HP 3450A linearity problems help/input requested

[valley001]

Hello All,

I have made a lot of progress on a basket case 3450A I have here. It needed its faulty reference and Peltier changed out, several bad caps, etc. 

I finally have it to the point where it is responding to inputs, and I have been attempting to calibrate it. 

1- Everything is fine until I get to the 100mv range and I cant bring it into spec, it stops shy of matching my 332D by 20 microvolts or so (reads 99.978mv and runs out of turns on the pot).  My 3456A reasonably agrees with my 332.   This fault dues not seem to be linear, (the following is approximate)  a 10mv input will read 010.008, 20mv will read 020.004, 30mv will read 030.001, 40mv will read 039.999 etc.  The 100mv range cal pot was reading 0-46 ohms instead of the specified 0-50 so I changed that out with no effect on the faulty calibration.

2- 1v and 10v ranges work fine and calibrate. 

3- 100v range reads way wrong (a 100v input from my 332D reads 114) but I have not gotten there yet, I really want to sort out the 100mv range.   

4- All of the amps, integrator etc zero out with my 845ar.  I can bring the reference and polarity amp to specified 10.00000v

I presume this may be related to a reed relay, some of them seem iffy but I need help narrowing down which one might be the culprit (removing them is an involved process). 

Thank you all in advance for any assistance. 

[valley001]

HI all,

I took some time and did measurements to show more clearly what is happening with this unit.  I am hoping that someone who understands the circuits better than I can possibly point to areas that should be checked.  Since replacing the 100mv range calibration pot I can just barely bring it up to 100mv but as you can see it is not linear. 

On the MV range this is what I will see, this is with the 100mv pot maxed out.   Voltages provided by my Fluke 332D.

332D   HP3450 readout
\/            \/
0v-  000.025
10mv- 010.021
20mv- 020.018
30mv- 030.015
40mv- 040.013
50mv- 050.011
60mv- 060.008
70mv- 070.006
80mv- 080.003
90mv- 090.001
100mv- 100.000

If I zero it to 50mv I ger

0v- 000.025
10mv- 010.020
20mv- 020.014
30mv- 030.012
40mv- 040.007
50mv- 050.001
60mv- 059.997
70mv- 069.994
80mv- 079.988
90mv- 089.983
100mv- 099.978

By contrast, this is what I see on the 10v range (the last digit will sometimes dither to one side or the other):

1v- 1.0000
2v- 2.0000
3v- 3.0000
4v- 4.0000
5v- 5.0000
6v- 6.0000
7v- 7.0000
8v- 8.0000
9v- 9.0000
10v-10.0000

What could cause this gross non linearity in the mv range?  I have checked

[Kleinstein]

The problem seems to a gain factor that is not right - due to the limit of the trimmer it can not be adjusted and an offset of some 25 µV. So the main problem would be the offset. There some extra adjustment for the offset, despite of the use of an AZ phase.

Is the transition to negative readings working OK ? As far as I see it there is an extra polarity switching at zero. This may also add some offset for one side.

[valley001]

The problem seems to a gain factor that is not right - due to the limit of the trimmer it can not be adjusted and an offset of some 25 µV. So the main problem would be the offset. There some extra adjustment for the offset, despite of the use of an AZ phase.

Is the transition to negative readings working OK ? As far as I see it there is an extra polarity switching at zero. This may also add some offset for one side.

Switching polarity with 0v output reads negative 26 microvolts. 

The really frustrating thing is I am discovering these faults are not constant.  It just now reasonably zeroed the 100mv range and then suddenly jumped 20ish microvolts low as in back to where it was in my post above.  It seems to do this as I change input polarity and possibly also as I increase input voltage from a lower voltage to higher in the 100mv range.

This is highly suggestive of a relay fault no?

[valley001]

Now I notice an audible change in the rhythm of the relays between faulty operation and correct operation where the 100 mv range comes into focus. 

Could a gain factor error be introduced due to poor relay timing? 

[Kleinstein]

A bad relay is definitely a possibility. Another common type of fault are the trimmers.

One may be able to find out if the problem is with the input amplifier or the rest of the circuit. If the problem is with the rest of the circuit there should be a comparable problem also with the other high ranges, including the 10 V range.
If the 10 V range is pretty stable, one has essentially narrowed down the problem to the input amplifier.

The input amplifier would still include the clock generation and control for the AZ relays, though I doubt this would be the problem. Synchronization to mains would be the main tricky part here.

Temperature control could also effect the amplifier.

