E3632A DC power supply intermittently jumps to sub-menus on startup

[jeraymond]

My E3632A intermittently presents the odd behavior of jumping to a sub-menu on startup. Normally after power on it briefly displays the GPIB Address then transitions to Output Off which I expect. However, sometimes (maybe 1/20 times) it instead to jump one of the function sub-menus. Sometimes it jumps to 'None 8 Bits' which is 3 levels deep in the I/O Config menu, other times into Level 32.0 V, the first Over Voltage menu. The supply seems to behave normally otherwise, even after the issue presents itself. It also always passes Self Test.

The problem to me seems to on the digital side of things. My first thought was something on the display panel board. However, I tried swapping in the display panels from another supply which works normally, and the issue reproduces only on the problematic E3632A. Next I tried replacing the MCU on the main board (U17), ASIC (U18), RAM (U14), and ROM (U13) and the problem persists.

Anyone seen this kind of issue before or have further idea on troubleshooting?

[m k]

Pretty nasty thing to troubleshoot in any way.

One can expect that the software has a subroutine for all settings.
So memory address is a one possibility but not a connection fault.

Maybe it's not a fault.
Maybe stored I/O or OVP settings are occasionally misread and defaults are used and indicated.
Misreading must of course happen between self test and final startup.

[HighVoltage]

I have seen lots of weird things with these power supplies but that indeed is a really strange problem.

I always start at the basics...
How about all the DC rails and ripple on them or intermittent lapses?

[IanJ]

Hi,

I've repaired a few Agilent PSU's on my YT channel and have seen your exact problem, right down to jumping to "NONE 8 BITS" but I can't remember which one it was (I had a quick look). I am fairly sure I recorded the fix.

See YT channel in my sig.

Ian.

[jeraymond]

I have seen lots of weird things with these power supplies but that indeed is a really strange problem.

I always start at the basics...
How about all the DC rails and ripple on them or intermittent lapses?

I checked the bias supplies. +15V is a bit low at 14.06V outside of the 14.25 to 15.75 range given in the service manual. The rest look OK at +5V/4.96, -15V/-15.12, -5.1V/-4.98 all within range. I don't see any significant ripple on any of the rails.

The +17.4V rail is at 18.10 and -17.4V at -18.42, but I think this is normal. The manual doesn't state a range for these and I've seen them be out this much online.

The 5V regulator had gone bad and I replaced it previously. At the same time also fixing the reversed reset and delay pins which are backwards on the E3632As. (Thanks to Ian's video #066).

Hi,

I've repaired a few Agilent PSU's on my YT channel and have seen your exact problem, right down to jumping to "NONE 8 BITS" but I can't remember which one it was (I had a quick look). I am fairly sure I recorded the fix.

See YT channel in my sig.

Ian.

Thanks for the suggestion. I saw in video #077 you repaired a similar power supply, model E3633A, that did jump to "NONE 8 BITS" and other randomness. The fix was replacing the dodgy microcontroller on the display panel. I don't believe the source of my issue is the same. I swapped the in a front panel from a good unit and the issue still reproduces on the problematic supply. The panel from the problematic supply works normally in the good unit. I've got a bit different timing of the symptom as well. For me, the issue only ever occurs once on power up (every 1/20 times or so), in the video the issue was quite persistent.

Other than this problem on startup the unit seems to behave normally. I've run it for hours at a time without seeing a problem, load tested to its max voltage and current ratings without problems. From the schematic, don't see the low +15V rail used in the floating logic, so I don't think it's related to the issue but I could be off base here.

[m k]

8N for RS-232 is one usual default setting.
If that is what you have there now put there something else if you can, 7 bits and even parity can be one possibility.
Then after the problem appears next time you can be wiser.

I'd say that this kind of a message is very specific and changing front panel MCU as a solution should be kept in mind.
It's also quite possible that panels for both models are pretty equal.
Front panel connection could be one direction to continue.

