My first attempt at amp design

[VEC7OR]

So this is my first attempt at discrete amp design...
Output stage is MOSFET (its hexfet and irf because those ones are cheap and readily available). I didnt want to use BJT because of secondary breakdown issues (kinda bored looking at my friends burned transistors, yeah they are cheap but its wrong anyway).

+/- 52V supply was used because I have a 40Vx2 450VA toroid laying around and want to design something like ~300W 1 channel into 4ohm.

I've decided to drive mosfet this way to get maximum voltage swing on the output (I havent seen this method used much, why ?)

So far I've managed to squeeze 3W out of it @ 0.04% THD and 300W with 8% THD with ~34dB of gain.

Now how can I track DC bias to 0 here ? (its simple with conventional amps) It's possible to set it to 0 in the sim but in real life it wont be easy.

How do I apply feedback ? (I havent studied it yet)

What else can be added to improve performance and stability ? (I think about adding input stage, current sources etc)

After reading the forum I've figured out that best looking pair of mosfets are IRFP9140 and IRFP240(I think about 2 or 3 per side of those), transconductance is similar, so symmetry will be better, only one thing is bothering me, one of them is rated 100V other one is 200V, do I risk using those ant get positive side blown because of across 104V PSU or I'm tripping ?

P.S. Most of the part will be others than shown (I dont have all the models at this time)

[AndrewT]

Hi,
3 comments:
Why are R9 & R10 different?
40Vac = 55vDC loaded = 58Vdc to 59Vdc unloaded.
Do you need a load between Q2 and Q3 emitters and to control their quiescent current?

[richie00boy]

I've a feeling that will be more of a firework than an amp. It will be too difficult to stabilise and get DC offset acceptable in real life with real toleranced parts.

[ChocoHolic]

...was thinking about something similar some months back...
(and still did not cancel the idea)
I would propose degeneration resistors for Q1 & Q4 + massive cooling (or some sort of temp compensation) + quite a careful
hand when adjusting the idle current in real life.
Feedback could be applied by a resistor between the output and the common emitters of Q2/Q3.
Hm... as soon as you apply feedback it could start to oscilate, but luckily it looks like your Q1 and Q4 output stage would form a pole
with their reverse capacities and the gate drive impedance...

...nevertheless: Stable operating points will be a specific challenge for you with that circuit....

It can easily happen that Richie's words are the short form of your
practical trials.

Anyway good luck!

[CBS240]

Maybe you could build 2 series zener regulators(for cost) or use regulator IC's, and make like +/- 15V or so. Then you could use an op amp circuit as a differential amplifier, and use negative feedback for error correction and get better than 8% TDH. No transistor is linear so almost all SS amps have some sort of feedback. Although with the wrong op amp, the slew rate will suffer and it would be better to build a descrete differential amplifier, but don't forget the DC blocking capacitor. AndrewT is right about emitter resistors on Q2 & Q3. This makes the effective Re difference much less, making the transistors act like they are better matched. Also another resistor from the output (drains) to the emitter resistors on Q2 & Q3(like a voltage divider with the 30 Ohms to GND) might help with stability. 500Ohms maybe??

Would there be a need for a Vgs multiplier (servo) for thermal stability in this type of circuit?

[mastertech]

looks good and should sound good

[CBS240]

A point about hexfet devices, they are designed usually for switching purposes and are not very linear. There are good complementary power fets for audio available.

[Leolabs]

Like the 2SJ/2SKs from Hitachi