Author Topic: Seeking help in troubleshooting Systron Donner M107 Precision DC Voltage Source  (Read 5658 times)

0 Members and 2 Guests are viewing this topic.

Offline intabitsTopic starter

  • Frequent Contributor
  • **
  • Posts: 319
  • Country: au
I have a faulty Systron Donner M107 Precision DC Voltage Source, producing little to no output. I believe I have found the main fault area, but before I start fooling around in there, I'd like any advice on likely culprits.

Brief overview:-
The digital section encodes the 3 most significant digits (MSD), and the 3 least significant digits (LSD) of the programmed voltage setting into two independent (though synchronized) PWM signals, each having a duty cycle range of 0 to 999, and a frequency of 1.15KHz.

These signals are passed to the analog section via two pulse transformers driven by U20, and the PWM signals reconstituted by discrete flip-flops Q1/2 and Q3/4.

I believe the problem is that these flip-flops do not flip (i.e. they are flops ;D).

The outputs of U20 appear correct, as their duty cycles vary as expected with changes to the digit settings on the front panel. All power supplies are correct.
I'm not sure what the output of the pulse transformers should look like, so I don't know if they are bad, or the flip-flops have a problem. (or if I'm barking up the completely wrong tree here)

Further operation, of academic interest only, while the above problem remains:-
The PWM signals from Q1-Q4 are used to gate the LM299's 7v reference voltage via FETs Q5-Q8 into a summing junction at R22+R23, which is then filtered by C15-C20 et al, to provide the control voltage to the output amplifier. The LSD signals going into the summing junction have resistor values 1000 times higher than those for the MSD signals, so that the MSD PWM value contributes 1000 times as much as the LSD signal to the control voltage.

Section of digital schematic, showing U20, the pulse transformer driver.


Section of analog schematic, showing the flip-flops, voltage reference, gating circuitry, summing junction and filter.


Photo of the above section of the analog board, showing the pulse transformers on the plate at the bottom.
The opto-couplers below them send range and polarity info to the analog section.


Parts overlay for the above.


The top trace in the following CRO shots is the output at U20-6 (MSD) used as a sync/reference.
(The analog ground has been tied to digital ground/chassis earth to ease probing about) 

The output of the MSD pulse transformer (LSD is similar, but shifted to the right a little)


The collector of Q1, turned off briefly and trying to float up to +7v via R13, before it turns on again.


The collector of Q2, turned on briefly, pulling down from 0v to -10.5V. (All the emitters are at around -10.8v, due to VR1)


The collector of Q3, turned on briefly, pulling down from +7v via R26.
(Why isn't R13 pulling Q1 up to +7v even harder, since R13 is a lower value than R26?)


The collector of Q4, staying on at -10v.


The transistors I guess are prime suspects, but could all 4 be failed?
And since both FF's are not working, one would expect a common part to be at fault, but what?
Could failed FETs be dragging the FF's down?

I was hoping that putting this post together might help me see the problem, but no light bulbs here...
Any insights or advice would be greatly appreciated...

BTW: The service manual can be found at:-
https://elektrotanya.com/systron_donner_m107_precision_dc_voltage_source.pdf/download.html


 
« Last Edit: November 30, 2017, 10:06:55 pm by intabits »
 

Offline intabitsTopic starter

  • Frequent Contributor
  • **
  • Posts: 319
  • Country: au

 :(

In the absence of any other ideas or suggestions, I went with my gut feel and replaced Q1 & Q2 (PN4275's) with PH2369's.

The old parts tested OK, though one had a lower gain (of 60 vs 100), so unsurprisingly, this made no difference.
I also touched up a few slightly sketchy looking joints on U9, but also no joy.

I've looked at every voltage in this section around the flip-flops, and all are as I'd expect them to be.
Which pretty much puts me at my wit's end on this.

I'm starting to wonder if I should try bypassing these pulse transformers entirely with some high speed opto couplers such as HCPL-2601 (10MB/s) to drive the flip-flops. 

And since I've tied the analog section ground to chassis/digital ground while troubleshooting, I might be able to just test the idea with some capacitors/resistors/diodes to get the signals from the digital side to toggle the analog side flip-flops.

Does anyone have any thoughts on that idea?
 

Offline intabitsTopic starter

  • Frequent Contributor
  • **
  • Posts: 319
  • Country: au
Something like this.



