Agilent 6812B AC source repair

[TheSteve]

I recently repaired an HP/Agilent/Keysight 6812B AC source that was purchased from a popular auction site.

I figured it would be good to share some pics and a few details of the repair as there is almost no information on them here.

Basic specs:
– 0-300 VAC
– 750 VA
– 6.5 A
– 0-1000 Hz
– rms, DC, AC + DC voltage and current
– Peak voltage and current
– Real, apparent, and reactive power
– Harmonic analysis of voltage and current waveforms providing amplitude and phase up to the 50th harmonic
– THD (total harmonic distortion)
– Triggered acquisition of digitized voltage and current

It seems with a very old software package Agilent made the output can even be FM modulated.

Link on the Keysight website - https://www.keysight.com/ca/en/product/6812B/performance-ac-power-source-750-va-300-v-65-a.html
Product series overview: https://www.keysight.com/ca/en/assets/7018-03206/data-sheets/5990-9309.pdf
It is now discontinued and has been replaced by the 6812C which appears to use the same power sections but has updated IO interfacing - instead of GPIB and serial it has GPIB, LAN and USB.

One bonus to the older gen unit is that there is an amazing service manual for this series of AC sources. It has excellent troubleshooting and a schematic.
https://www.keysight.com/ca/en/assets/9018-01131/service-manuals/9018-01131.pdf

First thing to notice in the manual is that when troubleshooting a full face shield and hearing protection is recommended. Seems we're in for some good times!

When I received the unit I didn't power it on, I took it apart(Dave is in my head).

Damage to the output inverter section was immediately obvious. The fuse was blown and so were several FETs, diodes and there was a hole burned through the PCB. I could see a previous repair had been performed. I don't know if it failed a second time or if the repair was not successful(leaning towards the latter).

The main output FET's were shorted, the diodes around them were also shorted. I proceeded to remove all confirmed bad components. I then checked the driver FET's and such and found that while they seemed to operate they appeared leaky - they just didn't quite seem right when probed with diode mode and ohms mode. As this is an AC driver there are two matching output sections so I was able to compare the readings between the two sides. Only one side appeared damaged. I noted that during the previous repair attempt none of the driver FET's had been replaced. In my book drivers should always be replaced when FET's blow/short etc. The next project was to find suitable repair parts. Most of the originals were discontinued or not in stock(some parts went out of stock while I was shopping). I found what I believe are suitable spares and placed my Digikey order. During the disassembly it became clear the previous person(s) to work on it were animals that had used a powered screwdriver of some sort to replace the screws used to bolt the FET's to the heatsink. They had cross threaded most of the stainless screws into the aluminum heatsink. The first several mm of each hole in the heatsink had no threads left and a few of the screws were very tough to remove and felt galled in place. All but one screw came out, the last broke. I pondered how to remove the screw and even tried cutting a slot but decided it really wanted to stay in place. I drilled right beside it and threaded a new hole. The FET is slightly offset but it's all good.

With the inverter board all cleaned up and new components in place I decided to just "go for it" and powered it on. The service manual actually allows for a procedure of testing the high power inverter section with a 0-60 VDC 3 amp current limited supply as it would be much safer(normal operation is over 400 VDC at plenty of current). I like living on the edge so instead I used a much lower value fuse for the inverter section. If I was to do it again I would have used the DC supply
However I got lucky(maybe I can pretend it's skill) and the unit powered up and all self tests passed. I gently enabled the output and it seemed to be operating as it should. Further testing over the next few evenings confirmed the output was working properly. The only issue I was having was some missing segments on the VFD. I was thinking a line or two on the driver IC's had failed so I ordered replacements. When I desoldered the VFD(drivers are below it) I found the VFD had a crack on the back side, amazing the vacuum wasn't lost. I replaced the drivers anyway as I already had them and resoldered the VFD. There was no change of course. So it was off to that same auction site to look for a VFD. There are VFD's listed for my exact model, but they want ~ 150 USD. However the VFD is the same model that is used on many Keysight power supplies and they are listed cheaper. So for $67 USD shipped I had a new VFD on the way. When it arrived I installed it and it looks great.
At the same time I did a lot of cleaning on the unit. The edges of the front face are quite damaged but it is 100% functional. Also both rotary encoders on the front were smashed and beyond repair. Replacements are cheap at Digikey.

