Revisiting my bench power supply project.

[Refrigerator]

Hey.
So a couple months ago i did my first bench PSU, but due to the lack of knowlege and parts i abandoned it.
Now i want to continue building my PSU and i have sorted out some things with the parts and gained some more knowlege. 
So far i made up a little schematic for the main voltage regulator circuit.
http://gyazo.com/369b707b723b4717acfbfbb91bcfdfaf
R7 will be a multiturn pot like this one:
http://www.ebay.com/itm/261185316819?ssPageName=STRK:MESINDXX:IT&_trksid=p3984.m1436.l2648
And the enclosure for the PSU.

The binding posts are for +5V, +12V, -12V and the variable voltage output.
This transformer has two ~15VAC and two ~8VAC windings ( if i remember right ), so i could use it to power the whole thing but i might fit another transformer in for the regulated 5, 12 and -12V outputs.
The last time i got my older circuit to work it managed to deliver 20V@5A which would translate to about 100W which is not that bad.
I want this to be a constant voltage power supply.
Any help or advice is appretiated.
Also feel free to criticize my circuit as much as you want.

[tautech]

Seems overly complex for just the variable PSU part.
Just use an LM338 and apply as per datasheet. 5 A & SC, OL & thermally protected. Easy
Presumably the fixed voltages are just tapped off the rectified secondary and use LM7*** regs?

[Refrigerator]

Seems overly complex for just the variable PSU part.
Presumably the fixed voltages are just tapped off the rectified secondary and use LM7*** regs?

I always tend to overcomplicate stuff and yes the LM7xxx regulators are tapped off the other ( lower voltage ) secondary but i want to add another transformer just to ease the work on the big one ( it's going to be working hard ) and because i have one laying around.
But i've already got most of the parts and i'd like to design one myself because i won't gain any experience if i just go the easy route and copy what is already done.

The further i got into this project the more i realised that my enclosure was getting too small but i think i'll gave a go at vertically attatched boards to save some space.
Edit: i'm thinking something like this :
http://gyazo.com/5f57de01eb5e6e41d76637dc46bf8d6a

[homebrew]

Hi Refrigerator,

my two cents on your circuit ...

1) The 7805 NEEDS bypass caps to operate stable. Just give it some 100n on both sides.
2) You NEED some large filter caps after the rectifier, especially if running at 5 amps. Without having calculated this I would recommend something larger than 22'000 uF ...
3) Why having the zener regulator for the opamp  and a separate one as a reference? Maybe you want to use a proper reference (compensated zener diode) instead.
4) Without any caps in your circuit I have my doubts that his design would be stable. Here is a really good explanation on this topic: http://www.ti.com/lit/an/snva020b/snva020b.pdf

And of course: Be careful with the mains !!!
Good luck!

[MrAl]

Hey.
So a couple months ago i did my first bench PSU, but due to the lack of knowlege and parts i abandoned it.
Now i want to continue building my PSU and i have sorted out some things with the parts and gained some more knowlege. 
So far i made up a little schematic for the main voltage regulator circuit.
http://gyazo.com/369b707b723b4717acfbfbb91bcfdfaf
R7 will be a multiturn pot like this one:
http://www.ebay.com/itm/261185316819?ssPageName=STRK:MESINDXX:IT&_trksid=p3984.m1436.l2648
And the enclosure for the PSU.

The binding posts are for +5V, +12V, -12V and the variable voltage output.
This transformer has two ~15VAC and two ~8VAC windings ( if i remember right ), so i could use it to power the whole thing but i might fit another transformer in for the regulated 5, 12 and -12V outputs.
The last time i got my older circuit to work it managed to deliver 20V@5A which would translate to about 100W which is not that bad.
I want this to be a constant voltage power supply.
Any help or advice is appretiated.
Also feel free to criticize my circuit as much as you want.

Hi,

The schematic in your first link looks like it has positive feedback instead of negative feedback.  It should have negative feedback.

