Keithley 2000 repair, only works in Hz mode

[Per Hansson]

Hi, I've been looking on eBay to get a Keithley 2000 for a long while, finally with the Corona shit decided to take the plunge with a dead unit to have something to bide my time
I picked this one as the seller only listed one error code: 100.2 however this is not so, if you sit and wait for a (long while) it continues and throws more errors.
I'm not one to complain though as the manual says you can ignore any other errors as they are invalid if you have errors in the 100.x range.
The problem as far as I have been able to see is that it locks up if I try to change into any mode, except for Hz mode, that works fine.
(If I input mains via a 12v transformer and then it showed 50Hz just fine).
The caps do not look like they are leaking, and I have not touched the thing with my soldering iron yet.
From the reference I do not have any voltage, however as I flipped the board to measure on the underside I did? The reference measured 7v just fine, but only once?
I powered off the thing to do some more stable tests but since then the reference never powered on again.
Right now I have about a dozen open tabs here on eevblog with different threads on this unit, this one in particular was of interest:
Because if I push "shift" during the self test in manual step mode at 100.1 then it shows: -1358692021 just like in that thread.
And he also had the problem of the unit locking up when changing modes.
At TP105 I get a rising voltage of up to 5v as the unit powers on, then this becomes -0.26v or so which is the same voltage as I see on the reference.
I did check voltages and they where off from the service manual, but then I found a post that described CR115 and that 15.6v / -15.6v would be normal.
My units 15v is 15.84 and -15v is -15.93, other voltages seem fine and they all have very little ripple.

I have the following in-circuit measurements on TF-245:

Code: [Select]

1: 0.9766k
2: 7.271k
3: 7.867k
4: 11.496k
5: 10.673k
6: 12.692k
7: 29.64k
8: 30.11k
9: 27.36k
10:0.9886k
11:8.255k
12:16.412k

On VR113/Q130 and VR112/Q129 I have the follwing:

Code: [Select]

VR113 -8.605v -14.953v
Q130:  3=22.2v 2=-0.47V 1=-8.605

VR112  8.232v  14.652v
Q129:  3=-22.500 2=0.40v 1=8.232v

[Per Hansson]

From the reference I do not have any voltage, however as I flipped the board to measure on the underside I did? The reference measured 7v just fine, but only once?
I powered off the thing to do some more stable tests but since then the reference never powered on again.

I attached a photo showing the soldering of the reference.
I have not fixed it yet because I would like others opinion first.
Would it be better to desolder it completely to do some tests with it out of circuit?
Or is it better to bet on the DG408DY A/D MUX since the unit locks up when I switch into the different measuring modes?

[Kleinstein]

I would not suspect the DG408 very much. The problem with switching the mode is more like that on switching the mode there will be initial extra measurements like the zero point and this fails it may not start with the new measurements.

With lead free solder it is not so easy to say if the joints look OK. Re-soldering the suspect joints at the reference is definitely an option.
The reference voltages should be there and there is not that much that can effect the reference generation. The CPU has not way to turn it off.

However a bad (open) solder joint at the reference would more like lead to too high a reference voltage, as it is a shunt reference.

The voltage readings more suggest a short, or another resistor open.  The short may also be at C167.

[Per Hansson]

Thanks for your reply!
Just to clarify I do measure the same +15.84 and -15.93 on pins 3 & 4 of the reference itself.
I tried taking some in circuit measurements on it:
Resistance measures 3kΩ between pins 1 & 2 on the reference, the capacitor C167 measures 1nF in circuit with it.
And in diode test mode I see 1.4v with minus on the AGND (pin 2) and 0.87v if I reverse the leads vs pin 1.
I'm waiting for my Flir One to charge up and hopefully it can show something interesting.

[Jwalling]

Re-soldering the suspect joints at the reference is definitely an option.

I thought that heating it could mess with the reference's short-term stability?

[Kleinstein]

The LM399 reference is relatively insensitive to soldering. It runs quite high in normal mode and the legs are usually long.
With a different reference, especially in a plastic case soldering would effect the reference.
Re-soldering the LM399 may void the calibration, but keeping it with poor solder joints is also not a real option.


The 2 pins with +-15 V at the LM399 are just the heater half - so just the supply. The other 2 pins behave like a little like a zener-diode, though not perfectly, especially not below some 200µA. So the resistance / diode test is not that relevant.

If the -15 V side does not have contact the whole reference may behave odd, as there may be substrate current.

