AC converter schematic--how does it work? (Fluke 5101B calibrator repair)

[bdunham7]

I have another old Fluke 5101B on the bench and it has a problem with runaway AC voltage output and a circuit I am unable to figure out.  At one time it appeared to be working, then I looked over and saw that it was cooking my DMM with 1345+ volts.  DC outputs are stable and work properly.  The problem has been narrowed down to the AC converter circuit on the analog control board.  I actually have two analog control boards for this and they both have problems with the AC converter circuit, although the problems are not exactly the same.  On one board, we'll call it board M, the output is occasionally OK but usually jumps up to around double what it should be.  The other, board F, occasionally works almost OK with some minor instability, but usually 'runs away', the output creeps up and then jumps to the maximum that the power amplifier can put out in whatever range it is in.  I stay out of the high voltage ranges now just to avoid any unpleasant consequences.

Here's the schematic:



The description given in the theory section of the manual is:



The way that it operates is that it receives a feedback input of 0.2 to 2.0V_rms from the ranging module (scaled from the output) and converts it to a 0.2 to 2.0V_DC output that goes on to the control circuitry which adjusts the output power amplifier accordingly.  The input is BUFFERED AC and the output is the upper left corner.  he ranging and control circuits are working properly--they work fine on DC and they produce the expected results from the wrong output of this AC converter.  So the problem is in this circuit.  My main issue at this point is that I do not understand how this circuit works.  There is a troubleshooting section and I've taken some of the suggested measurements, I can test components, etc, but since I work on these and have two broken units at hand, I'd like to actually understand the details of this circuit and as of now I'm a bit baffled--both as to how it works and why they designed it this way.  It needs to convert a known sine wave to DC as accurately as possible over the range of 0.2 to 2.0 volts, it doesn't need to do TRMS and it needs to be linear.  At first I glance I thought U42 was a comparator and this was a half-wave rectifier type AC converter, but it isn't that.  C39 and the description indicate that U42 is an integrator.  In normal operation, the signal levels at the gate of U36 are very low, so I don't see how it can be called a 'high-impedance' amplifier, as the feedback apparently cancels out the input at the gate.  I think that would be more like a transimpedance amplifier with the input resistance of 10k +/-20 ohms converting the input voltage to a scaled current.

I also am having some difficulty finding datasheets on some of the parts.  I'll list the part numbers I have and then I'll move on to a second post in a little while with my measurements and tests I've done so far.

Q36 is a N-JFET, Siliconix J2908, no datasheet.
U42 is an Analog Devices A3902 AD3092, no datasheet and I'd really like to know if it is anything special or weird...
U37 is an LM318H.
Q44 is a 2N3906 and yes they forgot the emitter arrow (should be on top).
CR37 is not really a diode and the 3mA and arrow are not just a helpful note.  It is a Siliconix J2901 3mA N-JFET constant current diode assembly.  It can be tested by looking for ~1.5V across R37 and that has always been good.
Q45 and Q46 are just some small-signal transistors MPS6520 and MPS6522. 
The various inductors are just ferrite beads or ferrites with a few turns of wire.  Q45 and Q46 have ferrites on their base leads to prevent oscilliation--and the circuit will oscillate at about 5MHz if you probe it the wrong way.
CR36/37 are FD7223 small signal diodes, but they state 'selected'.  That sounds ominous.

Any ideas on the theory behind this thing?  I'm especially puzzled that they connect the input to the output through C30 and R30, I have no idea what that accomplishes.

[bdunham7]

So here's the relevant part of the basic troubleshooting guide:



and my results:

WORKING:  55) yes, but not 100% stable  56)  RMS=1.00V, DC=0.9995V   57) yes   58)  U42-2: -0.8mV (adjustable to zero with AC zero pot) U37-2: 2.33V  U37-3: 2.32V  U37-6: 4.84V_DC with about 2V_{AC p-p}

NOT WORKING:  55: no, 24.2V_AC  56: RMS=2.42, DC=0.2804  57) no  58)  U42-2: -0.71mV (didn't try adjusting)  U37-2: 0.630V  U37-3: 32.7mV  U37-6: 1.185V.

