Agilent E3632A +/- 17.4V rails out 0.6 - 1.2 volts, this a problem?

[jeraymond]

Hello,

I have an Agilent E3632A power supply. I'm troubleshooting an intermittent problem where after turn on, the display turns on, all segments light up then goes black with a single beep. I hit this about 1/10 times turning the unit on. When the unit starts successfully, it seems to function normally. It can drive loads within its specified voltage and current range.

To troubleshoot I was measuring the bias supply voltages. For the +17.4V rail I get a reading +18 volts. For the -17.4 volt rail I get a reading of -18.6V.

Using the Bias Supply schematic as a reference, I tested the capacitors, bridge rectifier and diodes in this part of the circuit and they seemed OK. I suspected perhaps an issue with the resistors feeding into the ADJ pins of the LM317T and LM337T but the matched the values in the datasheet. Might both regulators be going bad or is the voltage being high by up to a volt really not a problem?

Service guide with schematic: https://xdevs.com/doc/HP_Agilent_Keysight/E3632A/pdf/E3632_service_sch.pdf

[alm]

The -+/- 17.4 V supplies are only for the display. I doubt they are that critical. It's telling that the troubleshooting steps don't tell you to measure those rails, but the +/- 15V that are referenced from them. So I would compare the +/- 15V rails with the values in table 5-1 in the manual.

I would also look at the ripple of the rails. Dry electrolytics are a common issue in older equipment, and excessive power supply noise could explain the intermittent operation of the digital circuits.

[jeraymond]

I get about 15 mV RMS ripple on the 4 bias supplies, which is within the input noise of my DP10007 probe in 10X mode. So I think the noise is OK.

I checked the bias supply voltages. All but one were within spec and very close to the center of their range. The +15 V supply measured 14.04 V below the bottom end of the valid range from 14.25 V-15.75 V.

The Zener diode CR9 which feeds the drooping +15 V supply has a temperature around 65-70 degrees C (according to a thermal cam). CR10 which feeds the -15V supply shows around 50 degrees C.

This CR9 temperature seems high to me like something is drawing too much current. My next steps was going to be to try to figure out what that may be.

Does this make sense? Any pointers on what to look out for?

[alm]

Sounds indeed like excessive current is drawing that one rail down. My first guesses would be either a shorted tantalum cap (I don't remember if these supplies have any) or an opamp that's oscillating for some reason. Is there anything else that appears excessively hot on the thermal cam that you wouldn't expect and is connected to the rail?

[jeraymond]

I tested all the tantalum caps on the board and they all seemed OK with no short circuit and they were not hot on the thermal cam.

There are 3 op amps that are some of the warmest things on the board. They are all around 45 degrees C. So not overly hot. I think and since they are all around the same temp, perhaps it's normal. (Op amps are U38, U41 pp 121 Power Circuit and Protection Circuit and U31 pp 123 DAC system).

There was also resistor R112 at 85 degrees C. It is a 200 ohm 2 watt resistor part of the Power Circuit and Power Protection Circuit on page 121. It's fed by the +15V and then goes into Q11 a p-channel MOSFET. I took the MOSFET out of circuit to test it and it checks out OK.

With the MOSFET out of circuit (no current through R112) I retested the +15V supply. It reads +14.44 volts (up from 14.04V). So perhaps the problem is near here. It's not clear to me what the purpose of this part of the circuit is for. Why use the MOSFET to connect the +15V to OUT- in this circuit?



Full circuit in page 121 of https://xdevs.com/doc/HP_Agilent_Keysight/E3632A/pdf/E3632_service_sch.pdf

[alm]

I'd say this is an over voltage protection circuit, though not the programmable one that's accessible from the front panel (I don't remember if this supply has that feature). If the output voltage from the pass transistors is more than 15V+Vth(Q11) - Vf(CR27), Q11 will start conducting, protecting the device under test, and possibly blowing a secondary-side fuse. So I'm suspecting a shorted pass transistor (Q3/4) or a problem in the voltage control circuit.

