F5 power amplifier

[jackinnj]

Originally posted by juma



Read my mail again
It's 9.9 cm x 3.5 cm (1" = 2,54 cm)

I'll print it into Photoshop CS2, and do a PDF at 100%, but you'll probably have to get rid of the jaggy's from going to BMP. I'll post a PDF here.

[jackinnj]

Here's the PDF to scale, but PM me for a PDF of the BMP which has no gray-scale.

[jupiterjune]

One dead bug channel of F5 is up and running.

Not having the correct devices on hand, I had to use some old junk-fets I had laying around, in this case a 2n4416A and a 2n5116. They had Idss's of 9ma and 12ma, closest match to the recommended I had on hand.

A 10khz square wave shows no overshoot, and looks very square.

Yes, that is speaker wire you see attached. After letting it warm up for about 40 minutes, it seemed very stable, so I hooked it up to a speaker and listened to one channel of Dido.

First impression was that the F5 does an amazing job of reproducing every aspect of her voice -- which is simply amazing.
Very detailed amp-I can't wait to do the other channel. (even though it is surely nowhere near one built with the right j-fets)

JJ

[jackinnj]

Too late to take a picture, and I want to sit down with the XYL and watch some TV (it's almost 11 p.m. here in the People's Republic of New Jersey), but using separate gate input resistors has had a most salutory effect.

[Tea-Bag]

Jack, Thanks for shinking this.
Jumo, thanks for doing the layout work.

[Nelson Pass]

Originally posted by jupiterjune
Not having the correct devices on hand, I had to use some old junk-fets I had laying around, in this case a 2n4416A and a 2n5116. They had Idss's of 9ma and 12ma, closest match to the recommended I had on hand.

A 10khz square wave shows no overshoot, and looks very square.

...(even though it is surely nowhere near one built with the right j-fets)

That the ticket. I'm sure that those JFETs work fine although they
do look to have a lower transconductance than the 2SK170/2SJ74.

[EUVL]

While your article describes the basic circuit brilliantly in layman terms, it contains little details on how it is designed for stability given the high bandwidth. After so many informative articles on the Zen series plus other First Watt designs, perhaps you could shred some light on how you go about designing for frequency related performances ?? And I guess F5 would make a great example for that purpose, specially in the light of the difference in peformance as measured by jackinnj than your own .....


Looking forward to your enlightenment,
Patrick

[jam]

..............enlightenment.

http://www.do-not-zzz.com/

Jam

Sorry! I could not resist.

[Nelson Pass]

Originally posted by EUVL
And I guess F5 would make a great example for that purpose, specially in the light of the difference in peformance as measured by jackinnj than your own .....

Why don't we start with his results?

[EUVL]

> Why don't we start with his results?

Sure.

To achieve decent square waves at 100kHz as shown in your article, one would guess that one might want to get the closed loop bandwidth at something like 1MHz plus. As he showed in his measurements, the frontend, with its low impedance gate and source impedances, would probably have a corner frequency beyond that. Thus, it is probable that the first RC dip in open loop comes from the 600 ohm drain resistor and the input capacitance of the MOSFET. If this is far enough apart from the corner frequency of the input stage, then we should not get 180 deg phase shift when closing the loop. Correct ? One would guess that f1/f2 wants to be larger than the amount of negative feedback (say > 35dB).

However, the F5 is essentially two amplifiers mirrored about ground, and the frequency response of each half is different. So I guess this explains why there are more than 1 peak in the frequency response. I have not worked out all the RC values yet (if I remember correctly, the MOSFET gate has a corner frequency of something like 30-40 kHz), but assuming the same devices, I wonder why he is getting such different response than you did ??


Patrick