My MOSFET amplifier designed for music

Administrator
Joined 2007
Paid Member
Hello PA,
To get the speed from FET's requires quite specialised drive circuitry. The gate is seen as essentially capacitive and it's how quickly you can pull charge to and fro thats the real problem.
Driving even 1nF at say 50 Mhz with a 10 volt swing takes some serious consideration.
Regards Karl
 
Mooly said:
Hello PA,
To get the speed from FET's requires quite specialised drive circuitry. The gate is seen as essentially capacitive and it's how quickly you can pull charge to and fro thats the real problem.
Driving even 1nF at say 50 Mhz with a 10 volt swing takes some serious consideration.
Regards Karl


You really dont need 50MHz in an amplifier, its more like to cause more problems than it fixes.

If it works to 50KHz at full bore then it will drive most speakers and tweeters.

I get away with whacking 12mA into 6 MOSFETS and my amp sounds great.
 
Mooly,
I suspect going back to a single ended input stage has a lot to do with it
In fact, that's exactly the reason!

Hello PA, To get the speed from FET's requires quite specialised drive circuitry. The gate is seen as essentially capacitive and it's how quickly you can pull charge to and fro thats the real problem. Driving even 1nF at say 50 Mhz with a 10 volt swing takes some serious consideration. Regards Karl
Now, how much current do you think is required to handle 600-900pF input capacitances? And how much is provided?
 
peranders,
Can you give me any numbers? An source follower built with lateral mosfets in a real circuit has a bandwidth of a couple of MHz. How much do you get for a vertical mosfet follower?
It would be more beneficial to start exploring the properties of the lateral and vertical structure. After that you may never think about using verticals in audio circuits again.
 
Administrator
Joined 2007
Paid Member
Hello Alex--- That looks fantastic to see it like that. I will have a much closer look later but it all looks good. Have you etched a board yet or is it still all in the machine as it were.
I see you have gone for a film cap for the input :) and a star ground on the PCB itself.
The only suggestions I could offer are to make sure the 220mfd rail caps are physically big enough on the PCB, on my test DipTrace layout I should have gone to the next can size up.
The other point is the print around R14, the 22 K feedback resistor. I learnt this from Doug Selfs articles some 15 years ago now, that the "correct" take off for the speaker and the feedback point should be as close together as possible. It sounds so obvious but so many amps fail on this one point alone. The current distribution through that print is anything but constant due to it's small but finite resistance, particularly so in a class AB amp where distortion from the rails creeps in. It's the most important connection in the amp. If you could move a probe around that print with the amp working the distortion characteristics would vary greatly. If you look at my layout that "spidery" connection is an attempt to get the two points as close together on the board as possible. The junction of R14 and R26 (are you putting your R26 off PCB, which is fine but again it's a vital part of the design) is the ONLY point in the amp where the output is corrected by the feedback and is a true representation of the input.
--- I see you have both a preset and R8. I put two R8's on mine so I could get a closer value to whatever the pot setting was. It's just me-- I don't like presets, particularly when there are a few milliamps going through the wiper. As I say -- you just don't know what they get up to when your not watching"
Fantastic job though.
Regards Karl
 
nigelwright7557 said:
I reckon it would have to be glued onto a MOSFET as the heatsink I use is huge and just never gets warm but the transistors do if worked hard.

In this circuit, the bias transistor should not be mounted in contact with the main output devices it should be mounted in contact with the drivers. This is because with the CFP type output stage, the local feedback loop takes the output thermal changes out of the equation. In fact if you sense the output temperature, the thermal stability will be very bad. Ideally the drivers should have their own little heatsink.
 
Mooly said:
Nigel,
You mention ground planes. I have never really looked into this as far as audio goes ( audio power anyway ). The technique does certainly have it's uses though.


You can eitehr add another layer with a ground plane or simply add a ground plane on the same layer as your current copper that connects to ground.

It costs nothing to do if using the same layer as your other copper, but decreases the ground impedance.