Ultra-High performance OpAmp-based MOSFET Power Amp

[tlf9999]

Got you, didn't I?

Well, enclosed is a conceptual design of a power amp using an opamp front end.

I have always wanted to see how well they will perform and how different they are from discrete designs. so here is my attempt and your comments are welcome.

Two issues that I had to deal with in this design:

a) insufficient voltage gain for high power applications: opamps run off low rail voltages. to solve that problem, I had to add a voltage gain stage consisting of Q1/Q2 and associated resistors.

Rather than driving them from the base, I decided to drive them from their emitter so that I can apply nested feedbacks.

b) stability: to get this thing to stable, I have to introduce local feedback and global feedback. thus the nested feedback here.

One really good thing about this amplifier is its clipping behavior: it is really soft, , un-SS like.

Also, most of the harmonies are 2nd and 3rd.

Now, the bad things:

a) significant phase shift: I haven't been able to find the cause of it. the amount of phase shift doesn't seem to change with the opamp used or mosfet's used.
b) relatively large thd: usually in the 1-1.5% range, unless you dial up the bias current to over 250ma.

Please let me know what you think of it and how you would improve it.

[tlf9999]

here is when it is overdriven.

[tlf9999]

on the issue of distortion, you can reduce it by a) running the driver stage at high current (how about 100ma?) and b) introducing some negative feedback here by using connecting resistors between base and collector of Q4/Q5. I used two 33k resistors there and it substantially reduced thd.

[amplifierguru]

Hi tlf9999,

long way to go down this path before it fits the title I'm afraid.

Let's see.

I like that R1 very practical. The OP amp gain is 10 in a local loop as long as a low Z source is connected. Why a local loop here, it wastes so much that could be applied globally and wastes thePSRR improvement of having all that gain before an injection point.

emitter drive is not so good for loading on the chip and current limiting following stages.

how about nesting the discrete part and save the rest for global NFB should give a far better result.

Cheers,
Greg

[CBS240]

You could try to eleminate Q3 & Q4 and turn the FET's around where the output is from the drains instead of the source, you can get rail to rail operation then, and eleminate a pole in the process. Place a pot across R4 & R6 to adjust bias? How does it work without R2 and just using global feedback? Any miller compensation needed? I'm sure other manipulations might also work. Shouldn't a ground be between V3 and V4?

[tlf9999]

Originally posted by amplifierguru
I like that R1 very practical.

yeah. so that I don't have to deleted it or add it in case I wanted to change its value. This way, a sufficiently large R1 just means it is not there. So are thos 1ohm resistors. convenience, nothing else.

Originally posted by amplifierguru
The OP amp gain is 10 in a local loop as long as a low Z source is connected. Why a local loop here, it wastes so much that could be applied globally and wastes thePSRR improvement of having all that gain before an injection point.

unless you can get it stable with just global feedback.

Originally posted by amplifierguru
emitter drive is not so good for loading on the chip and current limiting following stages.

that's why there is R5. the chip pushes about 4ma maximum, within its output capability I hope.

Originally posted by amplifierguru
how about nesting the discrete part and save the rest for global NFB should give a far better result.

only if I could get it stable.

Originally posted by CBS240
You could try to eleminate Q3 & Q4 and turn the FET's around where the output is from the drains instead of the source, you can get rail to rail operation then, and eleminate a pole in the process.

I have never had any luck with cfp type output stages, especially in a mosfet configuration. tough to get it to stablize.

Originally posted by CBS240
Place a pot across R4 & R6 to adjust bias?

bias is done via R7. R4/R6 in part controls the gain on Q1/Q2.

Originally posted by CBS240
How does it work without R2 and just using global feedback?

unstable.

Originally posted by CBS240
Any miller compensation needed?

probably.

Originally posted by CBS240
Shouldn't a ground be between V3 and V4?

no. you can, however, put two capaitors on the base of Q1/Q2 to improve the gain of Q1/Q2 and psrr as well. But it is not a have-to-have for this excercise.

[CBS240]

Hmmm how about a nested feedback resistor from output to where R4 & R6 are connected and a divider resistor from there to ground? or is this what guru was talking about. Don't know if that will help stability. Generally I haven't had much luck getting a lot of stability from an AB biased feedback pair drivers...Q2,Q3 : Q1,Q4 although it can be done, not my personal favorite. Lots o' gain though.

[amplifierguru]

Hi tlf9999,

"that's why there is R5. the chip pushes about 4ma maximum, within its output capability I hope."

Howsat? Looks more like 12mA to me given 15V supplies for the chip - a lot for the following stages to deal with and also your chip supplies.

Try grounding the junction of R4,R6 and running a feedback loop back to the top of each. Then go in from the chip to the bases for a high Z chip load. Then you need to work up the output nested loop and the global loop (having relieved the chip of it's local loop).


Your baby

Cheers,
greg

[CBS240]

If you do what Greg suggests, don't forget to take global feedback to neg input because the output phase will be reversed. and maybe Vin to the pos input.

[Eva]

That soft-clipping behaviour is just a side effect of very low open-loop gain.