Keithley 616 electrometer Jfet replacement

[Amper]

Hi!

I recenty came into the posesion of seven keithley electrometers. two were still working and show reasonable results compared to other meters, though in 5 of them there is problems. Obviously the input stage is in the most of danger, though i found problems on the second stage board as well. There is a matched JFET pair in cn package. Not selected by nude virgins like the input stage, but still unobtainium as it seems even with the part number (ITS30092).

Can anyone recommend a replacement? Would it be possible to replace them with some normal fets just to be able to test the rest of the instrument for basic functionality? I dont want to swap parts with the working meters to prevent myself from ruining them.

Attached is a screenshot of the circuit section in question, the manual is easy to find on the web.

Thanks in advance
A.

[42Khz]

I suppose you could drop any dual JFET in and it will work to some extent. There's only 5Vds so I doubt impact ionization would quickly be an issue with the 500k resistor.

I know of this monolithic pair cheaply available on ebay:
https://www.ebay.com/itm/4x-BFQ13-Dual-N-Channel-JFETS-in-TO71-NOS-Philips-Semiconductor-FET/112526759448?hash=item1a331e1618:g:udcAAOxyUrZS7QYx]

Alternatively, you could perhaps look at subbing the FET + OA for some nice modern OA with trim terminals, although I'm not sure exactly how the manual zeroing works... there's a trimpot in both the ADC front-end amplifier and in the electrometer front end, according to the schematic.

As for testing the unit, there's a x1 output on the back that's directly connected to the input stage output. The input stage does everything electrometer related and the rest of the unit is effectively just a DVM. Just keep in mind it can output over +/- 200V.

[floobydust]

Need to know the original part number or manufacturer, any markings.

Linear Systems makes precision JFETs and CMOS switches, some Siliconix, Calogic, Toshiba, Intersil etc. They change the part number prefix so LSK389 is a 2SK389, LS5905 is a 2N5905 etc. Disti in Europe too.


You could hand match a pair of JFETS, the circuit does trim (zero) out a little mismatch, not much.
If you measured (from a working electrometer) the op-amp's pin 2, 3 you could know IDSS required.
It looks quite low with ~62k resistors in circuit, estimate ID at 0.15mA ? Most JFETS are run at higher currents I think, so a low IDSS part I think is used.

[Amper]

Original part number is "ITS30092"    Sadly there is a lot of parts in this instrument that are either marked as "special" or are just so unusual that there is nothing on the internet.
The hint with the output is nice, thank you ! ill check there next. One of them has a missing input board, do you think its replaceable with modern parts or will this one be a parts donor now?

These things are ridiculusly sensitive btw. the maximum resistance range is 1e12, freaking T ... I connected a 1GOhm resistor in the 1e9 range and when i press together the 4mm connectors rubber (good ones) it drops by 100MOhm. These ranges are completely insane.

[42Khz]

Linear Systems make great parts, if you can get them. AFAIK the only distributor in europe is Micross and they don't carry all their parts.

I agree with the Id estimate, but I'm thinking they dual FET is likely used near its zero-TC point, so I reckon it's a reasonably high Idss part.
I think the most effective strategy would be to select a dual FET and find its zero-TC point (approximately Vgs = Vpinchoff + 0.6V according to the Graeme patent) and then sub the source resistors to get there. I guess you could install your own trim to remove any excess offset, although this will affect your CMRR to some extent. As long as the offset is within the trim range the thermal drift is one of the more important parameters as it is unchecked during measurement.

Replacing the input board is tough... the spec is < 500aA or 0.5fA.
I've tried to look around but was able to find almost no info on low-leakage MOSFETs. Nobody seems willing to spec gate leakage and I haven't found anyone who knows of parts on this order of leakage. I was only able to deduce that the 2N700X family leakage in the fA range, so this may be a candidate but you'd have to select for leakage somehow. Most MOSFETs seem to have a bidirectional zener clamp from gate to source which I suspect is the major cause of leakage. Unfortunately this clamp is often not shown or talked about in the datasheet.
If you go this route you also need a clamp of some kind. The 616 uses some obscure monolithic BJT array for this, I never quite figured out why they used a monolithic array instead of just two singles. Perhaps you could replace this with a pair of BFT25, they supposedly have fA level leakage.

It so happens that I recently experimented with some 2N3904s and I got about 30fA of leakage out of them in a similar configuration with a few 100mV across them. That's a far cry from the 500aA unfortunately...

Alternatively you could buy electrometer JFET pairs or an OA. I know LS has a JFET pair that's specced for a few 10s of fA but I don't know the part number off the top of my head. For OAs, the LMC6001 is specced for sub 25fA leakage and there's another TI fA level precision offset OA that I can't remember.

