Mosfet push pull as precision lab supply

[Spikee]

So I have been toying with the idea of making a -30 to +30V linear programmable power supply. I have tested the OPA549 High power opamp to do this thing but I was disappointed by the current regulation and the thermal performance (1.4 deg c/W JC-CS) thus with an extremely good heatsink (<= 0.1deg c/w) It was only able to sink / source 80W max.

Then there is the LT1970 circuit that is floating around. I have trown it in LTSPICE and have done some simulations. The problem that I have with it is limited voltage range and the high price of the part.

When Kevin.D answered in my previous OPA549 topic that it might be better to build a discrete mosfet push pull circuit I got interested... So what do you do when you want to find more info about such a method ?? Lets open The art of electronics third edition . So after a bit of researching I opened LTSPICE and started simulating some stuff.

After some  and  I simulated this:



In the simulation it works good and accurate on high and low currents and voltages, sinking and sourcing of current (four quadrant operation).\
Sadly I had none p channel mosfets to test it out on a breadboard so the simulation is all I can do at the moment.

I have yet to add current limiting to the ltspice simulation. I have ordered some P and N channel mosfets to do breadboard testing.

What do you guys think?

note: I'm not good at this stuff

ltspice file:
https://mega.co.nz/#!8MczBDTK!wTBD6CJnTTrEex-II5YXSoH-Y1BeaP4jr0anc4BG90w

The simulation software is ltspice. To use my simulation you need the PSPICE parts from IXYS (for installing search for LTSPICE third party .lib on google).

[mzzj]

What kind of frequency response you are expecting?
If you want more than say couple of hundred hz you probably better have buffer stage between mosfets and your opamp as current requirements rise pretty quickly. Also might be wise to consider biasing the output stage like class-AB audio amplifiers to reduce output impedance and improve frequency response.

[Spikee]

I'm net aiming for 400W worth of programmable waveforms just basic precision lab power supply. Around 100/200 Hz is kinda what i'm aiming for.

[wiss]

TDA7294? good for +- 35 V, 10A.
Or a current-boosted TDA2030? +- 15 to 20 V, infty A.

[Spikee]

those are some interesting chips, i'll take a look at it.
I'm wondering how good that will work as a lab power supply.

[dannyf]

One, you just recreated a linear power amp.

two, do an ac analysis on your circuit.

[Spikee]

One, you just recreated a linear power amp.

two, do an ac analysis on your circuit.

I know it is an linear power amp . I was just wondering if a audio amp that is designed for  4/8 ohm resistive load will work well with other loads.
Here is the AC analysis (1V amplitude, 1Hz - 100KHz 0,2Ohm load):


Edit:
I looked up the Ac analysis of the TDA7294 (at least i think this is the correct graph ...) and it looks much better  :

[MK]

You might find that a class G amp with multiple rails will let you do the same current range with a lot less heat sink.

[wiss]

And, you have to make sure it is stable with a capacitive load.
I have seen audio power-amps being stable with 8 ohm // 1 uF but unstable with 8 ohm // 10 nF load...
didn`t figure out why, just fixed it with a zobel-link.

[Spikee]

You might find that a class G amp with multiple rails will let you do the same current range with a lot less heat sink.

Thank you for the suggestion. My bench supply will also work as a dc load so it has to burn a lot of heat anyway. But it might be useful for the psu side of it.

Wiss:
Thanks for mentioning that, i ordered the most simple tda7294 board i could find locally to do some test.  That 'zobel link' stuff sounds interesting.
Do you have any circuit diagram of that?

[wiss]

It is just a capacitor in series with a resistor to load the output. The resistor should be about the intended load and the cap so small so it does not load the amp at the intended frequency-range. 10 ohms + 100 nF for a normal audio-amp, or so.

[rob77]

one of my next projects will be a 2 quadrant supply - and i'm considering the TDA7294 or TDA7293 as well - it's dirt cheap (the 7294 is 2 euros in single piece) and it should do the magic without significant issues