Video Tearown & Repair of an Agilent E3631A 6V/25V 80W Triple Power Supply

[Hugoneus]

In this episode Shahriar investigates a faulty Agilent E3631A 80W triple output power supply. The instrument powers on, display a fully illuminated VFD display before a long beep and dark screen. The disassembled unit reveals a two board construction. After an unfortunate finger burning experience, the fault is traced to pair of extremely hot voltage regulators and Zener diodes. The temperature of these components is measured using a thermal camera and reach 400C! The schematic of the power supply is examined and the appropriate components are replaced. The performance of the unit is verified after the repair.

Watch the video here: [37 Minutes]
youtu.be/vbjSWxWXurE

More videos at The Signal Path:
http://www.TheSignalPath.com

[free_electron]

replace the tantalum right next to the zener as wel as the two little tantalums about 3 centimtere left of the zener ( right next to the TO220 ). i've seen those blow up many a time. in newer versions of the supply those are ceramics.

the rotary encoders in these machines all die. ( same for 33120 6632 and all these other supplies using the same ALPS encoder. those things are pure shit. The geardisk oxidates .  i typically replace them by default.  if you can;t get them : simply bend back the top clamps  , remove the shaft, wash in a bit of alcohol, dip in a bit of deoxit and reassemble. avoid spraying deoxit directly on the board as the spray will leave greasy splotches everywhere ( on the display, under the membrane keys ). i simply spray a bit of deoxit in a bottle cap and dip the encoder disk in .

[bktemp]

Great repair job! You have been really lucky with your recently bought devices. A failed regulator and a vaporized ground plane, both easy to repair.
After seeing the hot regulator I totally expected some failed logic ics shorting the 5V rail, but was surprised when only the regulator and zener diode had failed.
There are 2 fuseable resistors before the bridge rectifier. Why was one significantly hotter than the other one? Did you check the other one if it is still ok? From the schematic both should share the equal amount of current and should therefore be at the same temperature.

[HighVoltage]

What a great repair video, thanks.

I bought one of these "used and broken" last winter and all it had bad was the encoder.
It seems these encoders have a design flaw, since many of my Agilent gear had them fail.
Even in one nice signal generator.
The good thing is, Agilent used almost only one type of encoders.

[Hugoneus]

replace the tantalum right next to the zener as wel as the two little tantalums about 3 centimtere left of the zener ( right next to the TO220 ). i've seen those blow up many a time. in newer versions of the supply those are ceramics.
the rotary encoders in these machines all die. ( same for 33120 6632 and all these other supplies using the same ALPS encoder. those things are pure shit. The geardisk oxidates .  i typically replace them by default.  if you can;t get them : simply bend back the top clamps  , remove the shaft, wash in a bit of alcohol, dip in a bit of deoxit and reassemble. avoid spraying deoxit directly on the board as the spray will leave greasy splotches everywhere ( on the display, under the membrane keys ). i simply spray a bit of deoxit in a bottle cap and dip the encoder disk in .

Great, will do!

[ElectroIrradiator]

Nice video as always, Shahriar. 

replace the tantalum right next to the zener as wel as the two little tantalums about 3 centimtere left of the zener ( right next to the TO220 ). i've seen those blow up many a time. in newer versions of the supply those are ceramics.

Thanks for the heads-up. My early Agilent E3631A will get on the operating table in the near future. Anything else I might want to have a look at, now the soldering iron is hot? The encoder seems fine for now.

I also have an HP E3644A as well yet no schematic for that one. Will try to have a glance inside and see if there may be similar offending Tants across the power rails.

[PA0PBZ]

Nice video again! 

I wonder what the 5 volt rail you repaired was feeding. The fact that it still beeped makes me think that the processor was still running.
I think that the 2 black wires have no connector because they had a problem or where worried about the contact resistance, as you can see the wires go to 2 0.2 Ohm current sense resistors. Also the silk screen indicates that there was no connector planned in those positions, the text is much closer to the connections there.

[Hugoneus]

Nice video again! 
I wonder what the 5 volt rail you repaired was feeding. The fact that it still beeped makes me think that the processor was still running.
I think that the 2 black wires have no connector because they had a problem or where worried about the contact resistance, as you can see the wires go to 2 0.2 Ohm current sense resistors. Also the silk screen indicates that there was no connector planned in those positions, the text is much closer to the connections there.

Yes, it is possible that the contact resistance was an issue for them. But on the other board, the same connection doesn't have that! 

[PA0PBZ]

Yes, it is possible that the contact resistance was an issue for them. But on the other board, the same connection doesn't have that! 

You mean the 6 volt board? I don't see a connector there at all...

[free_electron]

The chip shahriar replaced feeds the main logic (the 80196 cpu asics andother stuff)
The fro t panle has its own cpu , and the gpib also has its own cpu.
The frontlanel is voltage shofted 18 volts in respect to the main cpu due to the vfd drive voltage.
The gpib is chassis referred while optocoupled to the supply. The beeper is co trolled by th gpib section.
At powerup the display does the selftest without needing the main cpu. If no communication comes within a certain then the gpib block creates the long beep telling you : main cpu is out ... I dont get anything.  It is part of the startup and diagnostics block

[PA0PBZ]

At powerup the display does the selftest without needing the main cpu. If no communication comes within a certain then the gpib block creates the long beep telling you : main cpu is out ... I dont get anything.

Ah, thanks for clarifying that.

[free_electron]

same pinciple applied to 34401, 33120's and basically any machine that has the starburst vfd.  the displays are always running 18 volts shifted to the main system ground due to the vfd drive voltage.
the gpib is chassis connected and the main controller always floats.

the IO and the frontpanel initialise on their own. if they dont get comms from the main board : Beeeeeeeeep.

