Amber 3501A Distortion and noise measuring system

[innkeeper]

I'm repairing an Amber 3501A Distortion analyzer.

I have an operating/service manual off the internet but seems slightly different from mine as my unit has an analog supply instead of an SMP, if anyone has a different manual than the one I have attached, or has any info on this analyzer including failure points and upgrades, especially to the oscillator section to lower the THD of the oscillator,  please post.

I've read someplace else that someone had issues with shorted tantalums, so I want to proactively replace those while I am fixing it even if it's not the cause of my problem. Are there any recommendations on if I should use tantalums again or change them to something else?  They are all decoupling caps in the oscillator/filter section. 10uf 7.5mm pitch and 1.5 at 5mm pitch at 25v. 

I am considering using wima mks 2 pet caps as replacements.

Currently, there is 0v after the +15v lm78L15 regulator on the oscillator, so I might already have a shorted tantalum. I'll post what I find.

[1audio]

I have one of those as well. Mine has the switching supply and uses an external transformer/DC to charge the battery.

I would use good low leakage electrolytics to replace the dead tantalums. Maybe the polymer electrolytics. The mylar caps you referenced don't get big enough usually to replace the Tant's or the package is too big to fit. The ESR on the eletrolytics should be low enough to met the circuit requirements.

That's a very dense and somewhat difficult to work on instrument but has very good performance.

[innkeeper]

I have one of those as well. Mine has the switching supply and uses an external transformer/DC to charge the battery.

I would use good low leakage electrolytics to replace the dead tantalums. Maybe the polymer electrolytics. The mylar caps you referenced don't get big enough usually to replace the Tant's or the package is too big to fit. The ESR on the eletrolytics should be low enough to met the circuit requirements.

That's a very dense and somewhat difficult to work on instrument but has very good performance.

nice that you have one, I don't see much information about them out there nd they seem like they have good specs. I actually like the compact nature of it.
i really am looking forward to using it.

thanks for the insight on the caps. .. I realized they were too big after I posted. What's your opinion on mlcc?

if I can avoid putting tantalums in there again without reducing performance, id like to. I did notice they were selective about where they used the tantalums.

i did dig into it just a few min more, and the lm78L15 is dead, .. I have them on order... no dead shorts of the tantalums but could be leaky and so it begins. I probably have electrolytic on hand enough to get it operational if that's the issue.

[electro-56]

I did design work for Amber back in the 80's, mostly on the Model 5500. I still have Serial #0000, my development unit.

I did some design work on certain parts of the 3501a, though, including the optional frequency counter and battery power supply. There were several variants of the 3501a, often custom built for particular customers who would order reasonable quantities (e.g. 100s) of units. I have a 3501a Navy unit, with the main input and output connectors changed from the standard BNC to banana to fit the requirements of the US navy. The 3501a is indeed tricky to work on, but much better than its predecessor the 3500, which was considerably smaller with similar performance.

Tantalum caps were considered "good" at the time, so we used them where practical, generally in the teardrop packages. The parts we used were fairly low leakage, but since most of them were for supply bypassing, this was not very important. We did see shorted failures, especially if a tech transiently shorted a power rail while working on a unit - the tantalums apparently were damaged by the high peak currents. However, they were poorly specified for ESR, which is meaningful in this application. In my opinion, any decent quality, low-ESR aluminum electrolytic or polymer device would be a superior replacement. None of the integrated regulators were LDO with PNP outputs, so I doubt that there is any lower bound on ESR. Electrolytics were used to bypass TL431 references in places, and one should probably be careful to avoid values that could lead to instability.

It is a long time ago and I no longer have the schematic, but the charging circuit I designed for in the battery powered units did use a PNP output, and was slightly tricky to compensate. I don't know if very low ESR electrolytics could be an issue at its output, but it is possible.

For amusement, look at at the data sheet for the MPQ6842 that is used in various 3501a VCA circuits. This is a clock driver IC that was designed for the 2-phase clocks required on the original MC6800. The main analog guy at Amber discovered that both NPN and PNPs were quite well matched, and used them in linear circuits.

Cheers,
Howard

[innkeeper]

I did design work for Amber back in the 80's, mostly on the Model 5500. I still have Serial #0000, my development unit.