Simply put, the op amp inverts and the Q1 transistor inverts, and the other transistors do not invert, so that's positive feedback.
To put it another way, with a fixed reference on the non inverting terminal, when the output voltage increases the voltage at the inverting terminal increases, which causes the output to go lower, which causes Q1 to turn off more, which causes the collector voltage to increase, which causes the remaining transistors to turn on harder, which causes the output to rise even more still.  This causes a never ending ramp up until the output pins at the max voltage it can obtain.

One possible fix might be to make Q2 a PNP and change the biasing as needed, but i'll leave that to you.

[Refrigerator]

Hi Refrigerator,

my two cents on your circuit ...

1) The 7805 NEEDS bypass caps to operate stable. Just give it some 100n on both sides.
2) You NEED some large filter caps after the rectifier, especially if running at 5 amps. Without having calculated this I would recommend something larger than 22'000 uF ...
3) Why having the zener regulator for the opamp  and a separate one as a reference? Maybe you want to use a proper reference (compensated zener diode) instead.
4) Without any caps in your circuit I have my doubts that his design would be stable. Here is a really good explanation on this topic: http://www.ti.com/lit/an/snva020b/snva020b.pdf

And of course: Be careful with the mains !!!
Good luck!

Yes there will be caps, for now i didn't put them in my schematic, 6800µF 50V after the rectifier and 100nF all around the board.

[Refrigerator]

Hi,

The schematic in your first link looks like it has positive feedback instead of negative feedback.  It should have negative feedback.

Simply put, the op amp inverts and the Q1 transistor inverts, and the other transistors do not invert, so that's positive feedback.
To put it another way, with a fixed reference on the non inverting terminal, when the output voltage increases the voltage at the inverting terminal increases, which causes the output to go lower, which causes Q1 to turn off more, which causes the collector voltage to increase, which causes the remaining transistors to turn on harder, which causes the output to rise even more still.  This causes a never ending ramp up until the output pins at the max voltage it can obtain.

One possible fix might be to make Q2 a PNP and change the biasing as needed, but i'll leave that to you.

Now that i think of it it's true, i have plenty of PNP transistors so it won't be a problem, I think i'll go to LTspice and try to simulate my design.
Here i changed my schematic a little:
http://gyazo.com/ba4e5651779c089399d6c05348689187

[MrAl]

Hi,

The schematic in your first link looks like it has positive feedback instead of negative feedback.  It should have negative feedback.

Simply put, the op amp inverts and the Q1 transistor inverts, and the other transistors do not invert, so that's positive feedback.
To put it another way, with a fixed reference on the non inverting terminal, when the output voltage increases the voltage at the inverting terminal increases, which causes the output to go lower, which causes Q1 to turn off more, which causes the collector voltage to increase, which causes the remaining transistors to turn on harder, which causes the output to rise even more still.  This causes a never ending ramp up until the output pins at the max voltage it can obtain.

One possible fix might be to make Q2 a PNP and change the biasing as needed, but i'll leave that to you.

Now that i think of it it's true, i have plenty of PNP transistors so it won't be a problem, I think i'll go to LTspice and try to simulate my design.
Here i changed my schematic a little:
http://gyazo.com/ba4e5651779c089399d6c05348689187

Hi,

That looks a little better, but is that R6 a 10 ohm resistor?  Then i hope the base emitter of Q2 cant take 4.5 amps with that 45v supply.  Also, the power transistors may have to dissipate a lot of power.  At 1v output and 1 amp they have to dissipate 44 watts.  That's a lot.  You'll need a very big heat sink.

What is normally done for power supplies of this amplitude is a switching pre-regulator is used.  That lowers the power dissipated in the output transistors for the linear part.

I am not sure though what kind of output current you are looking to get.  Is it 1 amp, 2 amps, 5 amps, etc.  Even 1 amp though requires a big heat sink for the output transistors.

