DC high-current power supply-using PNP pass elements.

[Xnke]

So I have been collecting parts for a while, just random bits here and there over the last 20 years, and normally don't need to do much high-current power work. (I usually work with vacuum tubes at RF!)

I've been working on a silicon transceiver for a few years, and it will need a 13.6v, high-current, low-ripple/low noise power supply. The final amp has pretty bad PSRR, so the smoother, the better.

I've got a 16V center-tapped, 30A power transformer, a few 50,000uF/50V filter capacitors, and some 40A power diodes. All good there.

Going through the power transistor bin, I have:

3 each Fuji Electric 2DI 100Z-100-01 darlington bricks-Untested, not sure how to test this?

5 each RCA 2N6051 PNP TO-3 power transistors-these are new in the package

2 each ECG280 NPN TO-3 power transistors-these are new in package, audio finals

1 RCA 2N3055 NPN TO-3 power transistor-New

I have LARGE heatsinks available, rated for 1KW plus in industrial service, and the one that those three Fuji bricks came on was rated at 14KW. I am leaning towards the 2N6051's because they're available if I blow one up, but they are PNP so would have to put them in the ground rail-which is good, the heatsink would be grounded.

I haven't done a silicon power supply over 3A in a decade-and even then that was just a pass transistor around a LM7805. I've always had good bench supplies for everything I needed until I hit this high-current need.

Does anyone have any schematics to share, with insight on how I might want to change them for this application? Tim? I can't find your old schematic repository anymore, I think it moved a few years back.

[spec]

Hi  Xnke,

What is the maximum current load on the future PSU.

Do you want current overload protection.

[technix]

At that power level, I would suggest using a poly-phase switch mode design if possible. Each phase has a moderate current rating like 5A so you can use some higher switching frequencies in the high hundreds kHz to low MHz range so the ripple can be kept in check by an additional LC filter stage after all phases met.

[David Hess]

I did exactly this many years ago by wrapping a TO-3 NPN power transistor around a LM337 (negative adjustable regulator similar in capability to a 7805) and it worked fine.  Scaling up to 30 amps is feasible with parallel transistors and a 7805 style regulator can provide plenty of drive.  I used PNP transistors because I had (and still have) a tray full of TO-3 LM337s.

The poor power factor will mean only about 20 amps is available from the 30 amp secondary but with high surge current.  16 volts AC will yield about 20 volts DC peak through a full wave bridge so there is just about the right amount of headroom for a 13.6 volt regulator.

Assuming 3 volts minimum voltage drop, that leaves 3.4 volts maximum ripple from the input capacitor.  The 8200 microfarads per volt/amp rule means 48000 microfarads minimum to support 20 amps so I would use two of your 50,000 microfarad capacitors in parallel.  See below about the Darlingtons.

Today I would use the configuration shown below.  It has built in emitter ballasting through R1 so multiple transistors can be used in parallel while duplicating R1 and R2.  The diode can be the same transistor (not a Darlington) with the emitter and collector shorted together for better temperature tracking if you do not have a suitable power diode.  The regulator should be mounted on the same heat sink as some or all of the transistors.  If the thermal resistances are selected correctly as mentioned in the text, then the temperature protection built into the regulator will protect the transistors.

The 1.5 amp LM317 can be replaced by a 3 amp LM350 or 5 amp LM338 if you want more drive.  R1 needs to be pretty high power; I used 5 watt silicon wirewound resistors on mine and they turned red hot once although everything worked as normal.

Note that the positive regulator uses PNPs on the high side and a negative regulator uses NPNs on the negative side.  In both cases the collectors which contact the heat sink are all tied together.

Solid 12 or larger gauge wire works well for the high current connections around the heat sink because it stays in place.  Construction is more like plumbing than electronics when wires get this large.

Hmm, as to your collection of power transistors:

3 x 2N6051 12 amp 150 watt PNP Darlington
2 x NTE280 12 amp 100 watt NPN
3 x 2DI 100Z-100-01 IGBT?

The 2N6051s will certainly work but stability will be a little trickier.  Use two diodes in series to match twice the Vbe for good current sharing with the regulator.  I have a bag full of 50 amp 300 volt MJ11032 TO-3 NPN Darlingtons which would be good for this sort of thing but I never used them.

Note that using a Darlington increases the dropout voltage a bit but not enough to matter if you use two of your capacitors in parallel.

Fuji Electric's web site is broken so I could not find a datasheet.  If that is really an IGBT, then it is more trouble than it is worth in a linear power application.

