HP 3478A Troubleshooting

[bdunham7]

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

That actually seems less likely, but I suppose it is possible that a switch is leaking there or something.

Looking and your videos and your reported results seems a bit confusing.  If you have a display bouncing from 11 to 13 counts, I would call that 2 counts of noise and 12 counts of offset.  It's important to maintain the distinction.  Perhaps you could make a chart of the offset and noise in each range with AZ on and off.  So, after a set warmup--say an hour--go through the ranges from 30mVDC to 300VDC, a set time--say 15 seconds--in each range and write down your estimate of the min/max variation (noise) and average (offset).  After you do all the ranges, go back with AZ off and do it again.  I think this would really help me fully understand what you are seeing.

As far as what is normal and what is not, on my presumably working model AZ makes very little actual difference in noise, but it does appear to move faster because you see twice as many readings.  Letting it sit for few hours in the 3V range with AZ off resulted in 1 count of drift in that time.  If you have noise, say in U101, then AZ won't really help and may make it worse because each AZ reading will be affected by the noise differently--so your zero will be changing constantly.  At least with it off, the CPU is using the same value for zero all the time.  Short term, that could actually be more stable if the noise is high frequency--less than an A/D cycle-- and random. 

Normal offset on mine is a few microvolts--about 4.0 when first turned on, then +1.0 to +2.0 (10 to 20 counts) on the 30mVDC range when warmed up.  All the higher ranges are 0 or 1 counts.  Causes of a steady offset could be actual voltage on the inputs at any point prior to the junction of the input amp U101 and the AZ section (which I believe is S8DC inside U102) or an fixed offset error in the AZ circuit.  All of that would be inside U102 and probably would be an issue with the analog switches.  AZ should really take care of any other offsets anywhere else along the chain.  Of course there will be a residual offset and that is taken care of as a calibration constant, provided that it is stable.

You've probably read the manual pages 7-A-1 through 7-A-4, but they're worth studying closely.  Some of the tests, such as the part where you supply a precision 3 volts, short pins 2 & 6 of U101 and then make measurements with a precision voltmenter, probably require better equipment than you have.  You may have also seen the short part where they essentially tell you that if you have noisy readings in some ranges, replace U101 and if that doesn't fix it, U102.  Obviously nowadays that is easier said than done.  Try to make up that chart and I'll think a bit more about how you might be absolutely sure what the issue is.  U101 may be findable, U102 not so much....

[AVGresponding]

Just to provide you with an additional data point; when I first turn mine on it has around a 2.5uV offset with around 5 counts of noise on the 30mV range, 0 on every other range.
This rapidly drops to below 1uV after only 10 minutes or so, with the count noise dropping to 2 or 3.
Using AZ makes as you say little difference, at least at turn on. If I switch it off once it's warmed up for a bit it can make the offset and noise jump all over the place, to the tune of as much as 3uV +10 counts.

Ambient here is 20°C, and my short is a home-made one, using WAGO 4mm banana plugs and soldered together solid core 2.5mm² copper wire.

[Aphelion]

Here is what I have gathered for the DCV range Room has closed windows with 30 degC temp with a RH of 60%. I've let it run still for 5 hours before taking measurements. Cover is off, CAL ENABLE is in horizontal position.

Nickel plated banana plugs with soldered in 18 AWG wire for the shorting setup

[Aphelion]

Just to provide you with an additional data point; when I first turn mine on it has around a 2.5uV offset with around 5 counts of noise on the 30mV range, 0 on every other range.
This rapidly drops to below 1uV after only 10 minutes or so, with the count noise dropping to 2 or 3.
Using AZ makes as you say little difference, at least at turn on. If I switch it off once it's warmed up for a bit it can make the offset and noise jump all over the place, to the tune of as much as 3uV +10 counts.

Ambient here is 20°C, and my short is a home-made one, using WAGO 4mm banana plugs and soldered together solid core 2.5mm² copper wire.

Hmm, mine even after an extended warmup time has an oscilating -+50 counts of noise in the milivolts range

Offset often varies for some reason. Sometimes it zeroes out, sometimes it does not

[Kleinstein]

The noise seems to be at the amplifier input side and not from the ADC.
The offset is quite high, but this could be a calibration / software thing and in parts caused by the high temperature.
In the AZ mode the noise does not look to be so bad - depending on how it is measured (peak to peak or RMS).
The OPAs used for U101 may differ a bit, some may be more sensitive to temperature than others and the LF noise is also not always the same.

If the input hybird is part of the problem, one may also see some effect on the input side, like more input current or larger input current spikes. With the 100 K and cap from the filtering one may not see very much, but it could be worth to just use the scope to look the the input in votlage mode. For a slight votlage for the AZ switching to work against maybe also with some 2 or 3 V (e.g. 2 x AA battery) in series. There should be only minor switching spikes from the AZ swtiching visible.

