Repairing an HP 34401a DMM

[Corporate666]

So I picked up a non-working HP 34401a.  Schematics are at the end of this doc:

http://cp.literature.agilent.com/litweb/pdf/34401-90013.pdf

 It fails the self test, and when running through the individual self tests, it passes them all, except for #606, which gives the error "Rundown gain out of range This test checks the nominal gain between the integrating ADC and the A1U205 on-chip ADC. The nominal gain is check to ± 10% tolerance."

A brief review of the schematic looks like the voltage references goes to a few op amps and into a "hybrid resistor network" which is on a ceramic chip, which divides down the signals and then into the ADC of the main MCU.

I am not sure if some of my voltages are a bit out of whack.  I'm getting -18.198 on the -18V rail which seems OK, and I am getting 18.636 on the +18V rail, which seems a bit high.  The two 5V rails are within 10mV of 5, so they are fine.

The voltage reference in this is an LT1826-1249-5 and I am getting 6.8812V out of it.  I can't find the datasheet for this, but for some reason I recall it should be 10V, shouldn't it?

The +15 and -15V is generated by a couple of zeners (CR402-A and CR401-A).  The -15 one is getting -18.2V, but -14.386V on the other side, which seems too low.  The other is getting +18.6V and outputting +14.700 which also seems low. 

I should be getting a 5V reference out of U403, and I am getting 5.03V which seems close.  And I should be getting a 7V reference out of U400-B and 10V out of U400-A.  I'm getting 6.881V for the 7V and 9.829V for the 10V.  And I am getting -9.814 from the -10V reference out of U401-A.

Nothing is running particularly hot - several chips in the 120F range (50C) and the LT1826 reference is stable at 144F (62C).

It seems to me that the +15 and -15 supplies are too low, although they could be getting loaded down... but nothing really seems hot and it seems strange both +15 and -15 would be bad unless it was something else causing it.

Any ideas?  What's the LT1826 voltage reference output supposed to be?   Another EEVBlogger mentioned that often error 606 means the hybrid resistor network is bad.  I measured the resistances and they are all correct, except the 100k is measuring at 58k, but that's in circuit so not necessarily correct.  Hopefully it's not that part, as it's $191 from Agilent!

[Jay_Diddy_B]

Hi,

1826-1249 is an HP/Agilent/Keysight part number.

It is described as:



On the keysight website.

Linear Tech only makes two reference in the T0-46 package.

It is probably either a LM399H or LM399AH.

The datasheet is here:

http://cds.linear.com/docs/en/datasheet/199399fc.pdf

So 6.8812V is probably o.k.

Regards,

Jay_Diddy_B

[Corporate666]

Ah!  Thank you for the datasheet on the reference.


So what do folks think about the +/- 15V rails... am I wrong or they a bit far off from where they should be?

[Dr. Frank]

Depends also on the vintage of your instrument.
The +/- 15V are derived from +/-18V by ~3V3 zener diodes, and HP changed these throughout the years. Maybe the zeners have 3V9 instead of 3V3 in your instrument, so that may be fine.

Check for AC on the +/-18VDC.

The 6,8812V of the LM399 were attenuated/ amplified to the different reference voltages.
You can easily calculate the nominal values and compare to the measured ones,

The +7VREF is identical to the '399 direct output, that's ok.

The +5VREF should be ~ 4,9055V +/-2%, so 5,03V is too high => therefore check for AC components/oscillationon th eOpAmp output.

+10VREF is within tolerance, I assume, whereas -10VREF seems to be off, should be identical in value to the +10VREF.

In first instance, I really doubt that the resistor network fails,
There have been other, more ordinary  failures in the DC supply like defect diodes, tantalums, zeners, or defect OpAmps.

I don't know the other instrument you're checking the voltages with, hope it is good quality.

Frank


PS: I checked my 1992 manual of my own instrument (it's REV A)!
The +5VREF divider (R441/R442) is different and consists of 13k3 and 36k7, which gives nominal 5.05V +/-2%.

If your instrument is also that old (Rev A, B), then 5.03V are in tolerance.