My main suspects would be U139 and maybe the ohms part that may interfere with the signal at Pin4 of U139.
So it may be worth measuring the voltage at pins 3 and 4 of U139.

[Per Hansson]

The two solder joints that do look bad are for the heater pins on the reference.
My Flir One is unhappy and I was unable to get good pictures with it, but it reports a really high temperature of VR112 & Q129.
So I used a thermocouple instead and indeed they are over 50°C, being so small I think the reported temps by the Flir of up to 80°C might have been correct.
I took measurements of U139 with reference to AGND:

Code: [Select]

Pin1: AGND
Pin2: -15.89
Pin3: -0.20 (same voltage as on reference)
Pin4: -0.35
Pin5: -14.93
Pin6: 15.82
Pin7: 14.64
Pin8: -0.020

Note: these voltages are not absolute, in particular the reference begins around 0.26v or so then slowly goes down, the voltage on pin 4 seems to track it.

[Kleinstein]

The voltage readings would suggest hat U139 is not working correct: With Pin 3 more positive than Pin 4, the output of that OP (pin5) should be more at the positive end. There may however be a superimposed AC part that could cause the trouble and upset the readings. With the output close to the limits this is however not very likely.

VR112 should not get hot, unless Q129 goes into reverse breakdown on the BE junction. This may just happen at some 8 V, though it should not be the normal case. I still don't see where much current to the -14 V reference could go to to get a significant current draw to make it run hot.

Still I don't see the problem starting with the negative side. It is more like the positive side causes trouble (U139B acting odd) and the negative side only reacts to this unusual condition.

To test U139B without unsoldering it, one may remove the reference and VR113 and use instead something like a 6-12 V zener  (maybe VR113 if tested OK) from the base of Q130 (anode of VR113) to ground. This would give a makeshift reference and one could carefully (limit the current between pins 3 and 4) apply a test signal to pin 3 to test U139B.

[Per Hansson]

Thanks, if possible I would like to avoid to desolder the reference, but if it comes to that I will of course.
I redid the measurements with the Flir today, it was still not behaving so no photos to upload.
However with a third party freeware app and not Flir's own app I was able to set parallax and realized two hotspots I mistook from being radiated heat from the top of the board where infact not.
That is on the underside of the board resistors R218 & R214 are scorching hot too, but I do not understand their purpose from the electrical schematic. (They measure fine in-circuit).
And also I realized the difference between VR112/Q129 & VR113/Q130 was only about 10°C That is if VR112/Q129 was at 80°C then the other pair was ca 10°C below that.
The U139 itself sits around 75°C and that seems normal from these thermal pictures at xdevs, but just like you say the transistors and Zeners look cold in that picture.

[Kleinstein]

The reference amplification (U139-B) for some reason tries to output a negative voltage output instead of a positive. With the OP all the way to the extremes the zeners can conduct in forward direction and the transistor BE junction can reach the limit. This can drive the +14 V output to a slightly negative voltage.

R214 and R218 provide the base current to the transistors - the zener diode normally diverts most of the current and there is something like 5 V over the resistors. With the OPs all the way to the extremes there is way to much voltage to the resistors. So it is expected they get hot.

The root problem may the one half of U139. From the voltages measured for pins 3 and 4, the output at pin 5 should not be at -14 V, but more like +14.  So chances are high U139B is broken.
With the measured voltage at U139pin5 it looks like the ohms part is not acting up too bad and so should not damage the reference.
 So one could try to unsolder VR113 to isolate the broken OP output and than use an external OP+zener (e.g. OP07 or even 741 should be OK for a test) for substitute for the pins 3,4,5 of U139.

VR113 is not a critical part - the exact voltage does not matter (4-15V would be OK).

[Per Hansson]

Thanks! I found a similar part at a small Swedish electronics online store, OP27:
https://www.electrokit.com/en/product/op27gp-dip-8-single-op-amp-low-noise-2/
If I order it and am really really lucky perhaps it would arrive next Wednesday before the new years holiday.
Is there anything more you would suggest to purchase at the same time?
Also would you suggest any other tests meanwhile not requiring any esoteric parts?
Would the test maybe even be possible with something I could find on an old laptop board like a LM358?

[Kleinstein]

For a crude test the LM358 should be good enough. For a test also a 741 should work as an external bodge on OP to replace one half the LT1124. An OP07 may have comparable performance.

