Failure by Design – My digital power supply

It’s a Voltcraft DIGI-35 CPU. So far I’ve been quite fond of it and given its age of more than 15 years rightly so I think. The markings on the PCB read: “CTC/91 DIGI-35” and “18132 4350009427”. It’s the black thing with the two LC displays and the keypad in the picture below.


Yesterday I discovered that my car’s battery was 100% empty. Not only didn’t the starter move at all, but also the LCD clock was dead. Total lack of power. This has happened to me numerous times before and most of the time it was my fault (forgot to turn off the lights, heating for driver’s seat and so on), but this time all of it was off – I swear to “insert favourite deity”. The car was standing for 3 weeks though. Well… so I decided to recharge it with my trusty power supply. It’s a good thing I’ve parked the car on a sloping road, so I may be able to get it started by gravitational pull, if I can get the battery charged up enough to power the ignition system. I set it to current mode and 500mA and let it do its job over night. When I got up this morning, the voltage on the battery read something like 11.8V, the power supply displayed 0.0V and 2.8A… ARGH. Its heatsink was terribly hot. After removing the battery and letting it cool down it still was not behaving correctly. As soon as I turned it on, it internally shorted the output ;-(


The two MJ2501 PNP transistors couldn’t be the source of the problem. They can dissipate more than 100W each and there’s a temperature sensor on the heatsink as well. As it is a linear power supply the PNP transistors act as adjustable high-side resistors in series with the load, so there’s really nowhere they could short to. I knew from daily use that there are some relays (at least two) in there as well. One clicks as soon as you tell it to provide voltage to the screw terminals, the other one is used when switching into offline current adjust mode. It internally shorts + to – in this mode. At first I thought one of them had gone bad, so I removed it. But that wasn’t the problem at all.


Thankfully I found a Digi-35-CPU schematic on the web. One word of caution! There are several different versions of this device out there. While the schematic is quite similar to mine, it’s not quite the same. Some of the current setting potentiometers are called Rxx in the PDF and Pxx on the PCB. And as usual, pull the plug before you stick your fingers in there!

All in all it is a good device, but one aspect of it sucks though:

AC power voltage mode current mode relay 1 (NC) relay 2 (NO)
1 off off off closed / NE open / NE
2 on off off open / NE open / NE
3 on on off open / E closed / E
4 on off on closed / NE open / NE

NC: normally closed, NO: normally open, NE: not energized, E: energized

As you can see case 1 and 4 are identical as far as the relays are concerned, BUT in case 4 we have power applied to the transistors! In current mode the relays have the same switch-state as in case all AC power is gone. The same thing happens when two fuses are blown (the left ones in the picture below) which supply the logic with power, which is exactly what happened here. If the microcontroller and the relays don’t get any power, relay 1 shorts the output terminals (as if in current mode – 4) and all the power gets pumped into the transistors. Also the current shunt (R22, 0.39Ω, 5W) gets way too hot and burns its insulation.


I will have to replace this current shunt resistor. Some of the windings may touch anytime and confuse the current regulation completely.


The only advantage of having + shorted to – when the AC power is gone is that the smoothing capacitor is rapidly discharged and the output voltage drops fast. This may be desired as well, but it is definitely NOT desirable that a power supply shorts itself when the fuses for the microcontroller blow. They should have added another 5V relay to disconnect the main transformer winding from the power transistors, if the control logic is offline.


I assume that when a fuse inside a device blows, it should be in a much safer state afterwards. Here it was quite the opposite I think. The control logic board might have started to burn without the fuse, but 75W being pumped into the power transistors for several hours and a blazing hot heatsink is not exactly safe as well. WTF!?!

Oh, and while I’m at it, have a look at some of the solder joints…


I could have tolerated it for ordinary low current connections, but these wires are connected to the power transistors…


The rightmost pad is just sad. It should have been the same size as the other ones. You just don’t want to see what it looks on the other side. “Now how do we make this thick wire stick to such a tiny pad…”

This entry was posted in Electronics., Fix me. and tagged , , , , . Bookmark the permalink.

2 Responses to Failure by Design – My digital power supply

  1. Torx says:

    I have the same power supply and was looking for a way to use the serial port for ages. After swapping pin 2 and 3 I figured out how to send commands to the device. I documented the commands in a German forum “”

  2. robert says:

    Thanks for posting, I never tested the RS232 interface myself. I’ll have to replace the current shunt first ;-)

Comments are closed.