CR401, 402 are 3V3 types also. Seem not to be changed.
Therefore, one of the OpAmps may draw too much current, maybe U401A.
These AD706 like to fail after decades, i.e. breakdown of their +- input transistors.

Last issue (successful repair) was U400B, if I remember correctly.


Anyhow, strong AC ripple or oscillation on +/-18V may also be the root cause.

[Corporate666]

So just an update...

I traced the signal path... it's actually quite interesting as they are using a digital switch to flip between the front panel input, a "pre-charge" output to the measurement path, and an offset measurement.  So you get a square wave of amplitude relative to your input signal and frequency relative to the measurement speed and update rate.

The signal is fine through all the filtering and amplifying... but where it goes into the U102 hybrid resistor network... I see the signal going into pin 15, but absolutely nothing coming out pin 21.  I was bummed - another user mentioned in another post that 606 errors are usually the hybrid resistor network, and that part is $191 from Keysight 

There was supposed to be 100k ohms between those two pins, and I was getting 58k - but didn't quite make sense why I was seeing nothing on the output.  I guessed the hybrid resistor chip had failed internally... so I desoldered it.  And when I tested it, everything was fine - all resistances were dead on accurate.

So... it pretty much has to either be the U411 mux/demux chip or either the U402 or U420 op-amps.  I ordered all three parts and will just replace them all.



The parts I am working on/replacing are all located under the metal shield of the meter.  I've been touching various parts and now that I desoldered U102, there's flux residue and will be more once I swap out the other 3 parts.  Are there any special steps required to maintain accuracy after I button it all back up?  Of course, I will clean of the flux residue but is there anything else that needs done?  Or is it just a simple matter of calibration?

[Dr. Frank]

The signal is fine through all the filtering and amplifying... but where it goes into the U102 hybrid resistor network... I see the signal going into pin 15, but absolutely nothing coming out pin 21.  I was bummed - another user mentioned in another post that 606 errors are usually the hybrid resistor network, and that part is $191 from Keysight 

There was supposed to be 100k ohms between those two pins, and I was getting 58k - but didn't quite make sense why I was seeing nothing on the output.  I guessed the hybrid resistor chip had failed internally... so I desoldered it.  And when I tested it, everything was fine - all resistances were dead on accurate.

So... it pretty much has to either be the U411 mux/demux chip or either the U402 or U420 op-amps.  I ordered all three parts and will just replace them all.



The parts I am working on/replacing are all located under the metal shield of the meter.  I've been touching various parts and now that I desoldered U102, there's flux residue and will be more once I swap out the other 3 parts.  Are there any special steps required to maintain accuracy after I button it all back up?  Of course, I will clean of the flux residue but is there anything else that needs done?  Or is it just a simple matter of calibration?

No wonder, that you measured 58k instead of 100k.. I assume in unpowered state..
There are several resistors in parallel to the 100k+R440, namely the network R406, R403, R405, R407 and further, plus the chain R197, R196, and the 200k of U102-B, and several OpAmp parasitic outputs..

So it is no wonder, that the network miraculously was fine...

Who the heck claimed that the resistor network would fail (often)??
Please show the reference, or who this user is..

And I really doubt that, as it is not evident, how this network should be damaged, or by which mechanism it should fail..


And it's no wonder at all, that you don't see any signal on pin 21.. that is a virtual ground, i.e. always having precisely zero volt!!

Therefore, your assumption, that U411, U420 and U402 failed, is also erroneous..

Why do you think so??

You should see a triangular integration signal on pin 6 of U402-A.. then it's fine..
Only if you don't see that, that may be defect, but only if the other OpAmps, I've mentioned are ok.

I propose you better measure again the A/D and reference OpAmps, before you wildly replace components!

That also could greatly deteriorate the A/D performance, i.e. linearity and precision, as these components may have been selected.. and you are fiddling/soldering at a very delicate circuit part!

