Keithley 480 a la bodge

[Shimonu]

Took a chance on a Keithley 480 when it was lowered to $12 with the description that it doesn't power up. Sounded like it could be an easy fix but I'm definitely not thinking that after opening it.

So there's a somewhat crucial part missing, Q104, a pair of N-channel JFETs sitting on the input and that are matched with two resistors that are still on the board but had their legs cut. I'll attach a picture for reference.
They've also lifted a leg on U104 and I haven't looked to closely yet but I think it's connected to the zero check potentiometer with a bodge under the board. C111 has also been cut.

Maybe there is something wrong with U104 and they gave up when they weren't able to fix it.

Is there any chance of me replacing or fixing this with commonly available parts? I don't feel like looking for another donor unit as I don't expect to find one very cheaply.
Assuming the U104 op-amp is okay and the issue was Q104 broke, could this unit work with a new pair of JFETs and trying to hand-pick new resistors or bodge some potentiometers?

[MathWizard]

I would love to buy a few at $12 each. I'm guessing that's a rare custom part with not much info on it. I would email Keithley and ask them what info they have on it and if they can recommend something similar. I'm sure they are dual JFET's still for sale.

For fun I would make part of the circuit in LTSpice, and try some of the JFET's, maybe it's not that hard to get it working a bit. You can even try matching some single JFET's, and gluing them together. But yeah fine tuning it to work as it should, is unlikely. Maybe it would be a nA or uA meter, and it would need calibrating, and changing some digital data.

I would keep it, and redesign or mod some of it if I could never get the exact part. Some people would just sell it again tho.
I think old gear like that is beautiful, and a lot more repairable and modifiable, with today's modern IC's and MCU's.

[Shimonu]

There are dual N-channel JFETs that you can still buy in the metal can package so I'd happily try that, I just wasn't sure if it would be useless unless it was the exact same one. There's a youtuber that went through the circuitry and from what I understood these are essentially used to increase the input impedance for the op amp.

I could certainly send Keithley an email, couldn't hurt.

I think my last resort would be to redo the input with a modern op amp. I'd like if this was as close to original as possible. Not just for sentimental reasons but it's nice if the schematics and everything else stays valid.

[Uunoctium]

I don't know if this is any help - here is a link to the 1977 Siliconix FET Selection Guide. Includes a cross-reference table to most common types of the time, including dual fets.
http://bitsavers.informatik.uni-stuttgart.de/components/siliconix/_dataBooks/1977_Siliconix_FET_Data_Book.pdf

The LS830 might be a candidate for the 480 input stage due to its ultra-low leakage and drift. Price approximately in the 10$ region
https://www.linearsystems.com/jfet-amplifiers-duals/ls830-series

[Kleinstein]

The dual JFET is used in front of an OP-amp. In theory there would be the alternative to replace the OP-amp with a really low bias one. For the start even a TL031 could be sufficient. Not sure if the circuit would allow to reduce the supply for the OP-amp to get away with +-6 V or so, as than there would be much more super low bias choices.
Not having U104 may thus be a sensible change to the circuit.

[Shimonu]

I just had a try at turning it on and it looks okay at first glance which is nice to know for future mods. I haven't tried any measurements but it was showing 0 on all ranges. I don't have anything that will source such small currents but I could have a try at least with some reasonably accurate resistors and voltage source.

[Shimonu]

I've had a quick chance to try some measurements and I'm quite surprised how well it seems to be working, and this is still as I received it. I haven't been able to go to any of the lower ranges yet but 1.000 V across 100k 0.1% shows 9.97 uA. From the way the board looked I would have thought it wasn't working at all since it was also sold as for parts. I'm still not expecting too much from the lowest ranges since it likely doesn't have as high input impedance or the matched components. All I've done is use some contact cleaner on some sticky switches and changed it to 230 V.

I still want to try to restore it, I've ordered the suggested JFET pair and I'll see if I can get it to work with the original resistors. Otherwise I'll try to see if I can put in some new ones but I'm thinking it might need pretty specific values. There's some ugly writing on it with a permanent marker that I'll try to clean up too.