I wanted feet for the unit, it is 28 kg(62 pounds) so picking it up sucks without being able to wrap your hands around it.
I used this design here(special thank you to CatalinaWOW for sharing it) which is a very nice copy of factory HP feet:
https://www.eevblog.com/forum/testgear/replacement-knobs-feet-and-fittings-for-test-equipment/msg2318376/#msg2318376
I printed them in PLA and it works great but PETG may be better as it would allow the clip portion a little more flexibility. I do find removing the support material from a PETG print much more challenging vs PLA though.
The input and output AC connections are made using spade lugs. While the unit is deep and the back will out of the way I wanted the connections properly shielded. I'd hate to be trying to plug something into the GPIB port only to make contact with exposed AC. The output section uses the same plastic cover used on many HP DC power supplies. Luckily this has also been drawn and shared by an eevblog member(thank you to pigrew). It can be found here: https://www.thingiverse.com/thing:2476186
The input section of the AC source requires a different cover. HP has a rather massive cover that is included from the factory. I decided to adapt the cover pigrew had shared and give that a try. I narrowed his design by 20%, made it 20% taller and 10% deeper to fit the different set/size of terminals. With those changes made I printed both in PLA to test and was very pleased with the fit. I then decided to print the covers in translucent resin. While not totally clear I think they turned out great.

Examining all of the boards in the unit I noticed a Dallas DS1230 NVRAM for storing calibration data. This means there is a battery in it waiting to die. I desoldered the Dallas device, read its contents with a TL866 plus, and replaced with a FM1808 DIP FRAM(also programmed with the TL866 plus). It is a drop in replacement that seems to be working fine and removes any worry about a battery failing down the road.
Speaking of calibration, so far it looks just fine but I'd still like to go through the calibration procedure. It requires a substantial 20 ohm resistive load(1800 watts), a 4 wire 0.01 ohm current sense resistor and they want a 3458A DMM. I have the 3458A but need but the load and current sense resistor. I've ordered a bunch of power resistors to construct a load and will purchase a suitable current sense resistor. I will update the thread once I've run through the calibration procedure.

At this point I'm super pleased with the 6812B and plan to use it during future repairs and for testing/measuring AC powered devices.
If I can answer any questions or provide more information just let me know.

Here are some pictures:
(More and higher resolution pics are also available on my website - https://fmlabs.ca/6812b/ )

Repaired, cleaned and on the bench ready for duty:


Back side showing the input and output connections with resin covers:


Overview of the inside:
Power input is the upper right and then flows in a counter-clockwise direction through-out the chassis. It is a similar layout used by many SMU's.
Upper section is the AC input where it is converted to DC.
Front is control and measurement logic.
Lower is the inverter section.
Bottom is the processor and DSP.
Upper right small board is serial and GPIB IO.
Bottom right small board is the trigger control board.


Inverter board as found, damaged/replaced parts circled in red:


Inverter board repaired and reinstalled:

[TheSteve]

Additional pictures:

One side of the CPU/DSP board assembly, you can see the Dallas NVRAM.


Ramtron FM1808 FRAM installed in a socket replacing the Dallas DS1230Y:


Crack on the back side of the original VFD:


New VFD top, old VFD bottom, front side:


New VFD top, old VFD bottom, back side. The new VFD has the evacuation port on the back, no easy to snap off nipple:


3D printed replacement foot:


Resin(SLA) printed back covers:


This is what the original VFD looked like, missing a horizontal line on the main display, the "DC" indicator of the output is also not working:


Front panel PCB with original VFD still installed, rotary encoders replaced:


Front panel PCB with VFD removed, you can see the pair of PLCC packaged VFD drivers:

[TheSteve]

Part numbers used in the repair:

Primary power inverter FET's:
IXFH28N60P3 - MOSFET N-CH 600V 28A TO247AD (quantity 4)

Driver N channel:
NTD3055-094T4G - MOSFET N-CH 60V 12A DPAK (quantity 2)

Driver P channel:
IRFR5505TRPBF - MOSFET P-CH 55V 18A DPAK (quantity 2)

Diode:
MBRB4030G - DIODE SCHOTTKY 30V 40A D2PAK (quantity 2)