[jeraymond]

8N for RS-232 is one usual default setting.
If that is what you have there now put there something else if you can, 7 bits and even parity can be one possibility.
Then after the problem appears next time you can be wiser.

I'd say that this kind of a message is very specific and changing front panel MCU as a solution should be kept in mind.
It's also quite possible that panels for both models are pretty equal.
Front panel connection could be one direction to continue.

Good suggestion. I changed "NONE 8 BITS" to "ODD 7 BITS" and when the issue reproduces it now matches "ODD 7 BITS". Sometimes it still shows the Over Voltage menu as it did before as well. I previously swapped the MCU on the front panel and also tried the panel from the other unit so I don't think that is the issue. I tried replacing the RAM, MCU, ASIC and PROM on the main board and the issue still reproduces. So the problem must be somewhere else...

[floobydust]

At this point, could there be a problem with the GPIB controller and optically-isolated serial link, spewing out silly commands?
The front panel does have basic error checking, and upon power-up it looks for a rotary encoder to be present to determine its config for power supplies. Is the encoder OK?
The problem kind of reminds me of 34401a's with phantom key presses, but we never heard back what causes it.

[jeraymond]

At this point, could there be a problem with the GPIB controller and optically-isolated serial link, spewing out silly commands?
The front panel does have basic error checking, and upon power-up it looks for a rotary encoder to be present to determine its config for power supplies. Is the encoder OK?
The problem kind of reminds me of 34401a's with phantom key presses, but we never heard back what causes it.

I think the GPIB is OK. I'm able to connect to the unit over GPIB and issue commands that it responds to. The manual says the RS-232 config is only available from the front panel, so I take this to mean there are no SCPI commands that could be sent to jump to that config menu. I couldn't find any in the manual. The rotary encoder on the front panel works well, device self test passes too w/o issue.

[m k]

Since memory seems to be ok and front panel is serially connected only power and reset are reasonably left.
Reset is directly from main MCU so power has more options.

So, if mainboard connector check up and cleaning is not enough, some signal stability checks during and after power up are needed.
With luck a thin wire can be stuffed into the cable side connector from the back but then it can also strengthen the connection.

[IanJ]

Given the symptoms and what you have tried so far I would look at the supply's, especially the way the front panel derives its own +5vdc.

A noisy/bad supply could be tripping the display micro controller in much the same way it was faulty on on my unit......i.e. sending spurious commands back to the main board.

It all starts with the +/- 18vdc.......check that they are clean, and check that they dont get spurious kicks or dips.......which end up affecting the display +5vdc supply.

NOTE. Some of the E series psu's have the same display board firmware.

Ian.

[free_electron]

if the flash memory is a plcc and in a socket : take it out and resolder the socket. i've seen the sockets go bad on this generation of machinery. same for the AMI asic (if there is one).
other suspects are the FRAM ( if ther eis one ) 24c02 or 04 style but ferroelectric.

[floobydust]

What is the pin 4 and 5 swap on the LM2925? Delay C17 0.1uF on pin 4 should give ~200msec but it is X7R. Any chance C17 has a poor connection now.

For the POR timing of the three MCU's, I assume the main CPU comes up last. But the RS-232/GPIB MCU has a reset delay U1/U8/C1 tied to IG_TXD which might be a watchdog of sorts unless the N83C51FA has it built in.

"*IDN" what F/W rev is all this, ref service note https://www.eevblog.com/forum/repair/looking-for-agilent-34401-88842-io-processor-image/, Front Panel I think 888 13 was the last.

[jeraymond]

Since memory seems to be ok and front panel is serially connected only power and reset are reasonably left.
Reset is directly from main MCU so power has more options.

So, if mainboard connector check up and cleaning is not enough, some signal stability checks during and after power up are needed.
With luck a thin wire can be stuffed into the cable side connector from the back but then it can also strengthen the connection.

