Help needed PSU designing


I reacently accuired some parts and i going to built linear psu from them.

Parts whict i have:
230V/2x42V (60VDC unveighted) 1000VA toroid
20pcs 63V 12000uF EL-caps
10pcs of MJL4281A (NPN) and 10pcs MJL4302A (PNP) transistors
230W 15A 35Mhz 350V Hfe80-250

I`m trying to design psu with following properties

dual voltage out (+-50V adjustable)
Output current of about 8A
Adjustable current limiting (0-8A) with short circuit protection
Electrical overheat protection (fan or shutdown or both).

I would be greatful if someone could lend a hand give some hints
or maybe some kind schema for design.

Please help beginner.
John Linsley Hood did a Wireless World design article for an 80w Mosfet stereo amp many years ago (probably mid 80s).
This may be the Williams Hart that is still retailed.
It included the full design of a regulated dual voltage PSU.
The low current side was simple. The hi current side included fold back limiting and DC detection and both pole shutoff and I think there was some other detection built in there.
All the voltages were adjustable within the thermal limits of the pass fets and IMO they should transfer to a BJT amp without difficulty.
By the way he chose not to include the design in his book "The art of linear Electronics"


diyAudio Moderator Emeritus
2002-06-10 12:01 pm
AndrewT said:
fold back limiting solves that.

I think you miss my point.
To supply 50v, the input of the regulator will, in practice have to be at least 55v, so my 50v drop senario is not one of short-circuit (where of course you are right, that foldback solves), but in supplying 5V, not an unusual case. If we consider only 2 amps, the dissapation is still 100 watts.
Pehaps I should have been clearer.
What I am seeking to do, is to make the original poster aware that his wish-list is a difficult one, without taking some extra steps, or reducing his expectations. I know his transformer is a limitation, but there are even ways round this;)
Hi John,
I don't know if my brain has gone to sleep so please excuse me if I'm wrong but surely if the input voltage is 55v plus half the ripple say another 2 volts, then the average volts drop down to a regulated 50v is (55+0.5*2)-50 = 6v.
The max current is 8A giving power lost = 48w. Manageable.
When an excessive load develops the foldback reduces the output current to maintain a sensible power dissipation in the pass FET.

I am not advocating Reg output stage, just suggesting a solution.
I personally prefer reg supply to input & drivers and brute force PSU to the output stage.

I`v heard about that book but it`s expensive and not on my mother language so it takes too much time to study it.
I think that straight line current limitign is enough and i want to keep schematic simple as possible.

I will use multiple transistors in parallel to reduce power of them.
Yes power dissipation is still a problem but if voltage is +-50V and current is 8A power warms load and power dissipation of transistor is 25W (4 parallel).

Worst case scenario of dissipation is a short circuit of outputs
8Ax60V=480W (voltagedrop of transformer not included)
That means all 4 transistors have to dissipate 120W exremely much i think but i will never use supply with low voltages and high current i have no need for that.
There will be over heating protection because of that.
These transistors are bipolar

It is a bench supply for testing all kinds of stuff

I can design power supplly with these properties but not with
currentlimiting which will work allways no matter how load is connected.
(For example +V to ground, -V ground, +V to -V.)

Please correct if i`m wrong in some case

thanks for clarifying that you are building a bench supply.
I was on the wrong track.
All those nice transistors going to waste!
You could intercept each winding and take off lower voltage taps.
Then change the tapping as you change the output voltage.
My bench supply has relays that do the switching automatically. As I ramp up the voltage with the control pot the relays come in & out selecting the appropriate voltages.
This arrangement reduces the dissipated power considerably at lower output voltages and allows the use of a constant current control. The constant current control should be adjustable to protect a faulty circuit.


diyAudio Moderator Emeritus
2002-06-10 12:01 pm

Current limiting is normaly done by including a small value resistor in the current path, either in the +, - or ground lines, then measuring the voltage across it. (The voltage regulation is not affected by the presence of the resistor if the error voltage signal is taken after the resistor).
The voltage across this "current sense resistor" can be measured best by an opamp, but that requres external floating supplies. An simpler way is to use a small bipolar transistor which is biased by the voltage developed across the resitor.
Here is an example

Language: I believe that your grammar is rather different from European languages, so understand your reluctance to read Horowitz & Hill. However, if you can understand my muddled and sometimes confusing English, you will have no trouble to understand the clear and concise text of the book:)