Fast drivers used as output devices?

[AndrewT]

Hi,
I just ran some numbers on a 50W into 8ohm power amp output stage.

6pair of 2sb649/d669 can push 8r to 60 degree phase angle, 6r to 47degrees phase angle and 4r to 26degrees phase angle all within the DC SOAR and Tc<= 35degC. This would make a reliable and cheap 6 to 8ohm speaker amplifier.

This seems to indicate that a power amp with ALL devices faster than 100MHz is possible.

Is this a silly notion or could a fast amp offer some advantage in the sound stakes?

[richie00boy]

Stocchino would say yes, many others would say not really.

My take is that the faster the amp the more really high harmonics will be able to be dealt with (i.e. minimised by feedback). One problem though is that of stability and RF pickup.

[Samuel Jayaraj]

Andrew, if you had 2N5551 and 2N5401 as drivers and IRF9640/640 or IRFP9240/240 for the output stage, do you end up having an ultra fast amplifier with a very wide bandwidth. If the answer is yes, then look into the Stochino amp schematics.

The sound is fabulous, to say the least; the best I have heard till date.

[Samuel Jayaraj]

richie00boy, you just beat me to that reply.

[jcx]

faster output devices should allow higher loop gain

somewhere package parasitics start limiting performance

MOSFETs can be faster still - http://www.ixysrf.com/products/hf.html


Compensation can have a larger effect on how much loop gain you can have with global negative feedback

with single dominant pole compensation the loop gain can increase at the same ratio the compensation gain crossover frequency can pushed up before the limiting 2nd pole behavior sets in

even better improvement can be had from more complex compensation - Bode's "maximum feedback" compensation allows 30 dB/decade instead of dominant pole's 20 dB/decade gain slope

Self just touches on 2 pole compensation without enough analysis to show how to safely use it

Halcro seems to be trying to patent 3rd order compensation 60 dB/decade gain slope - this "conditionally stable" compensation is guaranteed to oscillate with output clipping unless you know a lot about this very obscure corner of control theory and add a fair amount of "extra" nonlinear compensation circuitry to control clipping behavior