I don't see a real need for a high speed OP for the reference circuit. Just from the circuit I would anyway more like expect an LT1013 instead of the LT1124.   There probably is some reason they did use the more expensive LT1124, though I don't see it (maybe better compensation of Zener noise from CR113,CR112). The LT1013 even has the same, slightly unusual pin-out as the LT1124.
The LT1124 is lower voltage noise, but the current noise is so high that overall the LT1013 may even be lower noise.

For a very crude test, one could even wire the anode side of CR113 (or a 12 V zener) to ground, instead of the OPs output. The slight problem is that the input of the LT1124 would be still connected to the lm399 reference. At least the other side should be relatively high impedance. With a lower reference level the ADC may work, but it could still give an error.

Of U139B is not totally bad, but only way too much offset, one could get away with a lower resistor (down to some 5-10 K) parallel to R315 to give the reference a higher start up current.

[Per Hansson]

Kleinstein, you're a legend!
I contemplated what to test first and finally felt that changing the resistance of R315 was the least invasive.
I tagged on a 8.2K resistor and powered on: the reference voltage came up and I can switch modes without the unit locking up!
With a 4-wire short the meter displays 0.02Ω and this also settles the DCV & ACV readings to around zero mV.
The instrument now passes on test 100.1 but still fails test 100.2 (noisy)
I tested to increase the value of the parallel resistor: up to 18K the reference voltage appears instantly at power-on.
With 22K and 28K it takes a while for the reference voltage to creep up to 0.85v and after that it comes up with 7v.

I'm posting voltage readings to compare with my OP and reply #6: (anode/cathode symbol seems reversed in the schematic for VR112/VR113?)

Code: [Select]

VR113 Cathode: 13.3v Anode: 7.2v
Q130 3=20.7v 2=12.7v 1=13.3v

VR112 Cathode: -7.1 Anode: -13.3v
Q129 3=-19.7v 2=-12.7v 1=-13.3v

Code: [Select]

U139:
Pin1: AGND
Pin2: -15.89
Pin3: 6.984v
Pin4: 6.5v (with 50mv ripple)
Pin5: 7.2v (with 100mv ripple)
Pin6: 15.82
Pin7: -7.10v (with 100mv ripple)
Pin8: -0.020

The ripple readings are not done with a scope, only estimate with DMM, if it would be useful I can take scope readings.

My Flir One continues to be finicky but at least it gives me temperatures, now they are:

Code: [Select]

U139 80°C
Q129 67°C
VR112 60°C
Q130 55°C
VR113 52°C
VR109 105°C

I glanced over the board and everything looks quite normal except for VR109 but looking through threads here on eevblog it seems others report the same so maybe it is normal that it runs so hot, it was the same without the parallel resistor mod to R315

Should the next step be to replace U139 or is there anything more worthwhile to check first?

[Kleinstein]

I would not expect the parts to get that hot. U139 may get a little hot, as it has quite some supply current and if off the normal operating point may also drive some load. VR112 and VR113 should not see much power: some 1 mA at 6.x V, so only 6.x mW. The added resistor parallel to R315 would not chance temperatures very much - a little more current to the reference, but not much.

There should not be much supply ripple or other AC part with the reference circuit.
The values look more like oscillation, so the DC readings can be misleading. It would definitely help to use the scope.
Oscillation with excursions to the limits can cause quite some odd current flow through VR112 and VR113.
I see a chance the reference voltage may be at some 14 V for some 95% of the time and near zero (slightly negative) for the rest of the time.

As a kind of brute force way to help start up and maybe hinder extreme oscillation one could consider a diode from Pin 5 of U130(output for the positive side = one side of CR113) to ground . There is no need for the OP to swing to the negative side. Just keep an eye on the temperature of U139 - it may get hot if there is fault that forces it to go negative.

[Per Hansson]

Thanks Kleinstein.
I had to look a little more at VR109, because it failed on me it seems.
Voltage reading on it was suddenly 22v with full ripple, and it no longer was 100°C degrees but at room temperature.
So I figured it must have died and desoldered it and put in a through hole zener of ca 18v.
After this the ripple went down here by allot, however issues remain.
I did do allot more scoping around but I want to look further before I say anything definitive.
But right now I am troubleshooting the unregulated 20v supply further because in some spots it is perfect.
Like after CR102 and now at VR109 after replacing it.
However in other spots there is crazy ripple, and it is so complex that my Scopemeter has great troubles to display it with decent performance.
I actually switched to my analog Tektronix scope briefly but now it is time to sleep on this issue I think
So it seems your hunch about strange ripple might be spot on

[Per Hansson]

