Replacement for old Intech or Fairchild op-amp.

[bdunham7]

This is a preamplifier for the low voltage range of a DMM.  The input range is 0-120mV plus some headroom, with a gain of 10X.  I haven't fully diagnosed the problem yet, but so far it looks like the op-amp is bad.  The requirements are low bias current, DC stability and compatibility with this circuit with as little modification as possible.

The service manual calls this out as a Fairchild ADO-26B, for which I found a brief release note but no datasheet.  The actual part installed is an Intech A-1001, for which I could not find any info.  I've attached the schematic, a photo, the release note.  Ideas welcome.

[Kleinstein]

There seems to be external pot for the fine trim. So one could get away with a low offset OP and ignoring the trimmer for the OPs offset.
The DC performance of the original does not look that great ( < 2 µV/K drift) so one could get away with a normal, non AZ op, like OPA140 or OPA145 (slightly slower, but may be higher offset to start with).  The change would be ignoring the ground and offset trimer for the OP. I would add a capacitor (or 2) for local decoupling, just in case.
The fast OPA140 might like some capacitance in the FB to slow it down.

The OPA140 may be available as DIP for an easy dead bug fix. A SO8 version could use a SO8  to dip adapter board and use wires from there instead of the pins.

I would first check the supply - just in case the zeners at the supply could have failed and this could damage a replacement.

[bdunham7]

There seems to be external pot for the fine trim. So one could get away with a low offset OP and ignoring the trimmer for the OPs offset.
The DC performance of the original does not look that great ( < 2 µV/K drift) so one could get away with a normal, non AZ op, like OPA140 or OPA145 (slightly slower, but may be higher offset to start with).  The change would be ignoring the ground and offset trimer for the OP. I would add a capacitor (or 2) for local decoupling, just in case.
The fast OPA140 might like some capacitance in the FB to slow it down.

The OPA140 may be available as DIP for an easy dead bug fix. A SO8 version could use a SO8  to dip adapter board and use wires from there instead of the pins.

I would first check the supply - just in case the zeners at the supply could have failed and this could damage a replacement.

Zeners are OK, the op amp has a total supply voltage of ~32VDC.  With inputs and outputs disconnected from the rest of the unit, the output is ~-13VDC with an input of ~+0.49VDC, which is just the output/10, limited by the input protection diodes.  So it looks like the op-amp is the issue.

I will probably have to use a modern device without the COM and TRIM connections, but I'm wondering if I'd be better off bodging something into the original design or just making new board.  The OPA140 actually looks pretty good except I think I need to lower the supply voltage a bit.  I'm wondering if its initial offset will really be good enough.  I might want to revise the circuit using existing components, perhaps using the balance pot in the existing zero circuit somehow.  My initial idea was to replicate the original hybrid design by adding JFET inputs to something like a 741, but that probably is silly.

[Kleinstein]

The moden JFET OPs like the OPA140 are really good - the only point missing is often the possibility to trim the offset. It would be very hard to get close performance (dirft and offset) with a old stype design with seprate fet and the low frequency noise is also not so easy to beat. It would likely need 2 step trim for offset and TC.

There is the external offset trim (likely from the front panel). With a divider 1.3 M and 2.7 Ohms this would not be very much trim range (e.g. +-30 µV), but it could be just enough if the offset is at the typical value and not much higher.

What is the resolution of the meter ? So how good does it have to be ?


There are also OPs (like OPA191/192) with lower offest, though slightly higher noise that could be used. 1.4 µV_pp noise for the 0.1 - 10 Hz range is still not that bad.

The OPA134 is an audio OP, but still relatively good LF noise and it still has offset trim pins. There may be others too - but the choice is not that large anymore.

Another possible option would be a small board with an AZ OP like ICL7650 or similar. This would however need a reduced supply (e.g. +-6 V).
A +-2 V ouput range would still be possible even with an AZ OP with only a +-2.5 V supply and thus a low voltage type.
There are AZ OPs (e.g. LTC2057, OPA187, ADA4522) than can run with a 30 V supply, but these tend to have a relatively high bias (more like 50-200 pA range).

The diodes on the board shown are accross the OPs inputs. So the actual input voltage can be higher than 500 mV.  To test for the amplifiers offset one should have a short at the input, not an open input.