This simulation worked as expected using the L1+L2 pulse transformer, and also after disconnecting it with "<SNIP>", and then bypassing it with the 12v zeners and resistors below it.

I'll give this a shot tomorrow...

(Is anybody actually interested in this? Are updates to this thread wanted?)
« Last Edit: December 02, 2017, 08:47:01 am by intabits »
 

Offline Ash

  • Regular Contributor
  • *
  • Posts: 161
  • Country: au
Hi Intabits,

Interested yes, able to contribute much.. dunno.. :)

Is there anything happening on the DC power rails on the analogue side? have you scoped them? You said that the voltages were correct, but do they stay correct while switching? There are a lot of tantalum caps in there by the looks..  :popcorn:

Good luck. Stick with it, you'll learn a lot getting it going. I know I did with my EDC 521 reference.

Ash.
 
The following users thanked this post: intabits

Online floobydust

  • Super Contributor
  • ***
  • Posts: 6924
  • Country: ca
I don't see anything wrong with the pulse receive section.
Amplitude at Q3 is about 1/2 that of Q1 as Q4 just stays on, Q3 is not pulling it low enough. Q4 might have failed short.

I think the problem is what's coming into it. This whole thing should be getting a gated HF burst, not just an edge, to generate control pulses for the reference modulation.
I would look at the pulse transformer waveforms, as the PWM carrier seems to be missing.
The schematic is clipped off, I think optos pulse transformers  -> U3, U4? -> U16 CD4042 and U14 driver; but you have U20?
 

Offline intabitsTopic starter

  • Frequent Contributor
  • **
  • Posts: 319
  • Country: au
Thanks for looking at this.

I don't see anything wrong with the pulse receive section.
Yes, me either. That's why I'm pulling my hair out and thinking about radical things like eliminating the transformers.

Quote
Amplitude at Q3 is about 1/2 that of Q1 as Q4 just stays on, Q3 is not pulling it low enough. Q4 might have failed short.
Yes, I also think that Q4 may be flaky, though I have seen a signal on it, so probably not a complete short.
Both Q1 and Q3 are pulled towards 0V by R9 & R28, but also pulled towards +7V by R13 & R26.
And on the schematic, R26 is 18K vs R13=3K9, so Q1 should be pulled up higher than Q3.
Trouble is, on the actual board, R13 and R26 are both 3k9, so the Q1 & Q3 should be the same (?!)
But I wasn't going to worry too much about the LSD flip-flop, until I got some joy with the MSD one.

Quote
I think the problem is what's coming into it. This whole thing should be getting a gated HF burst, not just an edge, to generate control pulses for the reference modulation. I would look at the pulse transformer waveforms, as the PWM carrier seems to be missing.
Are you saying that from the general principles of how these things should work, or from looking at this specific circuit?
Because I don't see how that would come about. (The five FF's in U4,5,6 ar not toggling at HF, if that's what you mean. They just form a 5 pulse delay line at the start of each 1.15KHz PWM cycle)
Of course, I could have that all wrong, but that's what it looks like to me,

The transformer inputs (eg. from U20-6, the top trace in the pics above) are nice PWM signals that vary in width according to the value selected on the front panel, so I'm fairly sure they are not the problem.

Quote
The schematic is clipped off, I think optos pulse transformers  -> U3, U4? -> U16 CD4042 and U14 driver; but you have U20?
Not sure what you mean there. Are you looking at the manual P7-21? That is the mother board A4, which has optos U2,U3,U4 to control the relays.
The U20 I refer to is on P7-19, the Digital control board A3 that plugs into the mother board. U20 is a 7404 that drives the transformers via 270R resistors, which go to the mother board via A3 pins 14 & 9, then to the PT's via A4 pins E1-E4.

BTW, from the A3 schematic it would seem that U20 drives both ends of the transformer primaries via R18-R20, but in fact R19 & R20 are not used. The primary windings go from R18/R21 to +5V, as shown on the A4 schematic.

 

Offline intabitsTopic starter

  • Frequent Contributor
  • **
  • Posts: 319
  • Country: au
Is there anything happening on the DC power rails on the analogue side? have you scoped them? You said that the voltages were correct, but do they stay correct while switching? There are a lot of tantalum caps in there by the looks..  :popcorn:
Just scoped them now.
All very steady with noise at 20mV P-P  & 5mV RMS.