Hi Howard,
That's great info, and glad to meet someone who worked at amber!

Thats pretty cool having unit #0000 of the 5500

I'm defiantly tracing down a shorted tantalum at the moment, did some looking today at it, but I am going to replace them all anyway, so, likely won't bother finding which one it is.... I'll have to wait till next week to get caps in because new years ... gives me a few days to make my cap choices.

my unit is pretty basic, with no balanced inputs/outputs, no battery no imd board, and an analog supply for which I don't have a schematic, thought did find a manual on the internet that included the schematic for everything else... it's an analog supply so, it is nothing that's not very apparent by looking at it.

on the tantalums, i did notice they only used them around components where it might help to have the lowest noise on the power rails.  I was assuming that the reason they went with tantalum bypass was to eek out a little less noise from the leakage current over the available electrolytics of the time.  With that in mind, I am leaning toward film, but also been looking into mlcc caps too.  though the only downside on those multilayer ceramic caps  I can see would be microphonics, though, I'm thinking that's likely not an issue in this situation.  I'm curious what the opinions of using mlcc's as alternatives to the tantalum is.

It is funny you mentioned the MPQ6842, I looked them up just because I was curious what they were, and didn't dig too deep at the time, but remember thinking wow they cheaped out and made it denser at the same time, but, didn't get that the whole reason was to get matched pairs cheap and id bet they thermal track well too .... that's brilliant. I'll have to re-look now with a bit more appreciation for what's going on there.

my example appears to be circa 1993 from the ic date codes and is labeled amber NEUTRIK INSTRUMENTATION INC on the back.

[electro-56]

Hi innkeeper,

I stopped doing design for Amber in 1987, but stayed friendly with them. They went through some changes: first they were purchased by Coreco, and eventually by Neutrik Instrumentation. I ended up designing an automated line card test system for Nortel with Neutrik in 2000, but a couple of years later they wrapped up their North American operation and abandoned the Amber brand. They are still active in Lichtenstein under the name NTI.

I was able to dig up some documentation for the 3501a, including a schematic of the line-powered supply. I have tried to attach it. This is the most modern one I have, and is likely correct for your unit. There were *many* revisions to some of the boards, but I doubt they made big changes to this circuit. The actual manual is a bit thick, but I would be happy to scan any pages you need.

I agree with 1audio that modern, high-quality aluminum electrolytics should be fine. I am not sure about the requirement for low-leakage, though. Does leakage across the dielectric really cause a lot of noise? In any case, low ESR is bound to be a good thing here.

Now that I have been browsing through the docs, I am no longer sure what the differences were with the Navy units. The manual is a bit confusing about the various options, but suggests that all of the 3501a variants were line-powered only, and had balanced inputs. The battery option on 1audio's unit was apparently only available on the 3501 (no a).

I will verify that the attachment got through once I post this

[innkeeper]

Hi Howard,

That looks like the right schematic, thanks!

The manual that is on the internet and making the rounds for this is issue 9 dated 1988 http://bama.edebris.com/download/amber/3501/3501.pdf

It would be great if you could scan a different version of the manual and schematics for everyone. If you do it you probably can't attach it here so send me a private message and we can figure that out, id be happy to post it in all the usual places people get free manuals.

Thanks for the input on the cap choice.
I also meant to thank you for the heads up on the tl-431's - I probably would not have noticed that and as our Ozzy friend says i may have come a gutser there.

is leakage noise an issue...it can be but is it in this case?  I have no clue.  Though it is a convenient additional characteristic for where the tantalums are placed.

here is what little I know on cap noise ...  this is something I ran into once myself when doing some work on a very sensitive low-level low-frequency instrumentation amplifier. The leakage current causes noise. the noise playes a factor there because it's not common-mode noise and therefore can't be as easily rejected by the opamps as common-mode noise.  Information on capacitor noise is not as easy to find as other information. mostly in research papers. but some cap data sheets do include info on it. basically, the less leakage, the less noise.

I think I have given up on the film cap idea, I can't quite find them small enough that they will fit in the space available, though really close. the one other thing I am checking into before I go low ESR electrolytic is MLCC caps.

