So the input capacitance discharges relatively fast.

The fast spike is likely due to capacitive coupling from the relay coil or drive part to the input in some way. It only need a tiny bit of coupling capacitance (0.05 pF range) to charge the 10 pA to -35 mV.
The parasitic capacitance may also be responsible for the dielectric absorbtion to cause the slow settling. To test this one could use different times of the relay turned on.

[iMo]

..The fast spike is likely due to capacitive coupling from the relay coil or drive part to the input in some way. It only need a tiny bit of coupling capacitance (0.05 pF range) to charge the 10 pA to -35 mV.
The parasitic capacitance may also be responsible for the dielectric absorbtion to cause the slow settling. To test this one could use different times of the relay turned on.

You are right again .. (.. how this guy does it )
Where to put the parasitic capacitance to simulate the slow part then?

[Qmavam]

Some will find this 19th century Galvanometer interesting.

                      Mikek

[Kleinstein]

The slow part could DA in the tiny 0.04 pF capacitor. The model would be something like a parallel  even smaller capacitor (ca. 0.1-1%) and a huge (Pohm range to get the right time constant) resistor in series.
I am not so sure where the slow part comes from, but loss in that parasitic capacitance is a candidate.
There are a few other effects to cause the recovery. There could be surfaces that are partially isolated and add extra parasitic capacitance to the input. These surfaces can slowly charge / discharge.
There is also the possiblity to have some thermal effects, e.g. from the relay cooling after turned off. The temperature may effect the FET - though the effect looks quite strong for this.

[iMo]

Ok, finally below 3 pictures:
1. a long run with relay contacts shorted, noise around 20-25uV stddev in quiet periods, it continued straight into the
2. overnight run with relay contacts opened - you may see the switching the relay at the very beginning
3. the same overnight run with relay contacts opened - but the detailed view after the voltage settled (at aprox -0.9mV)

After the 3. the DC voltage with relay shorted is -0.27mV aver. Thus the difference is aprox -0.27-(-0.9)=0.63mV.

PS: HDIG1030 source resistor 68k+5k6, Vds=2.108V, Ids=0.180mA

PPS: The slow part after the relay opening contacts is aprox 2 hours long..

[iMo]

Perhaps did it somebody here - a TIA calibration with ramp input voltage and capacitor generating pA level currents (see below the sim).
Not sure whether it is even feasible..

[ch_scr]

Shodan from ampnuts.com has built multiple revisions of this idea. Here is a demonstration video of his latest finished revision. More information can be found searching for GLIN on his blog. Reading the site with Chrome auto translate works well.
I have tried the idea with the Yokogawa 7651 as ramp generator, but at least my unit is not linear enough so there were periodic jumps in the current. Not sure if my unit is broken or out of cal.
Edit: Here is the github for GLIN v.2

[Kleinstein]

The calibration with ramp and capacitor is defintely possible. A point to observe is that the TIA may need some series resistance at the input. Not all TIA circuit are happy with a low impedance or capacitive source.

[bsw_m]

calibration with ramp input voltage and capacitor generating pA level currents

This technique of generation of ultra-low currents was used in at least two calibrators developed by MNIPI, these are ЕК1-6 (in mass production since 1979) and НК4-1 (in production since 1989, until about 2014).
The minimum current that these calibrators could reproduce is 1E-17A.
I began to describe the calibrator NK4-1, but as I did not meet much interest in this device - a branch stalled: https://www.eevblog.com/forum/metrology/nk4-1-low-current-and-high-resistance-calibrator-(made-in-belarus)/

[bsw_m]

Shodan from ampnuts.com has built multiple revisions of this idea.

Making a ramp generator is not such a big problem. The main problem is differentiating the ramp signal (ramp-to-current conversion). And this is primarily the design, not the circuitry, or the software. .

[iMo]

Ok, while reading above documents I get the method with the ramp voltage and capacitors works fine in the LTspice only
In the real life you would need special capacitors made of sapphire, etc.