[Refrigerator]

Hi,

The schematic in your first link looks like it has positive feedback instead of negative feedback.  It should have negative feedback.

Simply put, the op amp inverts and the Q1 transistor inverts, and the other transistors do not invert, so that's positive feedback.
To put it another way, with a fixed reference on the non inverting terminal, when the output voltage increases the voltage at the inverting terminal increases, which causes the output to go lower, which causes Q1 to turn off more, which causes the collector voltage to increase, which causes the remaining transistors to turn on harder, which causes the output to rise even more still.  This causes a never ending ramp up until the output pins at the max voltage it can obtain.

One possible fix might be to make Q2 a PNP and change the biasing as needed, but i'll leave that to you.

Now that i think of it it's true, i have plenty of PNP transistors so it won't be a problem, I think i'll go to LTspice and try to simulate my design.
Here i changed my schematic a little:
http://gyazo.com/ba4e5651779c089399d6c05348689187

Hi,

That looks a little better, but is that R6 a 10 ohm resistor?  Then i hope the base emitter of Q2 cant take 4.5 amps with that 45v supply.  Also, the power transistors may have to dissipate a lot of power.  At 1v output and 1 amp they have to dissipate 44 watts.  That's a lot.  You'll need a very big heat sink.

What is normally done for power supplies of this amplitude is a switching pre-regulator is used.  That lowers the power dissipated in the output transistors for the linear part.

I am not sure though what kind of output current you are looking to get.  Is it 1 amp, 2 amps, 5 amps, etc.  Even 1 amp though requires a big heat sink for the output transistors.

Power transistors will be mounted on a rather large heatsink, LM7xxx reguators will all be mounted on another, ( not so big ) heatsink and the whole thing will be cooled with a pretty powerful fan ( at 12V has almost enough airflow to lift itself ).

[macboy]

Don't use a 7805 as the voltage reference. They can vary quite a bit with temperature, input voltage, etc. You won't get a nice stable output. I'd suggest a TL431 as a basic precision reference. Of course you can do much better, but many power supplies depend on these. Create a basic shunt regulator (think zener regulator) after the 7805 with a few mA of current; the 7805 provides a stable enough voltage to ensure that the current through the TL431 is basically constant, providing good regulation. Your voltage reference will then be 2.5 V, not 5 V.

Don't forget some output capacitance. Don't go overboard. 100 uF is likely enough; 470 uF is a lot. You may need to put a few ohms in series with this capacitor for stability, since many regulators begin to misbehave with a purely capacitive load, and modern capacitors generally have much lower ESR than they did years ago. Without output capacitance, load regulation (response to changes in load) may be poor.

In testing you may see overshoot, ringing, or outright instability. You may be able to easily tame these problems with a small cap from the opamp output pin to the negative input pin. Testing will tell you how much capacitance if any. You can see and play with the effect of this capacitor in a simulator, but you can't determine the optimum value, since many real factors aren't accounted for, like the actual gain of your transistors, parasistics in PCB and wiring, and many other non-simulated variables.

[Refrigerator]

Don't use a 7805 as the voltage reference. They can vary quite a bit with temperature, input voltage, etc. You won't get a nice stable output. I'd suggest a TL431 as a basic precision reference. Of course you can do much better, but many power supplies depend on these. Create a basic shunt regulator (think zener regulator) after the 7805 with a few mA of current; the 7805 provides a stable enough voltage to ensure that the current through the TL431 is basically constant, providing good regulation. Your voltage reference will then be 2.5 V, not 5 V.

Don't forget some output capacitance. Don't go overboard. 100 uF is likely enough; 470 uF is a lot. You may need to put a few ohms in series with this capacitor for stability, since many regulators begin to misbehave with a purely capacitive load, and modern capacitors generally have much lower ESR than they did years ago. Without output capacitance, load regulation (response to changes in load) may be poor.