In the recent past, there was a crisis on tantalum and people were scrambling for alternatives to tantalum caps. one of those that were turned to was mlcc caps. but as you probably suspect the solutions were as varied as the use cases including low ESR electrlytics..

what I find interesting is mlcc offers about 10x less leakage than the traditional tantalums and about equal to modern high-reliability tantalums though have inverse temperature characteristics and low ESR. and that they age by losing some capacitance, but, not ridiculously bad, not enough to be of concern I think. id not put them in a high-vibration environment though do to the piezo effects. they will likely outlive any electrolytics and tantalums.

I have to say,  the engineers used some expensive brands on the boards.  I did notice many of the parts show up on mil-spec sheets, so id guess that was a factor, and as you pointed out they sold to at least the navy.  They didn't skimp on the quality of components but they also didn't over-spec stuff either.  In example like they used cheaper nosier TL opamps in places where the more expensive quieter option would not have made a difference. They split out separate power regulators in the oscillator area to isolate it, they could have cheaped out there but went with better performance.  So the design was defiantly done with some care for price/performance by the engineering staff that kept the performance characteristics.

some mental floss:


eevblog discussion - https://www.eevblog.com/forum/beginners/does-capacitors-cause-noise/
experiment - https://courses.physics.illinois.edu/phys406/sp2017/Student_Projects/Spring13/Anna_Czerepak_P406_Final_Project_Report_Sp13.pdf
paper on noise in tantalums - https://core.ac.uk/download/pdf/189570134.pdf
writeup on tantalum vs mlcc - https://passive-components.eu/the-basics-benefits-of-tantalum-ceramic-capacitors/
jim williams  application note 124 - some nuggates about low leage caps including an esamle of using a 400 dolalr wet tantalum- https://www.analog.com/media/en/technical-documentation/application-notes/an124f.pdf
math .. yea over my head -

[SoundTech-LG]

Long time 5500 owner here. Time to power it up again and see if everything is ok. Seems there was some minor issue last time. Maybe it was the meter. I will check my s/n as well. I also have the manual that includes schematics. Quite nice. Something you never see much of anymore.

[innkeeper]

Long time 5500 owner here. Time to power it up again and see if everything is ok. Seems there was some minor issue last time. Maybe it was the meter. I will check my s/n as well. I also have the manual that includes schematics. Quite nice. Something you never see much of anymore.

That's one I've not seen the manual available online as i actually went looking for it yesterday out of curiosity. if you could scan that too that be nice to get it online for everyone.

[electro-56]

I can see how the real components of impedance in a capacitor could be a source of noise, but am inclined to suspect that it is insignificant in these circuits. The R term in parallel with the C is going to be a pretty high value, and the noise will be heavily shunted in the audio range. The series R will be pretty low, and although I guess it could cause some trouble at high ripple currents, I would expect the ripple to be low after the regulators. I would speculate that both these effects would be drowned by the noise from the 3-terminal regulators here. I have not investigated this myself, but will keep in mind your experience and research should I ever face this in the future.

I do remember that tantalum or MLCC caps were measurably poor for coupling audio, but the whole business of selecting caps for audio descended into madness at one point. Cyril Bateman dug pretty deeply into the audio performance of various capacitor types, both with objective measurements and subjective evaluation. If I recollect, he found that PPS  and PP dielectric caps were generally very good for coupling.

I personally doubt that there will be any significant difference in performance of the 3501 circuits whether you use good Al or Ta caps for PS bypassing. There might be cases where adding an MLCC across an electrolytic could improve the high frequency performance. For example, there are some places in the designs where Ed isolated the the supply pins for an opamp with 100 ohms and bypassed them with 10uF electrolytics. If I have to do that to improve PSRR, I always add an MLCC right at the pin.

I now mostly use X7R MLCC parts for PS bypassing in my designs, radically derated to take into account the nasty voltage coefficient. It turns out that the voltage coefficient is worse for the hi-K dielectrics, even within a temperature sensitivity class (e.g. X7R), so inversely proportional to size. Smaller is good for bypassing very fast circuits where you want low ESL, like the pins of a BGA FPGA, but often gains nothing in terms of usable capacitance at bias. I also like to use polymer Al parts for bulk storage, but some stuff I work on ends up on the ISS and NASA is always uptight about Al electrolytics. So Ta parts are still on my list.

