3458A error code - internal failure 104 - repaired!

[Tracker]

Hello Forum,
I have a lovely 3458A 8.4 digit DMM which is rarely used.
I was planning on selling it, but as I was running through test validation routines it threw me an error message:

209 ‘ Hardware failure – Internal overload 104

I have checked the service manuals and all I can derive is that 200 type codes are hardware, the manuals don't provide much else.

Has anyone come across anything like this and has any suggestions?

Hopefully its nothing serious, the meter seems to read everything perfectly fine, all ranges seem to operate and measure accurately.

Thanks

[Tracker]

Understandably no replies... I didn't give much information to attract attention.

However...
I have done some more testing and narrowed it down using the front panel Auto Cal function.
I discovered the Auto Cal only fails on Ohms.
DCV and AC both pass the Auto Cal function.

Then when testing the Ohms ranges independently, ranges 10 ohm 1 Meg Ohm are good, but the 10 MOhm range is reading incorrectly, a 1 Meg resistor reads about 1.2 Meg ohms
The subsequent ranges are also reading wrong, this offset is maintained on the 100 Meg and 1G ranges.
So I'm concluding the issue lies in the 10 Meg Ohm range.

I plan to start more detailed schematic analysis, but any help or direction would me much appreciated folks!   

My initial thoughts are around the range switching circuitry, since it's not mechanical relay switched I'm suspecting its switched through high impedance IC devices....

Thanks in advance.

[Dr. Frank]

Hi,
the problem could probably be in the Ohm current source circuit, but also in the DCI shunt circuit.
As a first test, please use the 4W Ohm mode (sense shorted and connect to minus input should do the job) and measure all constant currents with another DMM, if they are off compared to the default values.. 20% too high should be easy to identify.
Then also check if especially the low DCI ranges are measuring correctly.

Btw.: if you get this error 209/104, is there additional information if you press <more> (arrow button)?
PS: What does the extended TEST reveal?
Frank 

[Tracker]

Hi Frank, thanks for taking the time.

Reading the currents produced in Ohms mode give the following
10 Ohm range = 9.7 mA
100 Ohm range = 0.971 mA
1K Ohm range = 0.971 mA ( same as previous range)
10k Ohm range = 97 uA
100k Ohm range = 48 uA
1M Ohm Range = 5 uA
100M Ohm range = 0.63 uA (this seems to be where the problem arises)
1G Ohm range = 0.63 uA (same as previous range)

And I happen to have a good 3458A to compare against, the readings there are 0.50 uA on the 100M and 1G ranges.



Then measuring low currents (using accurate 1v DC source and 1% 1 MOhm resistor)
100nA range: 100nA source reads 95.72 nA (small error may be due to my  primitive current source )
1 uA range: 100 nA source reads 0.09957uA
10 uA range: 100 nA source reads 00.100113 uA
100 uA range: 100 nA source reads 000.10014 uA

So these seem close enough to be reasonable don't you think?

What say you Frank, problem with Ohm current source on top two ranges?




 

 

 

[Tracker]

A little more digging and I'm starting to suspect this area... what you think?

The 0.5uA source must be common to both the 100M Ohm and 1 G Ohm ranges.

this diagram is from the Sentry Ohms Diagram.

[Kleinstein]

Too high a current is likely a problem with the current source. It is unlikely a defect in the protection part would cause to high a current.

For the cause there are several options. Worst case it would be the 6 M resistor, which is part of a resistor array - however this would be very unlikely to be 20% off.
My suspicion would be more with the transistor array U302, with one transistor not turning off. This would be relatively easy to check, by measuring the emitter voltages for the channels that are off. These should all be at the same 10 V level.

I don't think very much it would be the encircled part, as this part is working like 2 switches and the more likely defects (e.g. gate leakage at Q308) would tend to reduce the current, or bring it totally off.

[Dr. Frank]

There's definitely a defect in the 0.5µA current source circuit, see table.
I also measured on my 3458A that the nominal currents should be quite precise, to about 0.5%.
I used a 34465A.

10M, 100M, 1G : 498nA
1M: 4.98µA
199k: 49.8µA
10k: 99.6µA
1k, 100: 994µA
10: 9.94µA

What makes me wonder, why this too high 500nA current in the end affects the ACAL process.