[valley001]

Disregard the lined out portion, it was my 332D providing the intermittent output when switching from standby to operate.  I noticed when monitoring the voltage with my 3456a.    Another thing to look into some day...

I have narrowed down the intermittent fault to whatever detects an input voltage.  If I switch my 332d (set to 0v) between standby and operate I periodically see the offset change 20-40 microvolts.   

I dont think it is temperature, the jump happens abruptly.

I swapped the timing board with no change, but I have not measured the pulses directly.  The amplifiers, integrator, and zero detect seem to hold their calibration so I dont think those trimmers are the culprit necessarily. 

[valley001]

I spent time this evening and verified the problem is isolated to the 100mv range, it tracks very well in the 1v range.

Also, the voltage offset comes from my 332D.  Between the 3450A and 3456A there is a 5 microvolt disagreement in the offset. 

The 100mv range stubbornly refuses to adjust and match the reading of my 3456A, where on 1v and 10v ranges it will.

Do I understand the schematic correctly, that this problem would be related to A70 R2 10K or the A100 R3 100mv cal pot? 

[Kleinstein]

For the overall gain, there seem to be several trimmers to effect it: there is an adjustment for the reference voltage and individual gain for the x1 , x 10 and x 100 gain. However the adjustment for the x 1 gain seem to be marginal - not sure why it is there anyway, as it's essentially equivalent to adjusting the reference. It may more like a fine trim.

So the main gain adjustment for the 100 mV range is with the A100R3.
Trimmers are just notorious for intermittent problems. As these turned out to originate from the source, the trimmers may very well be OK.
Drift over time could be from other resistors (e.g A70R2 or A70R4 = 99.x K).

There is a principle problem with using relays for auto zero: the relays tend to get warm and can create thermal EMF, especially for reed relays.
Another possible point could be too much light getting to some diodes in glass case (e.g. CR1-CR6).

[valley001]

For the overall gain, there seem to be several trimmers to effect it: there is an adjustment for the reference voltage and individual gain for the x1 , x 10 and x 100 gain. However the adjustment for the x 1 gain seem to be marginal - not sure why it is there anyway, as it's essentially equivalent to adjusting the reference. It may more like a fine trim.

So the main gain adjustment for the 100 mV range is with the A100R3.
Trimmers are just notorious for intermittent problems. As these turned out to originate from the source, the trimmers may very well be OK.
Drift over time could be from other resistors (e.g A70R2 or A70R4 = 99.x K).

There is a principle problem with using relays for auto zero: the relays tend to get warm and can create thermal EMF, especially for reed relays.
Another possible point could be too much light getting to some diodes in glass case (e.g. CR1-CR6).

OK, so I am not too far off track maybe.  Could the auto zero relays affect the 100mv range only?

Cr1-CR6 are metal case.

A70R2 and 4 are custom HP wire wound resistors.  The ones I have checked seem to be in spec, I will check R4 specifically. What about caps A70 C3 and C4?

[Kleinstein]

The capacitors should not effect the gain very much. Usually there is quite some tolerance in the capacitors, so that a little drift of the caps should not make much difference. The capacitors would effect the settling after switching - one could check this with the scope.

The Az relays can effect the offset, but they should not have an effect on the gain. To have an effect on the gain it would need really poor timing or settling.

[valley001]

R2 and R4 are in spec. 

I noticed there is a note about adjusting jumpers on A5 if the the pulse transformer or integrator are replaced.  I did keep the A55, 56, and 57 assemblies from the donor unit since there are matched resistors on A55 to the reference A71. . 

Anyway, checking the "A5 gate open delay" as specified in the manual section 5-216, it does not read correctly by 50 counts.  I am wondering if this could be related. 

I guess I am grasping at straws here since I cant find anything grossly wrong that would account for the poor mv calibration. 

[Kleinstein]

The delay mentioned in the 5-216 section should have the same effect on all voltage rages. So if at all it should effect all ranges the same. The only way I could imagine a range dependent effect would be via some coupling via the supply, like current drawn in the ADC part effecting the supply (e.g. +-5 V)  and this somehow effects the amplifier.

A difference of 50 counts sounds like a lot of difference.
Still this would be more an ADC problem, and should this effect all ranges the same. As far as I understand it, the adjustment is to compensate for delay from the pulse transformers (both ways). As I understand it 1 count would be 1 µs or 1/6 µs. So 50 count would be way to much delay.
This would be more like an additional problem that would effect all ranges, especially the ratio mode.

[valley001]

It does sound like a lot since the manual mentions changing by 1 or 2 or 4 counts.  I swapped A5 boards and it shows the same effect with .001v input and 1/60 gate. 