[bsw_m]

All in all, it's not as complicated as it seems. But there are nuances, of course.
Here are pictures of the 100pF and 10pF air capacitors used in the EK1-6 and NK4-1 calibrators to get the 1E-12A and 1E-13A direct current ranges.
Lowest capacitor, that used in these calibrators is a 0.1pF. Highest capacitors that used in these calibrators is 11200pF teflon dielectric (2x parallel K72P-6) capacitors.

In general, to form the 1E-12A range, you can try to use PP type capacitor, of course with the leakage current selection and the minimum DA.

[David Hess]

Bob Pease mentioned that the problem with larger air dielectric capacitors is that they get charged from cosmic ray hits, so a dielectric like Teflon is required to reduce the volume.

[bsw_m]

The large air capacitor can to some extent be considered an ionization chamber. Therefore, indeed, the condensers should have the smallest possible dimensions. But with proper air capacitor design, up to 100pF air capacitors are the best choice.

[zrq]

I tested my 1 TOhm Victoreen glass sealed resistor more carefully today with a Keithley 617. It's marked with 2% tolerance, but the linear fit of the IV curve gives 885 GOhm, way out of tolerance. Per the fit residual, the voltage coefficient is up to 300ppm/V, which is also quite poor.

[ch_scr]

You've inspired me to try my hand at this, since I've had all the parts for it in the drawer already.
I've used GDR-made SMY60 (dual P-FET without gate protection zeners) as a input stage into an OP07, 1G KVM in the feedback.
The 3pF feedback capacitor is made of three 0805 NP0 10pF in series (is that a good idea?).
I've just now made and set it up:
- shorting the input (wondering why the output pegs  )
- measuring at the two inputs of the opamp
- then set the balance potentiometer of the input stage to get zero across -IN +IN.
To try it out:
- remove short, put 100Meg towards the bias voltage source on the outside
- feed in 1V, expect -10V out.
But out come only -7.42V. With -1V in, the output is 7.42V. With no input, output is pretty much zero
Possible conclusion to me:
- Feedback cap is leaky, shunting the 1G
- 1G is lower or dirty
- 100M is higher (metal film, this much drift up seems less likely?)
Is the feedback cap the likely culprit, or is it dirt on the 1G? Or something else entirely?

Edit: nevermind, just realised it's oscillating, 3pF seems too little.
Edit#2: with 100pF PS foil capacitor, after waiting a loooong time (for the capacitor to fully dry in circuit), it settles at 7.41V. So not the capacitor leakage?
Edit#3: These 100M all read high  The one used measures 120.8Meg. With 1.208V Bias, -8.96V comes out. (-1.208V -> 8.97V) Still "a little" short, but closer to reasonable.
Edit#4: At the same time / from the same ebay seller, I got other 1G, an 100M, etc. among other values. After quick rinse with IPA, the 100Meg reads low at about 90M, and the other 1G reads ~905M (even though I have less confidence in the latter measurement). Coincidence, or did these KVM all drift low with age? I should have measured the one I put in beforehand 
Edit#5: With 120.8Meg & -12.08mV Bias the output is 94.5mV, at +12.08mV the output is 85mV. Is that explaineable as ~5pA leakage?
Edit#6: With 0.57mV Bias on the 120.8Meg, the output is zero, so ~5pA input leakage confirmed?

[bsw_m]

Coincidence, or did these KVM all drift low with age?

High value (>1GOhm) KVM resistors most often drift in the direction of decreasing resistance.

[magic]

IME they drift down with age, but I only recall the 100G being more than 10% off.

I tested them by applying 100V DC in series with a 10MΩ multimeter.

[ch_scr]

Coincidence, or did these KVM all drift low with age?

High value (>1GOhm) KVM resistors most often drift in the direction of decreasing resistance.

IME they drift down with age, but I only recall the 100G being more than 10% off.

I tested them by applying 100V DC in series with a 10MΩ multimeter.

Even the 100MΩ is definitely down to 90.7MΩ. (Cleaned it more thoroughly and tried again) I have a lot more confidence in that measurement than the 905MΩ one. Did test them with up to 120V. Even going from 10V to 100V the voltage coefficient was noticeable. Not that it would matter in this usecase though.

