Path: utzoo!mnetor!uunet!husc6!bloom-beacon!gatech!mcnc!unccvax!dya
From: dya@unccvax.UUCP (York David Anthony)
Newsgroups: sci.electronics
Subject: Re: Class D Power Amplifiers?
Message-ID: <980@unccvax.UUCP>
Date: 9 May 88 17:30:15 GMT
References: <4110001@hpficad.HP.COM>
Organization: Univ. of NC at Charlotte, Charlotte, NC
Lines: 56

In article <4110001@hpficad.HP.COM>, stu@hpficad.HP.COM (Stu Bell) writes:

> Whatever  happened to Class D (digital)  power  amplifiers?  I remember that
> [Infinity] introduced one in 1978 . . .
> What's wrong with pulse modulation  (essentially  the digital  approach) for
> power amplifiers? . . .

     Nothing, really. Pulse width modulated switching power amplifiers
are commonly used in AM broadcast transmitters.  In one such system,
there are two valves connected in series, a "switch" valve and a 
"carrier" valve.  The switch valve runs at 70 kHz or so, and is 
connected to the anode of the "carrier" valve at operating frequency.
There is a massive transmission multipole filter between the two
valves.  The system is reasonably flat to DC.

     This is the kind of system which lends itself to precisely the
kind of application above (generating large amounts of audio power
efficiently).  No one needs a, say, 7500 watt audio amplifier for
entertainment use.

     Infinity introduced (supposedly) a prototype PWM audio amplifier
I think, around 1975 or so.  Power transistors were not wonderful
then by any means (power V-channel FET's also just coming out) and
I doubt there were many actually produced.

     The PWM approach for home use presents mostly packaging and
production problems, not to mention that regular linear amplifiers
can easily surpass PWM performance even in the most rudimentary of
specifications.  The fidelity of the PWM equipment is controlled
by not just the linearity of the PWM generator (trivial to do now
with ECL 100k logic) but by the tuning of the lowpass filter and
various traps.  It is difficult to get good supersonic suppression
and a flat frequency response AND deliver this to such a low impedance
load. (In the AM transmitter case, the load impedance for audio 
frequencies is a nice, comfortable 1000 ohms or so, and it doesn't
vary much-in a well designed transmitter- over the audio range). 
Speakers are notorious for interesting impedance v. frequency
characteristics.

     Yes, the switching waveform must be very well suppressed, and
it will fry speakers easily even with a few watts.  This is never a
problem in the AM transmitter, because you suppress the switched
waveform say, 50 dB, and the class "C" output stage with its various
tuned circuits takes care of any residual 70 kHz component.

     I suppose that one could use various lookup table schemes, 
high speed digital signal processing, or whatever, to predistort
the PWM signal so it comes out linear at the filter output, but
why would anyone want to?

      Unless you need gobs of audio power (or, for that matter, well
regulated DC), PWM just isn't the ticket; although some netter will
undoubtedly prove me wrong... 

David Anthony
DataSpan, Inc.