[bdunham7]

The moden JFET OPs like the OPA140 are really good - the only point missing is often the possibility to trim the offset. It would be very hard to get close performance (dirft and offset) with a old stype design with seprate fet and the low frequency noise is also not so easy to beat. It would likely need 2 step trim for offset and TC.

There is the external offset trim (likely from the front panel). With a divider 1.3 M and 2.7 Ohms this would not be very much trim range (e.g. +-30 µV), but it could be just enough if the offset is at the typical value and not much higher.

What is the resolution of the meter ? So how good does it have to be ?

It's an old Fairchild 7000A that I got from user Runco recently.  It really doesn't have to be particularly good because it is more a collectors item than anything else--although I may end up using it just because I like to use old stuff.  It is a 12000 count meter and this preamplifier is for the 100mV range, so the LSD is 10uV.  The mV offset is on the front panel and is a less-than-single-turn pot.  I could modify the trim circuit and use a multi-turn, I suppose.  I don't know how stable it was in the 100mV range originally, but on the other ranges it is surprisingly good for very ancient (late 60's) technology.  It autoranges smoothly and correctly measures the DCR of inductors (I had to check, of course). 

There are also OPs (like OPA191/192) with lower offest, though slightly higher noise that could be used. 1.4 µV_pp noise for the 0.1 - 10 Hz range is still not that bad.

The OPA134 is an audio OP, but still relatively good LF noise and it still has offset trim pins. There may be others too - but the choice is not that large anymore.

Another possible option would be a small board with an AZ OP like ICL7650 or similar. This would however need a reduced supply (e.g. +-6 V).
A +-2 V ouput range would still be possible even with an AZ OP with only a +-2.5 V supply and thus a low voltage type.
There are AZ OPs (e.g. LTC2057, OPA187, ADA4522) than can run with a 30 V supply, but these tend to have a relatively high bias (more like 50-200 pA range).

The diodes on the board shown are accross the OPs inputs. So the actual input voltage can be higher than 500 mV.  To test for the amplifiers offset one should have a short at the input, not an open input.

I assumed that since the input voltage was opposite the maxed-out output, the op-amp was bad.  You're right, shorted would be better as it would eliminate the possibility of me mistaking the polarity, so I tried again--same result of -13ish volts with inputs shorted.

I'm considering the LF155AH as a drop-in replacement.  It's old, but I think I can get one and these were designed specifically as a replacement for JFET hybrid module op amps.  The offset TC is a bit worse than what the ADO-26B was advertised at, but it might be worth a try--as long as I can zero it, I think it meets the original design goal.  Any reason it wouldn't work?

[Kleinstein]

A TO99 can could be nice from the optics. The LF155 still has a relatively large drift and also quite some noise.
The noise may be just visible (estimate some some 5-10 µV_pp for 0.1 to 10 Hz with mainly 1/f noise). The drift (up to 10 µV/K) may need a re-adjustment (at the front) from time to time.
The trim part would still need a seprate trimmer or changes to the board - this is also true for other OPs, unless they get away without trim (e.g. OPA191).

I would prefer just an additional trimmer. With the trimmer one should be able to zero it out - without the trimm range from the front will be very likely too small.

[floobydust]

I did find ads for the ADO-26B ED May 1969, not much mentioned and it didn't go into aerospace/military use.
2uV/°C
100,000 gain
20,000:1 CMRR 86dB
100uV/V supply rejection
5uV p-p noise

Shorting the inputs I guess rules out the feedback resistors/trimpot are not open-circuit. C1 1uF has some green at the edges, I thought that was corrosion and would check its leakage current. It has a manly 51k 2W carbon resistor feeding it, wow.
It might be possible to disassemble the old op-amp module, unless it's hard potted? It is lacking output short-circuit protection so whatever is downstream might need to be checked if the output stage has failed.

[bdunham7]

A TO99 can could be nice from the optics. The LF155 still has a relatively large drift and also quite some noise.

You're probably right that adapting a modern op-amp without a trim/balance terminal would perform better.  However, looking through my junk spare archaic parts trove, I discovered that one of the boards from an old HP spectrum analyzer has two 1826-0319 TO-99 op amps.  Depending on how and where you look, these cross over to National Semi LF156A (best possibility) or a Freescale LF356G (can't find a datasheet).  I think I'll just bodge it on there when I get time, and if everything works, I can see how well it works.  The meter is stable to 1 count in the 1V range, so there's no point in having a 100mV range where the last digit is just noise.