In addition I connected a scope to the calibrator output, the BUFFERED AC and various points in the circuit.  CH1 is connected to BUFFERED AC, CH3 to calibrator output in all captures.

In this capture, I forgot to turn on measurements, but it is not working and it putting out about 24 volts.  CH4 is connected to Q36/gate and CH2 to U37-2.



Here it is not working, CH4: Q36/gate, CH2: U37-3



Here it starting working mostly correctly (after sitting a while)  CH4: Q36/gate, CH2 U37-6.  Note that the signal at Q36/gate is very low, yet U37 outputs is 2V_p-p



And here it stopped working again.  CH4: Q36/gate, CH2 U37-6.  The feedback appears to be driving Q36/gate negative which puts U37 to the rails during part of the cycle. 



Any and all ideas welcome!  I strongly suspect the problem may be Q46, or perhaps Q45, but I'm not in a huge hurry to fix this and I really would like to fully understand what goes on here, especially how the feedback is supposed to work with the rectification and integration functions.

[JimmyJo]

Hey, I have repaired 5100B with an intermittent instability in it's AC output.  I finally tracked the problem to a failing capacitor, C30 in the schematic.  It would turn resistive if i warm it up slightly with hotair/ apply physical pressure, in the 70ohm range, and would recover if left alone.  I replaced it with a CDE 942C series 0.33uf.  Hope this helps.

[David Hess]

I can explain some of what is going on.

JFET Q36 buffers the inverting input of the U37 (the LM318 was the first "video" operational amplifier but suffers from high input bias current and poor DC precision) for low input bias current.  Only the inverting input needs to be buffered because the overall feedback loop is inverting.

U42 monitors the summing node at the gate of Q36 and adjusts the non-inverting input of U37 to force the gate of Q36 to ground.  The exact voltage is trimmed with R39.  This also removes any variation if Q36's Vgs and U37's input offset voltage.

Q46 sinks a constant current from Q45 which provides twice that current.  U37 drives emitter follower Q44 to steal half of that current from the emitter of Q45.  The result is that the voltage output of U37 is converted into a non-inverted current output at the collectors of Q45 and Q46.  The reason for using a current output is that there is no slew rate limit when switching the rectifying diodes CR46 and CR47.

CR46 closes the feedback loop on positive half cycles at the input producing no output, and CR47 closes the feedback loop on negative half cycles at the input, which appear at the output.

U51 is a method to make a low pass filter with no offset voltage because the input and output are AC coupled.

I would have tried driving Q45 and Q46 with the supply pins of U37 to make a current output but their method has the virtue of class-a operation so no switching distortion.  CR37 could have been replaced with a transistor current source driven by U42, with the non-inverting input of U37 grounded.  Or CR37 could have been replaced with the same transistor as Q36 plus a resistor.  Constant current diodes are pretty neat though.

U42 is an Analog Devices A3902, no datasheet and I'd really like to know if it is anything special or weird...

A3902 sounds familiar but I cannot place it.  Is it a normal looking IC or a module?  The application looks suitable for a LM308.  A modern precision JFET or superbeta bipolar part like the LT1008/LT1012/LT1097 could be used if that is the case.

CR36/37 are FD7223 small signal diodes, but they state 'selected'.  That sounds ominous.

I think they are selected for low leakage.  Just like now, back then tested low leakage diodes only came with a premium price.

Any ideas on the theory behind this thing?  I'm especially puzzled that they connect the input to the output through C30 and R30, I have no idea what that accomplishes.

The overall circuit is inverting, so I think R30 and C30 are selected to cancel feedthrough on positive input half cycles from imperfect cancellation of the input signal at the gate of Q36 and capacitive coupling through CR47.  In principle the gate of Q36 is a virtual ground and never changes but open loop gain is finite so there is some tiny variation.

On negative input cycles, feedthrough would appear as a gain error.

[bdunham7]

JFET Q36 buffers the inverting input of the U37 (the LM318 was the first "video" operational amplifier but suffers from high input bias current and poor DC precision) for low input bias current.  Only the inverting input needs to be buffered because the overall feedback loop is inverting.