[TizianoHV]

100Mohm shouldn't be that hard to measure. You can find much better (1% 100pm/°C) 100M resistors at affordable prices.
I dont like the look of these binding posts in the last photo, these could leak. A easy and accurate way to measure resistor up to a few Gohm's is to use a 100kOHM resistor as shunt and a 6dig multimeter to measure the current across the DUT while being careful to shield sensitive conductors.

Would be interesting to see if adding guard rings could help reducing voltage coefficient/leakage of these glass resistors

[bsw_m]

Would be interesting to see if adding guard rings could help reducing voltage coefficient/leakage of these glass resistors

My tests show that it will not help

I dont like the look of these binding posts in the last photo, these could leak.

If you mean the glass beads in the shielding box, they should not be a problem, the amplifier input has a potential of almost zero. No potential, no leakage. The box should be connected to the measurement ground.

[David Hess]

100Mohm shouldn't be that hard to measure. You can find much better (1% 100pm/°C) 100M resistors at affordable prices.

I have measured up to 1 gigaohm using a common 10 megohm input multimeter.  Use the most sensitive DC range with a 10 megohm input resistance as a current input and apply a fixed voltage to the resistor.  With a resolution of 1 millivolt, 1 millivolt across 10 megohms is 100 picoamps.  10 volts across 100 megohms is 100 nanoamps, so the measurement can be made to 1 part in 1000.

[ch_scr]

I dont like the look of these binding posts in the last photo, these could leak.

I've checked just now and the instrument reads 0.3nA while sourcing 202V into the open binding posts.

If you mean the glass beads in the shielding box, they should not be a problem, the amplifier input has a potential of almost zero. No potential, no leakage. The box should be connected to the measurement ground.

Box is not connected yet, good catch.
I've realised it's not particulary clear from the picture (the installed shorting link doesn't help either), but the DUT (resistor, low leakage diode or the like)
will be in the box together with the amplifier, and the beads only carry supply and bias voltages in and measured voltage out.

[magic]

Even the 100MΩ is definitely down to 90.7MΩ. (Cleaned it more thoroughly and tried again) I have a lot more confidence in that measurement than the 905MΩ one. Did test them with up to 120V. Even going from 10V to 100V the voltage coefficient was noticeable. Not that it would matter in this usecase though.

Similar experience to mine. I tested several specimens of different values, it was a few years ago, but I'm fairly sure all were found to be 5~10% off, always on the low side. Except for 100G, which was 72G. I felt like something may be wrong with my 100G measurement, but the fact that it came within 30% gave me confidence that at least the lower ones are likely right. No amount of cleaning the surface changed anything appreciably. I then obtained another batch of 100G and found them to be ~90G, kinda in line with the lower values, while my first 100G was still 72G. I also bought other 1G resistors, from Ohmite and a different Soviet type (not glass) and these tested in spec following my procedure.

Conclusion seems inescapable that KVM resistors just aren't that great. They may still be used if no other choice (e.g. 100G or 1000G - fairly exotic) but individual calibration is necessary. In the TIA picoammeter, one could insert a corrective voltage divider between the opamp and the feedback resistor. SNR will be somewhat higher than with a true 100G, but still better than with 10G or 1G.

Note that KVM are only rated for 100V.

edit
What I haven't done was to re-test at 10V to check voltage coefficient. Back then I didn't even know that I was already running them at their maximum rating, I assumed they would be good for at least a few hundred volts. So I don't know how much VC they have - could it be on the order of 5% at 100V and responsible for my results?

edit edit
Point 9 of the datasheet does specify "change in resistance as a function of change in voltage, no more than ±5%". So maybe it's partly responsible?

I'm attaching the datasheet. I have a scan of the original, a transcript found somewhere, and a machine translation of the transcript. Can't vouch for the accuracy of the latter two, but they seem about right at first glance.

Values up to 10G have specified TCR: ±2000ppm/K.

[iMo]

I've found 2 none-glass resistors - 300Meg and 500Meg in my junkbox, as well as an LMC6062 in smd (I previously wrongly reported as the 662). While being used in a T configuration - what could be actually achieved (noise regardless) in such a setup? Any formula known for the T gain calculation with such an asymmetric resistor setup?
Also - I have found couple of n-channel mosfets KF521 - I wonder how to replace that p-mosfet with them easily..