I did find ads for the ADO-26B ED May 1969, not much mentioned and it didn't go into aerospace/military use.
2uV/°C
100,000 gain
20,000:1 CMRR 86dB
100uV/V supply rejection
5uV p-p noise

Shorting the inputs I guess rules out the feedback resistors/trimpot are not open-circuit. C1 1uF has some green at the edges, I thought that was corrosion and would check its leakage current. It has a manly 51k 2W carbon resistor feeding it, wow.
It might be possible to disassemble the old op-amp module, unless it's hard potted? It is lacking output short-circuit protection so whatever is downstream might need to be checked if the output stage has failed.

I'll look again when I take it apart again, but I don't recall seeing any corrosion.  The 2W 51K is so that it can withstand clamping a 1000VDC input. 

I don't know if I can get it open.  I thought of sending it to Noopy if he was interested in a hybrid like this.  I could find absolutely no information on an Intech A1001.  I'm assuming this meter is an older version and that they switched to the Fairchild hybrid later on--the manual I have is from 1970 and doesn't mention the Intech.  But with no info and no dates on the meter, I can't tell.  It is serial # 1426, I've no idea of the total production volume.  If anyone else has one and wants to compare notes, I'm game.

[Kleinstein]

The LF356 is usually the consumer temperature range version of the LF156. So a bit odd to have a metal case for the LF356 at all - but it makes some sense that HP may use both for there internal number. Chances are it would be good enough noise wise, though the performance can vary between units.

[TimFox]

In the mid '70s, our lab in grad school used a lot of Analog Devices modular op amps in packages similar to your A1001.  The modules were not hybrids:  we de-potted an AD180 general purpose unit to satisfy our curiosity.
Inside, there was a TO-18 dual input transistor, a uA702 (!) monolithic amplifier, TO-92 transistors (including output stage), 1/4W carbon and film wire-lead resistors, etc. on a two-layer PCB.  I could not understand how the circuit (used at +/- 15 V) could use the uA702, which was rated for 21 V (usually used at +12, -6 V power supplies), but was extremely fast.
We did some high-voltage discharge tests (for use in precision several-kilovolt supplies), and the AD118s were a hell of a lot more safe against a capacitive discharge into the output terminal than was a monolithic uA741 or  PMI OP-05.

[5065AGuru]

Thought this looked familiar.

I have a PC card that has an Intech ADO-27B on it.

I can pull the module and you can have it for the shipping cost.

Not sure if it's any good or not!

Cheers,

Corby

[bdunham7]

Thought this looked familiar.

I have a PC card that has an Intech ADO-27B on it.

I can pull the module and you can have it for the shipping cost.

Not sure if it's any good or not!

Cheers,

Corby

Worth a shot!  PM sent, thanks!

[coppercone2]

I have a few of those op amps that I got from the trash, same package at least

Very interesting part. I would just buy a replacement on ebay because  as you know, if you do the leg work, you can get some monster amplifiers made with simple parts. I would get it from that guys card if the price is reasonable, you have the mechanical benefits and possibly its more resistant to high temperatures in the chassis and possibly supply transients etc.. you would need to check the rail to make sure its not too crazy for small modern sensitive parts. if you replace it wiht a new one be sure to do a power on test with a oscilloscope to see if the rails have over shoot and stuff like that

its also alot more resellable I think if you have the original parts without mods if you ever chose to sell it to a collector

[bdunham7]

Not sure if it's any good or not!

Installed and meter works, stable to 1 count (10μV).

Thanks!

[Runco990]

Nice to see you trying to fix it!  Another "Project" I just didn't have time for.   