*TLDR skip to TLDR section below*
Yesterday evening I did allot of scoping around.
The crazy ripple I mentioned was present on pins 4,5 & 7 of U139 as I had previously seen with my DMM.
I used freeze spray on U139 and that did seem to settle the ripple allot.
However since it is a densely populated board it could be another nearby component, or something on the underside...
I never got rid of the ripple completely though so in the end I started looking for a PCB crack or bad soldered component.
But touching everything on the top of the PCB had no effect, so I flipped the PCB over again.
As I mentioned in my OP and reply #1 the last time I did I had stable reference voltage once, but not this time.
I looked around the back for something that would affect the ripple but nothing did, and I was about to give up.
Then suddenly the ripple disappeared and the meter was able to autorange fine.
For example before if I measured a 9V battery it could not autorange to the 10v range, but I could select it manually.
But before the reading was so unstable due to the ripple that the meter could show basically any random voltage value.
Now it was dead on, only effected by the tempco of the battery itself.
I measured a 2.5v voltage reference and it came up a little low but still very good at 2.49985v
At this stage I figured I would have the same experience as in my OP: that is after powering off and on the device it would fail.
But it did not, so I put the board back into the case, and still it worked.
So I thought ok, I will leave it and come back tomorrow because I had noticed quite large differences before related to the temperature of the instrument.
But to my surprise today it powered on fine again, I also noticed that the reference voltage appeared instantly.
I wrote in my reply #12 that with the 18K resistor in series with R315 the reference voltage appeared instantly.
But after that reply I noticed that if I let the device cool down completely over an hour it would take 10 to 15 seconds for the reference voltage to come up.
So I desoldered the 18K parallel resistor I had added to R315 and the reference voltage still came up instantly!
And now my 2.5v reference measured dead on 2.50000v, which could be just a coincidence since it is a €20 eBay AD584-M, but still!
Finally I initiated a selftest and it passed all of the tests without errors!

TLDR: So with my now out of the blue working meter I am not sure I want to be happy or not, since the cause was never found.
But I added some scope shots below & put up a video showing the issue with the crazy ripple which is now gone after the work above.
After this there is still a pretty crazy waveform at pin 3 of Q129 & Q130 (the unregulated 20v input, infact this ripple did not change)
But this ripple does not exist in other places of the unit, for example at R168 it is the expected sawtooth waveform.
You can see in my previous reply when VR109 was dead how the ripple then was present there too, but now it is a perfect sawtooth at R168
I guess this is why Q129/VR112 & Q130/VR113 gets so hot...
I need to replace VR109 because my THT replacement looks very ugly and is also unable to create a perfect stable voltage: there is some AC ripple let through.
I'm actually contemplating replacing R168 270Ω with a 1KΩ resistor because I do not understand why it needs to be that low, it creates allot of heat in the zener!

[Kleinstein]

An intermittent fault is one of the more nasty ones.

The initial ripple looks a little like normal mains ripple. This could point to bad caps in the power supply.
Some of the Keithley meters tend to show bad caps in the supply - especially leaking caps.  If the meter is old and the caps not changes one may consider changing the usual caps before they leak.  Otherwise I am not a friend of just swapping caps with no reason, but here it may make sense.

With so much background on the supply the fault may be more with the supply and not so much near the reference.

[Per Hansson]

Agreed Kleinstein, intermittent faults are no fun
You are correct that it is mains ripple at Q129 & Q130, yesterday with my analog Tektronix scope I actually used the 50Hz mains as a trigger source in the video I linked on Youtube.
I do intend to replace the caps, especially after the really scary pictures posted on xdevs.
But I wanted to find what else I need to purchase at Mouser at the same time since my caps do look fine, no visible leaks and they test fine for ESR & capacitance in circuit.

But I added some scope shots below & put up a video showing the issue with the crazy ripple which is now gone after the work above.
After this there is still a pretty crazy waveform at pin 3 of Q129 & Q130 (the unregulated 20v input, infact this ripple did not change)
But this ripple does not exist in other places of the unit, for example at R168 it is the expected sawtooth waveform.
You can see in my previous reply when VR109 was dead how the ripple then was present there too, but now it is a perfect sawtooth at R168

