EEVblog #1017 - Enter The World Of Atto Amps

[EEVblog]

Dave tears down the Keithley 617 Electrometer, capable of measuring sub-femtoamp (attoamps!) resolution.




Video on the relay matrix:

[amspire]

Here is the service manual with schematics:

http://www.download-service-manuals.com/download.php?file=Keithley-5513.pdf&SID=9pl9n45mrf1eenarb06m74k6e0

[Cerebus]

There's an interesting design feature that got missed.

That input op amp and JFET have just a +/- 5V supply, labelled +/- 5V(B), but the feedback loop is driven by that +/-100V voltage source. That means that the op amp input can see +/- 100V which is enough to release the magic smoke from any run-of-the-mill op amp.

So how's the magic smoke kept in? The +/- 5V(B) supply has its '0V' centre tap actually set by the output so it's not +/- 5V from ground, but +/- 5V from wherever the input is at, or to put it another way, the input amplifier supply is bootstrapped off the output. There are two reasons for doing this, firstly you couldn't find suitable JFETs that combined both low leakage and high voltage ratings, secondly it boosts the input impedance of the amplifier substantially.

Also, if you look carefully, that 250G resistor Dave mentioned as connected across the input isn't across the input, it's connected (via R337 at 2.2k) to a common rail labelled 'B', which is the bootstrap centre rail (or if you prefer, the composite amplifier output - see the 'B' connection near K301). That 250G is in parallel with the feedback loop, not across the input. That's the perils of a circuit diagram that defies convention and slaps the 250G feedback resistor down exactly where you'd conventionally draw a resistor that bridges the input.

Here's a, deliberately didactic, question for anyone who's interested in how this particular circuit works. What are R340 and R341 doing? They are respectively, 4.02k from the +5V(B) supply to the drains of the input JFET pair, and 402R from that junction to 'B'.

[Kryten 2X4B]

For anyone interested in making the type measurements of the type that the Keithley 617 was designed for you should get a copy of the Low Level Measurements Handbook

http://www.tek.com/sites/tek.com/files/media/document/resources/LowLevelHandbook_7Ed.pdf

I have no affiliation with Tek/Keithley, just a fan boy.

[LapTop006]

I picked up a 614 a few months back (and then had to repair it as the battery pack got loose in shipping), interesting how much more complex the 617 is.

Digikey have the Pomona 2-lug triax cable available 4725PO-ND, although it's not cheap. Sadly the 2-lug to 3-lug triax adapters seem to longer be readily available.

[gardner]

get a copy of the Low Level Measurements Handbook

Thanks for this.  It's interesting reading.

the electrometer can measure the voltage on a 500pF capacitor without significantly discharging the device

Dave?  I think you owe us some demos on this sort of stuff.

[grouchobyte]

For anyone interested in making the type measurements of the type that the Keithley 617 was designed for you should get a copy of the Low Level Measurements Handbook

http://www.tek.com/sites/tek.com/files/media/document/resources/LowLevelHandbook_7Ed.pdf

I have no affiliation with Tek/Keithley, just a fan boy.

This is a very well written document....a keeper. Required reading for all EEs.
Thanks for posting the link! Reminds me of the good old days at TEK when they regularly published gems like this.

[Kryten 2X4B]

I have a 619 Electrometer/Multimeter and two 220 current sources, like 617 they have the older 2-lug triax. Some time ago I converted these to the 3-lug type as cables are readily available and you can't plug in a standard BNC. The most cost effective replacement connector is the Pomona 5219 (Digikey 501-1312-ND or Mouser 565-5219). I've also used the 5219 for calibration fixtures for these instruments and this part has also been mentioned on other threads in the forum.

[EEVblog]

Also, if you look carefully, that 250G resistor Dave mentioned as connected across the input isn't across the input, it's connected (via R337 at 2.2k) to a common rail labelled 'B', which is the bootstrap centre rail (or if you prefer, the composite amplifier output - see the 'B' connection near K301). That 250G is in parallel with the feedback loop, not across the input. That's the perils of a circuit diagram that defies convention and slaps the 250G feedback resistor down exactly where you'd conventionally draw a resistor that bridges the input.