This current is measured by the DCI circuit, and any deviation (in the 6M resistor, or in the 3V reference voltage) would normally be calibrated into the Ohm range gain.
Therefore there is a difference between the ACAL mode and the measurement Ohm mode, in first case measuring a lower (correct ?) current, in the latter case too high a current.

Anyhow, in a first attempt, the 500nA  source has to be investigated.

I also stumbled a bit about the other ranges, don't know if it's of relevance. I can't judge, how precisely you measured the other ranges, as their currents seems to be too unprecise, too: 3% vs. 0.5%.
But if you find similar values on your other, known-good 3458A, then it's only your test setup.

So let me think again, what's going on.

Frank

I suspect now U300, which should be a low bias type. If this OpAmp drains 100nA into the node, that would make a 20% effect in the 10M .. 1G ranges, but only a 2% effect in the 1M range.
In the 34401A, this similar OpAmp type (AD706) very often fails after years, with excessive bias currents.
The 34401A has many similarities with the 3458A, as it's derived from its design.

Therefore it would be interesting to measure the low currents more precisely, by means of your 2nd 3458A (lucky guy).

This topology of the current source was already designed for the hp3468A, hp3478A by a former hp engineer by the name of George Hnatiuk.
In the 3478A, an LM208 was used at this place of U300, which has typically 800pA bias @ 25°C.
That's a bit high, and I do not know, which type this U300 might be, otherwise.

[Kleinstein]

Normally ACAL could compensate for variations in the actual test currents. I don't think they have to be accurate to 0.5%, but there certainly is a limit set by the internal software and +20% is too much and thus the error message. The internal cal constants may not be updated and thus the wrong reading.

With the current source not working,  ACAL may also not work for the lowest current range - though not sure the smallest source is actually used for the calibration, they could use the 5 µA source. So there is a chance the 100 nA range is also not adjusted with ACAL since the defect turned up.

[Tracker]

Thanks guys,
Its late here so I'll measure currents more accurately tomorrow now that I know the importance.
I did run ACAL earlier on today, before taking my measurements so the Hardware fault has probably corrupted the calibration table. The ACAL routine quit with an error message mid way through the Ohms section.

I think were in the right area, if necessary I'll replace both suspect devices.
But I'm not keen to open my other 3458A for comparison measurements since its the Cal seals are still intact, it just seems wrong to me, its weird I know!!!

[Dr. Frank]

Thanks guys,
Its late here so I'll measure currents more accurately tomorrow now that I know the importance.
I did run ACAL earlier on today, before taking my measurements so the Hardware fault has probably corrupted the calibration table. The ACAL routine quit with an error message mid way through the Ohms section.

I think were in the right area, if necessary I'll replace both suspect devices.
But I'm not keen to open my other 3458A for comparison measurements since its the Cal seals are still intact, it just seems wrong to me, its weird I know!!!

Yes, I thought so, and so I try to define a non-invasive test method first hand.
The idea is to find out, if the 1M range (5µA)  is also affected, or the 10M..1G range (500nA) only.

If you reversely calculate the resistors from  the currents, by assuming 3.000V reference voltage you'll get identical digits, to about 0.2% precision:
3V/0.498µA = 6.024M
3V/ 4,98µA = 602.4k
.. 60.24k, 60.24k/2, 6.036k/2, 603.6/2

So at least in case of my unit, all these resistors are evidently trimmed to about 0.1% of their nominal values, or relative to each other.
You can check that also on your known-good 3458A, but also for the first 4 higher current ranges of your bad 3458A.
Therefore, if you measure these currents to say 4 digits resolution accuracy, and you already find that the 1M range (5µA) deviates by + 2..3% compared to the 10k and 100k currents, then the fault probably is not directly inside the 500nA switch, but more probably outside, like for example in U300.   
In this case, you can still overcome your inhibition to open the unit.

Frank

[Kleinstein]

I see no need to open the good unit, but I think one would not get around opening the broken one to repair it. A first point could than be to check the supplies, tough I doubt there is the problem.

The extra 100 nA could also be present in the 5 µA current. Compared to the other currents this may even be likely, as the others are some 3 % lower than nominal. However for the 5 µA current this would still be in the acceptable range and ACAL could compensate for it. So no error in the 1 M range.