U42 monitors the summing node at the gate of Q36 and adjusts the non-inverting input of U37 to force the gate of Q36 to ground.  The exact voltage is trimmed with R39.  This also removes any variation if Q36's Vgs and U37's input offset voltage.

Q46 sinks a constant current from Q45 which provides twice that current.  U37 drives emitter follower Q44 to steal half of that current from the emitter of Q45.  The result is that the voltage output of U37 is converted into a non-inverted current output at the collectors of Q45 and Q46.  The reason for using a current output is that there is no slew rate limit when switching the rectifying diodes CR46 and CR47.

Thank you very much for that!  I really was not understanding what was going on with Q44/45/46.  I think my main mistake was thinking that there was double-inversion in the circuit, but the output really is inverted.  I was able to get some more scope captures of the circuit when it is working so I see how the output section needs to have a very high gain and bandwidth during the transition from one rectifier to the other.  The need for this high gain explains at least some of the circuit.

CH1: U37-6  CH2: CR47-46 and Q45-46 junction  CH4: Q36-gate.

CR46 closes the feedback loop on positive half cycles at the input producing no output, and CR47 closes the feedback loop on negative half cycles at the input, which appear at the output.

And it is a half-wave converter, which I should have seen--especially since I was just looking at a half-wave AC converter in another instrument (the Fairchild 7000A) When I look back at that, I see that instead of being radically different, this circuit is actually remarkably similar--just drawn differently.  I think R30 and C30 are what threw me for a loop.  If you look at the schematic there is a '7' flag near R33 and one near R36.  If you look at the parts list, R33/36 are a tolerance and tempco matched set.  That makes it obvious what they are doing--using one half of the cycles only for 'waste' feedback to maintain DC balance, with the other side having a more precise set of feedback dividers.  Now that I see it it seems so obvious...

The description of U42 also confused me for a while until I realized that its description as an 'integrator' has little to do with the outputs.  It just keeps U37 balanced.  It's like there are two circuits operating over the same set of components--a high-speed one that produces the output and a low-speed feedback that keeps the DC balance.  And that may be leading me to the source of the problem, because what happens is that the DC balance goes away and the Q36 gate gets biased very negative.  That results in reduced to no output, which the rest of the calibrator responds to by increasing the power amplifier output.

A3902 sounds familiar but I cannot place it.  Is it a normal looking IC or a module?  The application looks suitable for a LM308.  A modern precision JFET or superbeta bipolar part like the LT1008/LT1012/LT1097 could be used if that is the case.

It is an 8-pin metal can, probably TO-99?  I'd really, really like to find some info on it--even a catalog listing or advert--because while it looks like a pretty jelly-bean level application, you never know.  I'm mostly concerned with stability--the Q44/45/46 section is highly prone to oscillation.  Just the capacitance of some probes makes them burst into a 5 MHz song.  I'm guessing the U42 section won't do that, but my repair experience tells me to find out as much as possible.  For example, A3902 AD3092 might be some magic temperature compensated op-amp made just for Fluke.

This capture shows the circuit not working, Q36-gate is biased very negative which should cause U42 to be nailed to the positive rail (or whatever it does with a full offset) since its positive input is fixed--and that should show up on  U37-3.  This shot is CH1: U37-6 CH2: U37-3  CH4: Q36-gate.  So I'm thinking Q42 is going bad.   I was unable to revive it with ice and it is difficult to probe some of the connections, so I'll probably try later to do both inputs and the output of U42 and U37 in sequence.

CH1: U37-6  CH2: U37-3  CH4: Q36-gate

[bdunham7]

My apologies--the op amp is an Analog Devices AD3092.   And I think it may be the issue, just waiting on one more cool-down cycle.

It also is apparently known by some other part numbers:  225961 (Fluke), 3140-225961 (??), AD3092, SG8023 (Microsemi) and NSN 5962-00-132-6389.  TO-78 package.  Still no info other than some very basic stuff from the NSN lookup, like +/-15V supply.