Thanks for the encouragement...
 

Online floobydust

  • Super Contributor
  • ***
  • Posts: 6924
  • Country: ca
The service manual has portions of the right side of the page missing only in Firefox 53.0 pdf viewer, so I got lost.

If you suspect the pulse amplitude is low on T2 /LSB, then I would measure and compare the primary drive on both transformers, you can also scope the transformer's current with the 270R resistors. That might help rule out the transformers. Seems unlikely to me they are at fault.

I see the 15pF C3, C13 which is quite small so think that it's not a blip triggering the one-shots, but instead is a pulse train, so you can toggle LSB, MSB for many microseconds. Are you thinking the 1usec pulse is working the analog board?

I think a clock is missing upstream. I see a four-phase clock, stretched, so a bit of work to figure out what U3 should be outputting. I would look at the signals there. I keep thinking U7 outputs a PW-gated burst, instead of a simple edge, maybe from U8.

If you think I'm out to lunch, that's ok. Not easy to figure out.
 
The following users thanked this post: guenthert

Offline intabitsTopic starter

  • Frequent Contributor
  • **
  • Posts: 319
  • Country: au
There has been some developments, but first I'll respond to your previous:-

If you suspect the pulse amplitude is low on T2 /LSB, then I would measure and compare the primary drive on both transformers, you can also scope the transformer's current with the 270R resistors. That might help rule out the transformers. Seems unlikely to me they are at fault.
Both transformers have identical size and shaped waveforms on both input and output, so I don't think there is a problem with them,
I just can't find why they aren't toggling the analog side flip-flops.

Quote
I see the 15pF C3, C13 which is quite small so think that it's not a blip triggering the one-shots, but instead is a pulse train, so you can toggle LSB, MSB for many microseconds. Are you thinking the 1usec pulse is working the analog board?

I think a clock is missing upstream. I see a four-phase clock, stretched, so a bit of work to figure out what U3 should be outputting. I would look at the signals there. I keep thinking U7 outputs a PW-gated burst, instead of a simple edge, maybe from U8.
I had avoided digging into the rather confusing details of the logic operation, but your gated burst suggestion has caused me to take a closer look. It appears that the D-type flipflops U5-5,U5-9,U6-5,U4-5,U4-9 form a shift register that is initially cleared by a pulse from one-shot U1-4, which is fired by the BCD counters U9-11 rolling over from 999 to 000 at at the start of each PWM cycle.
The shift register FF's are clocked in unison at 1.13MHz (except for an out of phase kink in the clock to U6, which I'm not sure about).  The first FF always a clocks in a 1, and this is propagated through the chain with each clock pulse, and outputs along the way are used to gate Various events, which again, I don't fully understand. I suspect it's all about avoiding possible logic glitches, and ensuring that the MSD and LSD outputs are slightly out of phase.

All this is described on pages 4-12 to 4-12 of the manual, but I still struggle to grasp the complete picture in detail. (I would like to get a logic simulator to dig into the detailed operation of this)

But to produce a gated burst of high frequency pulses, U7 would have to be rapidly cleared then clocked, which would be a strange way to accomplish that, and I don't see how their clock inputs could do it.
Paragraph 4-16 on page 4-2 says: "Pulse transformers T2 and T3 couple the start and stop transitions of the duty cycle controlled waveforms...",
which seems to indicate simple edges rather than pulse bursts.

So yes, I think the Q1/Q2 and Q3/Q4 flip-flops on the analog side are supposed to be set and reset by the negative and positive going pulses from the transformers, although as you say, the 15pf caps do seem too small for this. But it does work in the LTSpice simulation above. (I have tried adding 22pf in parallel, but no change)

New developments:-
Before trying any transformer bypass ideas, I thought it best to replace Q3&Q4 as well, especially since Q4 was suspect. But it tests OK, and has a gain of 83 (vs Q3=59).

And with the new Q3/4, the LSD flip-flop started toggling correctly!!! (so toggling on edges is confirmed)

But the Q4 problem had moved to Q1, that is, the collector was only showing a small glitch instead of at least trying to switch with a full but brief pulse, like the other 3 transistors.

