EEVblog Electronics Community Forum
Products => Test Equipment => Topic started by: Weston on September 02, 2022, 05:02:44 am
-
I recently repaired an Agilent U8032A power supply and I did not find much discussion about these power supplies online, so I figured I would post some pictures and repair info.
The Agilent U8030 series are relatively hefty power supplies, with two 180W channels and a 5V/3A channel The U8031A is 30A/6A per channel and the U8032A is 60V/3A per channel. Sadly they are hobbled by the lack of any digital control interface. I obtained access to a broken unit from the educational lab, channel 1 was out of regulation and the over-voltage crowbar was constantly triggering.
Lets start with a front panel view, after I had already repaired the unit:
[attach=1]
These units are pretty modular, and consist of a main board that hosts a PFC to provide a DC bus, the 5V output supply, some basic fan control, and the binding posts. It mainly serves as a mechanical host for the two "blades" for the two main output channels. Each output panel has a full bridge SMPS to provide an isolated output and then a linear post regulator. The SMPS is regulated such that it always provides 5V greater than the output voltage and has its own control loop. The linear post regulator controls the actual output current and voltage. The control circuitry for the linear post regulator is referenced to the positive rail of the power supply. The commanded voltage and current is set via an opto-isolated digital bus between a microcontroller on the front panel and a microcontroller on each channel. A small toroidal transformer supplies all the isolated supply rails for the front panel, main board, and each channel, as well as the fan power. Each "blade" is mounted on an aluminum plate that serves as a heatsink and this plate is screwed to the main board. PCB tabs on each blade protrude though slots in the main PCB and the boards are joined with solder, no fancy connectors.
Its not a great photo but here is a photograph showing the power supply partial disassembled as I was debugging it. The bottom of the main board is facing out and the top of the channel 1 PCB is facing up. Due to the limited wire length the transformer is precariously perched on a support. These units are a bit of a pain to take apart. You really need to de-solder each blade to be able to remove the transformer or get access to all the connectors.
[attach=2]
The main board uses a interleaved boundary conduction mode PFC with a UCC28061 controller. That can be seen in the right of this photo of the main board while the output 5V power supply can be seen on the left.
[attach=3]
Here are top and bottom photos the "blade" for each channel. The two channels have identical PCBs. The full bridge SMPS is on the left. The three power devices on the output side are the NMOS pass transistor ( IRFP150 ), a PMOS used for sinking some minor load current, and a SCR for output crowbar. The IRFP150 is connected to the metal backing of the blade, which was removed in this photo and I had a small to-220 FET on a small heatsink bodged in for testing. The SMPS uses a UC3825 controller.
The control circuitry for the linear pass regulator is on the right, the isolated power supplies from the 60Hz transformer are in the middle, and the SMPS is on the left. The feedback circuitry for the SMPS is in on the back side in the gold rectangle in the center of the PCB. It does not seem to share any connection with the linear pass regulator or any of the digital control over the feedback loop, it just regulates the SMPS output to be 5V greater than the output after the linear regulator.
[attach=4]
[attach=5]
The front panel uses a STM32 for the brains of the unit and some optos / digital isolators to communicate with each channel. I see a labeled TX/RX header. I wonder if you can control the power supply from that.
[attach=6]
Overall, the supply seems rather well built. It clearly has enough processing power for digital control over USB. Given it is Agilent, it must have been crippled for market segmentation reasons ::)
Now on to what was wrong with the unit I fixed! The NMOS pass transistor was blown and was short on all three terminals, which was causing the output to crowbar due to lack of regulation. Beyond the pain of disassembly it was pretty easy to replace. Just replacing the pass transistor made the unit regulate correctly, but I discovered a deeper problem which led to the transistor failing in the first place. The SMPS regulation had gone open loop so it was providing the maximum voltage, leading to excessive power dissipation in the pass transistor and its eventual failure.
This was traced back to a small signal schottky diode in the feedback loop that had failed ~30k in both directions. The secondary side SMPS regulation circuitry uses an opamp to drive the feedback opto-isolator such that the SMPS output voltage is 5V greater than the output voltage after the linear post-regulator. The output voltage after the linear regulator is sensed with a voltage divider where the top resistor is split in two, there is a antiparallel diode in parallel with the top resistor and and a small filter capacitor to ground at the midpoint. This allows the SMPS to track negative going transients after the linear regulator more quickly. However, there is a direct path between the small capacitor, the diode, and the output. It seems like repeated output shorts or fast transients could have caused the diode to fail due to the current required to discharge the filter capacitor? The capacitor its probably in the range of ~1-10nF but it can be charged to 60V. I replaced the diode with a 1n4148 and a 100 ohm resistor, which should protect it from failure in the future. After that it seems to work. Small chance, but if anyone else has issues with this in the future the diode is CR209
[attach=7]
-
Great post, thanks!
What's with that electrolytic flapping around in the breeze by the binding posts in pic number 3? Hope you also replaced those RIFAs while you had it apart...
-
Did not think / have the time to replace the rifas. They are only on the DC side, so hopefully they will be ok. Given those style of rifa caps are still made I assume they fixed the problems with them eventually? This power supply is pretty modern.
There have been two other forum threads about these power supplies, both show the same capacitors on the binding posts
https://www.eevblog.com/forum/testgear/agilent-u8032a-mini-teardown/ (https://www.eevblog.com/forum/testgear/agilent-u8032a-mini-teardown/)
https://www.eevblog.com/forum/testgear/agilentkeysight-u8031a-quickie-repair-mini-teardown/ (https://www.eevblog.com/forum/testgear/agilentkeysight-u8031a-quickie-repair-mini-teardown/)
Those capacitors are on the binding posts for both of the main outputs. Looking at the PCB for each channel, I don't see any capacitors after the linear pass regulator. I assume they are needed to provide stability to the control loop for the pass element. It seems like they should have been able to fit that on the main PCB, so I wonder if it was a bodge and they never decided to rev the PCBs.
Also, now that I am digging up old forum posts on these power supplies, I believe that the damage to the U8031A documented and fixed by ve7xen was caused by the crowbar circuitry. The burnt trace is going to the TO-220 SCR used for the crowbar. I am not sure what the exact criteria for triggering the crowbar circuit is, but I assume someone had triggered the crowbar circuitry by connecting the power supply output to another power supply or battery. The SCR then latched and conducted current until the trace fused. And if the external voltage was not too high it should not have caused any damage to other circuitry.