Saturday, October 7, 2017

GSC 20A DC Linear Power Supply Redo

This boatanchor was a find at the area radio store, as-is, think I paid 20 bucks for it, a buck per amp. Brought it home and did the usual lookover, found no issues, output was around 14V. Tried it with some VHF mobiles and ended up using it to power 3 at a time for monitoring and could transmit on one at high power with no issues. 

But not surprisingly, since this thing is old school brute-force (completely opposite from modern switching supplies) it just has a huge transformer, two diode rectifier (half-wave), a Zener diode regulator then a classic thee-transistor array to regulate the higher current flows. The transformer is simply bolted to the bottom of the case, the nuts had worked loose enough so it would buzz until i smacked the case. Plus these things tend to run hot, again, they are old school. 

Eventually I took the case apart again to analyze the design. Right away i realized most of the heat was coming from the two ECG5980 40A rectifier diodes mounted on a sheet of aluminum. This got me thinking how a full-wave would be nice to try in there, and make at least that part more efficient. The ripple tanks had been replaced from the original, for which there were still two ring mounts on the base of the cabinet, each measuring 1.5 inches diameter - like finding dinosaur tracks. We all know that vintage capacitors were much bigger than today and didn't age well, so it's not surprising these had been replaced with two 15,000μF axial lead in parallel. I also noticed there was a 510Ω bleeder resistor soldered across the outputs, so someone had done a decent job fixing it up. 

So in taking stock of what this thing offers, we have the dual-secondary coil that is presumably of 360 VA variety across both the windings. We also have the sturdy metal case that could be refinished, and the three TO-3 bays with heatsinks. Plus the main switch still worked fine, as does the fuse holder and rear binding posts. These would be the most expensive parts to start with from scratch, and so this makes it a worthwhile exercise to explore the PS theory I'd always read about. 

The panel light ran from the 120V side, and I believe the exact same lamp is available in the NTE line. But in this day and age there's nothing like the reliability of an LED, as long as the regulation works as it should. Plus, the LED acts as a bleeder circuit and helps indicate when things have gotten closer to steady-state.

For the rectifier I started getting a vision of the full-wave made of single diodes mounted on a breadbord that could be bolted to the case. Since this is dual-winding we'd go with parallel rectifiers, research shows nothing wrong with this. Quickly found the 10A diodes that would be decent to work with. 

As for the new ripple tanks, this part involved some of the biggest learning curve. I noted that my switching supply just had low capacitance and grounded two series ones at the middle, so that is what we tried first. Also at first I stuck with the original 15V Zener diode and replaced the original 270Ω 3W resistors with metal film equivalents, but wasn't seeing the regulation kick in. 

So I made a small breadbord tester with a 300mA transformer and some 3904 transistors, ran lots of tests, and concluded that a 7815 regulator would be the way to go. This would mean that all 3 of the 2N3055 power transistors would be in parallel and help share the current load, whereas in the original Zener circuit the first one sets the bias for the other two. 

But the regulation still did not live up to par, and just recently I got back to researching it. 

Realized it was time to consider high-capacitance for the ripple tanks. According to a couple formulas I found online it should come out to around 60-70,000μF across the regulated output. Well, the highest 25V capacitors I could find within a decent budget are the 22,000μF. (Note they need to be 25V since the regulators produce ~21VDC if memory serves.) Now a lot could be said about how these add up in the final circuit as wired, with all the series and parallel going on, but suffice it to say they seem to pass muster. 

When first test-firing it I noticed the panel LED stayed on quite a while, and realize i should have expected it given the massive boost in storage. Something made me check how much voltage that LED would take. The specs say 13.5 V max, and the nominal output is right around 14.2, so the finishing touch at this point is the 33Ω 1W resistor which brings it down to 13.25V according to my meter. Just to note I would have spec'd a 47Ω for the 2-volt drop @ 0.417mA but didn't have one. 


One also quickly notices the instant gulp as the transformer kicks in and loads up the capacitors, but I don't think it's enough to blow a lot of breakers.

So for the first real test I try it with the IC-706MKII running a packet station. The nominal receive draws 2A and 50W ERP transmit now and then will test the filtering. I let it run this way for almost an hour. The unit understandably got hot, but now it's mostly in the TO-3 units in the back and just a little over the transformer (which no longer buzzes). Granted, this is not a normal scenario, since the 706 at full power will draw 20A and I had never intended to use this PS for an HF rig. 

But, this seems to be a successful test, as I could detect no faltering or AC hum when transmitting at full setting when listening on a separate receiver. 


So, yeah, this can easily go back to being a workhorse PS for things that need bursts above 15A but cruise much lower, and that only need to run occasionally.

----------------------------- 

Update 2021: Gave in and upgraded the transformer to a new Hammond, and just as expected this leaves any remaining heat production to the semiconductors. Also drilled some 1/4" holes in the sides of the lid for a cross breeze and now I can trust it on a fairly shallow shelf below the desk. Could always add a quiet fan but there seems no need, this is what it should be, a no-drama and fairly efficient linear PS that has no qualms powering a 100W rig.