And then a few minutes later, it stopped!
Back to the original problem exactly. No LSD FF toggling, Q1 brief pulsing, Q4 only glitching.

Hmmm.  (heat? PCB hairline crack? some other intermittency?)
The glitch switch from Q4 to Q1 and back again, plus the ongoing non-operation of both flipflops indicates that something  they have in common is playing up. Wondering about that now...
 

Online floobydust

  • Super Contributor
  • ***
  • Posts: 6924
  • Country: ca
Is the voltage reference output reasonable, at U7? Guessing 6.95V at U8 LM299 and 13.90V output at U7. C4,C1,C3 may be old tantalums.
Because the two bistables connect to it at junction R13, R26.  C4,C1,C3 may be old tantalums.

Other than that, the -15V rail and VR1 are the only other common aspect I see. Unless a JFET switch is leaky, like Q7.
Perhaps checking downstream of the two bistables to see if their load is OK.


 

Offline intabitsTopic starter

  • Frequent Contributor
  • **
  • Posts: 319
  • Country: au
Yes, Voltage from the LM299 is spot on at 6.95V, and also out of U7 to R13+R26. (the 13.9V from U9 (not U7) is also correct, but it is used only to provide constant current (38mA) to the LM299 zener)
Voltages across all the tantalums in these circuits appear fine, and I have seen nothing to cast doubt on any of them.

Quote
Unless a JFET switch is leaky, like Q7. Perhaps checking downstream of the two bistables to see if their load is OK.

Yes, I think that's where to go next, remove the JFETs (I have replacements)
That'll be the final kiss goodbye to any hope of preserving any state of calibration...

Thank's for your ongoing interest and help.
 

Offline intabitsTopic starter

  • Frequent Contributor
  • **
  • Posts: 319
  • Country: au
Systron Donner M107 Precision DC Voltage Source - Major Progress
« Reply #11 on: December 05, 2017, 07:30:15 am »
 ;D

Remove/Replace the FETs:-
Remove Q6 -> Tested bad. MSD bistable oscantoggulates for the first time I've seen! and stayed good!
Remove Q7 -> Tests bad. But LSD bistable still not working.
Remove Q8 -> Tests bad. And LSD bistable now shows togosculations, and keeps doing so!
Remove Q5 -> Tested Good, as expected.

Install 4 new 2N4091 JFETs:-
-> Both bistables triggering properly.
-> Nice smooth level on R41 at end of filter chain, changes with front panel selections
-> Output terminals have voltages now, though wrong by 20-30%   
-> Ranges mostly working. Any given (but wrong) output voltage is scaled by exactly 10 when next range above is selected.
    Except 1000v range which is way off: 200V selected comes out as 630V   

So it would seem that (hopefully) only a calibration is now needed, except for the 1000v range which obviously still has problems, probably on the HV amplifier card.
Will do some more detailed testing, and look at the calibration procedure to determine if all other than the HV is fixed.

Thanks floobydust for your suggestions, which kept me prodding away at it.


 

Online floobydust

  • Super Contributor
  • ***
  • Posts: 6924
  • Country: ca
That's good news  :) discrete circuits are always a bit of challenge. I still don't see where the pulse-width (framing) is done, it's edges only unless a FF is done with the JFETs.

They want most of the CAL done on the highest V (1kV) range.
If the HV amp is not working then a bunch of cal would not adjust properly. A5/U6 has a 0-7V output so you can check what the commanded output voltage is.

With 1,500V on the A6 board, this is a little fun  :-/O
As Q5-Q10 2N3439 are supposed to split up the voltage drop, if one shorts it can cause the others to fail, in that series string. I kind of go after the high-failure rate parts before trying CAL. I would just pull the board and check for shorts on the HV pass transistors and drivers, diodes. Takes a couple minutes with a multimeter.
 

Offline intabitsTopic starter

  • Frequent Contributor
  • **
  • Posts: 319
  • Country: au
I still don't see where the pulse-width (framing) is done, it's edges only unless a FF is done with the JFETs.

You do seem to have been laboring under some misconception on this all along, I'll try to clarify the operation. It goes something like this:-

On the digital side, a three digit BCD counter is clocked at 1.15MHz, counting from 000 to 999 and then overflowing back to 000. So it divides the clock by 1000, giving the PWM frequency of 1.15KHz. The pulse transformer primary is energized at the 000 count, and de-energized when the BCD counter matches the programmed value on the front panel.

