FIXED: Weird E3631A power supply problem

[analogRF]

My E3631A is behaving strangely and I cannot quite figure out where I should be looking at for fault(s). The output voltage and currents are all OK and regulated and +/-25V are very accurate, too. The 6V output is off by almost 0.3V and 0.1A but I think it is due to lack of calibration because it is solid and regulated.
However the strange problem is that the readback V/I is totally out of wack (strange large numbers like -40.3V and 8.5A for +/-25 and OVLD and 8.5A for +6V)
When I go into the Limit mode (display mode) the set values are correct and I can change the V and I by knob and the output changes accordingly and accurately BUT when I go back into the readback mode the numbers are out of wack and turning the knob has no effect on the displayed numbers or on the output at all.
The ADC and DAC all seem to be ok and there are signals going into and coming from the display board of course.

Also all the front panel keys work fine except the +6V key! I checked the keypads and they all seem fine and similar. The resistive pads are there but that particular key does not work. Not sure if it related to the above display issue.

Any hint/help is appreciated. The circuitry is very very similar to many other E36xxA power supplies, so if you have had a similar experience with other units that is definitely helpful.

EDIT: the unit passes all the self tests with no errors

[HighVoltage]

Have you tried a calibration?

[analogRF]

Have you tried a calibration?

No I haven't but actually that's what I was also thinking of. Perhaps the EEPROM cal data is corrupted somehow
but why the output is so precise specially for +/-25?

The calibration requires a precision shunt resistor 0.1 ohm 0.01% for current calibration which I dont have and cannot find anywhere

Is it OK to just use the multimeter Amp meter connected to the terminals when calibrating the current?

[Samogon]

It is ok to use ampermeter to calibrate 5A, but it should be 6.5 at best and prolly will enough to use 4.5 dmm
I used fluke 8846A now i use keysight 34461A to calibrate this power supply
But frankly i doubt that it is calibration issue. If cal values wrong it would just sit still on wrong reading. I suspect ADC or sence circuit is funky.

[analogRF]

thanks, good to know I can use my Fluke 289 for the current calibration. It's a relief! I also need to calibrate an E3646Aa I recently repaired and I was wondering if that was possible

I have checked the sense and ADC and they look ok. If I understand correctly, these power supplies use the ADC-CPU-DAC to close the feedback loop of controlling/regulating the current and voltage, so if the ADC is wrong the output must be wrong, too, no?

At first I thought ADC is only used for readback but then I realized that's only part of its job. It's side job rea

EDIT: Problem solved (almost). Big mistake by me  I was completely wrong about the feedback loop on this regulator. See post #9 down below. 

[analogRF]

I also checked the GPIB connection and the readback values of V and I are exactly the same as the front display, completely wrong as I explained before but I can set the voltage and current through GPIB and the outputs work fine and regulated under load.

[wictor]

I have checked the sense and ADC and they look ok. If I understand correctly, these power supplies use the ADC-CPU-DAC to close the feedback loop of controlling/regulating the current and voltage, so if the ADC is wrong the output must be wrong, too, no?

I think ADC is used only for display. I had similar failure where display was showing maximum values, but output was working fine. There are few chips in ADC which often go bad.

[ddcc]

I think ADC is used only for display. I had similar failure where display was showing maximum values, but output was working fine. There are few chips in ADC which often go bad.

I haven't looked at the schematics for this model, but this is definitely true for at least the E3646A. I had a unit that was regulating correctly despite the display being completely off; the problem was a dead +5V voltage reference used by the ADC.

[SrS]

Also check all (green) resistors in the ADC section.

[analogRF]

As I have edited my previous post above, I was wrong about the feedback loop in the linear regulator. I dont know what I was thinking  That method would have been stupid specially with a 12MHz CPU and the loop would have been unstable under a dynamic load.
I had a closer look at the schematics and it is indeed a slightly modified version of the classic linear regulator namely reference voltage->error amplifier with feedback on the output voltage -> driving the pass transistor. However the feedback part is a bit confusing. I have attached the schematics for the +25V rail and you can see U26D does the voltage regulation on its own and only gets its reference from cpu. The only tweak that I still don't know what it is doing is "A" at R87. This voltage is generated by the DAC (see the other schematic). But still it can only contribute to an offset and the feedback loop is closed here and does not depend on the ADC -> CPU -> DAC (this would be a bad design as I mentioned above)

[analogRF]

....so last night I went back to troubleshooting the ADC. All voltages (reference and supplies) were perfect but I noticed U6 is saturated and there is almost 500mV signal on pins 14,15 of U5 while it must be zero. Took a short while to find out that R4 (yes a green 0.1% 25ppm TC resistor) was open circuit. I replaced it with a shitty resistor for now and, voilà  It is fixed! well almost.
All readback voltages and currents are correct (with some small errors and offsets) EXCEPT the +6V voltage reading. With regard to the +6V voltage readback I quickly found out that the optoisolator U15 is not working because the input LED was open, so I need to replace that.

I went about measuring some of the precision 0.1% resistors around the ADC and a couple of other places and some of them are way off like 1-2% so I am planning to replace those as well before doing a calibration.

I can use my 34401A for calibrating almost everything except the 5A for which I must use my Fluke 289. Do you think this would be an acceptable calibration?

[analogRF]

Also check all (green) resistors in the ADC section.

I actually ended up doing exactly that before I saw your post this morning 

[analogRF]

I think ADC is used only for display. I had similar failure where display was showing maximum values, but output was working fine. There are few chips in ADC which often go bad.

yes, that's right. In fact even the CC and CV reference voltages, that are given to the error amplifiers and are generated by the DAC, are derived from the voltage reference inside the DAC chip (AD1851) and not by the LT1021 (ADC ref). So even if ADC ref is dead at least +/-25V lines should work fine. ADC ref is used in biasing the feedback path of optoisolators in the +6V circuit, so if ADC ref is dead, the +6V line may stop working I think.