[magic]

Put them in series for 800MΩ, feed them from an 80% divider. Sensitivity will be 1mV/1pA as with 1GΩ. Not sure what else you need to know?

Unlike μA776, LMC6062 has low enough bias (10fA) that you could try it without external FETs.

[iMo]

So far I've made a sim with the T in the FB..
Aprox 10pA/V output..

PS: and with the 80% div in the FB

[ch_scr]

The "T" is often misunderstood. I know, because I did too when playing around with it in spice, trying to make sense of it.
R8 / R9 form a voltage divier, before going into the R2 feedback resistor. That's why there is no need to waste 300Meg on R8.
magic suggests to form an 80% divider with more regular values for R8 / R9, because it's just a voltage divider anyway!
Then have R2 as 300M+500M and take only a slight performance hit from dividing to 80% - but gain a convenient way to trim the output to 1mV/pA excactly.
I guess the 1G in my case beeing low is a blessing in disguise, having such a convenient way trim the output.
Compare that to the hassle of trimming a high one down with an excact parallel resistor 

[magic]

So far I've made a sim with the T in the FB..
Aprox 10pA/V output..

600μV/rtHz output referred noise from Johnson alone, 5mVpp or more in DC-1Hz bandwidth.

Does SPICE agree?

SNR is maximized by putting all your high resistors in series for feedback, until you have a problem with input current being too large for the output voltage swing of the chip. On second thought though, the difference between 800MΩ and 500MΩ will not be huge. But with several mV output noise there is perhaps no point having so much gain...

Note that your circuit is almost equivalent to a simple TIA with 500MΩ feedback followed by a 200x gain stage. (The only improvement is, noise and offset of one opamp instead of two).

I guess the 1G in my case beeing low is a blessing in disguise, having such a convenient way trim the output.
Compare that to the hassle of trimming a high one down with an excact parallel resistor 

I would divide the output of the TIA down by a few % or whatever it takes. But then it may need to be buffered somehow from external loads.

[iMo]

With the 80% divider for 800Meg FB resistor.

PS: Could it be in my Unknown probe the 39k resistor was actually a part of such an divider, such they were changing the probe gain from outside??

[ch_scr]

Some sort of addendum, here's my schematic for a "TIA with discrete dual-P-FET input stage". As we've learned above, a bit of divider action before the feedback would be advisable to be able to trim the output. Had it running half the day on sunday and didn't observe any noticeable drift on the offset.

[magic]

That's the high effort way. The lazy way is to plug them as source followers in front of a bipolar opamp, like in iMo's mystery probe.

Is there any reason for this gain stage? OP07 noise shouldn't be a problem (the FETs are probably worse) and its DC voltage gain is good enough too, I think?

edit
As before, I wonder if drain voltage has any influence on gate leakage current; perhaps there is a particular point where Ig drops to zero?

And regarding offset, there is a difference between offset voltage of the input stage and "offset" due to input leakage current. The latter can be hidden by temporarily shorting the feedback resistor. I think I would trim for zero voltage offset (with shorted feedback) in order to minimize leakage across insulation of the Hi-Z circuitry, input connector, any external cables, etc.

[ch_scr]

That's the high effort way. The lazy way is to plug them as source followers in front of a bipolar opamp, like in iMo's mystery probe.

Is there any reason for this gain stage? OP07 noise shouldn't be a problem (the FETs are probably worse) and its DC voltage gain is good enough too, I think?

edit
As before, I wonder if drain voltage has any influence on gate leakage current; perhaps there is a particular point where Ig drops to zero?

And regarding offset, there is a difference between offset voltage of the input stage and "offset" due to input leakage current. The latter can be hidden by temporarily shorting the feedback resistor. I think I would trim for zero voltage offset (with shorted feedback) in order to minimize leakage across insulation of the Hi-Z circuitry, input connector, any external cables, etc.

