Repair of a HP 6633A systems power supply

[mrnuke]

Situation:

• Display reads -0.7V, -1A
• Keypad is responsive, can set VOUT and IOUT values
• Output is stuck at -0.7V, with display reading a current of -1A

So the good news is that both the primary and secondary microprocessors are working. Whatever the fault is, it's most likely in the analog section, so that should be  fun to track down. I won't bore you with the wild goose chases with the ohms and diode functions, as they led nowhere. I will instead try to focus on the steps I took to track down the root case and fix the issue.

The first instinct is to follow the troubleshooting steps in the service manual. That points to the current monitor section (U106B, and U109B). Alright, let's take a look. U109B is set up as a x20 non-inverting amp, and U106B as a x1 inverting amp. Measure some voltages, and the opamps work correctly. It's not the current monitor section.

[mrnuke]

Next up, I've had the supply on for a while, and the power transistors were getting quite toasty. There's clearly more significantly than 1A of current going through. So where's our current going? The output voltage is negative, so it must be coming from the negative rail. The only devices connecting to the negative rail are the output pre-driver transistor(Q110) and the four PNP power transistors (Q105-108).

The emitters of the power transistors are pulled out from under the heatsink in convenient test points. With the emitter voltages and output voltage known, it's trivial to calculate the currents through the emitter resistors. Let's make a diagram.

AHA!



Looks like something is pulling the output hard towards the negative rail. We're expecting about 1.2A current from the output, which is very close to the -1A the display is showing. That current is coming from ground, through the current shunt resistor and the flyback diode, D101. At this point, I suspected it would be one of three issues:

• Ouptut transistor failed
• Q110 failed
• Control circuitry failed

[mrnuke]

Next in line is Q110. Is Q110 sinking the bias current for the power transistors? Using the same trick as before, I measured voltage drops across R110 and R142. There's about 20mA of current going towards Q110, through R110. Surprisingly, the current flows away from Q110 through R142, and through the BC junction of one of the PNP transistors and towards the negative rail. Q110 is completely off.



The current flowing through the BC junction of a PNP tells me that the junction is shorted out. Even if just one of the transistors is busted, it would bias the other three. To be completely honest, I did remove a good transistor before I realized it only takes one short to bias all of them. To track down which of the transistors was "providing" the short, I used a 3457A ohm-meter with a resolution of 10 micro-ohms. I was seeing around 0.7 ohm resistance, which got down to 0.5 right at the pins of the dead transistor. Pulled it out and the short was gone.

[mrnuke]

Now that I had identified a faulty component, I wanted to see if there was anything else wrong with the supply. Any single MJ15004 of the output transistors is more than capable of handling the current and power dissipation specs. With a missing transistor, I powered up the unit for some quick tests. It worked!

[mrnuke]

Now to complete the repair, I needed an MJ15004, which, as it turned out, is not particularly easy to get. I was able to find one with a date code of 1982 at a local surplus store. The main problem is that although it is the same transistor, being older, it come in a different-looking package. I take pride in my repairs, and I want the device to look like nobody was there. In this case, I had three choices:

• Get it from Digikey with a 6 week lead time
• Get one off ebay, but risk getting a counterfeit part
• Use the 1982 vintage transistor

Getting it off ebay was a no-go, since no seller could guarantee the parts would be genuine. I didn't like the six week lead time from Digikey, so I decided to use the vintage transistor instead. Here's how it looks like:

[mrnuke]

And of course, I wanted to check the thing now works (operation with the output shorted), and that it's still within spec:



That's within 5mA, so within spec. Don't even need to re-calibrate this!