My PSU design

[AkisTzortzis]

Hello, here is my PSU design, please comment/advise.

I have built a prototype and tested it with small transistors (not high power ones yet). It seems to work OK. My output transistor blew when I shorted the output (during testing). The current sense circuitry works well, but I had a capacitor at the base of the switch-off transistor (2.2uF) which probably made it too slow to protect the output transistor (doh!)

So my current testing/problems are:

(1) will the protection circuitry be quick enough to save the output transistor? Or do I have to select the output transistors carefully to withstand huge current spikes? Or do something else?

(2) the input sawtooth like ripple (sometimes 300mV peak to peak) hits the output transistors directly (without going through the LM317 first) and a capacitor at the exit completely eliminates the resulting output ripple. However, I am wondering if there is a better way to eliminate this ripple. When I say "eliminate" I mean what I can detect with my oscilloscope which goes down to 5mV per division. With the output cap I get a vague flat trace at 5mV, without it I get about half the input ripple (so from 300mV at the collector to 150mV at the emitter) and what's worse I get funny instability when I touch the collector. The output capacitor seems to completely eliminate these two problems, but I lack in theory so may be other factors are there unknown to me.


******************

The design went like this: take the unregulated DC from the bridge/filter and put a transistor in common collector configuration in series (so all current goes through it) and make sure you give it a stable voltage at its base no matter what the supply is. To obtain this "stable voltage" I could use a zener but with prices what they are I used an LM317 which also allows me to adjust the output voltage.

The LM317 is man enough to feed the output directly, so what's the point of the extra transistors? Well, the LM317 will do 1.5A and I want in excess of 5A, and what's more I want 5A at say 3V output - with a 35V DC input that means we would have to expend 32*5=160Watts somewhere.

So I decided to use the pass transistors but not in the way shown in the LM317 datasheet. In my configuration the LM317 does not feed the output directly, so the onus is on me to protect the output transistors from shorts as well as deal with the ripple voltage.

Edit: the output voltage is adjusted with a dual gang pot 10K each in parallel to minimise errors and again in parallel with an 11K resistor (all to minimise errors).

Edit2: I have a problem with the two output transistors: it seems one is bearing all the load whereas the other is taking less load. I have just almost burned my finger trying to see which one (if any) were hot during a shorted output test case (ok they are not mounted on any heatskink so they are not sharing the heat). How to solve this problem? Place very small resistors at the emitters to create some feedback which will "hold back" the runaway transistor? Or just use just one hefty output transistor?

[loudthud]

The circuit looks like it will basically work but regulation won't be very good because the emitter follower transistors are not inside the feedback loop. If you set the output to 3V with a 1amp load, it might drop to 1.5V with a 6amp load. Is that the level of performance that is acceptable?

Ripple: A good rule of thumb for 50Hz power supplies is that 4000uF will have 1V of ripple at 1amp. It doesn't matter what the voltage is. (For 60Hz use 3300uF) If you had 300mV of ripple at 6amps that would imply that you have 80,000uF of filter capacitance. Not very practical, your rectifiers will be very hot. Generally the ripple should be 10%-20% of the output voltage at the filter, around 6,000uF in this case. It's the job of a regulator circuit to reject ripple but there will always be some. 5mV pk to pk is not very much. The LM317 has good ripple rejection as long as the voltage between input and output does not fall below about 3V. The output transistors should be fairly immune to ripple on the collector except at very low currents, leakage may be too high. You might want to increase the capacitor at the output to 1000uF or more.

[AkisTzortzis]

So far I have been testing the operation of the protection circuitry (it works) :-)

My next test will be the regulation but I suspect I will need to construct using the real materials because so far my test harness uses a pair of 2N2222A at the output and receives 16V DC unregulated at the input. I am still waiting for a whole bunch of components to arrive (which I have ordered) and then I will do the tests.

Regarding what you said about 3V and 6A load - sorry but I do not understand. Why would the voltage drop if I demanded 6 amps? For what is worth, I am planning to use a 2x25V/300VA toroidal transformer and either 1 or 2 4,700uF capacitors as the main filter. That transformer is rated "25V at full load". So I reckon it would be able to supply near 6 amps without issues. I am planning to use the other 25V for a negative supply, which will be just like this one but mirror image.

[loudthud]

The base to emitter voltage drop of the output transistors will increase as the current increases as will the voltage drop across the 0.1 ohm current sense resistor. The output of the LM317 will remain constant thus the output voltage of your circuit will decrease.