The unit certainly is impressively sensitive! I really like how they use 0.1V across a 1e11 range resistor with a +/-200V input stage to allow measurement of up to 200 * 1e12 Ohm.

[Amper]

aA is insane, i didnt think it would be so little. Tough its the only option i have to replace them, otherwise a bunsh of the meters will be sad "trash". When  the broken ones are identified i will cut some open, im interested of there is something unusual inside.

 What i dont really get is why they have the second fet stage. There is a normal OP connected to the first stage anyways just for driving the guard. I have a sack full of lt1001, maybe i can adapt them somehow to make it work.

[42Khz]

Apparently I was mistaken, the "offset current" in ammeter mode is 5fA. I don't know where I got the 0.5fA from, I must have misread.
My unit seems to stick somewhere around 6fA, although I did have to soak the frontend in IPA before it measured right...

That might make a pair of 2N700Xs viable, if they go down that low. Some people over on the google groups seem to know a bit about them, perhaps some of them have done measurements. You could certainly try asking there.

That leaves the (BJT) clamp. I'll try to measure some BFT25 if I can find them back in my stash.

The JFET array I was talking about is the LS83X. The leakage is 80fA typical so it's not smashing.

I'm really not sure why they used that FET stage either, especially since it's an inverting stage. I guess it's because it was the best option available to allow them to use a fat input resistor to keep the input current down and simultaneously make the 10x range with the 5M to maintain low offset voltage.

I would love a teardown of the MOSFETs! I'm very curious what they look like inside. Although I guess the main important features will be hard to see.

[RobK_NL]

Linear Systems make great parts, if you can get them. AFAIK the only distributor in europe is Micross and they don't carry all their parts.

I have ordered LS parts from Ingenieur-Büro-Fluck. But it is also possible to order directly from the factory.

[Amper]

Than you for the link ! 
LS5905 seems to be the closest i found yet at 150fA, still a lot though.

[David Hess]

I would need to see more of the schematic to be sure but that does not look like it requires the lowest input bias current.  That configuration using a dual JFET and LM301A is typical from before integrated JFET operational amplifiers became available; they made a JFET operational amplifier by combining a dual JFET and a bipolar input operational amplifier.

You can likely get better performance and avoid dealing with the required dual JFET by replacing both Q301 and QA301 with a precision JFET input operational amplifier or perhaps even better, an improved super beta input bipolar operational amplifier like the LT1008/LT1012/LT1097.  Analog devices has some similar super beta input bipolar operational amplifiers as well but I don't know their part numbers offhand.  Super beta input parts have the advantage over JFET devices of lower noise, lower drift, and lower input bias current at higher temperatures.  In the past a part like the LM11 might have been used but we have even better now.

If you absolutely want to just replace just the JFET, Linear Systems makes suitable parts which are likely better than the original.

[42Khz]

Thanks for that! Do they offer affordable shipping?

The precision TI CMOS opamp I was talking is the LMP7721.
An incredible part, it even has fairly low noise!

Alternatively you could get the ADA4530 JFET amp from AD.
However, I don't see any advantage over the TI part for your application, other than the slightly lower typical bias current.

[David Hess]

The LMP7721 may be a very poor choice although TI cleverly disguises its shortcomings in its datasheet.  (1) This circuit has a low cutoff frequency extending to at least 0.3 Hz where the noise of the LMP7721 is very very high.  That is why I suggested a modern improved super beta input bipolar part which could perform more than an order of magnitude better.

Also, the LMP7721 is limited to a 5 volt supply.

(1) Notice that no 0.1 to 10 Hz noise specification is even given in the datasheet; it is that bad.  But if you look in the graphs, noise at 0.1 Hz is 300 nV/SqrtHz!  The LMP7721  is only low noise at high frequencies which is typical for CMOS.  The original JFET is probably about half that or even lower.

[Amper]

Maybe i will have some time this evening to put the lt1001 in. You told something about external offset comp. Wouldnt it be possible to just drop it in for the old one and bridge the jfet?

EDIT: Im talking about the second amp
EDIT2: I got it... Doesnt work the way i thought

[David Hess]

The LT1001 will be marginal with that feedback network at its inverting input.  I am not sure about the non-inverting input; I need to see more of the circuit.  It is good up to about 70 kilohms while the LT1012 is good up to 2 megohms.  Of course testing it won't break anything so try it.

There are of course lots of suitable precision JFET input operational amplifiers.  The offset voltage drift of the dual JFET they used is pretty easy to improve on; it couldn't be any better than 5uV/C unless they were selecting parts.