I cleaned the connectors and reflowed the solder, connections seem solid.

Other than power, any suggests on what to look out for? I could scope the data pins, but not sure what I'd be looking for other than seeing traffic (which there must be since the display and mainboard MCU are talking).

Given the symptoms and what you have tried so far I would look at the supply's, especially the way the front panel derives its own +5vdc.

A noisy/bad supply could be tripping the display micro controller in much the same way it was faulty on on my unit......i.e. sending spurious commands back to the main board.

It all starts with the +/- 18vdc.......check that they are clean, and check that they dont get spurious kicks or dips.......which end up affecting the display +5vdc supply.

NOTE. Some of the E series psu's have the same display board firmware.

Ian.

I scoped the 5V supply on the mainboard, reset signal that the MCU uses, the -17.4V that goes to the display and the 5V supply on the display board that uses -17.4 as COM. All of these seem to come up normally, are clean and have no spurious spikes. All seems clean when the issue is present and not present.

Seems to me to not be a power issue, but there could be something else I'm missing.

if the flash memory is a plcc and in a socket : take it out and resolder the socket. i've seen the sockets go bad on this generation of machinery. same for the AMI asic (if there is one).
other suspects are the FRAM ( if ther eis one ) 24c02 or 04 style but ferroelectric.

I've previously replaced the SRAM (U14) on the mainboard with IS62C256AL-45ULI it's soldered down. The MCU (U17), ASIC (U18) and EPROM (U13) were all soldered in. I've since socketed them and tried replacing each of these chips with no changes in behavior.

There's an EEPROM soldered in (U15) which I haven't touched. I'm guessing it's used to store changes in settings. Could it make sense it's causing a problem?

What is the pin 4 and 5 swap on the LM2925? Delay C17 0.1uF on pin 4 should give ~200msec but it is X7R. Any chance C17 has a poor connection now.

For the POR timing of the three MCU's, I assume the main CPU comes up last. But the RS-232/GPIB MCU has a reset delay U1/U8/C1 tied to IG_TXD which might be a watchdog of sorts unless the N83C51FA has it built in.

"*IDN" what F/W rev is all this, ref service note https://www.eevblog.com/forum/repair/looking-for-agilent-34401-88842-io-processor-image/, Front Panel I think 888 13 was the last.

Yup I replaced the LM2925 and swapped pins 4 & 5 which are backwards on the actual board & schematic. I also replaced the delay capacitor (C22, is this what you mean by C17?). I scoped it and the 5V rail and see the 5V rail coming up and being stable before the reset is enabled and the MCU starts. I played around here a bit before and temporarily increased the delay having the MCU come up much later and still was able to reproduce this intermittent issue.

I'm not following the info about the RS-232/GPIB reset delay. Are you referring to the items on the earth referenced logic section in the schematic (http://ridl.cfd.rit.edu/products/manuals/Agilent/power%20supplies/CD1/Service/E3632ser.pdf)?

This is the firmware info: HEWLETT-PACKARD,E3632A,0,1.4-5.0-1.0

[floobydust]

You've replaced all the usual suspects. Yes C22 reset delay cap. I was using the E3633/3644 schematic I had.
The EEPROM U15 93C66 stores calibration constants and store/recall values in it. It could be going senile   If you've upgraded the main F/W the Service Note advised to do a recalibration.
I can't find out which Front Panel F/W is compatible with which Main Processor F/W.
I thought maybe the GPIB MCU (earth-ref logic) is late coming up out of reset and/or spitting out junk to the main processor or an opto U7is noisy.
It would be good to narrow down where the commands are coming from. About how many phantom button presses are needed to go to the menu?

You can try a heat gun on components too, I would heat it up a bit and see if that aggravates things. Or it's a software problem.

[m k]

I guess I'd go back to that -18.4V and its doings.

Data out levels from panel are relatively referenced but data in and clock are different.
Maybe this and something else combined will do the trick.