The BAMA link is for a 3501 and does not describe the 3501a. However, it is very useful to compare the models, and gives me a chance to look at the design of the battery charger. It seems that I remembered the topology incorrectly, and it does not use PNP output stages. I remember having trouble compensating that, and must have simply switched to an NPN Darlington. It is odd to look at stuff I did so long ago.

I will remind you that the TRS jack(s) on the Amber equipment (3501 & 5500) are WE310/PJ051 type that can be damaged by inserting a 0.25" stereo plug. Pomona 2112 adapts Tip and Ring to banana jacks.

I have the October 1991 Issue 10 version of the 3501a manual - I will put scanning it in my queue. Probably best to publish it through BAMA as well. NTI technically still holds the copyright, but I seriously doubt that they will come after me for sharing it (grin). FWIW, my Issue 8 5500 manual from 1988 is about 2" thick with most pages double-sided and has section separators. This won't be fun to copy, but I will think about it.

All of the 35xx manuals were written by Wayne Jones, the founder and president of the company. I am still impressed by them. My partner Mary and I did a bunch of work on the 5500 manual, but Wayne brought it all together.

The core design team at Amber was very small. For many years Ed Meitner (emmlabs.com) did the low-distortion analog design, I did power, digital, and measurement circuitry design as well as system firmware and I shared systems design with Wayne, while Mary did applications firmware and math stuff. Mike Chang took over the analog work when Ed left, Santo Spinali did production troubleshooting and QA, Vince Goboyan did all of the drafting and PCB layouts, and there were other people involved in production assembly. Wayne took care of everything from marketing to sales to production to purchasing to the design of a database for component stock management. His partner Pat kept stuff organized. The staff expanded after I left, but there was never a large development team. Sorry for spelunking into ancient history, but it was a very formative time in my career.

@SoundTech-LG:
I started having some intermittent troubles with my 5500 about 8 years ago where THD measurements sometimes failed to null in the 1k-10k range. I looked into it quickly but could not see what was wrong. I did actually want to use it a few years ago so made a determined effort to figure out what was going on, and discovered tons of ripple on the -12V rail. The power supply is not fun to work on, but I quickly realized that C8003 was bulging. I was able to get the board out far enough without disconnecting too much to replace C8003 and C8004 with Panasonic ECO-S1VP153A parts available from Digi-Key. Way more capacitance and higher voltage than required, but they fit and the problems disappeared. The main filter caps on all supplies are under bias continuously when the unit is is plugged in, so it is actually surprising that there were no earlier failures after being powered on my bench for the better part of 35 years... The unit is now mostly ok except for problems with some of the input select relays that are a real nightmare to get at. This means that I cannot use certain combinations of input connections, but otherwise it is fine. My unit got many upgrades along the way, although the power supply and backplane are original. I suspect that retrofitting the input select relay assemblies involved some brutal hacking on my system - I was lucky that the Amber techs took care of it for me.

Out of curiosity, do you have the notch filter and oscillator with analog or digital tuning? The original design was analog because Ed was really into VCAs. Unfortunately, the thermal performance was not terrific and the frequency drift very noticeable on the counter. After Ed left, Wayne and Mike came up with a DAC-based design that was much more frequency stable.

[innkeeper]

I can see how the real components of impedance in a capacitor could be a source of noise, but am inclined to suspect that it is insignificant in these circuits. The R term in parallel with the C is going to be a pretty high value, and the noise will be heavily shunted in the audio range. The series R will be pretty low, and although I guess it could cause some trouble at high ripple currents, I would expect the ripple to be low after the regulators. I would speculate that both these effects would be drowned by the noise from the 3-terminal regulators here. I have not investigated this myself, but will keep in mind your experience and research should I ever face this in the future.

I've definitely fought that terminal regulator noise beast before, I had repaired a DBX286A and then used as a testbed to play with my DSA and found that all lm7815's were not created equal, find that adventure here https://www.gearslutz.com/board/geekslutz-forum/1168297-dbx-286a-modifications.html

I personally doubt that there will be any significant difference in performance of the 3501 circuits whether you use good Al or Ta caps for PS bypassing. There might be cases where adding an MLCC across an electrolytic could improve the high frequency performance. For example, there are some places in the designs where Ed isolated the the supply pins for an opamp with 100 ohms and bypassed them with 10uF electrolytics. If I have to do that to improve PSRR, I always add an MLCC right at the pin.

oh wow, I had a look at the 100-ohm resistors used for reducing PSRR after you mentioned it.  using MLCC's in that instance makes a lot of sense.