I've never implemented a discrete input stage before, so I worked from the LIS LS844 Appnote. Which made it sound like there is no downside (aside from low gain stability) to the "gain arrangement" compared to the "source follower".
I'll try the "shorted feedback" zero - I suspect it might be almost the same as now, with my ill guided "shorted input, measure -in to +in" zero attempt.
Will experiment with the effect of negative supply (and thus, drain voltage) on the leakage. I could also try to make the bulk/substrate substantially more positive than the source voltage?
I've taken it apart for now anyway; to add the "3x 100pF NP0 in series" 33pF feedback capacitor, an divider to get the gain up to 1mV/pA, mechanical improvement, more thorough cleaning (soapy water, DI water bath before IPA hose down, instead of only the latter) and measurement of the 1G resistor.

[ch_scr]

Similar experience to mine. I tested several specimens of different values, it was a few years ago, but I'm fairly sure all were found to be 5~10% off, always on the low side.
(..)
What I haven't done was to re-test at 10V to check voltage coefficient.
(...)
Point 9 of the datasheet does specify "change in resistance as a function of change in voltage, no more than ±5%". So maybe it's partly responsible?

The 1G I had used in circuit measures:
10V - 891M
100V - 857M
The other 1G specimen:
10V - 921M
100V - 911M
I'll try to use the latter now.
Edit: offset was found to be 150uV with the feedback shorted, so the "misguided" method worked fine as well, just higher effort.
Edit #2: the circuit is quite sensitive to the negative supply (and thus, drain voltage on the dual fet).
Not sure to measure the input leakage with open input or shunted over e.g. 120Meg? Did both:
Measured output:
state:   open        120Meg shunted
V-
-10V     600uV       8mV
-12V     0V            2mV
-14V    -700uV     -4mV

[MegaVolt]

Analog Devices provides a very handy tool for estimating bandwidth and noise for different amplifiers. Including for transimpedance.

Graphs show all components of noise including each resistor, input currents and op amp noise, etc...

https://beta-tools.analog.com/noise/

[ch_scr]

Analog Devices provides a very handy tool for estimating bandwidth and noise for different amplifiers. Including for transimpedance.

Graphs show all components of noise including each resistor, input currents and op amp noise, etc...

https://beta-tools.analog.com/noise/

They calculate 65uV RMS for a 1G / 33pF TIA. I've measured about 90uV RMS for my build, doesn't seem too shabby?

[MegaVolt]

They calculate 65uV RMS for a 1G / 33pF TIA. I've measured about 90uV RMS for my build, doesn't seem too shabby?

The output noise of the TIA can be affected by the input capacitance of the source. You can put a photodiode at the input and set its input capacitance and observe the increase in noise.

[bsw_m]

Current measurement ±1fA. Current source: NK4-1.
As a meter: ADA4530-1 integrating current-voltage converter
A bit of detail, a 10pF air capacitor is used as the feedback capacitor.
There is a differentiator at the output of the integrator. The output of the differentiator was measured with a voltmeter. No filtering/averaging was used.

[bsw_m]

Very simplified, the schematic of this I-V converter looks as shown in the attachment.

[dobsonr741]

About to build the classic one with LMC662 and 1G resistor, have 5x SO8 on cut tape. Should I bother picking the lowest input current one? In more details:

• What will be the symptoms of choosing a higher input current vs. lower in practical terms, how’s it going to limit my ability to measure leakage currents in the 10pA range?
• Given they are all from the same batch, should I expect huge variations in input current?
• Should i expect major input current variances between the two opamp on the same die, or not?

[magic]

I tried two chips and they were <0.01pA. I don't recall anyone here reporting a specimen with abnormally high leakage.

To measure yours:
- short across 1GΩ and measure offset voltage of the opamp
- remove the short, subtract previously measured offset voltage from the new output, divide by 1GΩ to calculate input bias current

[dobsonr741]

Wow, can not measure the input bias current. So low. Nulled out to <1uV offset when sorted. Short removed from 1G and it still stays within +/-2uV noise. Checked the 1G - it's really 1G. Air movements with open lid takes it 200uV away. Happy with the build.

Some of the measurements I did with it:
BAV199W leakage @5V: 100fA
CPC1025  leakage @5V: 700fA