Not bad for the price I think

EDIT: Really testing my equipment tonight. Useing my E3630 and trying to get it to output 10 mV which I measure with my Keithley 2000 and then running that through a 1% decade box set to 10M which gives me 1 nA which I'm actually getting on the 10 nA range. It's a little noisy but as long as you stay pretty still it won't change by more than 1 digit. The 1nA range however doesn't seem to work, it reads a negative pA current.

[Kleinstein]

The bodge looks like the dirctly used the OP-amp.  From the picture one can not see the type of OP-amp. If it is a good type, it could work just as well as the original solution with external JFETs. 

The source resistors at the JFETs would depend on the type and would be used to trim the offset voltage. Too much of an offset could be anoying as it adds voltage to the input terminal. So for current from a resistor and voltage source it would add the voltage.
For a test one could use an external meter an measure the voltage over the inputs - it should be reasonable small (like < 1 mV).

With the small currents main hum can be an issue and one may need some shielding.

[Shimonu]

I'm picking up the repair of this instrument a bit. I've undone the bodges and put in a dual JFET. There are still two JFETs missing that I have on order that I intend to put in. But the instrument seems to work pretty well already.
Some of the switches have been sticking so I've cleaned the power switch and zero check switch.

The only measurement I'm having trouble with is still the 1 nA range and I think this is just a dirty switch as well. I can measure on the switch and I get a very jumpy resistance, indicating poor connection. I've tried spraying contact cleaner and exercising the switch but it doesn't seem to do the job.
The other switches were pretty easy to clean once you got them apart but the ganged switches seem more complicated.

I know I've seen a video of someone taking ganged switches apart and cleaning them but I can't remember where I saw it or if it was this instrument. I'd really like to take them apart and do a thorough job.
Does anyone know if it's possible with these switches?

[Uunoctium]

This one mentioned here in eevblog two years ago?

Just to add...

In that videos comments it's mentioned they have silver contacts on that model he is repairing (so oxidization is possible). Some people were reporting that the solder joints may become intermittent due to stress as well.

I normally test intermittent joints by gently tapping (then pushing) on or around the component/pcb with the back of a screwdriver handle while looking at the display (or scope or multimeter measurement) in an attempt to see if I can reproduce it. You can also try visually inspecting the joints.

Here is a video mentioned in the replies, it's possibly a similar ganged switch but gives you an idea of what these could be like inside.

[Shimonu]

Unfortunately that's not it. They showed when they disassembled it and described how the ganged switches works mechanically. I wonder if it could have been IMSAI Guy, I've watched him quite a bit. It's just that this is not typically what the video is focused around so it's difficult to search for it.

[Shimonu]

I made a silly mistake before and it turned out the points I was measuring on were dirty and that's why I measured a jumpy resistance.

So I'm still having problems with the 1 nA range and what confuses me is all the other ranges are working so there isn't really anything jumping out as different about the 1 nA range.
There are 4 different shunt resistors, some of the ranges put in a 10x gain to achieve a lower range.
   1mA has a dedicated resistor otherwise the 10x is used as follows
100uA ->   10uA
    1uA -> 100nA
  10nA ->     1nA

Since the upper ranges work fine I would think the 10x gain is working as expected and since the 10nA range is also working and can for instance measure a 1 nA current correctly I would assume the shunt resistor, including the potentiometer, is within spec. That's why felt so sure it would just be a dirty switch but after emptying plenty of contact cleaner in it and measuring the switch as fine I'm now scratching my head.

I've attached a picture of the input section of the schematic for reference
I'm not sure I understand how the conversion works. Looking at TP8 from the op-amp with an oscilloscope it shows an alternating signal but it's the DC voltage of it that appears to be used and it's very precise. Since you can also read it from the analog output I've hooked a multimeter to it and the numbers match up perfectly with what the display shows. For instance, in the 1uA range a 1uA current will give a -1.000 V voltage at TP8 and that shows up as a 1.000 on the display.

When I try to measure a 1 nA current in the 1 nA range I see a small negative current of about -0.180 nA(corresponding with a 180 mV voltage) or so and if I increase the current the value will increase but not by the expected amount. If I get it to show 0.000 in the 1 nA range then the 10 nA range will show about 5.5 nA.

