SCR Regulated Power Supply

Introduction

I was on the lookout for a high current bench powersupply as my existing (linear) bench power supply with 3A per channel is often a limitation. I've been considering buying a 30A+ switchmode bench powersupply for some time but have kept putting it off as they aren't exactly cheap (being a student, I like cheap!).

This 30V, 30A supply came up on ebay for a reasonable asking price and the seller was local. I couldn't find any info about it and the manufacturer didn't have any either - it must have been produced in very low volume, custom built, a really old design or all of the above - component date codes suggest it was built around 1987 but the design could be much older. Condition was rough but stated as operational and I figured that if nothing else it would still be worth it for the 900+VA mains transformer and chassis which I could use to build a new linear supply.

Here is the supply as I picked it up, needing some TLC but with the promise that it was operational:

I gave it a quick load test and it seemed to work fine. The voltmeter did work despite having a cracked window. The power switch and current potentiometer could do with replacing. I assumed that the supply was linear regulated, although it did seem to have a fair bit of transformer buzz even with no load connected and the output voltage took some time to bleed off - must just not have any bleeder resistors on the output caps I thought. Inside I expected a big 1000VA transformer and probably a bunch of power transistors mounted to some substantial heatsinks. Boy, was I wrong. Here is what I found after lifting the lid:

So this isn't a conventional transistor regulated linear supply, it's SCR regulated.

SCR regulated supplies are highly efficient however they generally have poor power factor at low loads, high output ripple and poor transient response. They have fallen out of favour as high frequency MOS/IGBT switching supplies have become affordable, smaller, lighter and offering higher performance.

I've traced out the schematic for the main components of the supply to better understand how it works:

The SCRs are placed on the primary side of the transformer as this gives higher efficiency than placing them on the secondary side. The SCR gates are transformer-coupled to the gate driver circuitry on the control board which pulses the gate to latch the SCR on until the next zero-crossing of the mains AC voltage. Two SCRs are needed for full wave conduction. The chokes before/after the SCRs and the two 630V caps in series are to act as an RFI filter, filtering out the high frequency components of the SCR switching from the mains input.

On the secondary side are two 35A bridge rectifiers wired in parallel, since a single 35A bridge is not quite adequate for continuous 30A DC output. These haven't been wired in the most optimal way which is to wire each bridge as a half-bridge with two pairs of paralleled diodes. As the diodes within the same package are likely to have a closely matched I-V characteristic, they will share current evenly and you can construct a full-bridge which reliably has almost twice the current rating of a single bridge. Due to the way they have been wired in this supply, current sharing will only be good as the matching is between both bridge rectifier packages and if one has a lower Vf than the other it could experience thermal runaway under full load.

Following the rectifier is an L-C filter. Two E-core inductors are wired in parallel giving around 7mH followed by a bank of eight 22000uF 40V electrolytic capacitors. A random capacitor from the bank measured 27000uF being charged by the 50ohm output impedance of my function generator, and ESR measured 0.05ohm at 200KHz. I couldn't find a datasheet for them (Nippon Chemi-Con CEPW Series) but I guess this is pretty good for their age. Due to the amount of filter capacitance, the current limit isn't very useful for low power loads. There is 40ohms of power resistors paralleled with the capacitor bank to bleed down the supply voltage when it is adjusted down or the unit is powered off.

I haven't traced out the control board but it probably isn't too far departed from what you'd find in a linear supply. The only IC is a quad opamp, everything else is discrete. Power for the control board comes from a separate mains transformer, with center-tapped 50V and 100V secondaries. I imagine that the bulk of the control circuitry runs from +/-35V rails and the higher voltage rails are used for driving the SCR gates.

Light Refurbishment

As noted above, a few things needed fixing/replacement and the whole powersupply needed a clean out. I also took the opportunity to rewire some of mains side since it was somewhat questionable by modern standards. Apart from having exposed mains connections, at almost every place where a wire was terminated to a solder lug the wire was simply surface soldered instead of being wrapped through the eye of the lug. If a solder joint were to crack, there would be a live wire hanging freely. The same lazy technique was replicated on the low voltage side. Better practice employed was to wrap the wire through the eye of the lug, then solder and apply heatshrink tube over it. At the very worst if the solder were to crack then you only have an intermittent electrical connection.