[David Hess]

That is a tough application for an operational amplifier because of the required precision in the form of offset drift and common mode rejection, and the low input bias current; very few FET input parts meet the minimum common mode rejection requirement but selected ones will.  That Fairchild part uses JFET pairs selected for precision and you might very well have to do the same thing to get those kinds of specifications in an untrimmed integrated operational amplifier.  (1)

A super-beta bipolar part like the LT1008 or LT1012 is actually better but there are some barely suitable JFET input operational amplifiers.  I have been working on refurbishing a Tektronix DM501A and I have the same application problem.  Tektronix originally used the AD542 and its replacement is the AD711.  Barely suitable *36/40/44 volt DIP parts* include the OPA134 and OPA604 although there is a tradeoff in offset voltage drift between them.  The various precision replacements for the LF355/LF356 in the form of the LT1055/LT1056/LT1022 are also good candidates but none of these parts, except the super-beta bipolar ones, are always going to meet the required specifications.

I am planning on testing the LT1008, TLE2071, OPA604, and OPA134 for my DM501A rebuild.

The CMOS OPA191 and OPA192 should work very well but are not available in DIP packages and as mentioned, do suffer from higher noise so it may not be possible to take advantage of their higher potential precision which is why I am so interested in the LT1008 despite its slightly higher input bias current.

If cost is not important, then the OPA627 is excellent.

(1) The common mode rejection requirement is so stringent that if I designed a modern implementation of this circuit, I would seriously consider bootstrapping to reduce it.

[bdunham7]

That is a tough application for an operational amplifier because of the required precision in the form of offset drift and common mode rejection, and the low input bias current;

I'm quite relieved to have found an actual replacement for it.  I ran through the calibration procedure and ended up with 42pA bias current on the mV range (spec is <100pA).  There's not a separate way to control this on the millivolt preamp like there is with the main amplifier--that adjusted down to 10pA.

Thanks to everyone for your comments and especially to 5065AGuru for the replacement ADO-27B.

Now I need to figure out why it seems to be drifting a bit when I turn it off.  (nothing to do with the repair) 

[David Hess]

I'm quite relieved to have found an actual replacement for it.  I ran through the calibration procedure and ended up with 42pA bias current on the mV range (spec is <100pA).  There's not a separate way to control this on the millivolt preamp like there is with the main amplifier--that adjusted down to 10pA.

I posted my thoughts anyway because of my current DM501A multimeter project where I am actually looking to improve the performance.  The original design of the Siliconix integrating ADC used in the DM501A was flawed so suffered from non-linearity from low common mode rejection of the input amplifier.  Siliconix recognized this at some point and released a revised version of their ADC which supported an external buffer which allowed the use of an external JFET operational amplifier for a higher common mode rejection, but it is difficult for even the best JFET parts to meet the requirements.

The Intersil ICL71xx series solved this problem by cancelling the common mode error using the automatic zero cycle, which makes me wonder if Intersil patented it preventing Siliconix and others from doing the same.

[bdunham7]

I posted my thoughts anyway because of my current DM501A multimeter project where I am actually looking to improve the performance.

I'm not sure why the op-amps like these faded away instead of getting better.  The main input amplifier on this unit is a discrete high input impedance op-amp with separately adjustable balance and bias.  The mV preamp uses the ADO-267B which only has adjustable balance, the bias ends up being whatever it is.  Decades later Fluke was still building their top-of-the-line 8505A/8506A with a discrete op-amp front end with similar adjustments and I presume similar design.  If that ADO-2xB had been expanded and developed, it probably could have take the place of that whole discrete op-amp, since the performance appears to be quite good as it is.  Instead it appears to have gone away.  Perhaps it was very expensive?

I try to resist the urge to improve most of the old-tech things I work on, not because I'm a stickler for originality, but rather the realization that most of these things were made very carefully by very smart people.  Also, a lot of these things were quite advanced and expensive in their time and while it looks like you can outdo them with modern equivalents for pennies on the dollar, that's often not really true.

What is interesting to me about this old (1967 or so) DMM is how well it actually performs, or hopefully will once I get it straightened out.  It is only 12,000 counts, but it has high-impedance inputs to 12V, <10pA bias current (the max spec is <50pA, <100pA on the mV range) and the accuracy is +/- 1 count offset, +/- 1 count reading at full range and +/- 1 count per 5C tempco.  So essentially a max error of 2 counts over the entire range. 