I now mostly use X7R MLCC parts for PS bypassing in my designs, radically derated to take into account the nasty voltage coefficient. It turns out that the voltage coefficient is worse for the hi-K dielectrics, even within a temperature sensitivity class (e.g. X7R), so inversely proportional to size. Smaller is good for bypassing very fast circuits where you want low ESL, like the pins of a BGA FPGA, but often gains nothing in terms of usable capacitance at bias. I also like to use polymer Al parts for bulk storage, but some stuff I work on ends up on the ISS and NASA is always uptight about Al electrolytics. So Ta parts are still on my list.

This is the first time I really digging into MLCC parts, and discovered the bias voltage/capacitance curve and was taken back, I went through the process of seeing if i could replace the tantalums with them .. derating and still being able to fit them on the board. i can fit 1.5uf 100v in place of the 1.5uf 25v tantalums in most cases, though have to up the capacitance for at least one tl431 to keep it stable or might just bump them all to 2.2uf.  The best I could do was 10uf at 50v on the 10uf 25v tantalum, which works out to 8uf at 15V bias.   

The BAMA link is for a 3501 and does not describe the 3501a. However, it is very useful to compare the models, and gives me a chance to look at the design of the battery charger. It seems that I remembered the topology incorrectly, and it does not use PNP output stages. I remember having trouble compensating that, and must have simply switched to an NPN Darlington. It is odd to look at stuff I did so long ago.

Ah, that's what I have been using as a reference for my 3501A so far the schematic seems to match the boards, except for the power supply. But then again mine has no options.

its nice learning the background on all this who was on the team and who did what, 

[SoundTech-LG]

Well, I have to apologize, I did not even get the model # correct. Mine is a 4400, the multipurpose audio test set.
I won't be scanning that manual either, it's huge, and printed on both sides as well.

[innkeeper]

i found the postings on the 3501a electrlytic replacment of the tantalum and the difference in the distortion

https://www.diyaudio.com/forums/equipment-and-tools/309233-amber-3501a-dist-audio-analyzer-refurb-cal.html

posts #19 and #21

[electro-56]

The 4400 was a pretty cool design for its day, but it preceded my involvement with Amber and I did not participate in its design. The company did give me one as part of my compensation, though. Years later they desperately needed it for a good customer, so I agreed to return it. In exchange, they transferred ownership of my development 5500.

Howard

[electro-56]

@innkeeper

That is a very interesting discussion on diyaudio - some of these people have dug pretty deeply into the 3501 design. In post #14, Chamberman discusses problems he had with the tuning pot stack, and I suspect that is going to be hard to get around. Those pots were a custom order, and I remember that they were painfully expensive. As they wear, it is likely that the tracking will be imperfect, which will upset the notch filter and oscillator. Not sure how to get around that.

I have often wondered if there were modern drop-in replacements for the 5534s that would provide lower THD. It sounds like the LME49710 might have been a candidate, but they are EOL. Oh well, I don't actually do much analog audio these days, so not inclined to jump down that rabbit hole.

[precaud]

In post #14, Chamberman discusses problems he had with the tuning pot stack, and I suspect that is going to be hard to get around. Those pots were a custom order, and I remember that they were painfully expensive. As they wear, it is likely that the tracking will be imperfect, which will upset the notch filter and oscillator. Not sure how to get around that.

That is why I sold mine. The only way to get its full THD range was to measure at decade intervals with the stacked pot wide open.

I have often wondered if there were modern drop-in replacements for the 5534s that would provide lower THD. It sounds like the LME49710 might have been a candidate, but they are EOL. Oh well, I don't actually do much analog audio these days, so not inclined to jump down that rabbit hole.

In the 5500, Amber used the MA-332 in critical places (like the null circuit, I think), that would give another 4-5dB IIRC. I have a set of 5500 boards here, retrieved from a unit damaged in transit.