For finding leaky semiconductors it can help to apply local heat. Semiconductor leakage tends to go up with temperature quite a bit. So even some 10 K more would be well visible. If it is a leaky semiconductor there is also a good chance the error would be smaller when the meter is not yet warmed up. If the error would be with a resistor, the error is expected to be less temperature dependent.

Another possible failure candidate is Q305 (the P-JFET in the ohms current control loop, with possible gate leakage).

The suggested test against the 10 V level at the emitters would be against the internal 10 V level in that unit, to see if there is voltage over the resistors in RP303.  It is not about µV, more about possibly some 200 mV of error there that could lead to diode leakage in the 100 nA range.

[Tracker]

Thanks Guys, its really great to have this help 

Yes, opening the defective meter is a given. I'm just a bit reluctant to open the good meter at this point... but if needs must.

So returning to more accurate currents measurements using the good 3458A, keeping in mind the advice received:

New findings: the 500nA source has a noticeable drift over the first few minutes after switch on from cold.
I also can see quite a lot of unstable reading in the lower digits, noted with 'x' below.
The 5 uA range also drifts somewhat, but not nearly as much as the 500nA range.

10M, 100M, 1G = starts out 0.95xxxxx uA, dropping to 0.53xxxx uA over a few minutes.
1M M = starts out at 5.255xxxx uA, dropping to 04.926xxxuA over a few minutes.
199k = 48.63xxx uA, note least 3 digits are drifting
10k = 097.19xxx uA, note the least 3 digits are drifting
1k, 100 = 0.97182xx mA
10: 9.7291xx mA


Reversing the test, measuring 4 wire Ohms current from the good meter using the faulty meter:
10M, 100M, 1G = 0.4876xx
1M M = 4.8763xx uA
199k = 48.745xx µA
10k = 097.530xx uA
1k, 100 = 0.97523xx mA
10: 9.7524x mA
These readings have minimal drift over time, and in all cases only the least 2 digits are drifting/noisy (though I was using guard)

Given this fresh data, it does seem to be something common to all ranges, but impacting to a lesser degree on the higher currents. The dramatic drift from cold boot is also telling us something... does this seem like U300?

Thanks

[Dr. Frank]

On the defective 3458A, the leakage currents are +44nA for the 500nA source (10M..1GOhm), +100nA for the 5µA (1MOhm) and probably +30nA for the 50µA (100kOhm).

Therefore, that can be caused by U300, U301, CR314, or Q304 & Q305.
One of CR301..CR306? I don't think so, most probably it's U300, a Burr Brown OpAmp, or Q304, Q305.

The type of U300 should be readable if you open the defective unit, but this should be similar to AD706, i.e. a single OpAmp, bipolar type, low bias of about max. 200pA, low offset voltage.

AD has an OP97, which might fit into this layout, as it's DIP 8 originally.. to be checked.

Maybe T.I. still offers the Burr Brown OpAmp, if you can identify the correct type.

Frank

PS: I don't think that you have to open the working unit.

[Kleinstein]

The additional current in the 500 nA and 5 µA ranges is about in the same order of magnitude. This would fit like some additional leakage / bias current somewhere. It is however odd to see the extra current to go down with temperature / warm up after turn on.
The easy to check part should be U302 - this can be done in circuit with a simple voltage measurement.

There should be an easy check for Q304: measure the current with some additional drop (e.g. 1 or 2 diodes in series to the output). With higher voltage at the output Q304 can no longer deliver a positive current. Already the diode internally in the circuit should give enough of a voltage so that Q304 should not be able to provide a positive current.

I would also check the control signals : for some reasons failures in LM339 comparators are not that uncommon. A comparator not turning on all the way off (not reaching below some 5 V) could also be a (easy to fix) problem. It is not very likely, but easy to check.

U 301 and CR314 are likely not the culprits as the other ranges work. Leakage there should not directly effect the output current, as this senses the upper side of the resistors and this would add a constant factor to all ranges.

My main list of suspects is U302, Q305, U300, U303, U304,  Q310...Q314, CR302...CR306, dirt on the board.

[Tracker]

Ok, thanks guys.