[bdunham7]

Hey, I have repaired 5100B with an intermittent instability in it's AC output.  I finally tracked the problem to a failing capacitor, C30 in the schematic.  It would turn resistive if i warm it up slightly with hotair/ apply physical pressure, in the 70ohm range, and would recover if left alone.  I replaced it with a CDE 942C series 0.33uf.  Hope this helps.

OMG!!!  Silver bullet troubleshooting from 10,000 miles!  Thank you!

Here's what happened.  I have two analog boards, the one I've been working on and showing here, and another that sort of works, but the AC converter output is low by a bit more than 50% so the calibrator ends up putting out about double what it should.  I tested the C30 capacitor on the first board and didn't find any problem.  I ended up putting the second board in for some testing, now that I know more about how the circuit works and I just could not find any issue except that the output was 50% low.  So I tested the C30 cap and the damn thing had an DCR of 6 ohms, which I confirmed a few different ways because I couldn't believe my eyes.  I then swapped the C30 cap from the first board (F) to the second (M) and the whole thing now works properly.

So now in addition to the tantalum plague, all those white film TRW caps are now suspect.  And I still want to fix the first board, but I'm feeling a lot less motivated--and I don't want to chop up my working one. 

EDIT:  I was pretty convinced my problem on board F was in the area of U42 and after seeing the bad C30, I pulled out C38 and checked it.  It wasn't shorted exactly, but my LCR meter showed it to be a unstable and a bit weird.   temporarily put in a 0.033 mylar film cap, as well as a 0.33 version in place of the C30 I stole for the other board.  Success!  I now have fixed both boards.  Polypropylene film caps aren't high on my list of things likely to go bad, so all in all a bit surprising.  And a bit disturbing--there's a bunch of these in the unit.

[David Hess]

Thank you very much for that!  I really was not understanding what was going on with Q44/45/46.  I think my main mistake was thinking that there was double-inversion in the circuit, but the output really is inverted.  I was able to get some more scope captures of the circuit when it is working so I see how the output section needs to have a very high gain and bandwidth during the transition from one rectifier to the other.  The need for this high gain explains at least some of the circuit.

The LM318 is awfully fast, so I am not sure that they needed to make a current output stage for a faster slew rate.  Maybe there is some other reason.  Other contemporary designs used just the LM318 in a half-wave precision rectifier with one or another way to correct for offset voltage.

The description of U42 also confused me for a while until I realized that its description as an 'integrator' has little to do with the outputs.  It just keeps U37 balanced.  It's like there are two circuits operating over the same set of components--a high-speed one that produces the output and a low-speed feedback that keeps the DC balance.  And that may be leading me to the source of the problem, because what happens is that the DC balance goes away and the Q36 gate gets biased very negative.  That results in reduced to no output, which the rest of the calibrator responds to by increasing the power amplifier output.

U42 might also be called a servo amplifier but the common configuration for THAT function is an integrator.  The only signals it sees are the input offset voltage, and any changes at the virtual ground because of limited open loop gain of the main amplifier.

A3902 sounds familiar but I cannot place it.  Is it a normal looking IC or a module?  The application looks suitable for a LM308.  A modern precision JFET or superbeta bipolar part like the LT1008/LT1012/LT1097 could be used if that is the case.

It is an 8-pin metal can, probably TO-99?  I'd really, really like to find some info on it--even a catalog listing or advert--because while it looks like a pretty jelly-bean level application, you never know.

I searched through old Analog Devices databooks and did not find anything.

I'm mostly concerned with stability--the Q44/45/46 section is highly prone to oscillation.  Just the capacitance of some probes makes them burst into a 5 MHz song.  I'm guessing the U42 section won't do that, but my repair experience tells me to find out as much as possible.  For example, A3902 AD3092 might be some magic temperature compensated op-amp made just for Fluke.

U42 has nothing to do with oscillation of the voltage-to-current converter, and it is evident from the inductors in the schematic that they had oscillation problems, which is hardly surprising when an LM318 is involved.