Dang, missed that!

[EEVblog]

This is a very well written document....a keeper. Required reading for all EEs.
Thanks for posting the link! Reminds me of the good old days at TEK when they regularly published gems like this.

This is not a Tek document, it was written by Keithley before they got aquired, it's the "bible" of low level measurements.

[EEVblog]

the electrometer can measure the voltage on a 500pF capacitor without significantly discharging the device

Dave?  I think you owe us some demos on this sort of stuff.

Yeah, it needs proper setup and cables though. Die cast boxes etc

[Kryten 2X4B]

Even just for "simple" high value resistance measurement you may need a guarded shielded resistor, so you have a resistor in a die cast box connected to a guard potential and that box is inside another die cast box which is at the shield potential. The two boxes are insulated for each other and you have a triax connector on the outside, not to mention the cable to the instrument. There is some time, effort, and cost to set this up. I'm guessing that Dave doesn't have the time to go down that rabbit hole.

[LapTop006]

the electrometer can measure the voltage on a 500pF capacitor without significantly discharging the device

Dave?  I think you owe us some demos on this sort of stuff.

Yeah, it needs proper setup and cables though. Die cast boxes etc

I actually have the proper Keithley test enclosure for that stuff, just sitting at work over in Pyrmont.

[EEVblog]

Even just for "simple" high value resistance measurement you may need a guarded shielded resistor, so you have a resistor in a die cast box connected to a guard potential and that box is inside another die cast box which is at the shield potential. The two boxes are insulated for each other and you have a triax connector on the outside, not to mention the cable to the instrument. There is some time, effort, and cost to set this up. I'm guessing that Dave doesn't have the time to go down that rabbit hole.

Not for this video.

[Smith]

I know the video is relating to amps, but the device is also used a lot for voltages and high resistances. It's an electrometer. Perfectly for measuring voltages without loading your circuit on high resistance aplications. Also good for measuring resistances upto hundreds of TeraOhms. Third big plus are the low burden voltages on all ranges.

About the meter itself, don't forget the low leakage caps over the feedback resistors. Probably some polystyrene caps. Remember, their resistance should be way bigger than the feedback resistors and stable ofcourse.

Btw making a resistor enclosure isn't that big of a deal unless you go upto hundreds of TeraOhms. Mine even uses BNC because of costs. Only hard thing is making some easy to use internal connection to the resistor. I used some scrap teflon on plastic standoffs and some push to insert connector.

[eV1Te]

You can't use any triax cable, you need a low noise cable, preferably an ultra low noise cable (there is a difference and maybe a video in it self) 
Make sure it's the 2 lug version for your instrument as well.

Also the burden voltage should be close to zero (less than 1 mV on all ranges, on newer Keithley it's less than +-10 uV on the lowest range if I remember correctly), could be a nice comparison to your micro current box. Too bad you will need another electrometer to measure the burden voltage... unless you build your own high impedance amplifier with a lmc6001 or similar.

There is a very good reason why you need such low burden voltage, because otherwise the burden voltage would cause a leakage current in the triax cable and especially the connectors between guard and input terminals. If guard is always at same potential as input then you won't have a current flowing.

[HighVoltage]

Very nice video !
I have the Keithley 614 and have been looking for a good 617 for a while.
Like usual when Dave makes a video about some test equipment, now the prices will probably double on ebay.

[kridri]

According to a post from Free_Electron the matched jfet is a 2n59xx with presumable a 2n5904. These can be bought at Linear System as a LS830.

They use specially selected matched jfets in TO71 6 pin metal body. The type number is  2n59xx something. I have them. i think it's 5904

There is only one company that makes em and they are only made on demand. I bought the only 4 they had in stock a few years ago. 50$ a pop...
i since used 2 of them to fix broken electrometers.

edit: Source of type number http://www.linearsystems.com/lsdata/files/product_selection_guide.pdf

[alm]

From the text at 1:58: Fempto. Is that related to the Jiga prefix? Looks to me that one of the four contacts for the center pin is damaged in the input connector (see at 4:40). Possibly by some idiot trying to force a BNC connector into it. But you might want to replace it with a three-lug version anyway so you can use the more commonly available cables and interface with modern equipment.