So the primary current is a PWM square wave with a frequency of 1.15KHz and a duty cycle equal to PRG/1000, where PRG is the three most significant, or three least significant digits of the programmed value. The counter value is compared to both the MSD and LSD simultaneously, so we have two semi-independent PWM signals, and two transformers, etc, etc.

On the analog side, when the primary current starts at count 000, the secondary produces a negative going pulse that resets the bistable. And when the programmed count is reached, the primary current stopping causes a positive going pulse that sets the bistable, thus recreating the original PWM signals, but isolated by the transformers. (as shown in the first oscillograph above, which is for count=0, and zoomed in enormously to see the secondary pulse)

These PWMs gate the reference voltage via the FETs into the summing junction, but the MSD contributes 1000 times more than the LSD signal. The resulting "lumpy" control voltage is filtered to a smooth DC level for the output amplifier. (You can see the 50% duty cycle for a count of 500 in the next two pictures)

Back to the initial success, this is the erroneous output I was getting at first, after replacing the FETs.


And then scaled by 10 on the next range up.


Quote
With 1,500V on the A6 board, this is a little fun  :-/O
As Q5-Q10 2N3439 are supposed to split up the voltage drop, if one shorts it can cause the others to fail, in that series string. I kind of go after the high-failure rate parts before trying CAL. I would just pull the board and check for shorts on the HV pass transistors and drivers, diodes. Takes a couple minutes with a multimeter.

I have followed your advice to test the transistors on the HV board first. No obvious shorts found. Yes, the high voltage is scary, especially now that I have removed the many layers of foam insulation I had taped over it while I probed around nearby. I'm also really glad that the (probably unobtainable) U6, Analog Devices 52K module is working. That would have stopped this project dead.

This is the HV Amp board


But after checking out the HV AMP pcb, I reinstalled it to do some more testing, and it started outputting very precise voltages:-


That's accurate as far as I trust the calibration on that meter.
So wow, looking real good. And that's with the external covers removed, the internal covers removed, the shielding cans over the analog sections removed, no warmup, and no re-calibration after changing 8 transistors!

The 1000v range is still rubbish though, even for 1V programmed, it outputs 20V.  Which makes me suspect something in the range switching & feedback section rather than the A6 board, since that is obviously capable of producing 1V when properly commanded to.
« Last Edit: December 06, 2017, 07:01:12 am by intabits »
 

Offline intabitsTopic starter

  • Frequent Contributor
  • **
  • Posts: 319
  • Country: au
On the A6 High Voltage amplifier board, I removed all the transistors in the 1500V string (Q1 & Q5-10), and all tested good.
I tested all the other transistors and diodes in circuit, and they all seem to transist and diodate. But many of the high value 0.5W resistors were high by more than their 5% tolerance, so I'll replace them. (Especially the 180K resistors R20-R25 in the 1500V string, which were all 195K-210K) However, I feel that this board is mostly operating correctly, and that there is something wrong in the feedback network for the 1000v range, since that is the only section where that range uses different circuitry, switched via the relays K1 & K2.

I re-seated these relays - no change. Swapped them - no change. Then I noticed that the sockets for these were very poorly soldered, and it appears that they have been removed and replaced (why?!), and there are obvious scratchings on the PCB between the pins on these.

Not only is the output voltage very wrong on the 1000v range, but it slowly wanders up and down over time. This, together with the relay socket removal, makes me now suspect that there might be a break in the feedback loop, possibly a torn trace on the top side (now underneath the sockets). This will be a pain to checkout as the wiring is quite convoluted in this area, but hopefully, there is a break that can be fixed with a simple bodge wire.
 

Online floobydust

  • Super Contributor
  • ***
  • Posts: 6924
  • Country: ca
Carbon resistors do change value with age, humidity so I would expect the 10% drift.

The HV amp voltage feedback comes back through the A6 edge-card connector and then the two relays; so a long run and direct HV. The PCB, edge conn and relays would have to be clean, no flux. Someone else has been at this before, until they slipped and killed the JFETS lol. I'd look for any signs of dirt, carbon tracking etc.