I originally thought U42 would be an LM308 but the external compensation pins shown are 1 and 5 and the LM308 uses 1 and 8.  I could not find any operational amplifier which uses pins 1 and 5 for that compensation; maybe the schematic is marked wrong?

[coromonadalix]

seems the pin 1 and 8  on u42  not 1 and 5,   seems to be an lm308

[bdunham7]

U42 has nothing to do with oscillation of the voltage-to-current converter, and it is evident from the inductors in the schematic that they had oscillation problems, which is hardly surprising when an LM318 is involved.

When I connected a Tek 6139 probe to the Q36-gate on the problem child board, I would get the oscillation.  A different probe (also 10X, lower frequency rating) didn't cause it.  When I replaced the U42 feedback capacitor, C38, that issue has disappeared, although I can still get the Q44/45/46 gang to oscillate if I probe them directly.  I agree U42 probably won't be oscillating at 5MHz, but when you have something that is normally just barely stable, I suppose anything can happen.

I originally thought U42 would be an LM308 but the external compensation pins shown are 1 and 5 and the LM308 uses 1 and 8.  I could not find any operational amplifier which uses pins 1 and 5 for that compensation; maybe the schematic is marked wrong?

seems the pin 1 and 8  on u42  not 1 and 5,   seems to be an lm308

I checked the board, the compensation capacitor is connected to two sequential pins which by convention should be 1 and 8.  What I presume is 5 is NC.  They seem to have trouble drawing '8' here because I also mistook C38 for a C39 and I have a pretty good scanned copy.

So if I need to, can I just use an LT1008 as a direct replacement, provided I can ffind one in a can?

[David Hess]

When I connected a Tek 6139 probe to the Q36-gate on the problem child board, I would get the oscillation.  A different probe (also 10X, lower frequency rating) didn't cause it.  When I replaced the U42 feedback capacitor, C38, that issue has disappeared, although I can still get the Q44/45/46 gang to oscillate if I probe them directly.  I agree U42 probably won't be oscillating at 5MHz, but when you have something that is normally just barely stable, I suppose anything can happen.

There is no low impedance AC feedback path to the summing node at the gate of Q36 because it would limit bandwidth, and unlike most circuits of this type, a high impedance current output is used.  The high impedance current output drives resistive networks for about 22 kilohms to the summing node, and that makes the circuit sensitive to any capacitance at the summing node.  A 20 picofarad oscilloscope probe and 22 kilohms results in 45 degrees of phase lag at 360 kHz which is more than enough to make an LM318 oscillate.

Thinking about it now, I wonder if C30 and R30 were an attempt to stabilize it.

seems the pin 1 and 8  on u42  not 1 and 5,   seems to be an lm308

I checked the board, the compensation capacitor is connected to two sequential pins which by convention should be 1 and 8.  What I presume is 5 is NC.  They seem to have trouble drawing '8' here because I also mistook C38 for a C39 and I have a pretty good scanned copy.

That is almost too bad because if you could find a low input bias current precision operational amplifier from the 1970s which used a compensation capacitor between pins 1 and 5, then that could be it or at least a clue to what they used.  I spent last night looking through various old databooks and did not find any.

So if I need to, can I just use an LT1008 as a direct replacement, provided I can ffind one in a can?

Or pull the compensation capacitor and use an LT1012 or LT1097.  The largest problem will be finding something in a TO-5 metal can.  Considering how much better even an "old" modern part like the LT1012 is, I would make arrangements to use a DIP or SOIC part.

Note that whatever they really used, and I suspect it was a selected LM308 but maybe it was a precision JFET or precision JFET hybrid part, an LT1008/LT1012/LT1097 will be better.  The newer Texas Instruments OPA140 or OPA145 JFET parts would also be a good choice with their low noise and 1 or 1.5 uV/C offset voltage drift.  Beyond that any of the highest precision JFET parts are suitable.

The LT1008 and LT1012 do have the virtue that they were once available in the TO-5 metal can so could be a drop-in replacement, and with better performance than the original.  The TO-5 metal can versions can be found on Ebay but I would not trust them unless I had no other choice, or if I was prepared to test and verify their performance.