Dave, how did it take a low-current nut like you so many years to acquire an electrometer? They are available on eBay for affordable prices quite regularly, and are not particularly heavy to ship.

This is a very well written document....a keeper. Required reading for all EEs.
Thanks for posting the link! Reminds me of the good old days at TEK when they regularly published gems like this.

This is not a Tek document, it was written by Keithley before they got aquired, it's the "bible" of low level measurements.

I am pretty sure grouchobyte was referring to publications like the Tektronix Concepts series that discussed their technologies in more detail than any other publication. For example, try to find a discussion of scope probes with more technical information than their Oscilloscope Probe Circuits book.

[4CX35000]

I have a Keithley 617 which I bought damaged with MANY internal problems. It seemed to have been opened by someone with little or no knowledge of electronics who decided to cause as much damage as possible.

The damage done.

1. The main transformer was bent forwards as in the video and one of the brackets had broken off, this required reattaching to the transformer casing. The transformer is mounted on the main circuit board.
2. The Triaxial connector on the rear was the wrong type for the leads I already had and was damaged. Keithley supplied a Two lug connector which is similar to a BNC connector, but modern instruments use a three lug connector which is intended to stop someone forcing a BNC connector on and damaging the internals of the connector.
3. The triaxial connector coax wire which connects the triaxial connector to the measurement board inside the unit had also been badly damaged requiring replacement. I did this by botching the replacement using some dual core coax as I was unable to source the original or replacement.
4. The measurement board on top with the shielding on top had been removed in a violent manner damaging the plastic supports.
5. The front display/control panel is mounted on a circuit board with a flat wire DIL connector which connects the control panel to the main circuit board had also been damaged.
6. The power switch needed replacing. Damaged by violently removing the front panel control panel.
7. Two transistors on the main circuit board and used for the 100volt DC current source had been broken off and another transistor nearby replaced.
8. Replaced some of the larger capacitors on the main board near the front. Damaged.

In all I have managed to get the instrument working, but it does need a little more TLC to finish off the repairs.

[Dr. Frank]

Also, if you look carefully, that 250G resistor Dave mentioned as connected across the input isn't across the input, it's connected (via R337 at 2.2k) to a common rail labelled 'B', which is the bootstrap centre rail (or if you prefer, the composite amplifier output - see the 'B' connection near K301). That 250G is in parallel with the feedback loop, not across the input. That's the perils of a circuit diagram that defies convention and slaps the 250G feedback resistor down exactly where you'd conventionally draw a resistor that bridges the input.

Well, I directly thought, that this 250GOhm resistor would be used for the usual compensation of input bias current of the JFET. It has nothing to do with the feedback loop, as it's neither inside, nor parallel to the feedback loop.
Its purpose can be derived from the calibration procedure:

The +/- 5V from the potentiometer R348 is divided by 1000 to +/- 5mV (by double divider R339/R354 and R338/R337. This voltage drives an additional current into the inverting input node by R332 = 250GOhm, i.e. +/- 20fA. This bias compensation calibration limits the total input bias current to +/-1.5fA.

These 20fA are also the maximum allowable leakage current for the JFET pair Q308.

Very interesting circuit and video.

Frank

[(*steve*)]

Sadly the 2-lug to 3-lug triax adapters seem to longer be readily available.

I've been after a 2 lug male to 3 lug female for a while.

Beware of sites that claim to have availability but don't.  *Cough* Element-14 *Cough*

[pmcouto]

Keithley 6172 is a 2 slot male to 3 lug female Triax adapter.
But it’s not cheap…

http://pt.farnell.com/keithley/6172/triax-adapter-2slot-male-3lug/dp/2672724

I bought one of these adapters from Farnell a few months ago.
According to the information on Farnell's website today, they have 14 units in stock.