I would put an ohmmeter on the feedback line, from HV out through to the relays. I think you can do this with power off. Just to see if continuity is good. Also check leakage resistance, I think you will read 1M from RN1, and can remove A6 so it doesn't interfere with readings. What is C44 10uF doing on that line. It can do remote-sense on the HV ?

The feedback line might not have an intermittent high series-resistance.
Instead, a protective clamp diode, or something else got hit with HV when it shouldn't have.
Imagine K1 or K2 malfunctioning; the HV could hit the 1V adjust! Ow.
 

Offline intabitsTopic starter

  • Frequent Contributor
  • **
  • Posts: 319
  • Country: au
Carbon resistors do change value with age, humidity so I would expect the 10% drift.

I've replaced R20-R25 (180K), R14 (220K), and R15-R16 (430K) with 2x220K in series, all with Philips/Vishay metal film resistors.

My pump action desoldering iron can be rather brutal on PCB copper, especially just pads with no attached traces, which it just removes, possible along with the PTH.
So when reinstalling the the TO-39s (Q5-Q10), I had to solder to the top side traces on those legs that had the bottom pad removed. They have a 2mm spacer that just allows the iron to get in there, but care must be taken that the solder doesn't touch the underside of the case (which is also one of the terminals).
I must have got it right as it worked as before.

Quote
The HV amp voltage feedback comes back through the A6 edge-card connector and then the two relays; so a long run and direct HV. The PCB, edge conn and relays would have to be clean, no flux. Someone else has been at this before, until they slipped and killed the JFETS lol. I'd look for any signs of dirt, carbon tracking etc.
All very clean, no obvious damage apart from the scratches around the relay pins.

Quote
I would put an ohmmeter on the feedback line, from HV out through to the relays. I think you can do this with power off. Just to see if continuity is good. Also check leakage resistance, I think you will read 1M from RN1, and can remove A6 so it doesn't interfere with readings. What is C44 10uF doing on that line. It can do remote-sense on the HV ?
The actual traces on the PCB do not match with the relay pin numbers on the schematic, presumably they found the layout easier to do if they chose different sets of contacts for this revision of the analog motherboard.
I powered K1,K2 and K4 with an external supply to select the 1000V range, and all front panel terminals connect to what they should, so no (obvious) faults around the relay wiring.

Quote
The feedback line might not have an intermittent high series-resistance.
Instead, a protective clamp diode, or something else got hit with HV when it shouldn't have.
Imagine K1 or K2 malfunctioning; the HV could hit the 1V adjust! Ow.
I tested (roughly, in circuit with a little multifunction tester) every part involved with the output section in the redrawn schematic below, and could not find anything bad.
That's not to say they don't have a problem when higher voltages are applied, but even selecting less than 10V on the high range gives junk output.
The only thing that differentiates the 10v range from the 1000v range is which pins on the RN1 feedback network the output voltage is connected to by the relays. So it's scary what could happen to attached persons or equipment if the unit has a brain fart.

The complete output section (translated from the original convoliberrish)


For the moment, I've given up on the 1000v range, and put all the shields and covers back on.
Here it is with a couple of DMMs, after operating for 30 hours.


The readings were out by about 500uV initially, but have become better now. Which is pretty bloody good considering its age, work done on it, and no recent calibration. But it still jumps about by a few 10s of uV. I had hoped that even if wrong, it would at least be stable. Sometime in the next few weeks, I'll recalibrate it to agree with the HP3456A, and see how it drifts from there.

Thanks for your interest and help.
« Last Edit: December 10, 2017, 02:09:01 pm by intabits »
 

Offline intabitsTopic starter

  • Frequent Contributor
  • **
  • Posts: 319
  • Country: au
Update:
Over the next few days the output starting drifting, a lot. 10.0000V setting became 10.2V at one stage. Bouncing the power restored it to more like 10.001V, but next time I looked, it had slipped its moorings completely, putting out -4.somethingV. So I turned everything off for the day.

But worryingly, a few hours later, when I tried to see if the problems was still there, the HP3456A seems to have been affected. It now makes a buzzing/chirping sound (like a cicada), shows no voltage reading, and briefly says "Add020 0" at startup (not sure if that is normal or an error message, I haven't looked into it yet)

Back to the M107, I analyzed the net result of the range switching, and the ONLY difference between the 100V (which did work) and the 1000V (which never did) is that the 100V range shorts the feedback resistor network RN1 pin 8 to pin 5 (via a trimmer), and this is open on 1000V, which causes 10:1 change in divider resistance as might be expected. Similar happens for the 10V range.
I think I will need to have a closer look at error amplifier output control voltage to the HVAmp board in response to selected output value and ranges. And maybe I could inject an external control voltage to check correct operation of the HVAmp. So it's back to the drawing board on this one,  but at least I now have a much better understanding of how it works, and I have some things to try.