I was thinking what I said in the previous post about the rule of thumb. Actually it does matter what the voltage is. The numbers given are a good approximation for low voltage power supplies but don't hold up at high voltages. I recommend Duncan's PSU Designer II to give more precise results. Goto:

http://duncanamps.com/index.htm

Click on Software Downloads and download and install PSU Designer II. It will help you predict the amount of ripple your filter will produce.

[Steve Conner]

I take it you've seen the "High Current Adjustable Regulator" application circuit on page 16 of the LM317 datasheet? It's similar to your circuit, except for one crucial difference: the booster transistors are included in the voltage sensing feedback loop, so the output is actually regulated.

The LM195 was a "smart transistor" with onboard current limiting, but you'll have to provide your own limiting, unless you can find a LM195 in a museum somewhere Older versions of the datasheet had many more application circuits, all the way up to a precision current limit with a LM301 op-amp (now obsolete too)

[AkisTzortzis]

Yes I have been reading the datasheets, there is even more info in the LM105 specs. It talks about oscillation between the regulator and the pass transistor, overheating, current limiting, ways to prevent the oscillations etc. I will modify my design to take into account loudthud's comments, he is right about the current sense resistor being more prominent at low outout voltages. However I have no materials to play with, all my testing is done at 16V DC input and very little current capabilities using very small transistors. I am waiting for breadboards and other stuff to arrive to be able to build a proper prototype.

[Enzo]

I don't get the dual parallel pot thing? What "errors" are you talking about? I would turn the knob until the desired output voltage is set. Where does error enter the equation?

Ripple visible on the 5mv range of the scope? I made up the term "trace chasing" when I had a junior tech once who couldn't stop turning up teh scope until he found some sort of trace on the screen. Didn't matter what the context was, he'd manage to get something on the screen and then worry about it. I would not worry in the slightest if there were 5mv of ripple on a 30v power supply or even on my 5v for the digital stuff. This is not a laboratory environment.

Set your scope to a reasonable vertical scale for your circuit and leave it there. If you go chasing after a trace on tegh scren you are almost guaranteed to find one. Here, I can always turn the scope up enough to collect the waveform of the AM broadcast transmitter down the road a bit.

Parallel power transistors have to be made to share current. Left alone, one will almost always hog the current. You would need to add some low value ballast resistors in series with each emitter to encourage them to share equally. But it is only 6 amps. That is a pretty small powr supply in the grand scheme of things, and there are plenty of power transistors up to 6 amps. SO why use two? Look up the spec sheet on whatever transistors you chose, and scroll down to the "safe operating area" graph. For example that 250v 16 amp power semiconductor won't do both at once.

[AkisTzortzis]

I don't get the dual parallel pot thing? What "errors" are you talking about? I would turn the knob until the desired output voltage is set. Where does error enter the equation?

Ripple visible on the 5mv range of the scope? I made up the term "trace chasing" when I had a junior tech once who couldn't stop turning up teh scope until he found some sort of trace on the screen. Didn't matter what the context was, he'd manage to get something on the screen and then worry about it. I would not worry in the slightest if there were 5mv of ripple on a 30v power supply or even on my 5v for the digital stuff. This is not a laboratory environment.

Set your scope to a reasonable vertical scale for your circuit and leave it there. If you go chasing after a trace on tegh scren you are almost guaranteed to find one. Here, I can always turn the scope up enough to collect the waveform of the AM broadcast transmitter down the road a bit.

Parallel power transistors have to be made to share current. Left alone, one will almost always hog the current. You would need to add some low value ballast resistors in series with each emitter to encourage them to share equally. But it is only 6 amps. That is a pretty small powr supply in the grand scheme of things, and there are plenty of power transistors up to 6 amps. SO why use two? Look up the spec sheet on whatever transistors you chose, and scroll down to the "safe operating area" graph. For example that 250v 16 amp power semiconductor won't do both at once.

What I meant is when you turn the knob you are subject to the irregularities of the potentiometer which is a crude mechanical device, so by adding two in parallel and additionally a fixed resistor also in parallel I was hoping to make it less "jumpy" or "drifting" in operation. Just a thought, on my prototype I use a high accuracy pot and it is still not very smooth.

You are right about the 5mV - at that range the (old) oscilloscope seems to always pick up something - even if you short the probe :-)

And you are also right about the ballast resistors. I am not sure what value though. Probably small enough so that at 6-7 amps it does not drop too much voltage. So, say 100mV at full load?

[Enzo]

I'd probably use some fractional ohm value like .33 ohm or .47. Steve might have better ideas.

If the voltage drifts, it is probably not the pot doing it.