Maybe my method is too crude, as described earlier, I have a 10 mV source over a decade resistor set to 10 MOhm. But according to the datasheet the burden voltage should never pass 200 uV and I should have still had trouble with the 10 nA range in that case.

Is it related to the JFETs for the op-amp that were removed and I've now replaced with something inappropriate? I'm not sure I understand why that would only affect the 1 nA range like this though.

[Shock]

I think Dave or Simon Spiers had videos on ganged switches. There is a couple of them on youtube if you look around.

And congrats on the bargain, I thought mine was cheap, but can't go wrong at $12.

[Uunoctium]

I never assumed that the LS830 would work directly as a drop-in replacement for Q104 without redefining and adjusting resistors R115 and R116.
As I understand it, Q104B functions solely as a constant current source (same die, drift compensation) and sets the non-inverting input of U104 to U/2 = 2.500 V, since the manual states the instrument operates with both polarities.

My suggestion for alignment: temporarily replace R115 with a 10-turn spindle trimpot of suitable value (e.g. if the actual fitted resistor measures 478 Ω, use a 1 kΩ pot set to half). Adjust until the Q104B source voltage is 2.5 V (allow a few minutes for the instrument to settle after major adjustments). Then measure the trimpot value and refit a matching fixed resistor.

Next, connect the gate of Q104A to ground and repeat the same procedure for R116.

If no matching resistors from the E196 series are available, an alternative would be to modify the circuit as follows: add a trimpot and adjust for zero voltage between the two sources of Q104 (with gate A grounded)

[Kleinstein]

There is already the trimmer to adjust / trim the JFET offset. R115/R116 should be about the same - a slight difference if the offset is too high to trim out with the trimmer. They set the current at which the JFETs are operating - this is a balance between heat and resulting thermal effects, noise and the gate current that tends to be higher with more DS current. The current also effects the temperature sensitivity. There is no sharp optimum and thus no need for a very specific value.
With the OP-amp in the linear working range, the source voltages should be the same anyway. It would still make sense to check the source voltage an aim for some 2-3 V.

The 1 nA range seems to be the lowest range. So the issue could be just the noise / leakage currents.

[Uunoctium]

The trimpot R101 which looks like an offset trimmer and located in schematics at op-amp is the front panel zeroing adjuster!

[Kleinstein]

R101 should mainly be used for the OP-amps offset. This would be especially with BJT based amplifiers, but ideally also for others. Using this at the front panel is not ideal.
There is no need for a fine adjustment of both R115 and R116. The first could be chosen for a crude value and the 2nd than for a low offset (like < 2 mV) with both gates at ground. The pot at the OP can than do the fine part.
This input is a virtual ground and thus no need to really aim for exactly the middle of the 5 V positive supply.

[Spike]

Unfortunately that's not it. They showed when they disassembled it and described how the ganged switches works mechanically. I wonder if it could have been IMSAI Guy, I've watched him quite a bit. It's just that this is not typically what the video is focused around so it's difficult to search for it.

CuriousMarc made this repair video not long ago in which he disassembled similar but not identical switches.  Perhaps this?

https://youtu.be/djkySqUVgjg

[Shimonu]

Great, it sounds reasonable that it would be the JFETs. I'd just like to understand their function a bit better. For the moment I'll leave the switches alone.

Q104B has the gate tied to GND and since these are N-channel that would mean it's essentially closed. From reading about FETs there's an ohmic region here where it acts like a resistor for increased drain-source voltage and passes some current. Is that what's going on here, is it biasing the non-inverting input of U104?
I measured R115 and R116 and they are similar at about 1.22k, I had a try at measuring the resistance of the JFET and got about 1.6k but I don't know if that's representative of the function. What confuses me is that when I start the instrument and measure the voltage at the non-inverting input I see something like 0.25 V but I would have expected something in the 2 volt range as Uunoctium was mentioning.
For Q104A I saw a video describing it as it's essentially giving the op-amp increased input impedance by buffering the voltage.