I remember long ago grading TO-5 metal can operational amplifiers for input offset voltage drift and marking the top of the can with a Sharpie.

[Kleinstein]

The circuit is not that high in impedane (DC wise). So the OP for U42 is not that critical - if even an LM101 could be possible, though I would not recommend it. There is a trimmer, so that offset or bias could be adjusted.
A modern possible LM308 alternative is the OPA202 (SMD only).

R30 and C30 are an integral part of the rectifier - this is the path for positve half wave.

The diodes CR46, CR47 are likely selected / checked for low leakage. It is not easy find diodes that are fast, low capacitance and low leakage and still can carry quite some current (e.g. 10-20mA). No need for super low leakage in the pA range, but still probably better than an 1N4148.

[bdunham7]

The circuit is not that high in impedane (DC wise). So the OP for U42 is not that critical - if even an LM101 could be possible, though I would not recommend it. There is a trimmer, so that offset or bias could be adjusted.
A modern possible LM308 alternative is the OPA202 (SMD only).

I think the main issue is offset drift.  There is a 10,000:1 divider on the AC ZERO adjustment, so you can certainly dial it in--but clearly it needs to be stable.  This is why I worry that this is/was a special part manufactured or selected for Fluke.  There probably are better alternatives, but fixing these beasts is challenge enough--modifying them for better performance would be clearly beyond my skill level.  So maybe I'll try anyway!   

R30 and C30 are an integral part of the rectifier - this is the path for positve half wave.

The positive output half-wave derived from the inverted negative cycle goes through CR47 and is filtered for the output.  I just haven't been able to figure out what C30 and R30 are all about, so first I measured the voltages.  With the calibrator set for 10VAC and thus a BUFFERED AC input of 1.000V, the junction of R30 and C30 has 1VDC + 1VAC to ground, R30 has 1VAC across it and C30 has 1VDC across it, just like it should if you just look at the output end, which is 1VDC and the input, which is 1VAC.  What else could it be?  So then I disconnected C30, leaving the two components out of the circuit.  The result is that the calibrator appears to work normally without them.  Whatever effect they have is either extremely minor or addresses some issue that doesn't always occur.  So now I don't feel so stupid for not being able to figure out what they do.

The diodes CR46, CR47 are likely selected / checked for low leakage. It is not easy find diodes that are fast, low capacitance and low leakage and still can carry quite some current (e.g. 10-20mA). No need for super low leakage in the pA range, but still probably better than an 1N4148.

I actually crossed them to 1N3062, which lists a not-so-low reverse leakage of 0.1uA but a zero-voltage capacitance of 1pF @ 1MHz.  This section of the circuit is indeed not  high impedance and is shunted with 562 ohms, but the front side of the circuit is high impedance and with R47 on the output, the next stage has to be very high impedance as well or it will load down this output after R47.  There's no buffer.

https://pdf1.alldatasheet.com/datasheet-pdf/view/92443/CENTRAL/1N3062.html

[TimFox]

Note that common switching diodes such as 1N3062, 1N4148, etc. are in glass packages and will be light-sensitive (if reverse leakage is critical).  I usually wrap mine in small-diameter black polyolefin heat-shrink tubing if that is a problem.

[xavier60]

I think R30 and C30 are there to supply some feed forward to the output to reduce ripple.
As full wave rectification would I guess but it's AC coupled.

[bdunham7]

I think R30 and C30 are there to supply some feed forward to the output to reduce ripple.
As full wave rectification would I guess but it's AC coupled.

So of course I had tried it both ways and it made no discernable difference, but I was using a 1kHz output.  When I switched down to 50Hz (the minimum) I had ~120mV_RMS on the converter output with C30 disconnected and 40mV_RMS with it connected.  The output of the calibrator was about 0.04% low at 50Hz with C30 disconnected.  So, yes, it appears to do just that.

