[sdiy] Simple discrete Unity-Gain Follower ?

[Magnus Danielson]

From: Michael Buchstaller <buchi at takeonetech.de>
Subject: Re: [sdiy] Simple discrete Unity-Gain Follower ?
Date: Wed, 30 Apr 2003 18:48:18 +0200

> >That one will not really cut it for linear purposes, since there will a gray
> >zone where the input is between the points where either of the upper and lower
> >transistors is really biased.
> 
> Yes, i thought that this ultra-simple approach would not work as intended.

I think that for most of the case you can assume that if it where simple it
would be used. This is true especially when you look back in time and over a
varity of machines. You learn alot by looking at state-of-the-art designs from
60thies, 70thies and 80thies. Not everything directly ripples over to sound
design-decission today, since some things have changed where as others have
remained.

Personally I *really* enjoy studying measurementinstrument schematics/service
manuals. Things like that helps to bend the mind. Also, there are loads of them
if you care to look around.

PS. Don't throw manuals to old instruments. Make sure they get read and used.

> >Then naturally, you can always toss a feedback-loop onto it, but that brings
> >a whole new realm of problems.
> 
> I do not want an output that easily overshoots or oscillates, or behaves
> strange. So feedback shall be avoided, if possible.

The benefit of feedback should not be underestimated just because there is
additional concerns, it's only about appropriatly applying the technique.

> >However, a properly designed amplifier may very well be op-amp free,
> >no problems!
> 
> I never wanted to say that one could not design a very good amplifier without the
> use of Opamps. But without diving too deep in theory, using OP´s makes life much
> easier...

Indeed. However, since much of the stuff is there anyway, it later comes to
haunt you when you least expected it, and then you sit there anyway agonizing
over reading up on theory. Life is hard.

> But i do not really like the idea of an all-transistor discrete circuit just to have
> an Opamp follower at the output.

Why don't you build a transistorised op-amp? It's fairly easy to do actually
and should give you enought umph for the drive. With 6 transistors you can
build yourself a balanced op-amp having a class A output stage. What you do is
to use 2 input pairs, one of NPNs and one of PNPs. Each has a common emitter
resistor to one power line and separate collector resistors. Then you drive the
output by feeding the output transistor with collectors to power, emitter as
output and then base from the collector/resistor junction. Naturally you use a
feedback-loop such that you feed the design on the positive input and connect
the output to the negative input. 6 transistors (3 NPN, 3 PNP) and 6 resistors
and you have yourself an output drive for your all-transistors project.

If you want to make a little more elaborate exercise, replace the common
emitter resistors with a current source. It is recommended to use a current
mirror so that both positive and negative side operates on the same current.

There are naturally many ways to implement op-amps, but BC-550C and BC-560C
and common resistors get you going the right direction pretty darn quick.

> >The emitter-follower isn't that bad either, it have done marvels in the
> >computer field (CDC 6600, 7600, 8600 and then the Cray-1 to mention a few).
> 
> Yes, but for audio i would prefer something else. Especially when i think
> about the low load resistance that will be required to give low enough output
> impedance - i do not want to hear this output stage over the power
> connections everywhere in my synth !

Why not, make sure you make yourself heard! ;O)

Cheers,
Magnus