[jeraymond]

I tested both Q3 and Q4 out of circuit and they are both fine. This unit does have the OVP and OCP on the front panel. Its has a single output with sense terminals on the front as well.

Still scratching my head here. It's so weird. The unit functions normally most of the time, and the only thing I can find out of sorts is the +15V rail sitting at +14V. 

[alm]

What's going on in the circuit when it's malfunctioning? What's the voltage across the output terminals? What's the voltage across CR27? What's the voltage across R112? What's Vgs of Q11? What is the voltage across the pass transistors (Q3/Q4 pin 2 and the other end of R11/R17)?

What are the +15V/-15V rails referenced to? I can't see PCOM connected to any of the rails in schematic sheet A. Is this all a red-herring, and is the control of the pre-regulator failing because of a problem in the logic circuit?

[jeraymond]

These are the voltages I'm seeing when there is a malfunction:

Output terminals: -0.29V (vs 0V when OK)
CR27: 14.00V (vs 14.04V when OK)
R112: 10.17V (vs 10.22V)
Vgs of Q11: 3.8V (vs 3.8V)
Q3/Q4 to end of R11/R17: 42.5V on both (vs 14.8V on both when OK)

I've been using PCOM (e.g. on pin 2 of CR6) to measure the +/- 15V.

This could be a red-herring, and perhaps the issue is elsewhere. Something else I just noticed after turning the unit on/off a bunch to try to reproduce was that sometimes the unit would startup and act like a button was pressed (e.g. OVP menu came up). The issue seems most reproducible when powering on the device after it was just shut off (e.g. the fan in still spinning down and powering it back up).

[alm]

A closer reading of the theory of operations and troubleshooting might help, but to me the problem sounds rather in the digital section. Quite possibly the front panel. It could be either the power supply of these sections or the sections themselves, but that's where I would start looking. The manual has some details about what the different beeps mean etc. I believe a single beep is supposed to identify the start of the self tests? Then apparently it gets stuck there. What's the first self test it would run?

I'm confused. So there's 14V across CR27, which I'm guessing it means it's reverse biased, because a 14V forward voltage would mean a dead diode. How can Vgs then be a positive number? However, ignoring conflicting readings seeing the difference in voltage across the pass transistors, it sounds like in the "bad" case the pre-regulator is set to the maximum output voltage. Since the pre-regulator is controlled from the floating logic, I'm still suspecting a digital problem.

[jeraymond]

I had my leads reversed. Reading -14V across CR27 and Vgs of Q11 is - 3.8V.

I also have an E3631a which shares the same front panel as the E3632a. I tried swapping the front panels from both units. I was able to reproduce the issue on the problem *32 unit with the panel from the *31. So it seems like the front panel is not the issues. (I could not reproduce the problem the other way around with the *32 panel in the donor *31 unit).

I'll dig more into the theory of operation and into the digital section. Thanks.

[alm]

Nice test switching the front panels! I'd say that narrows it down to the digital circuit on the main PCB, or its power supply. Did you also measure the ripple on the digital power rails (floating and earth-referenced)?

[jeraymond]

I did check supply ripple and it all looked clean.

I was doing some more digging online and came across this gem. The +5V bias supply (U11) has its RESET and DELAY pins swapped. So the supply's DELAY pin is connected to the MCU's /RESET pin. Apparently this is a design issue with the E3632A supplies.

I repaired this by swapping the regulator's pins 4 & 5 and also added a 4.7k pull-up resistor to the MCU's reset pin. (Inspiration from IanJ's fix of his E3632A https://youtu.be/gZWbPhIBSgs?t=1011). I also put in a larger 1u capacitor (C22) on the delay pin (was 0.1u) to give the supply more time to stabilize before the MCU starts up.

Things are looking better now. I've been unable to reproduce the issue so far. I'm not totally convinced the issue is solved yet, but fingers crossed that was it. Will test some more and post back if the issue resurfaces.

Thanks for your help alm. Really appreciate it.