[Hugoneus]

same pinciple applied to 34401, 33120's and basically any machine that has the starburst vfd.  the displays are always running 18 volts shifted to the main system ground due to the vfd drive voltage.
the gpib is chassis connected and the main controller always floats.
the IO and the frontpanel initialise on their own. if they dont get comms from the main board : Beeeeeeeeep.

Excellent, thanks.

[edpalmer42]

Shahriar, you must have a pocket full of horseshoes!  I can't believe the good luck you've had.  You seem to come out smelling like a rose even when you do things that bother me.  I'm hoping that you do some tests off-camera that aren't mentioned.

It's too late now, but one thing that you should always do when a regulator fails is to measure the voltage on it's output.  Usually, it's the load that's shorted rather than the regulator itself.  If the output voltage is low, that points to the load as the problem area.  If the output voltage is high, it could mean that the regulator has shorted from input to output and has been driving the load at much higher voltage than is safe.

In this case, if the regulator shorted the input to output, everything on the load side might have been overstressed.  Knowing HP, things like capacitors might have been rated for 16 or 25 volts so they would be fine.  But the ICs are another matter.  Of course, the regulator could have developed an internal short to ground, in which case the output might have been low.  Or the protection circuitry might have still been functional and killed the output to protect the load.

We'll never know.

[Hugoneus]

Shahriar, you must have a pocket full of horseshoes!  I can't believe the good luck you've had. 

It must be luck, there is no other rational explanation. 

You seem to come out smelling like a rose even when you do things that bother me. 

Could be because I always wear this when I work in the lab.

...

All input/output voltages of that entire section of the schematics were measured before unsoldering (off-camera).
All voltages were measured after the damaged parts were removed (on-camera).
An external voltage source was applied to the 5V rail while the regulator was removed to measure current consumption, ensure proper operation and check for problems. (on-camera).
After replacement parts were soldered all voltages were again measured including thermal imaging to ensure proper behavior. (off-camera).
The power supply operation was verified once fully assembled (on-camera).
The power supply was calibrated (off-camera).
The power supply was put through an 8 hour random load-cycling using an active load in an automated setup to stress test everything (off-camera).

I never put any equipment back into my lab unless I am satisfied that it works.

We'll never know.

 O0

[edpalmer42]

All input/output voltages of that entire section of the schematics were measured before unsoldering (off-camera).

That's what I was hoping to hear.     I realize that it's almost impossible to include everything in the video.

Ed

[Hugoneus]

All input/output voltages of that entire section of the schematics were measured before unsoldering (off-camera).

That's what I was hoping to hear.     I realize that it's almost impossible to include everything in the video.
Ed

Great! As you said, covering every single thing is not possible and can get very boring for the audience.

[just_fib_it]

I also have an HP E3644A as well yet no schematic for that one. Will try to have a glance inside and see if there may be similar offending Tants across the power rails.

http://exodus.poly.edu/~kurt/manuals/manuals/HP%20Agilent/HP%20E3640A,%2041A,%2042A,%2043A,%2044A,%2045A%20User.pdf

Includes the schematics, but no complete parts list.

[dadler]

Here is the E3631A service guide with the parts list and schematics:

http://exodus.poly.edu/~kurt/manuals/manuals/HP%20Agilent/HP%20E3631A%20Service.pdf

The more recent guides from Agilent/Keysight seem to be deliberately lacking the schematics and parts lists.

[ElectroIrradiator]

http://exodus.poly.edu/~kurt/manuals/manuals/HP%20Agilent/HP%20E3640A,%2041A,%2042A,%2043A,%2044A,%2045A%20User.pdf

Includes the schematics, but no complete parts list.

Thanks, that will be hugely useful. I have the similar document dated 2009, but the schematics have gone *poof* in that one.

Unsurprisingly, given the similarities in designs across the E36xx family, the E3644A has several offending tant candidates across power rails, which I will check in the near future.

[View[+]Finder]

I have a 3631 that is in the process of being repaired. Bought on eBay, it arrived showing 'Error 625' and 'Unreg' on the 6V side. Testing was 'Fail', but the 25V side was working fine.

Thanks to the support on this forum--and the link to the schematics--I traced the problem to the opts-isolators U3 U4 (2211's). After I replaced U3 and U4, the 6V 'unreg' went away and both CC and CV regulation work on 6V.

The pesky 'error 625' remains, however. One post here implied that it was related to RS232 and GBIB, so I replaced U1 and U2 as well. That didn't help.

Does anyone have more information about fixing the cause of 'error 625'?

Thanks!

[ElectroIrradiator]

Does anyone have more information about fixing the cause of 'error 625'?

From the E3631A service manual, page 101:

625: I/O processor did not respond This test checks that communications can be established between U16 on the top board and U10 on the bottom board through the optically isolated (U1 and U2 on the bottom board) serial data link. Failure to establish communication in either direction will generate an error. If this condition is detected at power-on self-test, the power supply will beep and the error annunciator will be on.

[View[+]Finder]

Thanks for the reference to the service manual--I had replaced U1 and U2 in response to that suggestion--but, despite my efforts to do the SMD correctly, it is possible that one or both are not working as they should. I've checked the 'pin to pad' solder-joins and they are fine. The 5V power is correct for both chips, but is there any way to confirm that the input and output are OK? Or is it time to put the o'scope on the pins to see what is being transmitted?

Update: all OK, all fixed. I removed and replaced the U1 and U2 chips with new ones--I must have not been as careful for ESD on the first install--and the 'Error 625' has gone away. Testing = 'Pass' Yahoo!

Thanks to all . . ..