HP 3478A Troubleshooting

[Aphelion]

I've been trying to fix up a used HP 3478A that I picked up locally here in the philippines. Self test check out. So far, I have replaced the RIFA caps and proceeded to replace all of the electrolytics (main 3300uf filter cap is still in shipping)

The problem is that all of the ranges are jumpy with a dead short ie. At the 30mV range it jumps in between a range of values at +-00.0050 from cold start but sometimes after warmup and use it settles at +-00.0010 and returns to -+00.0010mV. Amount of noise is consistent on all dc ranges(for 30v and up the least significant digit is the only one that does not settle)

I've checked the 10V line at JM102 and it checks out. Unfortunately for the checking of the U101 opamp I can't check for noise for sure as I only have a UT61e to check for noise beyond 0.001v. Also, the manual didn't state how much noise I should look out for. 

So instead of getting a signal probe and sig analyser, I am going to try my luck and source U101, U765, U766, U467 and U550. Hopefully after replacing these I can remove the noisy readings of my meter.

U101, from searching the net is thought to be an AD544 BiFET, I got a AD711 which seems to be the closest I can get to the original spec of the AD544.

Problem is, I can't find any info on the rest of the ICs that the A/D Controller troubleshooting section suggest replacing before replacing the main IC. I can probably guess it is also like the input hybrid which is unobtanium in on itself (been searching for a month now with no luck)

I wanted to get another 3478A for parts, but the meter itself is almost unobtanium here in the Philippines by itself. Most US ebay sellers won't even ship to here.

And lastly, should I just bite the bullet and get a new multimeter instead of trying to revive this meter?

[bdunham7]

And lastly, should I just bite the bullet and get a new multimeter instead of trying to revive this meter?

I definitely think you should not try to 'fix' the meter by tossing in parts.  Given your description I'm not at all sure that anything is wrong with it.

I have an HP 3478A here with me.  I consider this meter to be in good working order.  I just plugged it in and put a 2-wire short (a Pomona dual-banana short, no wires at all) and I get 00.0047, which is slowly decreasing.  Going uprange gives me similar numbers--000.05 mV, 0.00001 V, 00.0000V.  That is 4.7uV of offset cold, which doesn't seem excessive to me.  I'll see how noisy it is after while.  During warmup you can have competing temp coefficients that will cause the offset number to wander around before finally settling somewhere.

So what you should do is make sure you have a very short short circuit on the inputs (no wires if possible) and that your room temperature is within 18-28C.  Then let the meter warm up for 90 minutes, check each range for reading and noise/instability and then compare that to the specifications.  If you still think the input is noisy--and you are absolutely sure that your short is perfect--then turn Auto Zero off (blue shift  key, INT/TRIG) and see how much it bounces around with AZ off.

[Kleinstein]

For comparing the noise / drift one should have it properly wamred up and protected from large variations in the room temperature. With a meter not used much for quite some time (which is normal for used meters) there can be some initial drift from humidity getting a new lower equilibrium. So it can initially take more than the normal warm up.

For the noise it depends which speed is used.  AFAIK it uses averaging over multiple conversions to get the full resolution. So don't expect it to be low noise at the higher speeds.  It seems to be reasonable working and no real need to mess around with the ADC. Some parts are special HP specific hybrids.
The ADC may need some of the noise so that oversampling works - it may get worse with to little noise.

[bdunham7]

So after an hour or so mine is reading between +00.0010 and +00.0020, bouncing around a bit.  Auto Zero on or off makes little difference, so this is simply a microvolt or so of input offset and noise.  The specification is +/- 4.0uV or 40 counts at zero, so well within spec.  My room temp is ~23C with reasonably low humidity and the meter has been in the room for weeks now with the temperature stable to +/- 5C.  Temperature variations, humidity, the quality of your short, etc may make some difference.

[Aphelion]

Here's a short clip of the meter with shorted inputs in the mV range. It has been sitting here turned on for more than 2 hours now. My shorting plug is no pomona plugs but I think this shouldn't be normal?.

https://youtu.be/uLai8bUrttI

[Kleinstein]

This looks a bit noisy indeed.
Is the current reading also noise ? internally it should use a similar setting of the input amplifier.

I doubt it would be the U101 to go bad with increasing noise. That would be a rather rare failure mode.

Candidates would be more like a wrong setting (e.g. 50/60Hz mains), supply ripple from a bad cap or a bad solder joint, poor relay / switch contacts and maybe resistor in the input path.
It sometimes help to excercise the front / rear switch a few times (e.g. 10 x) and see if this chances things. Also test if a short at the rear may work different.

