Weellll, I needed a good bench power supply, so I finally broke down and bought one. I got a Tektronix PS2521G. This doodad has three isolated outputs. This means that each of the three outputs can be wired in series or parallel with the others, and can float above ground or the other supplies. There’s two identical 0-20V at 0-2.5A outputs and a single 0-6V at 0-5A. Giving this supply a total of 130W it can output if all the outputs are going full bore.
Finding a triple output power supply at all was fairly difficult. There’s only really two good ones to choose from. This model and an Agilent/HP one (E3631A). The problem with the Agilent is you’re stuck with ground referenced supplies, and a mandatory +/- supply. This Tek one has similar ranges (and higher current) but the supplies are all isolated so you can connect them in any fashion you desire.
Anyways, on with the teardown!
Front Panel (After I cleaned it)
First a quick going over of the front panel. There’s three sets of output binding posts and another “earth ground” binding post that is connected directly to the chassis ground and 3rd pin on the AC line plug. The small 7-segment display on the left is the memory number (there’s 99 supposidly to remember various configurations), the current limit and the voltage limit for the selected channel. Buttons include a usual 10-key interface and buttons to set current or voltage. The supply has a cool feature where it can put supplies 1 and 2 into either series or parallel mode using relays, to get more voltage or more current out of them. In fact, all three supplies can be series or paralleled to get even more voltage or current.
Removing the case involves removing the 6 screws on the sides (3 per side) and removing the handle. Then the top just comes off. The very large transformer absolutely dominates the inside, along with the interesting heatsink/air tunnel doodad that is directly cooled by the fan. That transformer seems to have at least 11 windings on it for the various supply voltages.
Back showing the “wind tunnel”
What’s interesting about the “wind tunnel” heatsinking is that it’s actually made up of four separate heatsinks with small baffles between them for insulation and air tightness. This was done because the large TO-3 transistors are directly mounted to the heatsinks, without any insulating pads.
Here’s the inside view looking down from above.
There’s a GPIB board (I think this was an option you could buy) in the lower right of the rightmost picture, which allows someone using i.e. Labview to change the settings on the supply from an application on a computer (i.e. for automated testing). I think an RS-232 option was also available.
Right side, showing 6 regulators; three of them heatsunk
On the right side, there’s a riser board which contains no less than 6 regulators. The three on heatsinks are all 7805’s while the other three are a 7805, 7815? and 7915 I think for the 0-6V supply. The relays and big .3 ohm resistors are for current sensing and emitter swamping. The ribbon cable connects to the CPU board in front from the GPIB slot.
Left Side, supplies 1 and 2.
On the left, supplies 1 and 2 are hiding. The circuit for all three supplies is pretty much identical; There’s an AD7541 12 bit DAC for each supply, which appears to be analog sample and holded to get 3 or so separate outputs. This DAC costs $46 at Digikey and there’s 3 of them on here. The three outputs are Vset, Iset, and I think overvolt voltage. Interestingly I could not find any voltage sensing apparatus on here, for when the current regulation kicks in for example (or to show how many amps a load is drawing). I am wondering if they are doing some kind of self-made ADC using dual slope or similar since there’s an awful lot of 40xx analog multiplexers on here.
The CPU board is quite simple- just an 8031, 2K of NVRAM, 32K EPROM, a few latches/decoding chips and an 82C79 which is a single chip keypad reader / LED display driver. There’s 5 optoisolaters per channel too, for a total of 15. Notice how the ground/power planes are pulled back underneath them. I have dumped the EPROM and the sumcheck matched. I also socketed and dumped the NVRAM since it was made in 2000, and I don’t trust its 13 year old battery. If the NVRAM goes, I think the calibration information goes with it.
So that’s what’s inside the power supply. I was pleasantly surprised that they used NO custom parts at all, save the transformer which is unavoidable. Every chip and transistor inside is a standard off the shelf part, so fixing it shouldn’t be much of an issue, and since the NVRAM is dumped I don’t have to worry about that dying and having to recalibrate this thing.