[David Hess]

I'm not sure why the op-amps like these faded away instead of getting better.  The main input amplifier on this unit is a discrete high input impedance op-amp with separately adjustable balance and bias.  The mV preamp uses the ADO-267B which only has adjustable balance, the bias ends up being whatever it is.  Decades later Fluke was still building their top-of-the-line 8505A/8506A with a discrete op-amp front end with similar adjustments and I presume similar design.  If that ADO-2xB had been expanded and developed, it probably could have take the place of that whole discrete op-amp, since the performance appears to be quite good as it is.  Instead it appears to have gone away.  Perhaps it was very expensive?

I think they were just too expensive.

Integrated JFET and CMOS parts never achieved the performance of discrete ones but you can still get that performance and better today doing what you originally suggested; select a dual JFET to make a composite amplifier.  The integrated route using a part like the OPA627 or one of the other parts I mentioned is just much simpler and cheaper if its performance is acceptable.

It may not actually be necessary to match or exceed the performance of the ADO-267B in all aspects but this would take some reverse engineering to figure out what is important and the same list of viable replacements would come up anyway.

I have a number of old digital multimeters with discrete front ends which perform amazingly well even after 40 years, and improving on their performance with an integrated JFET or CMOS part would be difficult, except where a super-beta precision bipolar part could be used like the LT1008 or LT1012; they are like magic.

I try to resist the urge to improve most of the old-tech things I work on, not because I'm a stickler for originality, but rather the realization that most of these things were made very carefully by very smart people.  Also, a lot of these things were quite advanced and expensive in their time and while it looks like you can outdo them with modern equivalents for pennies on the dollar, that's often not really true.

In the case of the DM501A that I am restoring, the design was flawed and honestly I think the designers at Tektronix should have recognized the flaw in the ADC chipset they were using either during design or during testing.  Eventually it was identified and fixed, although Siliconix never owned up to it that I found, but the solution was less than ideal as I outlined above because it depended on the CMRR of a JFET operational amplifier.

When I do reverse engineering, I often find provable mistakes.

What is interesting to me about this old (1967 or so) DMM is how well it actually performs, or hopefully will once I get it straightened out.  It is only 12,000 counts, but it has high-impedance inputs to 12V, <10pA bias current (the max spec is <50pA, <100pA on the mV range) and the accuracy is +/- 1 count offset, +/- 1 count reading at full range and +/- 1 count per 5C tempco.  So essentially a max error of 2 counts over the entire range.

"Only* 12,000 counts - 12,000 counts with only 2 counts of error is a credible achievement and most modern comparable multimeters are not that good.

[floobydust]

Maybe x-ray the old module? I think there was a lot of labour involved to make them. Matching JFET's, calibrating, potting compound.

[Kleinstein]

I don't think the CMRR is really the problem with most JFET OPs. Even the old LF356 and related is not that bad. Definitely not a problem at the 4 digit level.
Some new ones like OPA140 are really good. I doubt the old  ADO-2xB would come only close.
The weak points with many JFET OPs are more the drift and low frequency noise.


Many old designs just used a pair of JFETs in front of an old (741 or similar) OP and this is a way to get a poor CMRR. The extra JFETs only help to get good CMRR if used with a kind of bootstrapping.

The super beta OPs get a rather good CMRR because they essentially need internal boostrapping for the input stage, as the super beta transistors only work with a low voltage. Some of the modern JFET OPs like the OPA140 use a similar bootstrapping, just with low voltage JFETs instead of BJTs.

With the relatively poor supply of matched dual JFETs it is today easier to bootstrap a whole OP instead of a JFET pair. I am quite pleased with the performance of a bootstrapped OPA145. As a buffer it is simpler than boostrapped discrete JFETs and with gain still OK.

[David Hess]

I don't think the CMRR is really the problem with most JFET OPs. Even the old LF356 and related is not that bad. Definitely not a problem at the 4 digit level.
Some new ones like OPA140 are really good. I doubt the old  ADO-2xB would come only close.

It is really making the best of a bad situation.  The CMRR of the JFET operational amplifier used in non-inverting mode limits the linearity of the following high resolution ADC.  There are parts that can do it for sure, but common ones do not quite make it with their minimum guaranteed CMRR.  I never heard of anybody selecting or grading them for high CMRR but I am sure some tried.

The 4.5 digit Siliconix chipset that I mentioned had a CMOS buffer with a CMRR of only 55 dB which is way to low.  I do not understand how they thought that could ever be sufficient.