I have not investigated this yet, but the 100 and 1000 volt ranges are grossly off, for example 1v input reads 1.144 on the 100v range.  Perhaps I should move on to figuring out this problem and then see where 100mv is after that.  I had focused on the 100mv range first since I was able to zero 1v and 10v ranges. 

I appreciate your input. 

[Kleinstein]

There is one more area that could cause trouble: the clamping part (input overload level) at the input. Too much leakage there would cause extra input bias current and this would add errors to the offset and also the gain, as leakage can depend on input voltage.
The 100 and 1000 V range have a common trimmer for the divider and than should work like the 1 and 10 V range.
There is always some chance the high values resistors in the divider did drift, or if the boards got dirty, they can be leakage on the boards.  14 % off is way too much for the trimmer, but would be only some  70 M in parallel to the 9.9 M resistance.

[valley001]

OK, so that is A68 R3-5 I will check those.  These was contamination in this unit (not so much on the A68-70 boards) but on some of the relays.  I cleaned it up and not much change, but perhaps I missed some. 

[valley001]

Just an update, R3 on A68 measures 3.65 M instead of 4.95.  I pushed on it and I can get it to bump to 4.95  I guess there is my problem with the 100v and 1000v ranges.  Now to decide if I should transplant an entire A68 board or just move the resistor over, I see "Matched" printed on them.

[Kleinstein]

If available I would swap complete A68 boards. R3/R4/R5 may be a matched for TC and ratio. R3 and R4 may be matched to compensate.  Soldering with those old resistors can also do some change.
  on finding the problem with R3.

[valley001]

The other A68 board is a different revision, differences in connector pin size etc. I ended up taking the risk and moving R3 and R4 over.  I used a heat sink on the leads while soldering.   Now 100 and 1000v ranges seem to calibrate.  Maybe I will move R5 at some point, HP does list R3-5 as a matched set so I think you are correct about tc etc. 

My solution for the 100mv range will be adding an additional A100 R3 in series.  I assume this will introduce instability in that range, but it is the only way I can see to allow it to calibrate.  It is about 20 microvolts low with R3 maxed out and I can not find any obvious reason for this outside possible drift in R2-R4 on A70 but even those seem to measure correctly. 

[Kleinstein]

Adding some series resistance to A100R3 makes absolute sense and should not cause major drift issues, as not much resistance would be needed (e.g. 33 Ohms) and thus only a small fraction of the 100 K in series.

Just in case CR6 or CR7 is leaky one could check the voltage at R6 - this should be rather low (e.g. < 100 mV, preferably less). If much worse one could consider another resistor behind CR3/CR6 to ground.

[valley001]

Voltage at R6 is about 500mv.  Would this indicate a fault in CR1-6? 

[Kleinstein]

500 mV would indicate a relatively high leakage current for CR8 and / or CR7. This could than lead to more than wanted leakage current through CR1-CR6. As a consequence this would lower the gain in the 100 mV range slightly nonlinear.

The kind of simple fix would be 2 additional non critical resistor (e.g. some 10 K ) from between CR2/CR3 to ground and CR5/CR6 to ground. So a second stage to divert the leakage current.

I don't know the quality of the zener diodes, the leakage could be OK, if CR1-CrR6 are really low leakage, but it could explain some of the trouble if the diodes are not very low leakage. 500 mV over 3 diodes may be enough for a little forward current in the 50 nA range.

[valley001]

 CR1-6 specs are 30piv 10pA, I cant find info for CR7/8. 

I will give this a try by clipping in some 10k resistors at those points and see the result. 

Just a note for anyone else who is trying to fix one of these cool meters, if you run into a lot of trouble with switching functions, ranging, or it locks up a lot, you should look carefully at the ribbon cable to the relay channel both top and bottom.  These traces are extremely brittle and break very easily and cause what can be a very frustrating intermittent fault. The fix is to solder a wire from male end of the damaged trace to its termination at the relay coil.  You can test each of these without even removing the cable by checking continuity between the bottom of the digital motherboard to the relay coil connection point.  Sometimes continuity is intermittent so if you have continuity you should lightly depress the cable to see if it looses continuity. 

[valley001]

Before I go clipping things to ground I just want to confirm that I am understanding correctly.  10K between CR2-3 and CR5-6 to ground as in below?

[Kleinstein]

Before I go clipping things to ground I just want to confirm that I am understanding correctly.  10K between CR2-3 and CR5-6 to ground as in below?

(Attachment Link)

This is my suggestion. It should divert most of the leakage current so that the remaining 2 diodes should see only a very low voltage.