If possible get some PC interface ( GPIB or RS232 if the 3478 supports) and record data to show as a graph. That is much easier to judge than a video.
Having a PC link is anyway a main incentive to get a bench meter.

[Aphelion]

I'll check if there are any more bad caps that I can replace. Measuring voltage between the main 3300uf cap has unstable voltage. That is so far my primary suspects along with U101.

Solder joints are fine from the last time I checked, and the mains switch is set to 60hz, which is correct for my country.

No idea how I can check the reed relays though, first time I saw these kinds of relays.

All of the carbon resistors along the input path that is 54k is 58k when I measured it, and R107 which is a 1 ohm 1% resistor is 0.68. might be a candidate for replacement. Can these affect the DCV ranges?

Also already tried using the outputs at the back and cleaning the input terminals, still no luck. All of the modes are noisy, even ohms and DCA. ACA seems to be fine though(000.001+- mAAC noise)

Gonna try to build the arduino GPIB interface if I can so that I can datalog it. I really hope it is not the input hybrid or ADC ☹️

[Greybeard]

No idea how I can check the reed relays though, first time I saw these kinds of relays.

You can try to knock on the relays by plastic or wood sticks, handle of screwdriver etc. while watching for jumping display.
Measure voltage drop over closed relay contacts.

[AVGresponding]

No idea how I can check the reed relays though, first time I saw these kinds of relays.

You can try to knock on the relays by plastic or wood sticks, handle of screwdriver etc. while watching for jumping display.
Measure voltage drop over closed relay contacts.

I think I'd rather use a magnet to test them than hit them with anything; that could easily damage them or the solder joints.

@OP, is there a SMPS or LED light within a couple of feet half a metre of your meter?

[Aphelion]

Here is a clip of the meter after 5 hours of settling time and turning off everything else in the room that I can see.
https://youtu.be/8NUr_PDUXHs

Bringing a hard drive magnet near the rightmost relay facing the front of the meter adds about 0.1 mv to the reading

Edit: forgot to mention that the reported noise is about the same even if I turn off all sources of EMF in the room as shown in the video.

[bdunham7]

Here's a short clip of the meter with shorted inputs in the mV range. It has been sitting here turned on for more than 2 hours now. My shorting plug is no pomona plugs but I think this shouldn't be normal?.

https://youtu.be/uLai8bUrttI

OK, that's an order of magnitude, or  decimal place, higher than what I first understood.

Turn off CAL ENABLE (turn the small slotted screw 90 degrees clockwise) and reset or power cycle the meter. 

I can't judge your shorting plug from here, but if it makes good contact internally it should be fine.

Press the AC volts button, give it  a minute and see what it says and if there is a residual reading, how steady it is.

[Aphelion]

At AC mode the reading is pretty stable, with only a deviation of -+0.005mVAC. It may be a bit larger now as the case is open to check the relays with a magnet which AVGresponding suggested to do

[bdunham7]

At AC mode the reading is pretty stable, with only a deviation of -+0.005mVAC. It may be a bit larger now as the case is open to check the relays with a magnet which AVGresponding suggested to do

OK, so you can look at the schematic and get some idea what that means as far as what input circuitry is bypassed in AC. 

The DC input system uses a high impedance input amp U101, with 3 gain settings for the 30mVDC, 300mVDC and 3VDC ranges and then a 100X divider for the 30 and 300VDC ranges.  So the U101 gain settings are the same (medium) for the 300mVDC and 30VDC ranges, and then the 3VDC and 300VDC ranges share the lowest gain setting.  Only the 30mVDC range gets the highest gain.

You might be able to determine whether the noise is before, from or after U101 by comparing the ranges with the same gain setting.  So look at the 300mVDC range and see what the noise looks like, then go to the 30VDC range and compare it.

[Aphelion]

Ok, after poking a while in the meter, only off things I can say is that all of the 51k resistors reads 56-58k, the capacitors C102 and C101 reads at around 480-ish pF. Also, R110 reads 1.2k. Might get replacements for those carbon resistors.

300mV range has noise at the range of -+000.010mV, and 30v range has about the same amount at -+00.0010V.

[Kleinstein]

The 50 K resistors are there only for protection. So the actual valeue does not matter. An important point is that they can withstand high pulse voltage - here modern thin film resistors are often worse than carbon.