[electro-56]

In post #14, Chamberman discusses problems he had with the tuning pot stack, and I suspect that is going to be hard to get around. Those pots were a custom order, and I remember that they were painfully expensive. As they wear, it is likely that the tracking will be imperfect, which will upset the notch filter and oscillator. Not sure how to get around that.

That is why I sold mine. The only way to get its full THD range was to measure at decade intervals with the stacked pot wide open.

Chamberman suggests replacing the pot with switched resistors. That would probably address tracking and contact resistance issues, but requires a lot of effort

I have often wondered if there were modern drop-in replacements for the 5534s that would provide lower THD. It sounds like the LME49710 might have been a candidate, but they are EOL. Oh well, I don't actually do much analog audio these days, so not inclined to jump down that rabbit hole.

In the 5500, Amber used the MA-332 in critical places (like the null circuit, I think), that would give another 4-5dB IIRC. I have a set of 5500 boards here, retrieved from a unit damaged in transit.

Even though the MA-332 is very old ('81?), I don't think that Amber was using them in production until after I left in '87. I might have this wrong: I was not directly involved in the design of the low-distortion signal path. You are right, though, they did get used at critical places in the new DAC-based filter and oscillator for the 5500. My unit got fitted with the newer boards when I brought it in for cal in '91. 

[innkeeper]

so I've decided to try mlcc's

so the rough plan is to replace all the tantalums with X7R MLCC's 50V at the same value - considering the lower ESR and ESL, i think that will be ok..works out to 1.38uf for the 1.5'uf caps and around 8uf for the 10uf caps at 15v bias. all the documentation shows you can get away with much lower capacitance values for the same / better filtering, but mostly that's in digital situations and smps.

once i see it is alive then I will replace the electrolytics with 10000h low esr caps where possible.  Anyplace were there using 100-ohm resistors with electrolytic decoupling caps for PSRR reduction i will change them to MLCC's.   i'll leave the ceramics film and mica caps alone and then go through the cal procedure.

only 1 value I am questioning one cap value and that's the c2059 which is using a 4.7uf cap on a tl431, and as best I can tell that still be in the unstable range, so I might try bumping that to 10uf.

i also noticed that u2004 which is a ca3280 a has 100 ohm resistors on the 15v lines but, has no caps..odd...u2021 which is also a ca3280 also has the 100 ohm resistors but with decoupling caps. might just be a schematic mistake, if not there easy enough to add.

hopefully, when done, it will be a long time before anyone needs to touch the caps again.

oh i found something interesting, not something i had ever seen before, that is an elapsed time indicator on the power supply that's not mechanical. It's a glass unit that looks a lot like a fuse or a thermometer.  I did a little research and i think this might have been a mercury filled units.  it counted up to 5000 hours. looks like it's likely gone past that ... pictures attached.  the company that makes them seems to still be in buuisness, i reached out to them to see if they had any information on it they could pass along.  I don't believe there still made.

@electro-56,  do you know why they had those? maybe to keep track of when re cals were needed??  or maybe some lifespan of something?

there's a video on how they work here:


pics:

[precaud]

Chamberman suggests replacing the pot with switched resistors. That would probably address tracking and contact resistance issues, but requires a lot of effort

On the other hand, how often does one *need* THD at some unusual frequency? I probably should have just lived with measuring THD at 100, 1k, 10k, and 100k. For most audio electronics that is sufficient to characterise it. I was probably short-sighted to sell it...

Even though the MA-332 is very old ('81?), I don't think that Amber was using them in production until after I left in '87. I might have this wrong: I was not directly involved in the design of the low-distortion signal path. You are right, though, they did get used at critical places in the new DAC-based filter and oscillator for the 5500. My unit got fitted with the newer boards when I brought it in for cal in '91.

Nice that you were involved with the design. I'm pretty sure these boards are from an older unit. I bought seven 5500's from Bell Canada in the mid-2000's. This one got damaged, and was the oldest of the group. The other six I was able to fix to meet spec, but not this one, it wouldn't track and null for THD. It did everything else well, though.