[Xnke]

Total draw on the power supply will be at *least* 14A during transmit mode. This will depend on the efficiency of the final, mainly due to heatsinking consideration. The reason for a linear supply is to eliminate as much RF interference as possible, ANY noise on the supply line will likely get transmitted out as intermodulation. I have to maintain a second harmonic at least 43dB down under 50Mhz, and at least 60dB down over 50Mhz. Less crap in, less crap I gotta handle in the output filters. I already have a switchmode converter generating -8 volts for the low-voltage analog and gain control elements, and I have to shield and filter the crap out of it already or the 350khz carrier can be seen in the 1mW feed to the RF amplifier.

The Fuji electric are a pair of three-stage NPN darlington BJT's configured as a totem pole. They claim 100A/1200v on the side of the brick, but I have not found a complete data sheet yet, just the diagram on the side of the package explaining the pinout. They were used in linear mode as a true-sine-wave variable frequency motor drive. (three phases, three bricks.)

Other equipment in the shop also operates at 13.6V, so having a supply capable of 20-25A is ideal. I figure 20A is a good output current max limit, although it would be handy to have an adjustable current limit down to very low current for troubleshooting.

Let's see, I have five of the PNP transistors that are supposed to be capable of 12A continuous current, but the datasheets don't have an SOA graph-so I'll derate them to 10A continuous. The way I see it, that would let me get by with two in pass duty, assuming excellent heatsinking. 3 of them would be safer as far as reliability is concerned.

[Ian.M]

There's a better 2N6051 datasheet with SOA & derating curves here:
https://4donline.ihs.com/images/VipMasterIC/IC/HRIS/HRISD005/HRISD005-2-73.pdf?hkey=EF798316E3902B6ED9A73243A3159BB0

You could replace the diode D in the schematic with a Vbe multiplier using any non-darlington reasonably high gain power   transistor that can handle the regulator IC's worst case short circuit current (before its SOA & thermal limiting kicks in) which would allow you to trim the current at which the external  pass transistors start to 'kick in'. 

[Xnke]

Thanks, Ian. That sheet shows that a better operating point (at least a known safe operating point) would be anything less than 10A at 10Vc-e, so long as I have enough heatsink to keep the suckers are room temp (I don't). It also calls out a linear deration per degree C over 25C, for current-so some napkin scratching later, if I can keep the cases at 75C, each device can handle 7.5A.

That's stinking hot. I'd prefer to keep them around 40C, where each device can handle 9.5A or so. As far as heatsinking goes, a 400mm x 200mm section of heatsink with 80mm tall fins spaced 10mm apart with a 10mm thick base is available, and was used in the 14kW linear VFD that I pulled those fuji transistors from. It did have a 220v fan blowing on it at that level, though.

[Ian.M]

I wouldn't go over 80% of the derated SOA if you want reliability.  Also you have to design for worst case ambient temperatures which can be quite high in an equipment rack in a summer heatwave (assuming you don't have server room grade air conditioning in your shack), so 40 deg C case temperature is unrealistic.   As the transformer is good for 20A DC continuous, IMHO you'll probably need three of the power transistors, operating at 60 degC max case temperature (which derates them to approx 8.3A Ic_max).   

I'm betting you'll need active cooling, though you could use 12V DC and PC fans so you can easily make them variable speed and temperature controlled, shutting off below 50 deg C heatsink temperature to minimise the annoyance factor when its lightly loaded or the environment is cool enough.  As you want the main output to be as clean as possible, its probably worth using a separate transformer or PSU for the fans and system monitoring.

[David Hess]

My big 50 amp Astron uses 8 output transistors under similar conditions without active cooling but the heat sinks and tops of the TO-3 transistors are exposed on the sides.

The one I built used a 20 watt heat sink with a fan mounted directly over it and an operational amplifier (probably an OP-90) based temperature controller with a 1N4148 epoxied to the heat sink for temperature control.

[Xnke]

Alright, quick topology check here, gotta start with something. Only working out voltage regulation at the moment, the adjustable current foldback will come later. (when I figure out how the heck to do that.)

Transformer with the chunky stud diodes and all the capacitance I can bolt on it (339,990uF) puts out about 13.5VDC loaded at 40A output, coming up to 15.5VDC at 30A, and 16.2VDC at 20A. It has a few taps that I can work with to move the voltage around, but it will definitely need active cooling on the transformer if I pull more than 20A continuous.

Here, my load resistor is the 690R on the far right-I should have labeled it better but missed it while watching the hockey game (Preds-Blackhawks)



I've definitely made some mistakes in the basic topology here, and none of the components are anything but representitive. I'm probably going to be using something like an LM324 or NE5532 op amp, I've got a LOT of the latter pulled from old broken audio gear. (Lots of the old Realistic brand CD players with dead transports, all had 8 NE5532's and all of them were socketed)

Pick that apart and let me know what's gotta change. If anyone has and good methods of applying an adjustable (preferably with a 10K pot) current limiting, that behaves such that when the limit is hit, the power supply just folds back and the voltage drops, it is much appreciated.