When witch came.......

[hugobors]

Hello,

Let explain the problem.......
I'm trying to make a buffer with only one stage, and very low output impedance (shematic attached).
In static, no problem, I can vary my courant source, all voltages are OK.

The problem is if I connect anything on the output (yes, just a cable), the minus pole of my lab power supply just pass in the current limiting mode........... but the positive stay OK and the current source still work.
So I tried to connect a loudspeaker, and it work........
Maybe witches.......... but I'm sure there is some ghostbusters that can help me.....

Thanks
Hugues

[Wavebourn]

If you connect a scope I suppose you can catch up your witches flying very fast...

Try to increase resistors in gates. Also, I'd recommend you to use BJT in a CCS:




The CCS in my amp follows input signal, so idle current (as you may expect, is defined by current through resistor in emitter multiplied by beta of the power transistor) is only half ot the peak output current. It will save you electricity and money.

[hugobors]

So you think the problem can be hf oscillation???

[lineup]

Originally posted by hugobors

I'm trying to make a buffer with only one stage,
and very low output impedance (schematic attached).

In static, no problem, I can vary my current source, all voltages are OK.

The problem is if I connect anything on the output (yes, just a cable), the minus pole of my lab power supply just pass in the current limiting mode........... but the positive stay OK and the current source still work.
So I tried to connect a loudspeaker, and it work........

Maybe witches.......... but I'm sure there is some ghostbusters that can help me.....

MOSFET or HEXFET it is

But what transistors exactly?

And if it is a buffer,
which really looks like it have potential to work,
I get it is 2 P-MOS.
You have gate resistors 200 Ohm ... Very good!

How much current do you use, like?

That resistor after the Output Cap, what value has it got?
I would suggest you use like 1.0 - 2.2 kohm as a dummy load,
if nothing else is connected.


Regards
lineup

[hugobors]

Thanks lineup,

It's some hexfet.

The load after the condenser is 200K but just to try (little condenser to achieve bandwith), the definitive version will have something like beetween 200 and 500 Ohm to drive.

I did resolve the problem this afternoon by adding a 100nF condenser beetween base and drain of the fet down (not the current sourcing one). So problem seems to be oscillation. It is working but I have to listen now.........
Do you think that this condenser can be bad for sound??? and is there an other way to stop the oscillation? (if there is some but I only have a handled scope with something like 30Khz bandwith)

Thanks

[Wavebourn]

Originally posted by hugobors
So you think the problem can be hf oscillation???

Yes.

[lineup]

this is a good method
when adding compensation caps:

1. Start with adding 100nF or some cap that will make amp stable.
2. Decrease, lower cap value and try with 47nF
3. If still stable, try next 22nF and so on
4. Until amplifier starts being un-stable.
Say oscillation will happen when you come down to 4.7nF
5. Then use 10 nF cap ( 2x4.7nF) to have a fair margin

This above is the procedure I use in my spice simulation
and when doing final trim of amplifiers I build.
--------------------------

100nF seems very much. So either you can do with less
or you may find a better, more effective, place to add some compensation capacitance.
You should defenitively try
From GATE to Ground of the same transistor.
Depending on what HEXFET you use
you can also try to increase value of GATE stopper resistor.
Maybe 470 Ohm .....

--------------------------

You should also not forget to add a little cap from input to ground
to reduce the amount of HF noise that can enter amp.
(This is almost the same as from GATE to GROUND.)
This will limit the upper freq of amplifier a bit.
By doing so, you may not need as much compensation inside amp.

Something like 220pF - 1nF is often used.
In my amps I mostly use 1nF film caps.
It is BETTER and much easier to stop problems to enter a circuit
than to try to remove them, when they are already in circuit.

lineup

[Wavebourn]

Increase the resistor value in gate. You don't simulate for The Matrix, you build the real thing, so have to face the reality.
Your cap screws down frequency response of your amp.