So now I have two units to fix, but can't spend much time on them just yet. Have to do some proper work...
 

Offline Vgkid

  • Super Contributor
  • ***
  • Posts: 2710
  • Country: us
For the 3456a replace ALL the caps in the power supply , and check the big cap on the inboard/outboard???(not sure which one) for ripple. When the caps started going , thevrelays cycled nonstop (not good) , and before that would throw errors at random.
If you own any North Hills Electronics gear, message me. L&N Fan
 
The following users thanked this post: intabits

Offline kerryb33

  • Newbie
  • Posts: 1
  • Country: au
Update:
Over the next few days the output starting drifting, a lot. 10.0000V setting became 10.2V at one stage. Bouncing the power restored it to more like 10.001V, but next time I looked, it had slipped its moorings completely, putting out -4.somethingV. So I turned everything off for the day.

But worryingly, a few hours later, when I tried to see if the problems was still there, the HP3456A seems to have been affected. It now makes a buzzing/chirping sound (like a cicada), shows no voltage reading, and briefly says "Add020 0" at startup (not sure if that is normal or an error message, I haven't looked into it yet)

Back to the M107, I analyzed the net result of the range switching, and the ONLY difference between the 100V (which did work) and the 1000V (which never did) is that the 100V range shorts the feedback resistor network RN1 pin 8 to pin 5 (via a trimmer), and this is open on 1000V, which causes 10:1 change in divider resistance as might be expected. Similar happens for the 10V range.
I think I will need to have a closer look at error amplifier output control voltage to the HVAmp board in response to selected output value and ranges. And maybe I could inject an external control voltage to check correct operation of the HVAmp. So it's back to the drawing board on this one,  but at least I now have a much better understanding of how it works, and I have some things to try.

So now I have two units to fix, but can't spend much time on them just yet. Have to do some proper work...
Hello intabits

Sorry to be coming into this one so late, but I was wondering if you eventually managed to fix your M107.  I've also got one of these and would be interested in knowing the solution for future reference.  My M107 is currently in working condition, so if it would help to make any comparison measurements related to your 1000v range problem, please let me know. 
 
The following users thanked this post: intabits

Offline intabitsTopic starter

  • Frequent Contributor
  • **
  • Posts: 319
  • Country: au
Hi kerryb33,

No I haven't got back to it yet, since the HP3456A problem made the overall job somewhat larger and more complicated, and I have to do other things first.
(BTW: The HP has come good again, all by itself. But I want to go over it anyway, before leaving on for long. I'll be checking for obvious problems, replacing bad caps, backing up the ROM etc)

I know that some of my re-soldering of the transistors on the HV amp pcb was dodgy, and may have failed. If the FET PWM outputs are still good, the HV amp soldering will be my main suspect.

...if it would help to make any comparison measurements related to your 1000v range problem, please let me know.
Thanks. That could be a big help. I'll keep it in mind.
 

Online floobydust

  • Super Contributor
  • ***
  • Posts: 6924
  • Country: ca
I was left wondering if the 1,000V relay's insulation and clearances could actually take it. It seems to be a lower voltage part. If you switched that relay with HV, then I could see it getting damaged.
For the intermittent wandering and such, I would take a plastic (pen) and tap things to rule out a bad connection.
 

Offline intabitsTopic starter

  • Frequent Contributor
  • **
  • Posts: 319
  • Country: au
(sorry, meant to reply earlier)
IIRC, the relays are not shown in the parts list, or the manufacturer number was missing, or something, (the pulse transformers also). But I could find similar based on the relay markings (I don't have them at hand right now).
I was just wanting the pinouts, not the specs, so didn't look into that. But you raise an interesting point, though I hope/presume that SD took voltage ratings into account during design.
I did swap the relays around while trying to sus out the problem, and that had no effect. But when I return to the repair, I'll try tapping things to uncover any dodgy connections.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf