[net.audio] Tube sound ???

wjm@whuxk.UUCP (12/02/83)

John Wray's comments about tests showing that people prefering the sound of tube
amplifiers in double-blind tests are interesting.
I've heard about similar tests and suspect that the difference is in the
nature of the harmonics created by the active devices.  Two amplifiers can
have the same TOTAL harmonic distortion (say .01%) but one may be preferred
by most people due to the specific harmonics contributing to the distortion.
Psychological research seems to indicate that people find even-order harmonics
sound "mellow" and less offensive than the "grating" sound of odd-order ones.
Bipolar transistors tend to have greater third-order distortion while tubes
(and their solid-state kin, FET's) tend to have more second-order distortion.
In addition, most tube and FET amplifier circuits tend to clip more gradually
when the input signal sends the amplifier stage to the power supply reference
while bipolar circuits tend to clip sharply, with a rapid rise in distortion.
Since bipolar transistors are subject to thermal runaway (the device resistance
DECREASES with increasing temperature, causing the heat producing current to
rise with temperature, making the device hotter and hotter until it self-
destructs), elaborate protection circuits are required and these circuits
can cause distortion of the audio signal.
On the other hand, tubes and FET's are self-limiting since their resistance
INCREASES with temperature, so these protection circuits are not required.
Also, bipolar transistors are usually subject to crossover distortion when
the signal voltage goes from positive to negative or vice versa (which happens
40,000 times a second at 20KHz) since a bipolar transistor will not start to
conduct until the base-emitter voltage reaches 0.7 volts (polarity depends
upon whether the transistor is NPN or PNP - also this value is for silicon
devices - it is 0.2 volts for germanium ones).
Clever bias circuits have been developed to minimize this distortion but
the only way to totally eliminate it is to:
1. Use true Class A bias - in which the transistor conducts 100% of the time.
This causes the device to get quite WARM, so Class A amplifiers are quite
large and quite expensive.  Also, this bias current has a rather significant
impact on your electric bill.
2. Don't use bipolar transistors as the active elements.  Until 1980 or so,
the only other option was to use tubes.  Unfortunately, tubes have their own
drawbacks - they require power for their heaters, they are physically large,
they have to be replaced periodically, and tube amplifiers generally have to
use output transformers - that cause tube amplifiers to be expensive and power
hogs.
Fortunately, it is now possible to obtain power FET's (field-effect transistors)
which provide a reasonable solution to this problem, since they operated on
the same principle as vacuum tubes (using an electric field on a control element
to control the power-handling output circuit) they have the same general
characteristics and advantages.
However, they have many advantages over tubes, since they are solid-state
devices.
They are long-lived, do not require heater power, compact, and unlike tubes
come in both polarities (p-channel and n-channel) so output-transformerless
circuits can be used.
Personally, I've been quite pleased with the sound of my FET power amp (a Hafler
220) and have heard generally favorable comments about many other FET designs.
Bill Mitchell
Bell Laboratories, Whippany NJ
whuxk!wjm

emrath@uiuccsb.UUCP (12/07/83)

#R:whuxk:-34200:uiuccsb:5700016:000:502
uiuccsb!emrath    Dec  6 17:54:00 1983

I don't buy that argument that you shouldn't use bipolar transistors
because of crossover distortion (unless you run them class A).
Tubes and FETs are usually used in a class A-B configuration for
power amps, too, and they will suffer from the same kind of distortion
as a result of the configuration (maybe not to the same degree).
A good class A-B with 50-100mA of bias on the output stage suits my
purposes, but I wouldn't use tubes because of the output xformer and
unless it doubled as a toaster.