I realized one thing today after my post yesterday, relevant part quoted above.
The crazy ripple seen in the Youtube video and Scopemeter shots yesterday are infact only present after resistor R302 for Q129 and R303 for Q130.
So I uploaded Scopemeter screengrabs today of all unregulated 20v areas I could easily reach to paint a better picture, the screenshots are named with the measured components name.
I think of concern is the ripple measured at R302: This is not present in the other -20v traces (CR102- & Q115)
Meanwhile the +20v has a bit more ripple after CR102+ that is on Q116 and R168 both show a little bit ripple.
I am missing R303 (I could not figure out where to get to it topside as it is routed under other components) but then the unit decided to fail again before I could take the measurement.
It is now back to the same behavior as in the OP, which is all the well, because I did do some more tests with the freeze spray:
I am now about 75% certain that it really is the cooling down of U139 that brings the reference voltage online again.
So it will need to be replaced if I am right, would that affect the calibration of the unit?
At least I have some numbers from it now when it was working to compare against...

Also for completeness sake here are some voltage readings I took yesterday and today when it was still working perfect.
These are without the extra 18K parallel resistor on R315:

Code: [Select]

VR113 Cathode: 14.610v Anode: 8.395v
Q130 3=20.3v 2=13.966v 1=14.610v

VR112 Cathode: -8.284 Anode: -14.574v
Q129 3=-19.4v 2=-13.966v 1=-14.576v

Code: [Select]

U139:
Pin1: AGND
Pin2: -15.89v
Pin3: 6.982v
Pin4: 6.982v
Pin5: 8.382v
Pin6: 15.82
Pin7: -8.269v
Pin8: -0.0007

[Kleinstein]

Changing U139 may effect the calibration a tiny little bit, as the offset will effect the reference current. The main part of the scale factor is still from reading the 7 V ref or divided down 1 V or 0.1 V signal. So it does not include U139 in the path. The Keithley meters tend to do the check of the ADC scale factor in real time in parallel to the conversions - the newer HP meters tend to use this only occasionally (e.g. during ACAL) and otherwise rely on the resistor ratios.

The other parts that sometimes fail in such DMMs are LM339 comparators used to control JFET switches. Not sure why they fail but seem to be a part with limited lifetime and other meters are effected as well. I don't see such a failure here, but it would be a part to maybe have at hand. 

The ripple really looks like from mains and for the high frequency part seen in some pictures like some higher frequency oscillation. The capacitors usually don't fail very often, but failure at C220 or C219 could cause oscillation. The exact value there does not matter. So it could be a try to replace those 2 caps, possibly also with a larger value for an initial test.

[Per Hansson]

Thanks for the advice about the comparators, I will be sure to add a few to my shopping cart.
I tried to measure the SMD caps in circuit but got different readings with all meters I tried.
The Agilent U1733C made the most sensible readings, it showed around 300nF at 100Hz and and 30nF at 1Khz
I did try bodging on a cap with a value of ca 20nF and another with 47nF but it made no difference.
Is there a complete parts list somewhere or do you know what type of capacitors these are so I can buy a couple?
As I was writing the rest of this message the device started to work again.
So attached to this post is the missing R303 scope screengrab, the noisiest up the path to Q129/Q130 so far.

I also need to add some parts for VR109 that I mentioned failed in my reply #14 & #15 and consider replacing R168.
The partnumber of VR109 is known from the quoted thread below, but I looked more at this today and got really confused:
All that VR109 seems to do is supply U123 an AD706JR Dual Picoampere Input Current Bipolar Op Amp with its required 750µA Supply Current.
But it's negative supply comes from the regulated -15v supply, so why did they not just use the regulated +15v supply for the positive supply then?
And why in the world would they pick a 270Ω resistor for the 17v zener diode if so little supply current is actually needed?
Ontop of that the Op Amp is only rated for 18v maximum, and in my meter I measured 18.7v on the Zener, MetraCollector measured the same on his & flolic measured 18.6v.
The switching transistor Q118 only drops this by ca 0.6v so we are right at the maximum here...

Thanks for the reply!

I've actually made some progress on this; I'm waiting on some parts to arrive before I'll know more.  But what I know is that the zener I put in was running way too hot.  The resistor in series with it (R168 iirc) is only 270 ohms, and that put my zener at its max wattage. Maybe you should check yours? Curiously, my other k2k has a 1k for R168, even though the repair manual specifies 270.

[E-Design]

If you are still looking for MFG part #'s next week, I can provide them once I return to the office (next week)

[Kleinstein]

In the retraced circuit C220 and C219 are shown as 470 pF. this would than likely be C0G type, though the exact type should not matter.

The AD706 is specified for up to +-18 V and thus 36 V total. -15 V and +18 V would be only 33 V, so still OK.
It looks like they need to OP to work up to 13.3 V at the input and for this seem to need a little more than the +15 V.

