Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!utgpu!water!watmath!clyde!rutgers!husc6!think!ames!aurora!labrea!decwrl!ucbvax!ucbcad!eros!max
From: max@eros.UUCP
Newsgroups: rec.audio,sci.electronics
Subject: In defense of solid-state devices
Message-ID: <1895@ucbcad.berkeley.edu>
Date: Sat, 26-Sep-87 22:46:59 EDT
Article-I.D.: ucbcad.1895
Posted: Sat Sep 26 22:46:59 1987
Date-Received: Sun, 27-Sep-87 20:44:24 EDT
References: <729@alliant.Alliant.COM> <4124@pyr.gatech.EDU>
Sender: news@ucbcad.berkeley.edu
Reply-To: max@eros.UUCP (Max Hauser)
Organization: University of California, Berkeley
Lines: 109
Xref: utgpu rec.audio:2920 sci.electronics:1282

In article <4124@pyr.gatech.EDU> kludge@pyr.UUCP (Scott Dorsey) writes:
>
> ... You hit something.  It's trivial to design good sounding audio
> equipment with tubes.  It's damn near impossible to design good 
> transistor gear ...

Arghh! This is exactly the argument I heard from my friend John,
18 years ago when we were in high school and building lots of 
amplifiers, he usually with tubes and I usually with transistors.
His argument was, essentially, "when I just throw something 
together using tubes, it more-or-less-works; but if I do the same
with transistors, it doesn't." (Here I am speaking of basic
amplifiers like preamps and transformer-coupled power amps).

Not even that statement is true, however, given -- and do NOT
overlook this qualification if you respond to me -- equal 
background in the two technologies. John, in fact, had been 
introduced to electronics through vacuum tubes and had developed a
solid inventory of prejudices and rules-of-thumb that didn't apply
to transistors. He therefore concluded that transistors were more
difficult to use. But if you understand the different devices in 
their own right, then it is actually much EASIER to get bipolar 
transistors, in particular, to bias up and exhibit predictable gain,
because their critical parameters (transconductance and DC voltage 
bias) are largely device-independent, unlike tubes, whose circuits 
must sacrifice performance in order to tolerate the "personality"
of each part number and each device within that part number.

But a much stronger statement in favor of transistors is possible.
If you actually tighten up the definition of "design" to mean not
"produce something that more or less works" but rather "understand
precisely what is going on," then there is no comparison between
transistors and tubes as basic amplifier elements, at least among
design engineers. Now I realize that these comments may largely be 
wasted, or misinterpreted, in rec.audio, but nevertheless let me 
try to state carefully the reason. Bipolar transistors in particular
exhibit nonlinear large-signal terminal characteristics (Ic vs. Vbe)
that are probably the most predictable and consistent nonlinearity
known to electronics (as Gibbons and Horn first pointed out 
eloquently in 1963). Every competent designer is aware that, as a
result, a bipolar stage with 1 mA collector current at 27 degrees C
will have a gm of 38.5 mmho and a base-emitter (DC input) voltage of
very nearly 0.6 or 0.7 volt. Vacuum tubes also exhibit a large-signal
nonlinearity of the same critical kind, but its very *shape*, let
alone numbers, depends on device construction, unit-unit variation,
and age. It is not even possible to predict stage gain and DC bias
in the same way as with bipolar devices, and therefore, unless you
lard the circuit through with trimmers and align it periodically, 
you actually are *stuck* at the level of "producing something that 
more or less works" -- brute force, not true design. This is why,
if you move beyond the level of circuits-that-more-or-less-work, 
there is much less argument in favor of tubes, even in simple 
amplifiers.

There are also, of course, whole genres of audio circuits that can
be constructed with bipolars that have no antecedent in tubes, 
owing to the precise predictable I-V characteristic of the 
transistors. I am speaking of things like "translinear" circuits
(which can give a fundamentally linear -- not just approximately,
or corrected-with-feedback) large-signal input-output relationship;
or the precise variable-gain amplifiers (a subject dear to my heart)
on which depend such products as Dave Blackmer's (the db in dbx)
companding and noise-reduction systems. I'm responsible for a few
other such circuits myself.

The converse is not true, since most anything you can accomplish
with triodes or pentodes can also be done with JFETs or MOSFETs.
These devices share some of the gain and DC-bias predictability
problems of tubes, but much less so, because the I-V characteristics
are mathematically simpler, can be better controlled, and do not
change with age.

True, it is harder to get output swings above about 50 volts p-p,
with common small-signal transistors than with tubes, but then you
rarely need to.  Since you can get transistors with audio-band 
voltage noises less than 100 nV RMS (which is to say, 140 dB down
from a one-volt line level), you can still achieve a lot of dynamic
range with a "mere" 50-volt output capability.

Kludge continues:

> ... With modern tubes and design methods microphonics and 
> self-oscillation aren't anywhere near the problems that 
> they used to be.  

I question this: the 12AX7's, 12AU7's, 12AT7's, 6L6's etc. that
are so dear to tubophiles now are not only the same devices used
in the 1950s and 60s, but in many cases actually were themselves
built in the 1950s and 60s. Microphonics in particular are a function
of the tube design.  Perhaps Kludge designs his own tubes;
if so, he's the first such designer I've met (on the other hand,
I design my own transistors: it's cheap and pretty easy).

I would not for a moment question someone's preference for the sound
of a tube preamp or power amp (even if it is more expensive, less
reliable, and has an uncertain future as existing VT stocks get burnt
out, or used by enthusiastic suburban kids as BB-gun targets).
Also, I am not talking about the relative ease of designing good
transformer-coupled tube amps and direct-coupled solid-state amps
(except to note that direct-coupled TUBE amps are also tricky).
However, the broad statement that "transistor amplifiers are harder
to design than tube amplifiers," unless qualified in the sense of
"harder *for me*," is naive and not supported by hard facts.

Incidentally, my friend John, who argued that tube circuits were
"easier to design" 18 years ago, recanted that position after 
broader experience (he now designs analog ICs for a living).

Max Hauser / max@eros.berkeley.edu / ...{!decvax}!ucbvax!eros!max