Upgrading a Hewlett Packard / Agilent 6632B System DC Power Supply with front panel binding posts and correcting an original design error.
My local makerspace recently had a donation of a HP / Agilent 6632A and a 6633A System DC Power Supply. I was not familiar with this series of supplies, so when I returned home I started doing some research on them. What I found was very interesting.
The 663X series are “two quadrant” power supplies, meaning that they can source current and also sink current at one voltage polarity.
After a quick search on Ebay, I was now the proud owner of a supposedly non-working 6632B power supply, which is the newer version of the 6632A.
The auction description stated “When attempting to connect a device to the power supply, the power supply boots into safe mode. I was unable to get a power output from the device.“
The HP / Agilent 6632B has a rating of 20 volts at 5 amps for a 100 watt output, and also has output current measurement capability in the microampere range when measuring less than +/- 20 mA. It has a current sinking capability up to the maximum rated output current of 5 amps.
The unit arrived in great shape, with only a few scuffs and scratches from normal use. After a quick cleanup and some basic safety checks, including the fuse, transformer, leaky capacitors, and visual damage to components I was about ready to power it up.
In preparation for attaching a multimeter to the output terminals on the back of the unit, I noticed that the screws were very loose including the ones for the sense terminal jumpers.
After tightening the screws, and connecting the multimeter the unit powered up with no errors. After trying various loads across the operational range of the supply, the unit operated perfectly with almost no issues. The only problem that I could find was a loose voltage adjustment shaft on the front panel rotary encoder.
My guess is that the loose screws on the sense connections, were the source of the Ebay sellers issues with the unit.
To access the rotary encoder the front panel control board must be removed. Gaining access is not very hard, after removing power from the unit, first remove the knob and then the T10 torx screw on each side of the front panel. These are hidden under a thin plastic cover that is attached with adhesive. The cover can then be removed by detaching the ribbon cable and four power connectors on the switch. Document the cable positions before removing. The front panel circuit board is attached with one T10 torx screw, 10 sliding plastic fingers, and one tab near the screw. Gently bend the circuit board finger to clear the tab, and slide the circuit board towards the switch. Once the plastic fingers are cleared the board should easily lift straight up.
On my unit the four small metal fingers that hold the rotary encoder together were bent out causing the loose shaft. A quick re-crimping with needle nose pliers solved this problem for me.
You may have noticed the strange looking cutouts on the back side of the front panel. This was for a factory fitted option #020 front panel terminal. This option provided a pair of binding posts for the positive and negative outputs in parallel with the back panel connectors.
Unfortunately this was just an extension cable from the back of the unit with no voltage sense connection at the front panel terminals. Under load this can cause a voltage drop at the front terminals, along with the inability of using remote sensing without connecting to the rear panel sense terminals.
I wanted to add front panel output and sense terminals, along with a chassis ground terminal on my unit.
In a production environment which these units were designed for this would not be an issue, as these supplies would most likely be rack mounted and attached to a system wiring harness for use.
I was able to find a 3D printed adapter on Thingiverse.com for HP 66xxA 65xxA series power supplies by NearFarMedia. This insert also works for the 663xB power supplies. They show how a HP 6644A supply was upgraded to utilize front panel binding posts.
This was very similar to what I wanted to accomplish on my supply except for the front panel button which I left out.
I ordered the recommended Sato Parts #T‑45 binding posts on Ebay, and was happy with the quality of the binding posts, even though they were a bit pricy $$.
I was able to find pictures and schematics for the original option 020 circuit board, and crafted my own enhanced version of it.
The interface board consists of a high current ferrite core and a pair of film capacitors for each positive and negative terminal.
On the 663xB series of power supplied there is another option for a relay board, that was not populated on my unit. The connector footprint for this sits just in front of the option 020 connector on the A1 board. Unlike the 4 pin option 020 connector, this connector footprint J320 also includes the output sense connection and the negative trace shield connection.
I decided to disable my back panel connector by removing 4 jumpers R456 thru R459, this also removes the rear terminal capacitors from the J320 connector. There were a couple of notorious RIFA capacitors that I replaced with Y2 safety capacitors, even though they are not used in my configuration.
I used a Molex #26–61-4090 vertical 9 pin header which required some modification to clear an existing resistor and jumper. I also removed jumpers R451 and R454, which connect the sense circuit to the rear panel terminal. No need for the extra noise on the sense lines if I won’t be using that terminal. The cable side connector housing used was a Molex #09–93-0900 along with #08–65-0115 crimp connectors for 18 AWG wire.
A pair of 18 AWG shielded three conductor cables were used from J320 to the front panel terminal interface board. I connected the red and black wires of cable #1 (red) to pins 1 & 2 of J320 (+OUT_P) and the white wire to pin 6 (+SENSE), no connection was made to the shield. On cable #2 (black) the red and black wires were connected to J320 pins 3 & 4 (-OUT_P), with the white wire to pin 7 (-SENSE), and the shield wire connected to pin 9 on J320 with no connection on the terminal interface board end of the cable shield. I color coded the two cable ends with red and black heat shrink for easy identification. By connecting two 18 AWG conductors in parallel, I ended up with the equivalent of a 15 AWG conductor for the positive and negative connections, which is more than enough for the 5 amp. rating of this supply.
Before finishing up the project there were several small things to complete. I needed to finish replacing the remaining problematic RIFA safety capacitors located on the power input RFI board. These capacitors are always connected across power and ground when the power cord is plugged-in, even if the power switch is off. The original RIFA capacitors were a metalized paper type that were susceptible to humidity, which caused cracking of the insulation over time. After being stored for many years these capacitors are notorious for exploding or catching fire shortly after having power applied.
My RIFA capacitors looked to be in excellent shape, but it’s only a matter of time before they fail, so they must go.
The other issue that needed to be addressed, was an original design defect on the A1 board. When the engineer drew the original schematics for the 663xB series, they left out a junction dot where two filter capacitors on the +/- 15 volt regulator inputs crossed ground.
This error was initially discovered by Dr. Frank on the EEVblog.
The fix is a simple jumper from where C496 and C497 connect to a nearby ground on U304 (+ 15 volt regulator) pin 2.
The original circuit works OK, until one of the tantalum capacitors starts to become leaky, placing a high percentage of the total +/- unregulated supply busses (52 volts) across one capacitor, exceeding it’s rated voltage of 35 volts. Typically tantalum capacitors in this type of circuit fail with a breakdown short circuit or explosion when their voltage rating is exceeded.
On a side note, I did try replacing the noisy high-airflow heatsink fan with a much quieter Noctura 60mm fan. Unfortunately the Noctura didn’t produce enough airflow to cool the heatsink even at full speed. It was worth a try though.
Below are some extra images of internals and externals of the 6632A:
(click the images to enlarge)
EagleCAD V7.7 Terminal interface schematic and board files Rev. 1.1