Resistor(original was working but damaged):
CRCW25121R60JNEG - RES SMD 1.6 OHM 5% 1W 2512 (quantity 1)

Thermal pads:
SP600-90  - THERM PAD 21.84MMX18.8MM GREEN (quantity 12)

Inverter fuse:
5HF 10-R - FUSE CERAMIC 10A 250VAC 5X20MM (quantity 1)

Rotary encoders:
PEC16-4025F-N0024 - ROTARY ENCODER MECHANICAL 24PPR (quantity 2)

VFD drivers:
HV5812PJ-G - IC DRVR 20CH 80V 28PLCC (quantity 2)

VFD:
Purchased VFD listed for usage with a 6632B / 6633B on ebay (quantity 1)

Knobs(unit was missing both knobs):
0370-3238 - Keysight knob 6mm (quantity 2)



I will add additional pictures and details on the calibration once it has been performed.

[Swake]

Thank you very much for this detailed write-up.
I'm currently looking out for an AC power source and while this beast is nearly perfect and having specs well beyond anything I would ever need, it is unfortunately also not a all in line with the 'affordable' criteria on my list.

[TheSteve]

Well a little over a year later I decided to have a go at calibrating the 6812B.

For the load I used ten 2 ohm 100 watt resistors in series mounted to a very large heat sink(from 300-500 watt 900 MHz RF amp).
The 4 wire 0.01 ohm current measuring shunt was also mounted to the heat sink.
Shunt part # Ohmite TGHGCR0100DE
The heat sink worked out very well, I didn't have to drill a single hole, I was able to use existing ones for all resistors and the shunt.
The meter used for all measurements was a 3458A. Ideally you're supposed to use a 30:1 ratio transformer for the AC volts but the 3458A seemed to work fine connected directly.

I was also to verify 100% for certain that the FM1808 FRAM replacement for the Dallas nvram reads/writes flawlessly.

Once calibrated I compared its indicated current to my 1147B current probe and it matched within 0.01 or 0.001 amps depending on the range. Overall I'm very pleased. I don't think it needed to be calibrated but it was still a fun experience.

I should add max power output during cal is about 700 watts. I bought the load resistors for less than $20 shipped for the 10 of them from aliexpress and they seem decent.

[danils]

Hello There,

I have a 6812B and I was going to perform a calibration on it.
I was wondering why a 8.5 digit multimeters is required to calibrate an instrument that has nowhere that precision.

Do you think the calibration is feasible with a 6.5 digits one?

[TheSteve]

I'm thinking it would likely be fine.

[heize0]

Hi! Have you back-up the firmware in EPROMs? I have this one but unfortunately it could not power up and only can display shift if I pressed the shift button. I'd like to repair it. I suspect the program is corrupted.

[jgorsk]

heize0, have you checked the communication interface between the front panel and the controller board?
The connection goes via RJ connectors and they may need just sone cleaning up.

[TheSteve]

I didn't pull any of the PLCC packaged firmware devices during my repair.

[heize0]

Thank you.  I have seen the service manual. I checked the signals followed the instruction and found that there was no UART signal from the DSP. I tried to switch on and off many times and there was only one time the display segments were all lit up. More situations were either the shift could light up or the keyboard had no response.

[anatech]

Hi Steve,
Thanks for the amazing writeup!

I bought a 6812C right from Keysight, pre-owned but certified, repaired and working.  It failed after being used less than 10 times on very light loads.  It is a wonderful supply.  Just out of a 90 day warranty and they won't lift a finger, nor will they release the manual for this model, but someone from Keysight did send the "B" version manual to me.  Needless to say, this is not the "HP way" experience I am used to having.

It goes into protect immediately.  Resets and cycling power didn't change the behavior.  Clearing error codes didn't help either.  All fuses are good, and the switching and driver transistors tested fine.  I have to put it back on the bench at some point, but I demoed my back lifting a 125 pound instrument.  Stupid, I shouldn't have as I'm in my mid-60's. 

It was running at about 120 VAC, 2.5 amperes for over an hour, then just shut down.  It has clear air, in a rack and nothing was running below or above it.  It didn't seem to be running hot and this is well within it's ratings.  This was the one and only time it was delivering over 1/2 an ampere.

I will update once I am back into it.

Thanks again Steve, Chris