I wish the OPA140 was available in a DIP package; I may have to make an exception.  Its precision specifications seem too good to be true; I wonder if there is a hidden gotcha.

The weak points with many JFET OPs are more the drift and low frequency noise.

Those are problems as well.  The drift is potentially cured by placing the buffer within the automatic zero loop but the flicker noise is practically intractable and can only be avoided by not generating it.

The super beta OPs get a rather good CMRR because they essentially need internal boostrapping for the input stage, as the super beta transistors only work with a low voltage. Some of the modern JFET OPs like the OPA140 use a similar bootstrapping, just with low voltage JFETs instead of BJTs.

I wonder how the old JFET precision parts managed it, although they were only incrementally better.

With the relatively poor supply of matched dual JFETs it is today easier to bootstrap a whole OP instead of a JFET pair. I am quite pleased with the performance of a bootstrapped OPA145. As a buffer it is simpler than boostrapped discrete JFETs and with gain still OK.

Bootstrapping would be my first choice but an alternative is to use the automatic zero loop to also correct the CMRR.

[Kleinstein]

I still don't see how the automatic zero function can help with the CMRR. It helps inside AZ OPs, but this is a different point as they internally may work like a bootstrapped amplifier.
The auto zero switching can help with flicker noise if done fast / frequent enough. With a JFET input it makes a difference if the switching is with 20 ms or 200 ms integration. However the effect on popcorn type noise is a bit limited.

So far I have not found a major gotcha with the OPA145 or OPA1641  (I have not used the related OPA140).  For the linearity there is an effect of cross over distortion of the output stage (seen with the OPA145), but this should be similar with other OPs, including super beter ones. Biasing the output to class A can avoid this problem.  If needed one could start with a SO8 to DIP adapter in a socket.

I have looked at a discrete FET input and was surprised to see quite some popcorn noise with 2N4391 and J113 FETs. So discrete JFETs are also not without a problem and matched pairs are expensive now. Unmatched pairs are cheap, but need extra work and for manual matching the SMD cases are a slight problem and not sure if the still match well after soldering.

If you like super beta OPs: there is the relative new and affordable OPA202 - not the absolute lowest bias, but low noise.

[David Hess]

I still don't see how the automatic zero function can help with the CMRR. It helps inside AZ OPs, but this is a different point as they internally may work like a bootstrapped amplifier.

When the automatic zero measurement is executed, the common mode voltage is usually ground (zero) but there is no reason this has to be the case.  If the common mode voltage is set to the input voltage to be buffered, then the automatic zero measurement will include the common mode error so that gets corrected as well.  This is what allows an integrated CMOS buffer which otherwise has terrible offset, offset drift, and CMRR to be used to buffer a high resolution integrating converter without limiting performance.  Of course this does nothing to improve flicker noise.

So far I have not found a major gotcha with the OPA145 or OPA1641  (I have not used the related OPA140).  For the linearity there is an effect of cross over distortion of the output stage (seen with the OPA145), but this should be similar with other OPs, including super beter ones. Biasing the output to class A can avoid this problem.  If needed one could start with a SO8 to DIP adapter in a socket.

My short list of parts to consider also includes the OPA191 and OPA192 which use "e-trim", however that is implemented, but they have awfully high noise, albeit better than the originally recommended AD542 which was replaced by the AD711.

I have looked at a discrete FET input and was surprised to see quite some popcorn noise with 2N4391 and J113 FETs. So discrete JFETs are also not without a problem and matched pairs are expensive now. Unmatched pairs are cheap, but need extra work and for manual matching the SMD cases are a slight problem and not sure if the still match well after soldering.

Because of their intended application, I doubt the 2N4391 and J113 are even tested for any noise, and their specifications only give a typical value for high frequency noise.

If you like super beta OPs: there is the relative new and affordable OPA202 - not the absolute lowest bias, but low noise.

Even if I considered the OPA202 competitive with the LT1008/LT1012, my application places the buffer inside the automatic zero loop so speed on the form of settling time has some importance, and the LT1008 supports feedforward compensation.

One thing that is often overlooked when considering super-beta bipolar parts is that in practical circuits they may have lower input bias current because unlike JFET parts, their input bias current does not double every 10C.