Regards,
Pedro Couto

[(*steve*)]

According to the information on Farnell's website today, they have 14 units in stock.

I asked to be informed when they were in stock.

I received an email from them saying they had them in stock in Australia.

Their web site said they had 16 in stock.

I ordered one.

I am informed there is a 5 week delay whilst they source them from Tektronix.

Their web site still said they had 16 in stock.

[mk_]

Sadly the 2-lug to 3-lug triax adapters seem to longer be readily available.

I've been after a 2 lug male to 3 lug female for a while.

Beware of sites that claim to have availability but don't.  *Cough* Element-14 *Cough*

something like this?
http://www.helmut-singer.de/stock/-148524350.html

They do have a lot of triax-stuff, don`t know if they fit.

[Cerebus]

...
It has nothing to do with the feedback loop, as it's neither inside, nor parallel to the feedback loop.
...

Oh, it most definitely is across the feedback loop. Read the circuit diagram very carefully, paying attention to all the common terminal symbols marked 'B'. You'll find that 'B' is connected to the output of the class B output stage (at the same place that all the other feedback connections start from at a star point), one end of that 250G resistor (via an insignificant 2.2k, where it also connects to that trim circuit) and the centre rail of the 5V bootstrap supply. The other end of the 250G resistor is attached to the amp's summing point.

[ocw]

I have both Keithley 610C and 616 electrometers.  I have a triaxial test cable to use with the 616.  Earlier this year I built a current to voltage converter using an ADA4530-1 IC.  It has switch selectable feedback resistors which included every decade between 1k and 1T ohms.  It has 100 zeptoamp resolution--none of this coarse attoamp stuff!

I searched for the switch with the best insulation resistance specification.  I didn't find anything with the desired figure significantly above 1T ohms.  I used Johnson/Cinch banana jacks as a first trial and found them adequate for at least attoamp measurements.  I used Teflon insulated wires and standard insulation and cleanliness precautions.

Attached are pictures of my ADA4530-1 A-V converter.  The second one shows it with my calibration box connected (before I labeled the ADA4530-1 switch).  That's the only way which I found to make extremely low current measurements without having noise contamination problems.  Inside of it is a 1.5 volt N cell battery which feeds a megaohm voltage divider resistors which has a 150 uV output feeding a 1T ohm 10% accuracy resistor (it has an actual resistance of about 950G ohms).  That produced the 160 aA current which I measured with 0.1 aA resolution.  That was to the point where I needed to take the ADA4530-1's bias current into account.

Attached is a view of my similar measurement of 1.6 fA.  No correction to the IC's bias current was made for this.  Longer tests had what's shown on the attachment continue.  I kept the recorded measurement short due to the precautions required to minimize noise.

While my Keithley meters can be handier to use, at other times the ADA4530-1 converter can provide better results.

[Dr. Frank]

...
It has nothing to do with the feedback loop, as it's neither inside, nor parallel to the feedback loop.
...

Oh, it most definitely is across the feedback loop. Read the circuit diagram very carefully, paying attention to all the common terminal symbols marked 'B'. You'll find that 'B' is connected to the output of the class B output stage (at the same place that all the other feedback connections start from at a star point), one end of that 250G resistor (via an insignificant 2.2k, where it also connects to that trim circuit) and the centre rail of the 5V bootstrap supply. The other end of the 250G resistor is attached to the amp's summing point.

Nope, in Ampere mode, Ground "B" is 'bootstrapped ground', not the signal output!
Signal output is created as explained on page 6-6.
It's really a bit tricky, to understand, how this is accomplished, as the whole driver circuit is intended also for volt and ohm circuit which may float +/- 210V. Carefully check the relays settings for the different modes.
In U, R modes, this amplifier output works differently, than in current and charge mode.