Another simple test is the input bias: e.h. some 10 M accross the input and than read the DC voltage. The input bias should be less than 100 pA and thus less than 1 mV, ideally less than 0.2 mA (20 pA).
Another way to test the input is to measure the drift with a 10 nF cap (low loss like polypropylene or polystyrene type)  accross the inputs. 100 pA would give you 10 mV per second as a drift rate. 

[bdunham7]

So your offset scales with U101 gain and with the 100X divider, but the noise more or less scales only with the gain of U101.

The offset may just be a calibration issue and may not even be a problem, since it seems fairly stable.  You might take your other DMM and check the voltage at the junction of R110 and R103 on the various ranges just to see what is there.  However, as far as the noise goes, I'm not sure I can offer you better advice at this point than what the manual says--replace U101 first, if that doesn't fix it it needs U102. 

You can try shorting pins 2 & 3 (inputs) of U101 with a very short wire (perhaps on the back side of the PCB) and then see if there is any noise and if it scales with gain.  You can also try the procedure in the manual that involves connecting pins 2 & 6 so as to set the gain to x1, but I think the noise level is just low enough that you won't be able to see it this way.

Unfortunately I'm fearful that this is a U102 issue, but I can't see quite how to prove it at this point.

[Aphelion]

Initially I kind of came to the conclusion that U101 is defective as ACV takes a different route through the input hybrid instead of the DCV mode that goes to switching FETs and directly to U101. And ACV has almost no noise (+-000.001mVAC)

As far as replacements go, I've only read up a single thread on the internet that indicated that it is a AD542KH

https://www.electronicspoint.com/forums/threads/crossreference-for-a-ad542kh.172618/

Is an AD711 a suitable replacement to even just check whether or not the problem lies with U101? Once I confirm that it is the case I might get an OPA145 as the AD711 I ordered is still in transit as Kleinstein suggested it is a better replacement for it. The AD711 is what the analog devices page suggested as a replacement for the AD547

[Aphelion]

I've also tried the "short the pins 2 and 6 of U101" from the manual. While the reading in the 3478a is no longer near the 3.5v of the battery that I hooked up, the noise in the reading basically disappeared. I don't know whether or not that matters though

[bdunham7]

I've also tried the "short the pins 2 and 6 of U101" from the manual. While the reading in the 3478a is no longer near the 3.5v of the battery that I hooked up, the noise in the reading basically disappeared. I don't know whether or not that matters though

Well, it reduces the gain of U101 to x1, which means it attenuates any input noise.  If it ranged down into the 3V range, that's 3.33X, so you'd have to judge whether the noise reduced by more than that.

Is an AD711 a suitable replacement to even just check whether or not the problem lies with U101?

I can't see any reason it won't work.  I was unable to find any specs for U101 or to confirm that 1826-1056 crosses to the AD542, but supply voltage, input bias current and so forth all look acceptable. 

[Kleinstein]

Shortening pin 2 to 6 reduces the gain from some 3 to 1. So the noise would also go down by a factor of 3 and thus more or less get invisible.

As far as I understood it, the 3478 uses the average of 10 shorter conversions in the 5.5 .digit mode. So in AZ mode this would be something like 23 conversions short per second. So the relevant frequency should be around 20 Hz and the effective bandwidth for the ADC at about 3 Hz.

So the expected noise from U101 should be the noise density at some 20 Hz times  sqrt(3 Hz)  times sqrt(2) 2 from the AZ mode using a 2 conversions.
for the AD711 this would be about 30 nV/sqrt(Hz) * 2.5 Sqrt(Hz) = 75 nV (RMS)   or some 0.45 µV peak to peak.

So this should be considerably (e.g. 10 x) better than the current state for the AZ mode. The bias may be higher, but that is OK for a test.
So the AD711 should be OK (low enough noise) for a test.

The non AZ case may however be relatively poor, as the 0.1 to 10 Hz noise is suppose to be quite high (6 µV).  Modern parts may be better than the specs, but one never knows for sure with the LF noise.

[jeffjmr]

I know this post is ancient history, but folks with 3478As with similar problems (like me) might benefit from something I found in a newly acquired HP3478A.

U101 on this unit is an AD18261921 8 pin DIP. There are signs of resoldering on some of the caps and the battery has been replaced, but U101 looks like it came from the factory. My other, slightly older 3478A has the specified 1826-1056. Somewhere along the line HP changed from the CAN U101 (AD542KH?) to the DIP, but the parts list still shows 1826-1056.

Supposedly some chip houses have the 1826-1921in stock.

I have not yet compared the 1921 data sheet to the AD542KH that other 3478A threads mention as a possible sub.

Has anyone else found an AD18261921 in their 3478A?

Jeff

Edit: No dice on the data sheet for the 18261921. Still awaiting a quote though.