Depending on the zener type for VR109, it may need some current to reach it's voltage.  If the voltage is sufficient a little less current may be OK. VR109 / Q118 not working may also upset the reference part if the positive supply to U123 goes below some 5 V.

It may be worth checking the ohms ranges too. To high a voltage in the ohms ranges may have damaged VR109 and maybe also more parts at the input (e.g. Q120).

[Per Hansson]

Thank you both for your replies.
I learned something today: the numbers in the schematic are capacitor codes, not capacitance ratings.
So the "470" for C220 & C219 means they are 47nF capacitors.
I have also been lazy and just searching for "C220" in the repair manual, but that turns out to have been a mistake:

So I will add a few Kemet C0805C470J1GAC and AVX 08051A470FAT2A caps to my order.

And thanks about the idea to check the ohms ranges, as mentioned the device passes the built in self test.
But I made sure to also test it with the resistances listed in "Table 2-9 2Ω/4Ω reference switching" of the service manual and it seems ok.

As for U123 the AD706JR Op Amp, it was exposed to the unregulated +20v when VR109 failed, because it failed shorted.
I do not think it was a low impedance short though, because it measures around 100Ω at room temperature with a DMM.
That said I have some AC ripple at the input of U123, I attached scope shots of it, for sure my current 18v through-hole zener is the wrong part.
So we will see what happens when I put in a MMBZ5247BLT1G, I will also experiment with R168 when I do.
It appears like I have the same problem as Galaxyrise in this quote now:

Since this isn't the only bad VR109 thread, I figured I should check the ones in mine.  Of the two K2k on my bench right now, one had a good VR109, the other seemed bad.  So I replaced it with what seemed to be the same thing (and then promptly lost the original, doh!)  But now the meter is "noisy" (fails 100.2), and indeed, compared to the working k2k, the voltage has some AC coming through it.  (In both  cases, there's a strong sine wave component coming into it.) Since it wasn't failing that test before, it seems likely that I used the wrong part, but, what was the correct part, then?

(The repair manual calls it a "DIODE, ZENER 17V, MMBZ5247BL (SOT-23)")

[Kleinstein]

Of cause 470 stands for 47 * 10^0 pF  = 47 pF, the normal cap coding. 
No need for a special type. Any C0G or similar should be OK, +-20 % probably OK too.

The unregulated +20 V with  the -15 V would still be likely within the 36 V range for operation, so likely no damage to u123 from that.
The critical case would be high ohms, like 1 M or 10 M that may be effected by extra leakage. The test is just in case as this may be a path to damage U109 or surrounding parts.

It is somewhat odd to see the zener Cr109 fail more often - maybe a not so reliable batch. Chances are 17 V zeners are not used very often and a roll may last quite some time. If one wants to use less current there it may be better to use a 18 v zener. At a reduced current it may give you 17-18 V.

There should be not much AC ripple around U123. Even if the supply from the zener has some ripple this should not effect the signal much (e.g. measurable with a normal scope).

Anyway, the reference amplification looks odd with the fast OP. There is no fast variable load to the reference and thus no real need for fast amplification. One ideally wants some low pass filtering to get rid of 10-50 kHz range noise of the reference, so more like some larger caps at the OPs to slow things down even more.  Other dmms use more like slow OPs like AD706, LT1001 or OP177 in a comparable circuit part. So if you need to replace U139 and are a little adventurous, one could consider a LT1013 instead of the LT1124.
Noise wise it should be slightly better, because of lower current noise.

[Per Hansson]

Thanks Kleinstein.
The unregulated 20v is around 22.3v depending on incoming line conditions, so in theory U123 could have seen 37.3v.
But more than likely VR109 did not fail in a complete short as I mentioned so I agree and think it is still ok.

I just had a thought about LT1124, the part marking on mine is just "1124" which confirms it is a LT1124CS8 according to the datasheet.
Actually in the repair manual Keithley have called it LT1124C58 but I am sure this is just a typo, the Keithley reference is: IC-955.
As I have mentioned mine runs at around 75°C to 80°C and if we look at xdevs IR shots his ran at 90°C.
But according to the datasheet the part is only rated for up to 70°C operating temperature.
So should we not be using LT1124IS8 / LT1124AIS8 rated for 85°C or even better LT1124AMPS8 rated for 125°C?
I ask because yes I will replace it, but I don't feel adventurous enough to try the LT1013, or do I?