[m3vuv]

I once had issues with a lm317 powersupply rail in a hp 54503A scope psu,it was caps had pissed there pants and the electrolyte was conducting to the adj pin of the vreg,replacing the caps and cleaning the board fixed it !,maybe worth a look if one rail is out of spec!,just a thought.

[alm]

Wow, that's odd. There must have been many thousands of these supplies running in the field without any issue. Clearly something happened during the life of this supply that changed the timings slightly, and now it's failing without a functional supervisor. If you have any more problems, it might be worth checking the voltage on the reset pin, this post (linked from one of the comments to that video) suggests that the internal pull-up resistor for the reset pin of the supervisor IC might not be sufficiently low.

[jeraymond]

I once had issues with a lm317 powersupply rail in a hp 54503A scope psu,it was caps had pissed there pants and the electrolyte was conducting to the adj pin of the vreg,replacing the caps and cleaning the board fixed it !,maybe worth a look if one rail is out of spec!,just a thought.

I pulled/tested all the electrolytic caps on the board. No leaks and they tested OK with an LCR meter (low ESR and capacitance was in spec). I also cleaned the board with isopropyl so I think it its OK in this respect. I may still replace them all to be on the safer side.

Wow, that's odd. There must have been many thousands of these supplies running in the field without any issue. Clearly something happened during the life of this supply that changed the timings slightly, and now it's failing without a functional supervisor. If you have any more problems, it might be worth checking the voltage on the reset pin, this post (linked from one of the comments to that video) suggests that the internal pull-up resistor for the reset pin of the supervisor IC might not be sufficiently low.

I did add a 4.7k pull up on the reset pin of the MCU when I rewired things. Thanks for the suggestion.

[m3vuv]

On my hp scope psu pcb as mentioned,in the end i had to grind/burr/dremmel most of the way thru the pcb around the vreg,it seemed the cap juice had leeched into the pcb substrate and made the whole lot slightly conductive,had to rebuild the copper tracks i ground away after with wire,wasnt pretty but fixed it!.mind you the caps had pissed there pants big time.

[jeraymond]

On my hp scope psu pcb as mentioned,in the end i had to grind/burr/dremmel most of the way thru the pcb around the vreg,it seemed the cap juice had leeched into the pcb substrate and made the whole lot slightly conductive,had to rebuild the copper tracks i ground away after with wire,wasnt pretty but fixed it!.mind you the caps had pissed there pants big time.

Beyond the visuals of the capacitor electrolyte on the board, how did you test for the issue? Were you able to measure the conductivity on the top of the PCB with your multimeter?

[SoundTech-LG]

On my hp scope psu pcb as mentioned,in the end i had to grind/burr/dremmel most of the way thru the pcb around the vreg,it seemed the cap juice had leeched into the pcb substrate and made the whole lot slightly conductive,had to rebuild the copper tracks i ground away after with wire,wasnt pretty but fixed it!.mind you the caps had pissed there pants big time.

Yep, those Boschert supplies piss all over the place. I had the exact same issue with that area of the board although nowhere near what you went through. The supply voltage adjust pot got the electrolyte wicked up inside and became intermittent. On another scope Boschert, the transformers wicked up electrolyte and shorted (leaky) primary to secondary. This was blowing the switching transistor. It also seemed like there might have been some slight leakage between primary and secondary on the board itself, but changing out the transformers fixed the issue.

[jeraymond]

I've power cycled the device a couple hundred times by now and no more black screen on startup with single beep (yay!).

I have one remaining issue. Every 20-30ish power-ons the seems to device jump to the wrong section of code after boot. It ends up on the 'None 8 Bits' screen as if I'd navigated via the menus: I/O Config -> RS-232 -> 9600 Baud -> None 8 Bits. Less frequently it ends up in the first Over Voltage menu (Level 32.0 V adjustment screen). In this state I can interact with the device via buttons and Adjust knob, and once I get back to the main screen the device seems to work normally. This isn't a sticky button issue, I reproduces with the original display board and the one from my other power supply.

My thoughts are perhaps the uController, SRAM, or perhaps EEPROM is going bad. Any idea which is more likely cause the issue, or idea on what else it could be?