Of course it makes sense in hindsight--the positive input would add some current when CR47 is not conducting, then take some away when it is on the negative input cycle for a net smoothing effect.  The net contribution to the DC current at the output will be zero, unless C30 leaks. This seems to be just an additional band-aid sort of thing, as if the original design just wasn't quite good enough and so they added this.  Additional output filtering would work as well.  One issue I saw is that with C30 connected, the output ripple is distorted while without it the ripple is a sine wave, albeit 3X larger.  C30 has an impedance at 50Hz of about a third of R30, so there is a phase shift at low frequencies--thus no cancellation will be perfect.  I wonder if there's any downside to increasing the size of C30...

[David Hess]

The circuit is not that high in impedane (DC wise). So the OP for U42 is not that critical - if even an LM101 could be possible, though I would not recommend it. There is a trimmer, so that offset or bias could be adjusted.
A modern possible LM308 alternative is the OPA202 (SMD only).

I wondered about that.  There is not much need for such a high impedance but the LM318, especially with no input bias current cancellation, is something else again so it makes sense to buffer its input.

The LM308 back then had an outstanding input bias current compared to other bipolar parts, but was comparable to modern precision parts like the OP-07/LT1001 which always surprised me, so that puts it into the nanoamp range and 10s of kilohm impedances which is right about where the circuit operates.  The parts that replaced the LM308 like the LT1008 have an order of magnitude lower input bias current, less than half the noise, and a fifth of the offset voltage drift, so they are a sure replacement. (1)

What they did *not* trim out was the offset voltage drift of U42 which is a major error term in a DC coupled design, and the LM308 has 1/3rd the offset voltage drift of general purpose parts like the LM301A, which is why I suspect they were using a LM308 selected for low offset voltage drift.  Several times I have replaced "high performance" parts comparable to the 301A/741 with a 308 only for better precision.

(1) In the 1990s I used a lot of National LM11s which are essentially improved LM308s with internal compensation but even higher input voltage noise.

[Kleinstein]

The normal full wave rectifier circuit has no C30 (just a resistor) and virtual ground / inverting stage at the output.
In that circuit die direct path is the path providing one half wave.
Here there is C30 as the ouput filter sees a DC voltage and the DC part thus need to be suppressed. Without C30 / R30 the amplifier still works as a half wave rectifier, only looking at the negative half wave. With a symmetrical sine this does not change much, as C30/R30 add the contribution from the other half wave, but also reduce the gain by a factor of 2. The nice point is that the accuracy of C30/R30 does not directly enter - after all they only provide some AC signal to fill gaps and reduce ripple.

A larger C30 may not have much downside, maybe slightly slower settling, but the main filter caps are 1 µF and thus still a bit larger and no series resistor.
It may still effect the low frequency calibration a little.

[David Hess]

The circuit is not that high in impedane (DC wise). So the OP for U42 is not that critical - if even an LM101 could be possible, though I would not recommend it. There is a trimmer, so that offset or bias could be adjusted.
A modern possible LM308 alternative is the OPA202 (SMD only).

I am still trying to figure out U42.  If they had wanted to use a 301, then there are several easy ways they could have added external input bias current cancellation and they had to know about it because the 301 application notes showed them all.

But if they did that, they could just as easily have added input bias current cancellation to the 318 removing the need for the JFET buffer.  That makes me think it really is a 308.

What options other than the externally compensated 301 and 308 were there?  Except for weird exceptions, there were not a whole lot of externally compensated parts even back then.

I hate to do an error analysis to figure out what specifications were required but maybe that is the best way.

[Kleinstein]

Just bias compensation at the LM318 may not be good enough, as there is still the current noise.
The OP should be some relatively low bias precision OP with less than about 1 mV offset. The LM308 would be a possible choice, though not the only one.

[David Hess]

Just bias compensation at the LM318 may not be good enough, as there is still the current noise.
The OP should be some relatively low bias precision OP with less than about 1 mV offset. The LM308 would be a possible choice, though not the only one.

But the 318 is not that much worse than a 301, and a 318 with input bias current compensation would be better.

Input bias current cancellation would always increase the input noise.

[easvcinfo]

Thanks for information about C30 in Assembly A14 of FLUKE 5100B.
So far this is what I found.