The following was probably the most alarming thing in the supply - the mains connection to the RFI filter PCB. The holes drilled in the PCB are too small to accept two wires twisted together so the assembler has surface soldered them on the pads. The PCB is single sided (no through hole vias) so the pads could be easily ripped off the PCB by the strain of the cables. I drilled the holes in the PCB larger and inserted the wires from the top. I also added some slow setting epoxy to every wire terminated to this board to give them some strain relief.

The exhaust fan had a very restrictive mesh grille on it which worked great as a dust filter and had become completely blocked. It doesn't make any sense to have a dust filter on the exhaust side of an exhaust fan so I assume the purpose was to prevent large debris from getting in to the powersupply. I don't see that as being a problem for my uses so I swapped it for a less restrictive wire finger guard and installed the slowest reasonably priced 240Vac fan I could find. It's not whisper quiet but a good improvement over the old fan.

The mains power cord/plug, power switch and neon indicator were past it so they were replaced. The fuse holder seemed ok but I shattered the plastic housing trying to tighten it so I replaced that too with the newer 'safe' style. Almost all of the mains connections are made on the following terminal block and all the earths meet on the chassis screw next to it

The current setting potentiometer was very scratchy (current would max out intermittently while turning). I cleaned up the contacts inside the pot which restored it's electrical operation but the shaft still bound up hard in the middle of it's travel - I assume it has taken a big impact at some stage and the bore is out of alignment with the guide on the back plate. The old pot was 100ohm with a 100ohm resistor in parallel - seemed like a strange thing to do as you can prevent a wiper failure just by wiring the unused terminal to the wiper pin. Maybe they wanted a slight log taper. I just replaced it with a new linear wirewound pot, connecting the wiper to the unused terminal.

The voltage set potentiometer is a Bourns 10-turn wirewound - i'm not sure if it's original but it still works perfectly so I'm happy to leave it in. I wired the wiper to the unused terminal as a precaution. I also added heatshrink tube to the connections on the CV and CC LEDs.

Fixing the broken window on the voltage meter was more difficult than I had imagined. The original panel meters have a raised window which fits flush inside the 100x52mm front panel cutout. I had planned to just cut the broken raised part of the window out and glue a new rectangle of 3mm acrylic in however it proved difficult to do this and achieve a clean finish.

Plan B was to either find a new voltage meter as a drop in replacement, or a different type of meter (current, different scale, etc) with a compatible window/cover which I could install onto the old meter. Even though 100mm is a somewhat of a standard size for panel meters, the actual sizing of panel meters are anything but - I found meters advertised as being 100mm were as small as 93mm and most were not designed for flush mounting. The needle adjustment screw is not guaranteed to be in exactly the same position either. In the end I bought a cheap $8 (inc shipping) meter off ebay which was not designed for flush mounting but had an outer width of 98mm. I cut the top half off the new meter window and machined some acrylic to create a small frame which was then bonded to the window with acetone. The bottom part of the old cover was retained and the two parts were glued together with slow setting epoxy. The result is not perfect but most of the glue joints can't be easily seen once it is mounted.

Testing the supply after putting everything back together:

Ready to be put to use:

The lid could do with a new hammered paint job, but otherwise it's done.

Ripple Measurement

The following is the measured output voltage ripple at various output loads

5V, no load:

28.8V, no load:

28.8V, 3.6A:

17V, 10A:

Transient Response Measurement

Connecting and disconnecting 8ohm load while set at maximum voltage

Ultimately it is perhaps not as useful as a linear supply as a sudden or accidental short will see the capacitor bank discharged into the circuit under test - by the time it enters constant current mode the circuit is likely already damaged. I have found one good use for it though - it's perfect for performing inductor saturation testing as the voltage on the capacitor bank barely drops when dumping an 80A+ pulse into a several mH inductor.