But we're discussing Ampere mode, anyhow.
The feedback loop, or in other words the shunt resistors, are R312, R322, R331, or R330, i.e. 100 Ohm, 100k, 100M or 100G. These create a voltage, proportional to the input current.
This amplifier output is at last created over RL, see figure 6-7 on page 6-6!
The Ground B you're referring to, is left to RL, and is signal output for Volt and Ohm mode only, like in figure 6-6. This bootstrap circuit accounts for 200TOhm input resistance in Volt mode, for all ranges up to 200V, which is very special.

R332, the 250G resistor, has absolutely nothing to do with with this current measurement / feedback loop.
You can check this, if you carefully read the bias calibration paragraph 7.4.9 on page 7-5, and following the schematic in parallel.

I have to admit, that this part of the circuit is not well explained, like the whole "Theory of Operation" chapter is very bad. No fun to read, in contrast to other (HP) manuals.

Frank

[Cerebus]

...
It has nothing to do with the feedback loop, as it's neither inside, nor parallel to the feedback loop.
...

Oh, it most definitely is across the feedback loop. Read the circuit diagram very carefully, paying attention to all the common terminal symbols marked 'B'. You'll find that 'B' is connected to the output of the class B output stage (at the same place that all the other feedback connections start from at a star point), one end of that 250G resistor (via an insignificant 2.2k, where it also connects to that trim circuit) and the centre rail of the 5V bootstrap supply. The other end of the 250G resistor is attached to the amp's summing point.

Nope, in Ampere mode, Ground "B" is 'bootstrapped ground', not the signal output!
Signal output is created as explained on page 6-6.
It's really a bit tricky, to understand, how this is accomplished, as the whole driver circuit is intended also for volt and ohm circuit which may float +/- 210V. Carefully check the relays settings for the different modes.
In U, R modes, this amplifier output works differently, than in current and charge mode.

But we're discussing Ampere mode, anyhow.
The feedback loop, or in other words the shunt resistors, are R312, R322, R331, or R330, i.e. 100 Ohm, 100k, 100M or 100G. These create a voltage, proportional to the input current.
This amplifier output is at last created over RL, see figure 6-7 on page 6-6!
The Ground B you're referring to, is left to RL, and is signal output for Volt and Ohm mode only, like in figure 6-6. This bootstrap circuit accounts for 200TOhm input resistance in Volt mode, for all ranges up to 200V, which is very special.

R332, the 250G resistor, has absolutely nothing to do with with this current measurement / feedback loop.
You can check this, if you carefully read the bias calibration paragraph 7.4.9 on page 7-5, and following the schematic in parallel.

I have to admit, that this part of the circuit is not well explained, like the whole "Theory of Operation" chapter is very bad. No fun to read, in contrast to other (HP) manuals.

Frank

I think we're going to have to agree to disagree because I can see the diagram in front of me and I can quite clearly see the  (unswitched) path from the output of the preamp's class B stage, back through that 250G resistor to the summing point. We can hope that it's perhaps a difference in terminology rather than understanding.

Talking of terminology, I don't think it's helpful to keep using the word 'ground' for several things that aren't - the "bootstrap ground" isn't a ground, it's just the common voltage of the preamp's bootstrapped +/- 5V supply, which is at the same potential as the preamp output. There's a time you can play fast and loose with the the terminology of grounds and common points, but that time is not in describing a circuit like this, with a handful of separate common points, none of which is ground in the proper sense.

Just to add to the confusion, the published circuit diagram has failed to label the JFET pair sources as powered by -5V(B), and the +/-5V labels on the offset trim circuit ought to say +/-5V(B). Given these, there are probably other errors lurking waiting to be discovered.

[Alex Nikitin]

Oh, my favourite electrometer  . I have one at work and one in my home lab (modified with an opamp in place of the input FETs) , as well as it's perfect partner in the Keithley 263 Calibrator/Source.

250G resistor is only there to trim the input current as close to zero as possible, with some thermal compensation as well. The circuit will work just fine without that resistor,  with a somewhat increased input bias current.

Cheers

Alex

[bjcuizon]

Hmmm...Interesting. I never knew we made chips in the Philippines.