I didn't have much time on this today.
Opened the faulty meter, located the PCB area. Everything looks spotlessly clean. The meter in general has very little sign of dust build up.
PSU voltages look ok, except -10v which reads -9.712v but it doesn't look to be utilized in Ohms current generation.

U300 is an 8 pin metal can type marked with makers logo BB (r). part number 1848.
Took a look at temperatures in the area of interest but no components are ruining hot, infact all were very cool (room temp) except for U400 which read about 27 degree C.

I didn't get a chance to take any more measurements, hopefully tomorrow evening.

Any tips on where to start would be appreciated.

Thanks

[Kleinstein]

I would start with measuring the voltages at the anode sides of CR302-CR306 which is also the emitter sides of U302. This could give a hint if one of the other resistors / ranges is not fully turned off  (still not clear if it is from U302,CR30x, or the Q31x but it would show which path is possibly at fault). This measurement can be done in circuit with the ohms source in the 500 nA mode.
Similar the control signals are easy to check (setting for 500 nA) - though an unlikely fault.

For a first test on U300 one could measure the current and give a little heat (some +20K) to U300. If there is significant change, U300 gets more suspect. 

[TiN]

Does meter correctly read overload on 1G range, when nothing is connected at the inputs?
I had an issue with leaky ohms protection network on my first meter before. Likely not related to issue here, but easy to check.

[Tracker]

Hi Tin, thanks for your input,  yes meter reaches OVLD on all ohms ranges.
I plan to start measuring voltwhes as advised tonight.

[Dr. Frank]

As the failure current seems to be higher at lower temperatures, the quickest way might be to cool the suspicious components carefully and one after the other.
Frank

[Tracker]

Hi Frank, yes thats a good idea, can of freezer spray is at the ready!

But before that here is an update from some debug work earlier this evening.

Control logic checks out good, driving full logic levels to U307 and U304. 
U302 voltages are consistent in all ranges, Base is 13.7v, collector is 13v and delivers 13v on emitter when enabled by logic circuit.

Moving to U300 area...
Performed some easy checks first... like resistor values.
Then stumbled on R304 which is supposed to be 2.43k but reads 476.5k in circuit, but when powered on the potential divider appears to be working fine, delivering 0.1947v to Q304.
So this may be a false lead. PS...I'm not sure what function Z300 transistor array has?

Then I measured output voltage using a standard 10 MOhm DMM and found an interesting difference between the good and bad meter.
On the 10M Ohm range the good meter outputs 4.924v but the bad meter outputs 7.641v with drift. This is odd since all tree ranges share the same logic and driver circuits (U304B -> U302A )
I can also hear a relay click between 10M and 10M ranges which implies there is some other form of switching going on for the 10M range.
Good meter           Bad meter
1G = 2.449v      3.648v
100m = 2.449v      3.648v
10M = 4.924v      7.641v
1M = 8.438v      8.423v
100k = 8.013v      8.002v
10K = 7.920v      7.913v
1K = 7.872v      7.807v
100 = 7.872v      7.807v
10 = 9.844v      9.423v

I'm not sure where this is taking me, but the difference did surprise me.
Wonder if it is due to the calibration table to correct ?

[Tracker]

Folks,
Having thought about it some more, the difference in output voltage on Ohms ranges is probably due to  the AutoCal attempting to correct.

So I started temperature debug with hot/cold air as Dr Frank recommended and found a suspect...

C302 is a 27pf ceramic providing feedback on U300 and is extremely sensitive to temperature change
At room temp the output current (10M 100M, 1G range) reads 0.6x uA as per previous readings, but when C302 is heated to about 50 degrees the current rapidly rises to over 1 mA, then when cooled with a small blast of freezer is instantly falls to below 0.5 uA. These responses were virtual instant.
I took great care to isolate/screen all other components from the heat/cooling using credit cards! 

I tried to measure ESR in circuit but with such low PF value I'm not sure its valid.
My 4 wire LCR meter reads 180pf and ESR 211k and these values didnt change much over the same temperature range, but I'm not sure this reading is valid since the capacitance is probably swamped by the other circuit components.

What do you think guys? Snip one leg and measure out of circuit? Or have a fallen into a hole here?

   

[Kleinstein]

The measured external voltages with a 10 M meter just reflect the high current. So noting really significant.