[electro-56]

so I've decided to try mlcc's

so the rough plan is to replace all the tantalums with X7R MLCC's 50V at the same value - considering the lower ESR and ESL, i think that will be ok..works out to 1.38uf for the 1.5'uf caps and around 8uf for the 10uf caps at 15v bias. all the documentation shows you can get away with much lower capacitance values for the same / better filtering, but mostly that's in digital situations and smps.

once i see it is alive then I will replace the electrolytics with 10000h low esr caps where possible.  Anyplace were there using 100-ohm resistors with electrolytic decoupling caps for PSRR reduction i will change them to MLCC's.   i'll leave the ceramics film and mica caps alone and then go through the cal procedure.

Makes sense to me.

only 1 value I am questioning one cap value and that's the c2059 which is using a 4.7uf cap on a tl431, and as best I can tell that still be in the unstable range, so I might try bumping that to 10uf.

The TL431 datasheet suggests that there is no upper bound on the bypass cap

i also noticed that u2004 which is a ca3280 a has 100 ohm resistors on the 15v lines but, has no caps..odd...u2021 which is also a ca3280 also has the 100 ohm resistors but with decoupling caps. might just be a schematic mistake, if not there easy enough to add.

hopefully, when done, it will be a long time before anyone needs to touch the caps again.

All the original schematics were hand sketched, and sometimes the layouts were done from them. The schematics for the manuals often came later, so I would hardly be surprised if there were errors either in the implementation or documentation. I would certainly make sure that all OTAs and opamps are bypassed if they have a decoupling resistor - please let the thread know if you find any that are not.

oh i found something interesting, not something i had ever seen before, that is an elapsed time indicator on the power supply that's not mechanical. It's a glass unit that looks a lot like a fuse or a thermometer.  I did a little research and i think this might have been a mercury filled units.  it counted up to 5000 hours. looks like it's likely gone past that ... pictures attached.  the company that makes them seems to still be in buuisness, i reached out to them to see if they had any information on it they could pass along.  I don't believe there still made.

@electro-56,  do you know why they had those? maybe to keep track of when re cals were needed??  or maybe some lifespan of something?

Some government contract required a means of tracking power-on hours, but I don't remember which one. But yes, it is a Fredericks device, although I don't immediately see it in any of the schematics I have here. I cannot say for sure that this is true of the ones in the 3501, but the attached datasheet from ~2001 states that they are Hg-free. Note that all models use identical capsules but with different markings, and are scaled using external resistors. I have a vague recollection that Wayne asked me to scale the one in the 3501 for 50kh - the actual series resistor and excitation voltage will prove it:

[electro-56]

I just found a schematic showing the time recorder: it does seem to be scaled for 5kh:

[electro-56]

Sorry, the time recorder is in both versions of the schematic I provided - it has been a long time and I just missed seeing it before. However, I am not overjoyed about seeing those 22uF caps on the 317 and 337 reference pins. Might have helped with noise, but seems bound likely to hurt load transient response. 

[innkeeper]

Some government contract required a means of tracking power-on hours, but I don't remember which one. But yes, it is a Fredericks device, although I don't immediately see it in any of the schematics I have here. I cannot say for sure that this is true of the ones in the 3501, but the attached datasheet from ~2001 states that they are Hg-free. Note that all models use identical capsules but with different markings, and are scaled using external resistors. I have a vague recollection that Wayne asked me to scale the one in the 3501 for 50kh - the actual series resistor and excitation voltage will prove it:

Ah good to know it is Hg free, I was worried that the mercury leaked out.  The schematic you posted of the analog supply for the 3501a does have it in the schematic. it looks like mine is a 5000h one and the resistors match the schematic. .. doing the math looks like 14ua which matches the current for 5000h.   I can see where just making it 1.4ua would make one 50k hours, pretty cool how that works.

[innkeeper]

Sorry, the time recorder is in both versions of the schematic I provided - it has been a long time and I just missed seeing it before. However, I am not overjoyed about seeing those 22uF caps on the 317 and 337 reference pins. Might have helped with noise, but seems bound likely to hurt load transient response.

I've seen this before with that exact same 22uf value at 15V output.  it's for ripple rejection. ..Application Notes says 10uf at 10v output gives 15db of ripple rejection at 120hz.

it is a design tradeoff I suppose, not expecting transients when doing critical measurements.

[electro-56]

Right you are about the cap - I have not looked at that data sheet since dirt was new. I don't use the 3x7 parts much anymore, and now am more often concerned with loads interacting.
Thanks for the catch