The divider is 1:100. So 0.01 mV and with the divider 1 mV noise corresond to the same level of noise.  The difference is in the input path before the amplifier. So either both paths are bad, or more likely both are OK, or at least kind of OK.  This kind of excludes the 50 K resistors at the input.
It could be still the switching part for the auto zero mode.  So one may be able to compare the non AZ and AZ mode (I suppose the meter has this option not sure though).

It is still hard to tell if the noise is more from before or after the amplifier. For this one would have to compare 2 different gain settings, like the 3 V range and the 300 mV or 30 mV range.  Noise from before the amplier would stay the same. Noise from after the amplifier would be scaled with the gain.

[Aphelion]

The divider is 1:100. So 0.01 mV and with the divider 1 mV noise corresond to the same level of noise.  The difference is in the input path before the amplifier. So either both paths are bad, or more likely both are OK, or at least kind of OK.  This kind of excludes the 50 K resistors at the input.
It could be still the switching part for the auto zero mode.  So one may be able to compare the non AZ and AZ mode (I suppose the meter has this option not sure though).

It is still hard to tell if the noise is more from before or after the amplifier. For this one would have to compare 2 different gain settings, like the 3 V range and the 300 mV or 30 mV range.  Noise from before the amplier would stay the same. Noise from after the amplifier would be scaled with the gain.

Thank you for explaining each part to me clearly.  Kind of learning as I go while I troubleshoot the meter so it already paid for itself even if it turns out to be a dud 😂.

Already probably probed everything I can test before U101 that can affect it except the input hybrid itself. Haven't tried measuring the voltage drop in the relays though, gonna have a crack at that later.

Also can't check whether or not the ADC is the one at fault, reason why I was going for scorched earth tactics and replace that might be replaced easily. Don't have a logic probe/signature analyzer/pulser etc.

bdunham7 already suggested me to check in between 30v and 300mv ranges, and it scales( 00.0005V-0.0010V noise on 30v range, and 000.005V-000.01V for the 300mV range). So if I am correct this points to the ADC section that has the problem?

[Kleinstein]

The comparison between the 30 V and 300 mV range does not tell if the problem is more before the ADC or more at the ADC. bdunham7 was wrong on that, though the general idea is good:  The main difference between the 300 mV and 30  range is only the input that. So it kind of checks parth of the input hybrid and they seem to be similar.

To see if the problem is more with the amplifier (or before ) or the ADC one would need to compare different gain settings, like the 3 V and 300 mV range.

A logic analyzer would likely not be needed. The instrument of choice may be a scope.
Even with limited test instruments one can still get a good hint on where to search for the noise and not swap part withut a good reason.

The hum measured at the supply could be a hint for example.  If lucky, just the still missing caps will fix things.

[Aphelion]

Replaced the main 3300uf electrolytic, still no dice ☹️

[bdunham7]

bdunham7 already suggested me to check in between 30v and 300mv ranges, and it scales( 00.0005V-0.0010V noise on 30v range, and 000.005V-000.01V for the 300mV range). So if I am correct this points to the ADC section that has the problem?

No, I think that test points away from the ADC being the problem, at least when combined with the results from your other ranges.  Just to be clear, you've corrected the CAL ENABLE issue and you are observing those number as noise, not a steady offset?  Can you compare the 3VDC and 300VDC ranges?  I know those noise levels were low, but it still might be handy to see them. 

The comparison between the 30 V and 300 mV range does not tell if the problem is more before the ADC or more at the ADC. bdunham7 was wrong on that, though the general idea is good:  The main difference between the 300 mV and 30  range is only the input that. So it kind of checks parth of the input hybrid and they seem to be similar.

Me wrong?     Actually I didn't say that, I was suggesting using the comparison of two scaled ranges to see where the difference was relative to U101.  Comparing them all might help too, but I wanted to start somewhere and Aphelion indicated that the noise was very small on the highest range, while the lowest range/greatest gain at 30mVDC doesn't have an equivalent with the 100X divider switched in.

So now we know that it has ~10 counts of noise in the 33.3X (medium) gain setting regardless of the input divider setting, and the noise apparently scales up in the higher gain setting and down in the lower gain setting--the former is on video and the latter I've just asked Aphelion to check again.  Since all of the ranges use the gain of the input to keep the FS of the ADC to be in a -10VDC to 10VDC range, there is no additional scaling or digital magnification which would cause the ADC to read differently from one range to the next--the displayed counts are the (ADC output * calibration correction * a constant) with the decimal point set separately.  So to me, the ADC doesn't seem to be the issue.  That and the apparent absence of noise on the AC input range.