It would be rather unusual for a 27 pF capacitor to fail. At least it would be one of the cheapest parts to replace. So if the part is suspect one could just replace it with something new.  Chances are that removing it from the circuit could effect the defect so that measurement of the removed part does not tell that much about the current behavior. So cutting one leg has limited value, but also would not hurt very much.

It could also be something near by the is also effected by the heat, may be indirect through the leads. Anyway this narrows down the search essentially to C302, U300, Q305 and maybe though unlikely CR307.

For testing U302 the critical case is when the current path is turned off. Really turned off the emitter should read some 10 V. With more than about 10.2 V there could be leakage current going through the diodes.

Z300 is the protection of the ohms source against too negative a voltage at the input. As there is no positive voltage involved it can not provide extra positive current.

[Tracker]

Thanks Kleinstein,
The Off voltage on U302 is 9.7v, so well and truly off.

Anyway, good news to report.   
I lifted one side of C302 27pf and it reads 30pf with resistance of about 3M ohm, heated it falls to 1M Ohm and when cooled it rises to 50 M Ohm so this definitively is leaky.
The leak was probably causing a current flow irrespective of range setting.

I have tacked in a 30 PF ceramic (hand selected to 28pf which is as close as I can find) and re-run the AutoCal which now passes   
Self test also now passes.

Also, my decade resistance box shows matching readings on both meters now. All the way up to 100 M Ohm

So this meter is well an truly repaired, thankfully no major surgery needed.
I'm not even going to lift the PCB and risk contamination, I will replace the cap from above leaving the old legs in place, much safer to do.

Who would have thought it eh! A leaky ceramic cap!

Many thanks to Dr Frank, Kleinstein and TiN for invaluable direction.
And it was impressive how close we got before even opening the meter.

Cheers guys, this forum is amazing.

[Dr. Frank]

Folks,
Having thought about it some more, the difference in output voltage on Ohms ranges is probably due to  the AutoCal attempting to correct.

Certainly not, as the ACAL routine does not influence the analog circuit and its values at all.

So I started temperature debug with hot/cold air as Dr Frank recommended and found a suspect...

C302 is a 27pf ceramic providing feedback on U300 and is extremely sensitive to temperature change
At room temp the output current (10M 100M, 1G range) reads 0.6x uA as per previous readings, but when C302 is heated to about 50 degrees the current rapidly rises to over 1 mA, then when cooled with a small blast of freezer is instantly falls to below 0.5 uA. These responses were virtual instant.
I took great care to isolate/screen all other components from the heat/cooling using credit cards! 

I tried to measure ESR in circuit but with such low PF value I'm not sure its valid.
My 4 wire LCR meter reads 180pf and ESR 211k and these values didnt change much over the same temperature range, but I'm not sure this reading is valid since the capacitance is probably swamped by the other circuit components.

What do you think guys? Snip one leg and measure out of circuit? Or have a fallen into a hole here?

 

That C203 is intended to suppress oscillation of the U300.
That would be a rare failure, but can well happen with ceramic capacitors.
These might create cracks inside, mostly by mechanical stress (bending), so  affecting or even destroying the dielectric / isolation layers.
Especially on high capacity MLCCs, we have to provide safety measures, if these are used directly parallel to the car battery.

At first I thought that the 50°C  might create excessive thermo couple voltage on the - input of the OpAmp, but that can't be that severe, as it's on 10V, versus maybe µV or mV of t.c. voltages.

So yes, try to clip it off, so that you may afterwards solder it together again.. or if you have a spare 27p COG type.


The voltages difference between the good and bad unit is simply a sign that the current source has a leakage problem, so no real new information out of that.
If you calculate the flowing currents, you'll get the same ballpark as your direct current measurements.
The only difference are the higher (different) threshold voltages over your 10M test resistor. This shows, that this error obviously is also dependent on the output voltage of U300, so that's another hint that either this C203, or FET Q305 is the root cause.

Frank

Ups.. too late.. you've overtaken me.. great that you solved the problem.
Make sure, that your new COG part is low leakage, also!

[Tracker]

"Make sure, that your new COG part is low leakage, also!"
Yes indeed Dr  Frank, its low leakage type and also checked it to be > 1G ohm.

A rare fault indeed, very glad to have it repaired so simply.