diyAudio (
-   Solid State (
-   -   Power amp under development (

quasi 24th October 2004 01:35 PM

Power amp under development
2 Attachment(s)
This is a schematic of my latest amp. It is a quasi-complimentary design running mosfet outputs.

The prototype running on my bench with the rails noted delivers around 200 watts into 8 ohms. The amp is very quiet and seems to be very low in distortion (I have trouble measuring below 0.1%).

With the right power supply and FETS it could deliver maybe around 800 watts or more into 2 ohms. The protype is running IXYS26N50 FETS and 4 or 5 of these per rail would be up to it.

The one planned for my Hi-Fi will run IRFP450's. These seem to be more linear and approach clipping better.

Want more power - add more FETS (and power supply)

Anyway feedback and ideas are welcome.


Lars Clausen 24th October 2004 01:45 PM

quasi: Hi, and congratulations on your design!

There are a number of things that you should probably take a closer look at:

The MJE350 needs to build up about 14 Volts on the 100 Ohms in the bottom, this gives you 140 mA @ 70 Volts drop before the IRFP's kick in. 140 mA @ 70V is 10 Watts, i think well over the max limit of the MJE350.

Also to run the BC550's in the input stage with a 70 V collector potential, is a bit over the safe limit.
There are also a few other thing i would do differently, but then, we are all different ;)

All the best from


quasi 24th October 2004 02:16 PM

Hey thanks Lars;

In the prototype the input and driver stages are running the same rails as the output stage.

In my next amp the rails will be seperated as shown in the circuit. This is probably why I haven't blown up the MJE350, so your feedback has already been valuable. In the current setup it only develops about 4 volts @ 40 milliamps on the 100 ohm.

But with the rails tied the +ve rail FET clips before the -ve rail and this is my main reason for going to the higher driver stage rails.

I wonder if I can leave the negative rail tied and just lift the positive rail for the driver stage?

The BC550 is a mistake from a previous drawing. The input stage runs BC546's.

Thanks again & cheers

MikeB 24th October 2004 04:49 PM

Hi quasi !

I think even for the bc546 70v is a bit much, maybe you should
cascode them ? (The same for Q1)
The feedback-resistor (33k) should be same size as inputres (22k),
this reduces dc-offset and might eliminate the need for VR1, and
replace it with good metal-resistors.

You might consider adding basestoppers (~1k) to q1/q5, i had once
the experience that this kind of circuit can flip at startup, providing
full supplyvoltage at output. This can happen if the amp clips to
negative side, driving q5 into reverse bias. In this case q5 "sucks"
out all current via it's base from the 2diodes, shutting down q1,
keeping the amp in it's clipped state. This is some horrorscenario, but
possible. This destroyed one of my testspeakers. Only switching off
the amp pulls it out of this state again... (or q5 burns up)
To be really sure, use two separate ccs.


Zero Cool 24th October 2004 05:40 PM

Your driver stage is very different then i have seen. Most have a resistor tied between the 2 driver transistor and the gate drive is taken from either side of that resistor and its associated drive transistor.

I have never seen a drive system where the gate drive is taken from a V drop across a resistor to the speaker line and V- system before.


AndrewT 24th October 2004 09:36 PM

Hi Quasi
The upper FET Vgs accurately tracks the voltage drop across Re of q7. OK here, all quite conventional.
But the lower arm FET Vgs is trying to track the voltage across Re of q7 plus the voltage difference between driver stage -Vrail and current stage -Vrail. Any slight variation between these two voltages will modulate the current through the output FETs and thus the output voltage.
I believe the quasi output driver voltage must accurately follow any droop under load of the hi current rail. Unless the designer adopts a special circuit to accurately track the -Vrail voltage then the driver MUST run on a common -Vrail with the current output.
Comments from the knowledgeable most welcome.
regards Andrew T.

quasi 25th October 2004 12:06 AM

Thanks MikeB;
I have never experienced the fault you have described but I do see your point. I always have DC detect on the output of my amps and this saves the speakers. Yes I would probably lose bits of the amp. Seperate voltage references for these current sources is quite easy.

Thanks Zero Cool;
Transistor Q8 and it's collector resistor provides the bias voltage for the negative rail output stage. It is similar to a quais-complimentary bi-polar output drive but instead of driving the base of the output stage directly it provides a voltage ref. for the output FET.

Thanks AndrewT;
You are of course correct. The prototype on my bench has both the positive rails and negative rails tied together at +/- 60v and the reference for the output FETS is the same on both sides.

I want to seperate the rails and run a fully regulated supply for the input and driver stages at a higher voltage for two reasons.

1. I hope to achieve less ripple on the input and driver stages.
2. I want to be able to drive the FET on the positive rail fully and avoid it clipping some 5 volts before the negative rail FET does.

Note I haven't done this yet and indeed advice from Lars has probably saved at least 1 MJE350 and who knows what else.

I think I have 2 maybe 3 options;

1. Leave it as it's running now on common rails, forget about the assymetric clipping and be happy.
2. Change Q8's collector resistor to 350 ohms to provide 14v across it. I.e. 4v+(70v-60v). But then this stage would be assymetric.
3. Leave the negative rail tied to the 60v rail and provide regulated +70v to the other side only. This seems ok but I'm worried that I have missed a consequence of this.

Thanks all for your input.


quasi 25th October 2004 02:02 AM

Thanks MikeB again (forgot this bit);

Yes, the BC546's are probably just hanging on with around 58 volts across them (currently 60 volt rails). So yes a cascode will be used. Under this arrangement I'll drop around 50 volts across cascode BC546's and leave 10-20 volts across BC550's (they're quieter) for the diff. pair.

The voltage across Q1 is about 45 volts and is currently a BC546. This should be ok.

Your advice about the feedback resistor and input resistor is noted and I'll probably change the input resistor. I still want the amp to have a gain of around 33. I'll probably leave VR1 there so I can get near enough to 0.00v offset.

All of the fixed resistors in the amp are metal film (except for the output stage).

Thanks again for your valuable feedback.


Workhorse 25th October 2004 11:12 AM

Commentable thoughts
Hello Quasi I have seen ur Circuit which has following Mistakes in Practical world.

Input Transistors VCE limit is outreached.
The Voltage accross them are more than there specified limit instead of it use 2N5551/2N5401 Trannies.

The lower side Driver stage resistor is not tied to lower side mosfet source. A fatal mistake for destruction of ur mosfets.
No use of emmiter resistor with lower side driver.
You have to tie the driver stage with VCC's of mosfets and only the Voltage amplification stage is to be tied seperatly at +-70VDC.

Take a look at these mistakes, Hope u eradicte these problems
Workhorse Technologies.

MikeB 25th October 2004 12:31 PM

Hi quasi !
I think VR1 is not the best idea if looking for low noise/distortion.
Quality of pots is a catastrophy compared to good metalfilms.
As you replace REs of diffamp, this might be an issue, these resistors
are some of the most critical resistors.


All times are GMT. The time now is 12:10 PM.

vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2016 DragonByte Technologies Ltd.
Copyright 1999-2016 diyAudio

Content Relevant URLs by vBSEO 3.3.2