[bjcuizon]

Oh, by the way..Is this a triax-bnc adapter or an adapter and a feedthrough terminator at the same time?
The triax is the other way around though.
Tektronix Adapter/Terminator

[Dr. Frank]

I think we're going to have to agree to disagree because I can see the diagram in front of me and I can quite clearly see the  (unswitched) path from the output of the preamp's class B stage, back through that 250G resistor to the summing point. We can hope that it's perhaps a difference in terminology rather than understanding.

Talking of terminology, I don't think it's helpful to keep using the word 'ground' for several things that aren't - the "bootstrap ground" isn't a ground, it's just the common voltage of the preamp's bootstrapped +/- 5V supply, which is at the same potential as the preamp output. There's a time you can play fast and loose with the the terminology of grounds and common points, but that time is not in describing a circuit like this, with a handful of separate common points, none of which is ground in the proper sense.

Just to add to the confusion, the published circuit diagram has failed to label the JFET pair sources as powered by -5V(B), and the +/-5V labels on the offset trim circuit ought to say +/-5V(B). Given these, there are probably other errors lurking waiting to be discovered.

Sorry, that's not a case of different views or terminologies.
It took me a while to fully understand the current vs. volt circuit, but in the end, I'm right.

Alex Nikitin just confirmed my understanding about this 250GOhm resistor..

Also, think about the conversion factors for current shunts..
If you have a current of 10mA, 1mA, 100µA, 10µA, down to 1 pA , the A/D converter (which is obviously a 2V F.S. type) needs a conversion factor of 1 * 10^X, or a voltage of 1V for each of the above mentioned ranges and measurements.

If this 250GOhm would be used in the feedback loop, as a current shunt, it would give 2.5 * 10^11 as a conversion factor, or 0.25V for 1pA input, which would be amplified additionally by a factor of 10. A f.s. input current of 2pA would give 5V for the A/D, and overrange it.
That makes absolutely no sense, as it would be usable for this single range only.

Please, check if your view makes any sense under these arguments.

Frank

[Alex Nikitin]

Here is a somewhat simplified circuit for the voltage and current modes of operation. R7 in the current mode  circuit is the feedback resistor.

Cheers

Alex

[razberik]

Perfect !
If anyone is interested how works modern, low-end K6485 I attach simplified diagram of main loop.

If anyone is looking for cheaper triax -> koax adaptor, it is Pomona 5300, mouser:
http://cz.mouser.com/ProductDetail/Pomona-Electronics/5300/?qs=sGAEpiMZZMv8kklI404QlX%2fxYBqdL8C9

[free_electron]

chips were not made in the phillipines but assembled ( injection molded ) in the phillipines

as for that fet int to99 : that's the protection diode pair (two matched npn's used as diode).
the actual fet sits elsewhere.

And yes, if you flurp that fet, it is game over. i got the last 3 remaining new ones (new old stock)  in the world >  2N59-something and yes, they are hand made , on demand only.

Last time i needed those to repaired these machines they only had 5 in stock. i bough all 5 of em. I've use two so far. They are (were, they are gone off their website now) made by a small company called Linear systems ( not linear technology ! )

an other note : if you solder in that area : it needs to be cleaned with freon and you need to wait a few hours for the surface charge to dissipate .. BEFORE you attempt calibration.

don't touch anything with pokenfingersptizen! the grease on us bipedals is deadly for this things precision. And when it is running : don't touch it either. that top board makes +120 and -120 volts for the floating amp...

During calibration : the shield must be installed. There is a small 2 pin jumper for calibration : after flicking that jumper you need to wait a few hours before attempting anything.

Another nice note : current does not really flow through a sense resistor. this machine has no burden voltage ! the machine creates an output voltage , sent through a very high ohmic resistor , into it's output. all the amplifier does is keep the balance between in and out equal to zero. so for every electron sent in to this thing , the high voltage amplifier forces an electron out ( direction doesn't matter, hence the+120 to -120v drive capability.