These are just rough guides to the noise issue:

http://www.analog.com/media/en/reference-design-documentation/design-notes/dn015f.pdf
http://www.analog.com/media/en/reference-design-documentation/design-notes/dn355f.pdf

[MadTux]

Try  2N5909 aka LS5905 as replacement JFET pair, they worked great as replacement on my Keithley 617s.
If you get leakage problems, it's probably more likely due to bad reed relays than due to leaky JFETs.
https://www.eevblog.com/forum/testgear/fun-with-low-leakagebias-current-femtompere-electrometer-keithley-617/25/

[42Khz]

The LMP7721 may be a very poor choice although TI cleverly disguises its shortcomings in its datasheet.  (1) This circuit has a low cutoff frequency extending to at least 0.3 Hz where the noise of the LMP7721 is very very high.  That is why I suggested a modern improved super beta input bipolar part which could perform more than an order of magnitude better.

Also, the LMP7721 is limited to a 5 volt supply.

(1) Notice that no 0.1 to 10 Hz noise specification is even given in the datasheet; it is that bad.  But if you look in the graphs, noise at 0.1 Hz is 300 nV/SqrtHz!  The LMP7721  is only low noise at high frequencies which is typical for CMOS.  The original JFET is probably about half that or even lower.

Perhaps I am missing something, but the low frequency noise given for the LMP7721 (which I would assume to be bandlimited from 0.1Hz to 10Hz) is only about twice as bad as the LT1008.
My suggestion for the LMP7721 was for the front-end amplifier - a super beta part won't do here.

I did indeed miss the supply rail limitation for the LMP7721. I suppose you could cascode it although that would be somewhat inconvenient.

[Gyro]

The LMC662 is a better (and cheaper) choice. Alex Nitkin repaired his Keithley 617 with one it the thread that MadTux linked - First page, Reply #5.

[42Khz]

The LMC662 is a better (and cheaper) choice. Alex Nitkin repaired his Keithley 617 with one it the thread that MadTux linked - First page, Reply #5.

That's actually a nice choice as the offset is trimmed out manually anyway! The only downside is that the maximum leakage spec is very high (2pA), but I guess you have to be very unlucky to end with a bad one and you can afford to weed them out as a hobbyist.

[Gyro]

Yes, you typically get in the very low fA region unless you are really unlucky.

P.S. If you do a forum search for LMC662 you'll find quite a lot of hits for its use in commercial pA range instrumentation.

[David Hess]

I'm not sure which part in the schematic you are thinking of but the LMC662 is a dual part and none of its cousins mentioned here have offset null pins.  The LMC6082 is a precision version of the LMC662 and the LMC6081 is single version of the dual LMC6082 and my first choice for electrometer applications but as Gyro points out, you need to grade it yourself because it is only tested to have an input bias current of 4 picoamps or less.  (1) Personally I have never tested one with a leakage above 10 femtoamps or so and 2 femotoamps was typical.  If you want to avoid testing, then the LMC6001 is a LMC6081 which has been graded at the factory for a maximum input bias current of 25 femtoamps.

Something to beware of when using these parts is that they are built on an 16 volt CMOS process and they are not kidding.  Even a 15 volt supply is marginal for operating them reliably.

(1) This is not a design issue but a testing issue.  Common automatic test equipment cannot test lower currents and specialized equipment which is what is used for the LMC6001 takes 10s of seconds for each test which costs money.  Test time is tens of cents per second.

[Kleinstein]

For the second stage amplifier a mode JFET based OP (e.g. OPA170, OPA145) would likely be easier than extra JFETs in front of the old OP. A super beta type would also work and might have less 1/f noise.

For the input stage Alex Nikitin suggested a circuit that replaces the FETs with OPs (thus using a dual OP). However the FET at the non input side does not need to be super low bias. So a single low bias OP (e.g. LMC6081,..) and an LT1001 should also work and be slightly lower noise than with a dual CMOS.

[free_electron]

ITS30092 is Intersil hand selected 2N5907

linear systems has them. if they cant supply : i bought at one time the 3 they had in stock. i have em in my garage.

[floobydust]

From Linear System's datasheets, assuming this is a most important spec for electrometer use:
IGSS gate leakage current:
50pA 2N5911,12
3pA SST/U424, U425, U426
2pA 2N5905,6,7,8,9
1pA SST/U421, U422, U423
0.2pA LS830, LS831, LS832, LS833


I thought Keithley part number TG-98, is Intersil ITS30092 Dual N-ch JFET TO-71.
I did find GE/Intersil IT500 dual cascoded N-ch JFET, a real odd part, but high IGSS 100pA.

[landau]

I own a Keithley 610C. It is in a fully working conditon. However, I am wondering whether the input FETs are replacable or not. Without removing the heatsink connecting the two FETs together, I have managed to take a photo from them. The part number is ...TS3...009... According to the first comment in this topic this might be an early version of the ITS30092??
On other forums they refer to this FET as HD1G1030 in a Keithley 600B. Despite it is a different instrument, the photo from the input FET looks exactly the same (but there the part number is not readable).