🚀 go-pugleaf

In article <1990Oct21.171736.9918@wsrcc.uucp> wolfgang@wsrcc.uucp (Wolfgang S. Rupprecht) writes:
>>In article <271c8c45-4b1.2sci.electronics-1@vpnet.chi.il.us> cgordon@vpnet.chi.il.us (Crash Gordon) writes:
>>>Hardcore audiophiles claim that a tube amp sounds better than solid state.
>
>In sci.electronics you write:
>>Yeah, it's so hard for a transistor circuit to reproduce all those warm
>>fuzzy nonlinearities.
>
>;-)
>
>Actually the truth is probably that it is much easier to design a
>*bad* transistor amp than a bad tube amp.

What does this mean?  It takes a lot of knowledge to correctly design a 
reasonably complicated audio amplifier, regardless of the devices used.

>The real problem is that a transistor stage with high gain and no
>local loop feedback (ie. no emitter resistor) is *very* nonlinear.
>Much more so than a tube stage constructed similarly.

Sorry, but analog circuits used for operational amplifiers are considered
weakly non-linear otherwise they would not be useful for many applications.
Have you ever seen what happens if you run a vacuum tube amplifier into
clipping?  Try it, it's educational.  Also, connect to a spectrum analyzer
if you have one and you will see the rich collection of harmonic content.

>Many of cheap transistor amps run each stage at close to open loop
>gain of the transistor.  They then fix the strong nonlinearities with
>a feedback loop around the whole amp.  This is what I call the Phase
>Linear approach - shove a 741 in there to get .00000000000000001 %
>distortion.

One could only wich the distortion could ever be so low.  What are you
trying to say anyway?  The whole *idea* is to use the *high* transconductance
of bipolar devices in op-amp designs.  This way, the loop transmission can
be large leading to small closed loop errors.  

>It works quite well for carefully selected test
>conditions, that is ones that leave the every stage biased in a linear
>enough region of its curve (eg. 1khz sine wave).  Unfortunately the
>first stage, the one that does the error difference subtraction, is
>very easy to overload if we task it with the chore of multiplying any
>error by 10**5.  How does one overload the input stage?  Simplest case
>inputting a square wave.  For the time that the output slews from low
>to high the amplifier is *not* modelable as a linear system.  Any
>small change in input will not result in *any* change in the output.

Use a larger slew-rate.

This is 1990.  Stop using the 741 for everything.  Open your eyes, and your
ears, and LEARN what modern integrated-circuit processes can now accomplish.

Also, check out the large signal square-wave response of a vacuum tube 
amplifier sometime into a purely resistive load.  A non-ideal step
response will be readily apparent.  Try this with a reactive load also.

>This is the "transistor" sound.  The sound of a bone-headed designer
>that just learned which end of a transistor went to which rail. ;-)
>
>The solution?  Do a good job on feedback on each stage, and keep each
>one of the stages linear by using lots of local feedback.  Don't try
>to patch it all up in the end.
>

Use a larger slew rate.  Use modern op-amps.  Do not use lateral p-n-p's.
Do not compensate your op-amp for unity-gain if the closed-loop gain is
40.  Do not use a 741 for every single op-amp job in the world.

>Tube amps all have very limited gain in all the stages, and little or
>no global feedback.  This is all becaues tubes have very limited gains
>at best.  Notice how they force you to do a good job.
>

They do no such thing.  Vacuum tube amplifiers sound more melodious than
solid-state amplifiers and that is *all*.  

>You still need to add the "tube distortion curves" is you want the
>transistor amp to sound like real tube amp while clipping. You can
>probably use a large bag of diodes and resistor ladder to model this. ;-)
>

Why bother.  Use vacuum tubes and you will be happier as will I.  Personally, 
I prefer not to clip my audio amplifiers.  If you use solid-state op-amps,
get one designed recently that has a reasonable slew rate so the input stage
worries are not important.  This does not mean that 2000V/usec slew rate is
needed for audio, unless the tweeter has to settle out within 25ns.  

>-wolfgang
>-- 
>Wolfgang Rupprecht    uunet!{nancy,usaos,media!ka3ovk}!wsrcc!wolfgang
>Snail Mail Address:   Box 6524, Alexandria, VA 22306-0524

Andrew Karanicolas
MIT Microsystems Laboratory
ankleand@caf.mit.edu