HP6236B Power Supply Repair

[Shuggsy]

Hi all,

Just wanted to talk about a little repair project I've been working on the past couple weekends. I scored a "BROKEN, FOR PARTS" HP6236B power supply on eBay for cheap and figured I could either fix it or harvest it for parts if the fix proved too costly or involved extremely hard to find parts. Maybe this will give insight into the workings of this supply or at least be interesting.

The HP6236B is a 6V@2.5A and +/-20V@0.5A triple-output power supply. These things are quite serviceable, are conveniently laid out for testing on the board, and are pretty hardy instruments in general. These supplies are at least from the late 70's and can still operate quite well today. A quick google search landed me with a full copy of the service manual complete with schematics, board layout and description of the layout, and some troubleshooting sections that could get me started.

The unit was described as looking in good condition but that the output was fixed at ~3V no matter what you turned the knob to. Oddly, the description also said that the voltmeter on the front panel responded to turning the dial to change the voltage... a quick look at the schematic shows that the voltmeter is in parallel with the outputs through some calibration resistors, so I figured that the output should be changing in line with the voltmeter reading. Maybe all was actually well with the unit. Well, only one way to find out.

When I got the unit in the mail, I opened the lid to look for anything wrong staring me in the face but couldn't see anything out of the ordinary aside a little dust and wear from time. After plugging up the supply and turning it on, I noted that the 6V output did indeed move the meter and seemed to function correctly. A quick check with my multimeter showed it seemed to be right around where the meter indicated. When I switched to the 20V output however, I did indeed see the meter indicate just over 3V. The multimeter confirmed this at the output and no change of the 20V adjustment made any difference.

So, I set to find out what the issue was. I started with the troubleshooting section of the service manual and started probing some of the test points. All the reference voltages looked fine, but I was getting some odd voltages at the output of a diode OR-gate that switches control of the output between voltage or current-limiting. After isolating some of the resistors and capacitors the troubleshooting section listed, I was down to the driver BJT for the big power output BJT or one of the diodes in the circuit. The driver BJT seemed to be at least somewhat driving the output since it was stuck around 3V, so I started with the diodes -- no dice there, they seemed to work as usual. I did, however, notice some odd outputs from the op-amp IC used to compare the voltage and current sensing to references to limit the output. The op-amp output was swinging from around +0.7V to -10.2V with no appreciable change in the output. The output didn't change linearly between these values either, it just swapped between them at one point on the 20V adjustment which landed about where I expected the 3V setting to be.

At first, I thought the op-amp might be acting like a comparator -- a broken feedback path somewhere had it operating open-loop -- but then I checked the inputs to the op-amp. Per the description of the voltage-limiting function, the non-inverting input of the op-amp is set to -0.016V via a simple resistor divider off one of the reference voltages. During normal operating, the non-inverting input is tied through a resistor divider between the output voltage and a reference voltage with part of the divider consisting of a potentiometer mounted to the front panel. As you adjust the pot, the voltage out of the divider changes and the op-amp servos its output to drive the driver BJT into the output BJT to change the voltage out of the supply until the two inputs to the op-amp are at the same voltage (-0.016V). Well, a quick look at the non-inverting input showed the voltage out of the reference voltage divider to be ~+0.7V. This is rather odd since the divider was between Common and -12.4V... I figured at that point that the op-amp IC might be hosed. I lifted the leg on the IC for the inverting input and sure enough the reference voltage divider put out -0.016V when the op-amp wasn't in the circuit.

The part listed in the service manual was HP part no. 1826-0092 with a possible commercial replacement listed as a "CA 3458T" which is an old RCA dual op-amp IC. Well, some searching around the net and the parts on digikey/mouser/etc. landed me on the LM1558H from National Semiconductor (now part of TI). The H model actually had the same metal can package with the same pinout as the original IC so no odd kludge to get the thing on the board was needed -- a definite plus! I went ahead and ordered one and replaced the bad IC on the board then turned on the supply again. Success!... kinda. Now I could adjust my 20V output from about 2.8-3.3V. Outside the region, I got no response from the circuit, but when I was in the 2.8-3.3V adjustment range on the 20V adjustment everything changed nice and linearly.

Now, all the inputs to the op-amp seemed fine, but the output drive to the base of the driver BJT seemed a little odd compared to the 6V circuit which was working fine. I desoldered the driver BJT and put my meter across it... the collector/emitter had about 50 ohms across them and the base was just about shorted to the collector. So now there was a bad BJT. I drove up to a semi-local electronics shop and found 3 somewhat similar transistors to try. I couldn't find a direct replacement for the one in there and I wanted to try to fix the supply without ordering another part online. I overspecced the collector/emitter current requirement as the original BJT had a pretty funky heatsink on it. After putting the new transistor in the circuit, the 20V output now went over the full adjustment range without any issue! However, the -20V output seemed slightly off the 20V output. The -20V output is made to track the 20V output within 1% (unless you change the tracking ratio...), but was almost constantly about 2.5V higher (lower? inverted, of course) than the 20V output. Thinking I might have the same issue on the -20V output that I had with the voltage comparison op-amp, I went ahead and checked the input voltages. The reference voltage (this time set to Common or 0V) seemed fine, but sure enough the non-inverting input was being kept to ~+0.7V rather than 0V by the op-amp. I had to order another replacement IC online (I ordered another just so I would have a spare this time...) but replacing it got the entire supply fully functional again. Awesome.

I toyed with the idea of putting it back on eBay and getting a little money out of the thing, but I think I'd rather keep it around my small but growing lab. Never can have too many power supplies, after all. Now I just need to start expanding my multimeter collection. Some pics below... don't know if the Fluke or the Extech meter is off, but when checked with the same meter the supply outputs are much closer. I tend to trust the Fluke a bit more, but the Extech meter is newer.