So it looks like the noise is present on the inputs of, or internally to, U101.  Since the input paths are mostly separate up to that point, and since the inputs to U101 are effectively shunted with ~100K in either case, it appears to me that the issue will be U101 or the closely associated parts of U102. 

[bdunham7]

Replaced the main 3300uf electrolytic, still no dice ☹️

I just had a thought--is your power supply frequency switch set correctly?  On the back, underneath the EXT TRIG connector, there are a series of switches for the HPIB address setting, but the first one is up for 50Hz, down for 60Hz. 

[Aphelion]

Replaced the main 3300uf electrolytic, still no dice ☹️

I just had a thought--is your power supply frequency switch set correctly?  On the back, underneath the EXT TRIG connector, there are a series of switches for the HPIB address setting, but the first one is up for 50Hz, down for 60Hz.

Yep, made sure of that as that was the first thing that was on the manual. Even tried toggling the DIP switches to account for dirty contacts.

Strange thing is that I hear some buzzing inside the case that I can't seem to pinpoint. Checked all of the caps, even near the input hybrid and there seemed to be nothing out of spec (all reading 460-480 pf). Maybe it is just the transformer humming.

And yep, all testing done was with the CAL ENABLE switch in horizontal position.

[Aphelion]

bdunham7 already suggested me to check in between 30v and 300mv ranges, and it scales( 00.0005V-0.0010V noise on 30v range, and 000.005V-000.01V for the 300mV range). So if I am correct this points to the ADC section that has the problem?

Just to be clear, you've corrected the CAL ENABLE issue and you are observing those number as noise, not a steady offset?  Can you compare the 3VDC and 300VDC ranges?  I know those noise levels were low, but it still might be handy to see them. 

So it looks like the noise is present on the inputs of, or internally to, U101.  Since the input paths are mostly separate up to that point, and since the inputs to U101 are effectively shunted with ~100K in either case, it appears to me that the issue will be U101 or the closely associated parts of U102.

Here is a clip of the 3v and 300v range noise
https://youtu.be/PMRGZhfuDbk

And yes, the cal enable has been in the horizontal position with all of the tests.

Gonna probe U101 with a scope to know once and for all to know whether it is faulty or the input hybrid is. Is a hantek dso5102 fast enough? Currently thinking of probing pin 2 and 6.

[bdunham7]

Gonna probe U101 with a scope to know once and for all to know whether it is faulty or the input hybrid is. Is a hantek dso5102 fast enough? Currently thinking of probing pin 2 and 6.

It's going to be very problematic to probe this with a scope.  The input noise floor of the scope will be much higher than the noise you are looking for unless you use a very specialized low frequency preamplifier.  You could try shorting the input pins together and then seeing if the noise is still there and changes with range settings. 

Am I seeing that right--the noise (in counts) is higher on the 30V range than the 3V range?  If true, that might be a big clue right there.

[Aphelion]

It's going to be very problematic to probe this with a scope.  The input noise floor of the scope will be much higher than the noise you are looking for unless you use a very specialized low frequency preamplifier.  You could try shorting the input pins together and then seeing if the noise is still there and changes with range settings. 

Am I seeing that right--the noise (in counts) is higher on the 30V range than the 3V range?  If true, that might be a big clue right there.

Well, that is one option out.

As for the 30v range having more noise than the 3v range, yes it seems that way with autozero off. It is more apparent when you turn autozero on.

Well, rechecking the input paths it may really be the input hybrid that has a problem. I have already checked all of the resistors and capacitors that is attached to the input going to the hybrid.

So basically if 3v is noisier than 30v range, there is a chance that the precharge amp section of the hybrid is dead?

[Kleinstein]

The noise in the AZ mode looks good. It is normal to get quite some extra noise in the auto zero off mode. For the 5 digit mode I would not consider the AZ off mode to be of any use, maybe except in the 3 V and 300 V ranges. The AD544 was a good OP for it's time. There are better OPs now. If you really want to change it, my choice would be an OPA145 - though it would need an adapter board ( SO8 to DIP). Not sure if the lower noise is really needed - it may help a little in the 30 mV range.
 
The speed of the scope is not a problem. The interesting point would be the test-point at the output of U101. The point would be mainly if the settling after AZ switching and the precharge phase is OK and to see if there is significant hum ot other signal that is not supposed to be there.

P.S. The precharge part could just have too much offset. AFAIK the same input hydrid is also used in the 3457 and there they have additional offset trim. One could still had an manual (one time) offset trim if needed.