[Alex Nikitin]

chips were not made in the phillipines but assembled ( injection molded ) in the phillipines

as for that fet int to99 : that's the protection diode pair (two matched npn's used as diode).
the actual fet sits elsewhere.

And yes, if you flurp that fet, it is game over. i got the last 3 remaining new ones (new old stock)  in the world >  2N59-something and yes, they are hand made , on demand only.

Last time i needed those to repaired these machines they only had 5 in stock. i bough all 5 of em. I've use two so far. They are (were, they are gone off their website now) made by a small company called Linear systems ( not linear technology ! )

an other note : if you solder in that area : it needs to be cleaned with freon and you need to wait a few hours for the surface charge to dissipate .. BEFORE you attempt calibration.

don't touch anything with pokenfingersptizen! the grease on us bipedals is deadly for this things precision. And when it is running : don't touch it either. that top board makes +120 and -120 volts for the floating amp...

During calibration : the shield must be installed. There is a small 2 pin jumper for calibration : after flicking that jumper you need to wait a few hours before attempting anything.

Another nice note : current does not really flow through a sense resistor. this machine has no burden voltage ! the machine creates an output voltage , sent through a very high ohmic resistor , into it's output. all the amplifier does is keep the balance between in and out equal to zero. so for every electron sent in to this thing , the high voltage amplifier forces an electron out ( direction doesn't matter, hence the+120 to -120v drive capability.

1) The special FET can be replaced by a cheap modern opamp, with good results.

2) The supply range for the input amp is +/-210V, so the preamp output can be up to that value (actually in Amps mode it is limited to about 20V, but in the Volts and Ohms it could be the full 200+ Volts)

Cheers

Alex

[technogeeky]

Dave:

I spent about an hour trying to find what TG-162 is.

Some facts:

• TG- is the internal Keithley designation for transistors
• These designations are universal among all Keithley products and don't change over time (with one or two exceptions)
• Some TG- designators are actually selected versions of others (e.g. TG-161 might be a selected TG-135)
• Keithley products one or two generations earlier than the 617 will list complete manufacturer details of all parts (but this generation does not)

Unfortunately, I scanned as many instrument manuals of that era as I could find, and none of them anywhere use TG-162. There is one false positive in searching which is a TG-182 that OCRs as TG-162 for some reason.

So unless some insane person is going to decap one of those, or unless someone knows one of the instrument designers, we'll never know!

Oh well, I tried.

-tg

[Kleinstein]

In case one would need a replacement for the input FETs, there is no real need for the part number. It is a low current JFETs. So look for a pair of JFETs with suitable low input currents. One might need to select some, or if the parts is not that good allow for more compensation (e.g. change the divider at the 250 G resistor).

It is a little odd they use a 10 M series resistor just for protection. So don't expect really low voltage noise. I would have more expected 2 or 3 lower value resistors in series, to make sure not to exceed the voltage limits.

[free_electron]

like i said : it is a hand-selected matching pair 2N5912 from Linear systems.

[Cerebus]

Sorry, that's not a case of different views or terminologies.
It took me a while to fully understand the current vs. volt circuit, but in the end, I'm right.

Alex Nikitin just confirmed my understanding about this 250GOhm resistor..

And when I look at Alex's diagram I see exactly what I described. I see a 250G resistor, from the output of the preamp feeding back to the input of the preamp, in every circuit configuration. You appear to see something else. Those two views appear to me to be irreconcilable as I'm just describing what I see, literally the physical topology of the circuit.

That's why I suggested we agree to disagree, but you seem to be determined to insist that someone's 'right' and someone's 'wrong' and refuse to accept that we are probably both essentially saying the same thing but misunderstand what the other is saying.  I really do see no point in me repeating variants of my description ad-infinitum in the hopes that you'll 'get' what I'm saying; that generates no enlightenment for anybody and I do not enjoy mud wrestling*. Hence this is my last word on the subject.

* Anybody who doesn't get this, just Google "engineer argue pig mud".

[Cerebus]

chips were not made in the phillipines but assembled ( injection molded ) in the phillipines

I think his point was that everybody else spells it "Philippines"...