Frank

 

[Corporate666]

The signal is fine through all the filtering and amplifying... but where it goes into the U102 hybrid resistor network... I see the signal going into pin 15, but absolutely nothing coming out pin 21.  I was bummed - another user mentioned in another post that 606 errors are usually the hybrid resistor network, and that part is $191 from Keysight 

There was supposed to be 100k ohms between those two pins, and I was getting 58k - but didn't quite make sense why I was seeing nothing on the output.  I guessed the hybrid resistor chip had failed internally... so I desoldered it.  And when I tested it, everything was fine - all resistances were dead on accurate.

So... it pretty much has to either be the U411 mux/demux chip or either the U402 or U420 op-amps.  I ordered all three parts and will just replace them all.



The parts I am working on/replacing are all located under the metal shield of the meter.  I've been touching various parts and now that I desoldered U102, there's flux residue and will be more once I swap out the other 3 parts.  Are there any special steps required to maintain accuracy after I button it all back up?  Of course, I will clean of the flux residue but is there anything else that needs done?  Or is it just a simple matter of calibration?

No wonder, that you measured 58k instead of 100k.. I assume in unpowered state..
There are several resistors in parallel to the 100k+R440, namely the network R406, R403, R405, R407 and further, plus the chain R197, R196, and the 200k of U102-B, and several OpAmp parasitic outputs..

So it is no wonder, that the network miraculously was fine...

Who the heck claimed that the resistor network would fail (often)??
Please show the reference, or who this user is..

And I really doubt that, as it is not evident, how this network should be damaged, or by which mechanism it should fail..


And it's no wonder at all, that you don't see any signal on pin 21.. that is a virtual ground, i.e. always having precisely zero volt!!

Therefore, your assumption, that U411, U420 and U402 failed, is also erroneous..

Why do you think so??

You should see a triangular integration signal on pin 6 of U402-A.. then it's fine..
Only if you don't see that, that may be defect, but only if the other OpAmps, I've mentioned are ok.

I propose you better measure again the A/D and reference OpAmps, before you wildly replace components!

That also could greatly deteriorate the A/D performance, i.e. linearity and precision, as these components may have been selected.. and you are fiddling/soldering at a very delicate circuit part!

Frank

user free_electron is a well respected EEVBlogger who has repaired dozens of pieces of test equipment.  When I search the 606 error, I came across a post (in reference to a 34970A, but that uses the same conversion mechanism and hybrid resistor) and his experience is that rundown gain 606 errors are typically due to failed hybrid networks.  The fact that Agilent sells that part separately also suggested it's something that people are replacing (which matches free_electron's experience).  And the fact that the A/D signal path is tested linear up until the U102 chip suggested it was a likely culprit for being bad.

But of course it's not simple to test a resistor network in-circuit so I desoldered it and sure enough, it was OK.

The service manual for the 34401a explains the theory of operation of the unit and in the troubleshooting section for inaccurate readings, it indicates to test the linearity of the A/D input which I did and it's fine.  Then to check the +/- 10V, which are both within sped.  And then to look at U411 (the mux chip) and op-amps U420 and U402.

When you get to the U411/402/420, the output from that portion of the circuit goes right into the ASIC... other than the mux and op-amps, there isn't much else.  I was ordering parts from Digikey anyway and the mux and op-amps were about $4 so I think it was reasonable to order them

I've checked the linearity of the A/D signal up to U102 and it's spot on.  What A/D and reference op-amps are you suggesting to check?  I've checked U103, U105 and U106 with no issues.

[Corporate666]

I noticed something else.

While testing the unit, I heard the occasional beep, but my desk is a mess so I thought things were just pushing the buttons.  However today I noticed that it is resetting.  It's very sporadic, but every few minutes it just resets.  Every segment on the display turns on.. I get the long beep and "RS-232" shows on the screen then it goes back to the reading.

Sometimes it will also lock up.  Normally (with nothing connected), it will read -1.786VDC.  Occasionally it will show "overload" on the screen and sometimes it switches to a higher range.  Sometimes it shows just underlines instead of a reading and it only comes back when I cycle the power.

All the voltages seem fine though.

[rx8pilot]

Did you ever get this thing up and running?

[shaohu.tang]

I have repaired a 606 error failure of an Agilent 34401 multimeter. After replacing u420 and u402, it is ok.