[BC] BE 4M Modulation (from other list)

[Hershberger, Dave]

This may be a breach of list protocol, but the broadcast.net list server
keeps rejecting my posts as spam, so I'll try to reply to Harold over here
at radiolists, since Harold is here too!

The message rejected by broadcast.net follows...

-----------------------------------------


> Harold 
> Hallikainen writes:

> I just read the article in the December RG about the new BE 
> AM transmitter
> with "4M" modulation. From what I understand from the 
> article, it MAY be
> similar to an idea I had 20 years ago or so.
> 
> I had always thought that you could back down the pulse width 
> driving the
> grid of a class C RF amplifier to back off the power without adversely
> affecting the efficiency. The relationship between output 
> power (or output
> amplitude) would not vary linearly with the pulse width, but you could
> probably predistort audio to drive the pulse width generator 
> such that the
> RF amplitude linearly followed the modulating audio amplitude.

I did my MSEE thesis on RF pulse width modulation back in the mid 70s.
That's what 4M is - RF duty cycle modulation. There is a nonlinear
relationship between RF duty cycle and envelope amplitude, and it's a sine
function. In other words, the amplitude of the output is proportional to the
sine of the conduction angle. So, the predistortion must take the form of
the arcsine function. This is easily done by comparing the modulating signal
to a sinewave, much the same as conventional PWM is formed by comparing the
modulating signal to a triangle function.

Unfortunately, this technique does not work very well with tube type
amplifiers. RF PWM (aka "4M") is only slightly more efficient than simple
linear amplification or grid modulation. It's easy to understand why. When
the tube's output amplitude is reduced, then the plate voltage is not
swinging close to zero when plate current is flowing. So by Ohm's law you
get a lot of dissipation in the tube when you cut back to lower amplitudes.
There's a lot of voltage across the tube when the current is flowing. It's
the same mechanism that limits efficiency in linear amplifiers and in simple
(non-Doherty) grid modulation.

> 
> I discussed this with someone at NAB. It MAY have been an 
> engineer from
> AEL. I was looking at his new FM exciter at the time, I 
> think. Anyway, he
> said that dealing with such narrow pulses would be difficult 
> and suggested
> generating two pulse trains and vary the overlap between 
> them. This would
> be the same as phase modulating the two pulse trains in opposite
> directions. The two pulse trains would drive the grids of two tubes in
> series so the series combination only conducted during the 
> overlap of the
> pulse trains giving the desired variable RF pulse width "class C"
> amplifier.  I commented that this sounded a lot like Ampliphase. He
> agreed.

Yes, it IS a lot like Ampliphase. According to Jerry Westberg's NAB
presentation, it's done with a bridge amplifier. Each side of the bridge is
driven with 50% duty cycle squarewaves. The phases of the two sides vary in
opposite directions with the arcsine of the modulation function, and the
output is taken differentially. The result is a series of pulses of
alternating polarity, of varying width. That amplifier topology results in
something that is a combination of RF pulse width modulation and Ampliphase.

So just as you described, this is in fact two amplifiers in series with each
other and with the load.

The hard part of getting this to work is to make the transistors switch
efficiently (and without blowing up) when load current is flowing. Each side
of the bridge amplifier sees a load with a phase angle that varies as an
instantaneous function of the modulating signal. The voltage and current in
each side of the bridge amplifier are only in phase at the maximum positive
modulation point. Below that amplitude, one side sees a capacitive load and
the other side sees an inductive load. Jerry told me that BE has done
something to make each side of the bridge see a better load. Whatever that
is, that's the major engineering accomplishment that makes RF PWM work.

> 
> Anyone have any details on 4M? It sounds interesting!
> 
> Harold
>

Yes, it's very cool. The 4M PA topology is basically a balanced modulator.
Modulation can theoretically pass beyond -100%, causing the carrier phase to
flip 180 degrees as the envelope folds over. If BE has solved the "reactive
load" problem then it opens up a new possible way of directly generating IQ
signals such as COFDM, SSB, ISB, DRM, etc. Rather than using envelope
elimination and restoration, it may be possible to directly generate IQ
signals at high level by phase offsetting half the modules in a transmitter
by 90 degrees, creating a highly efficient IQ modulator. The 0 degree
modules would be modulated with the I channel signal and the 90 degree
modules would be modulated with Q. (Or to equalize device dissipation, one
set of modules could be modulated with I+Q while the other set is modulated
with I-Q.) The whole transmitter would turn into a high efficiency arbitrary
(but narrowband) signal generator. There would be no spectral
growth/cancellation process the way there is in envelope elimination and
restoration systems. This approach could carry a penalty of up to 3 dB in
power or silicon (depending on the ratio of Q channel energy to I) but it
would still be high efficiency. And it should be very clean too, again
because of not having to split the signal into phase modulation and envelope
components, with the bandwidth and time delay cancellation issues that
ensue.

Maybe we can all learn more from the Boston station that's using one.

Dave Hershberger 
Principal Engineer 
AXCERA
Tel: 530-272-HDTV (4388)
Fax: 530-272-4505 
dhershberger at axcera.com <mailto:dhershberger at axcera.com> 
AXCERA. Axcessing the new era of digital communications. www.axcera.com 
960 McCourtney Road, Suite C, Grass Valley, CA 95949-7423