[Kleinstein]

The 250 GOhms resistor is in a kind of feedback path. However up to the point from where the 250 G resistor feeds back to the input the voltage gain is very close to 1 (I would estimate something like 0.99999). So the 250 G resistor is not very effective in setting the input impedance or the amplifiers gain. So despite of the resistor, the input impedance can be significant larger than 250 G Ohms. The propose of the resistor is to compensate for a possible small bias current of the FETs and protection diodes.

The 2 pA (FS) range has a separate resistor for the feedback.

[HighVoltage]

Analog Devices has the ADA4530-1 Electrometer OpAmp

[Alex Nikitin]

And when I look at Alex's diagram I see exactly what I described. I see a 250G resistor, from the output of the preamp feeding back to the input of the preamp, in every circuit configuration. You appear to see something else. Those two views appear to me to be irreconcilable as I'm just describing what I see, literally the physical topology of the circuit.

That's why I suggested we agree to disagree, but you seem to be determined to insist that someone's 'right' and someone's 'wrong' and refuse to accept that we are probably both essentially saying the same thing but misunderstand what the other is saying.  I really do see no point in me repeating variants of my description ad-infinitum in the hopes that you'll 'get' what I'm saying; that generates no enlightenment for anybody and I do not enjoy mud wrestling*. Hence this is my last word on the subject.

* Anybody who doesn't get this, just Google "engineer argue pig mud".

I have to point out that this resistor is connected from the output to the non-inverting input, hence it only can be considered as a positive feedback. Fortunately, it is connected with the feedback coefficient just under 1, so no oscillations, and it is used to provide a DC bias to the input while following the output exactly, so the effective value of this resistor referred to the input is hundreds of Teraohms!

Cheers

Alex

[alm]

If anyone is looking for cheaper triax -> koax adaptor, it is Pomona 5300, mouser:
http://cz.mouser.com/ProductDetail/Pomona-Electronics/5300/?qs=sGAEpiMZZMv8kklI404QlX%2fxYBqdL8C9

Be careful using adapters that short the inner and outer shield with equipment that uses a driven guard.

[razberik]

Definitely. One must exactly know what he is doing.

[Alex Nikitin]

If anyone is looking for cheaper triax -> koax adaptor, it is Pomona 5300, mouser:
http://cz.mouser.com/ProductDetail/Pomona-Electronics/5300/?qs=sGAEpiMZZMv8kklI404QlX%2fxYBqdL8C9

Be careful using adapters that short the inner and outer shield with equipment that uses a driven guard.

In the Keithley 617 the guard output is driven through 150K resistor and it is relatively safe to short it to the case ground (that happens if you try to engage the guard with the adapter) . The problem could be if you connect anything but the case ground to the "COM" terminal with this adapter in place. There is a fuse in line with the COM terminal but it is better be careful in that respect. You need to remember that with the adapter you lose the isolated analogue common and measure anything in relation to the case ( = mains)  ground. If the shorting link is connected externally anyway (as shown in Dave's video) it doesn't matter.

Cheers

Alex

[alm]

I prefer adapters that either connect outer shield to BNC shield and leave the inner shield (often guard) floating, or connect inner shield to BNC shield and leave the outer shield floating. The former is good for higher current measurements when leakage current is not an issue. The second is good if the inner shield is close to ground or if the leads are inside an interlocked case.

I do not have part numbers handy, but Trompeter makes them, and Keithley also sells them at probably a substantial mark up.

[JonM]

I built a 617 into a cyclotron mass spectrometer system in the 1980's. The electrometer was used to measure ion beam current injected into a superconducting magnet. I started out with a 614 and upgraded to the 617 because GPIB current read-out and range changing were needed to implement auto tuning for the ion optics.

Memories of the brown boxes.... The system also used a Keithley switching chassis.

Jon

[coldmama9]

Sorry, but I have fears that the 617 is no longer calibrated at the femto-amp level, isn't it? Nulling wouldn't work too.