In testing you may see overshoot, ringing, or outright instability. You may be able to easily tame these problems with a small cap from the opamp output pin to the negative input pin. Testing will tell you how much capacitance if any. You can see and play with the effect of this capacitor in a simulator, but you can't determine the optimum value, since many real factors aren't accounted for, like the actual gain of your transistors, parasistics in PCB and wiring, and many other non-simulated variables.

I'm thinking of using the LM7805CV, which ( if i remember correctly ) has better stats than the conventional LM7805 ( better stability, higher input voltage, etc...), I do have other regulators, such as KA78R12 which is a 12V LDO regulator and from what i measured it outputs 11.98V under load and didn't seem to change even when relatively warm.
Although it will need a pre-regulator ( like the one on the op-amp ) since it won't tolerate 45V ( the LM7805CV will, though ).

[Refrigerator]

I've just edited my schematic a little:
http://gyazo.com/b42b822ea9a4b192b9341afd3f603954
I added the KA78R12 as it seemed more stable than the LM7805, and i've also added a bias diode and resistor on Q1 yet i marked them with "?" because i don't know if i should leave them in.
For now i won't put any decoupling or filtering caps since there will be many on them and i want to save some space in my schematic intill i'm done designing it.
As usual, any help or advice is appretiated.

[MrAl]

I've just edited my schematic a little:
http://gyazo.com/b42b822ea9a4b192b9341afd3f603954
I added the KA78R12 as it seemed more stable than the LM7805, and i've also added a bias diode and resistor on Q1 yet i marked them with "?" because i don't know if i should leave them in.
For now i won't put any decoupling or filtering caps since there will be many on them and i want to save some space in my schematic intill i'm done designing it.
As usual, any help or advice is appretiated.

Hi,

It is starting to look better now :-)

As the other reader pointed out, these kinds of circuits almost always need some kind of compensation.  This will probably become more noticeable when there is some capacitive loading in addition to resistive loading.  A fix is to connect the small capacitor from output to inverting input, but also place a fixed resistor like 10k in series with the inverting terminal.  The resistor and cap make the LM358 look more like an integrator, which then gives you an integral control which usually works pretty well.  You want this RC time constant to be as low as possible but still prevent oscillation with various load types.

Another improvement once you get everything working well is to replace the LM358 with a faster type op amp.  This will give you faster regulation for step changes in load.  The compensation capacitor can then be reduced too which means it speeds up the circuit.

I am assuming that you have the right resistor values for the transistors so that the gains allow for the max output at full load current but dont overdrive any of the bases.

[void_error]

You might find a few ideas in this thread. However, only some things have been tested.

[Refrigerator]

Thanksfor help. 
Now i'm gathering various parts from parts bins and making a little list of what i've gathered.
Here's the transistors i'll be using:
Q1- BC457
https://www.fairchildsemi.com/datasheets/BC/BC547.pdf
Q2- 2SA1011
http://lib.chipdip.ru/239/DOC000239506.pdf
Q3- C2073
http://www.ic72.com/pdf_file/-/c2073.pdf
Q4, Q5, Q6- i had them somewhere, i'll post them soon.
Q7- C2229
http://www.audiolabga.com/pdf/2SC2229.pdf
Q8- C1815
http://pdf1.alldatasheet.com/datasheet-pdf/view/30084/TOSHIBA/C1815.html
Q9- C1815
Q10- C2229
Updated schematic here:
http://gyazo.com/430c7766682b2983e458ed92f995ebe0
I added some capacitors, and a LED circuit, which will be connected to a bicolor LED to indicate when there's no voltage on the main power supply ( +45V ).
The purpose of that circuit is to indicate when the main transformer/rectifier in not functioning ( blown fuses or diodes ), for example, green when there's voltage after the rectifier, thus everything is ok and red when there's no voltage, thus the fuse or the diode rectifier is blown, it will also indicate when if i get no voltage out but the light is green my regulator circuit would have failed, which makes it easier to find the fault.
That 20 or so volts will be coming from another tap/transformer, which will be supplying the voltage for the regulators.