[shebu18]

Nice writing, great explaining of the debug methode.
What is the tolerance if the supply now?
I find the idea great to buy something markes as broken and then repair it.

Sent from my GT-I5800 using Tapatalk

[Shuggsy]

Nice writing, great explaining of the debug methode.
What is the tolerance if the supply now?
I find the idea great to buy something markes as broken and then repair it.

Sent from my GT-I5800 using Tapatalk

Over the range of the 20V supply, the -20V supply tracks within spec until very small (~0.1V) outputs. At that range, the outputs track within 1.2%. Outside of the low regions (5V+), the outputs track arorund 0.13% of each other which is well within spec. I hadn't let the supply warm up before taking those measurements though, so performance may improve if it's left on. Or it could degrade! Only one way to find out.

On that train of thought though, I haven't yet put the supply through its paces to see if it meets its specifications for load/line regulation, ripple/noise on the ouputs, and drift. I may get to that in another couple weeks, but that seems much less exciting than the circuit debug efforts.

[andyturk]

Nice work, Shuggsy! I've been nosing around at power supplies on ebay and while a high tech 3631 would be great, an older 62XX would meet my needs for a lot less money. It's good to know that they're fixable with a little bit of effort.

[SeanB]

If you have the same opamp in the 6V supply then change it as well, it is common for a batch to exhibit failure with time if there were contaminants on the wafer, or if the can batch leaked. It might work now as it is run at lower voltage, but probably will fail later.

[Shuggsy]

Yep! I love old HP supplies or the new Agilent equivalents. I haven't looked much into what's changed in the newer variants though.

Unfortunately, when I started putting my "repaired" supply through its paces, I found that the 6V output wasn't as good as it seemed. Rather than limiting the current at ~2.75A as it should, it limits around 1.1A. I've tried the current limit adjustment pot, but that didn't work.

If you have the same opamp in the 6V supply then change it as well, it is common for a batch to exhibit failure with time if there were contaminants on the wafer, or if the can batch leaked. It might work now as it is run at lower voltage, but probably will fail later.

Posted right before I was about to update this! Good idea, and that's one of the big reasons I got a spare on my second part order (the other big reason being if there was a fault somewhere else in the line that blew up the other ICs...) but... replacing the voltage/current comparison op-amp IC (as was the main issue with the other outputs) had no effect... which I kinda guessed since the input voltages made sense as I was probing around, but I was hopeful it would just be that easy!

I removed the current limit adjustment pot to make sure it hadn't been damaged/corroded, but it checks out fine. My next step is to check the other resistors in the current foldback circuit. I'm very hopeful that the resistor in the nice matched resistor network isn't the culprit. If it is, I'll probably clip the leads on the matched resistor IC and get a 1% or 0.5% resistor in its place. Not sure just yet what the issue is, but I'll update this thread when I get it back to full spec.

[SeanB]

Check the supply capacitor is fine first, and then check the jumpers and those diodes ( replace rather than check, they are cheap enough) and the capacitor C23. Not likely to be the thin film network, unless the resistor has failed open, where it would make the current limit slightly lower, not stick at the bottom end. CR44 is a likely bet as being leaky, zener diodes often are.

[Shuggsy]

Well, after several hours of isolating nearly every component in the current limiting circuit, output capacitors (and a couple resistors), and figuring out exactly what CR44 was (it is a Stabistor, NOT a zener!), and a lot of head scratching I finally believe I have the answer and it lies between my keyboard and my chair.

Unfortunately, when I started putting my "repaired" supply through its paces, I found that the 6V output wasn't as good as it seemed. Rather than limiting the current at ~2.75A as it should, it limits around 1.1A. I've tried the current limit adjustment pot, but that didn't work.

Turns out I should have kept putting it through its paces. That current limiting number of ~1.1A was when I was short-circuit testing the outputs. I figured the short circuit current limit would be the same as the max current limit... well, upon reading up on the current limiting functionality it seems that there's a linear relation with voltage and current limit on the output. It starts at ~2.75A when the output is at 6V and decreases to a minimum of 0.85-1.15A (varies from unit to unit) during a short circuit. DOH. TOTAL FAIL. Ah well, live and learn.

Silly engineer, why RTFM? To make myself not look like a noob, that's why!

I'll be bringing the supply into work soon and putting it on a proper electronic load to see if I can verify the current limiting up to its maximum on all the outputs. It does, however, lead me to a question for you all: how would you make a low voltage, "high" current, variable load in your home lab if you didn't have an electronic load? Throw a bunch of resistors together? Make a constant-current load like the one from Dave's video (DIY electronic load)? Use that giant 100W potentiometer that's a big resistor with movable metal ring across the middle of it that you happen to keep in the corner of your lab? I've got decent test gear but very little in the way of spare parts lying around to toss a quick circuit together but I figure some of you all should have some interesting suggestions.

[robrenz]

I'll be bringing the supply into work soon and putting it on a proper electronic load to see if I can verify the current limiting up to its maximum on all the outputs. It does, however, lead me to a question for you all: how would you make a low voltage, "high" current, variable load in your home lab if you didn't have an electronic load? Throw a bunch of resistors together? Make a constant-current load like the one from Dave's video (DIY electronic load)? Use that giant 100W potentiometer that's a big resistor with movable metal ring across the middle of it that you happen to keep in the corner of your lab? I've got decent test gear but very little in the way of spare parts lying around to toss a quick circuit together but I figure some of you all should have some interesting suggestions.

I recently learned about a very usefull bit of kit, the bipolar operational power supply.  Check out this data sheet showing all that these things are capable of.  My thread on the two that I got. These are capable of being a constant voltage or constant current load as is. With external circuitry they can be CR and CW loads.