[Kleinstein]

The JFETs work as source follower. The votlage at the source should follow a change in the gate but shifted a little, like 1 V or 2 V.  How much shift depends on the threshold of the FETs and value for R115/R116.
The LS830 would likely need way higher values for R115,R116 as they are for rather low current. E.g. some specs are for 30 µA of operating current.
So a suitable resistance would be more like 33 K - maybe even 100 K. One may still need a 2nd iteration as the threshold can vary a lot.  This is typical for many JFET types and thus the R115/R116 chosen suitable for the JFET pair and no fixed value. One may still not get 2 V, but this would not be an issue.

U104A is for the input signal and U104B with the gate at ground is there compensate for the shift and drift of the shift.

With a better working point, that is a chance to get a litte less noise.

[Uunoctium]

Beg your pardon - i'm totally wrong with my suggestions for setting an  operating point for the source voltage of 2.50 V. Didn't study the datasheet before writing.
The recommendation of @Kleinstein is much more close to reality, than mine. Also his classification of Q104A as source follower ie impedance converter.
The LS830 is in specs a magnitude away from popular  j-fets. Idss max is given with typical 400µA, cut-off Voltage -Ugs at -2V.
At Id ~ 200µA the voltage at source will be probably in the region of 0.5V and with 90…100µA around 1V. Corresponding resistors 2.500 … 15.000 Ohms, at lower current ~ 30.000 Ohms or higher.

(For interested  i add an informative paper from Siliconis related to fet-bising)

[Shimonu]

I think you guys are on to something. I don't have an appropriate potentiometer yet but I did find some higher value resistors and I put in a couple of 47k as a quick test. It's still not giving me a correct reading but instead of -0.180 I'm seeing about -0.080 which I think is an improvement. The voltage on the input has gone up to 0.86 V. So I'll likely need a lot higher resistance, maybe up to 100k or thereabouts as suggested.

This meant I also had to zero the meter after the new offset, I hope there's enough range in the adjustment but I assume there must be since the offset I'm trying to achieve would be expected.

I'd like to see if I'm understanding this more in terms I know. We're biasing the input to the op-amp with the JFET acting as a current source but now the JFET is passing a smaller current compared to the previous one and giving a smaller voltage drop over the resistors and that's why larger values are needed, would that be correct?

[Kleinstein]

0.86 V at 47 K is not so bad. This would be some 18 µA. This already on the low side. The offset specs are for 30 µA. So I would more consider 27 K or 33 K. There is no need to get 2.5 V - with the given FETs one will likely not reach this as the threshold voltage is too low. The OP-amp worked also without the FETs and thus seem to work all the way to 0 V.
The FETs are specified to have G in the 100 µS / 300 µS range depending on the current. This corresponds to a resistance of 10 K / 3.3 K and if possible one should stay well higher of that resistance to not loose too much of the signal amplitude.

If needed one can consider 2 slightly different resistors to get less offset.

[David Hess]

The dual JFET and U104 combination form a low input bias current inverting amplifier for a transimpedance, current in and voltage out, amplifier circuit using 1978 technology, when good integrated JFET operational amplifiers where not yet available.

Replacing the dual JFET is certainly possible, and I could find a modern available part number, but I think better would be to replace the dual JFET and operational amplifier with something like an OPA140 and move the input offset adjustment to drive the non-inverting input with a resistive divider.

Shown below is basically the same circuit, but with a fixed 100 picoamp range and 10 femotoamp resolution.  I need to either build something  better, or buy a dedicated picoammeter to rebuild.

[Shimonu]

Positive update. I recently got a Keithley 225 current source and had a new go at testing the ranges and this time the 1nA range looks great, I think there must have been something I did in my setup before.

That means I put in the LS830 and the 47k resistors and that's working fine as far as I can tell.
The 225 isn't calibrated but it's clear enough that the increments are correct even if there's an offset sometimes. I've also tested it on several multimeters.

I feel happy and the 480 is screwed back together and will go on the shelf.
I'll soon get a Keithley 261 Picoamp source and I'll be sure to compare results with that as well.