[Amps] MOSFET amp filtering - was: auto-tune

[Manfred Mornhinweg]

Jim,

>> But is it in an RF amplifier?  Really?
> 
> Spend a bit of time on any ham band with a spectrum analyzer that has 
> good frequency resolution and you'll learn in the first 30 minutes that 
> this is true.

Unfortunately that test isn't able to check what I want. My question 
was, specifically, whether waveform distortion such as clipping a 
two-tone signal at the zero line, which certainly causes intense 
harmonics, will or will not cause objectionable IMD.

> Or take a look at pictures from mine looking at CW and SSB signals. 
> http://k9yc.com/P3_Spectrum_Measurements.pdf

Nice and interesting tests, but not applicable to my fundamental question!

> The splatter MUST be created in the power amp -- AF harmonics outside 
> the SSB filter will be suppressed by that filter.

Yes, that's crystal-clear. Nonlinearity before a good SSB filter will 
create distortion within the passband of the transmitter, but not 
outside it. But still doesn't answer my question! :-)

So, given that nobody clearly answered my question, I had to do a 
supreme effort, and start thinking! :-)  Actually, in addition to 
thinking, I did some simulations and tests to understand the case. Here 
is the result:

I first created a high quality two-tone signal, and put it through a 
spectrum analyzer. The result, as expected, are the two peaks, and very 
low levels of stuff around them. So far, so good.

Then I hard-clipped this two-tone signal at the zero line. This is the 
same as passing it through a half-wave rectifier, or through a single 
ended class-B amplifier that doesn't have any frequency-selective 
filtering. I put this extremely distorted signal into the spectrum 
analyzer. The result was this:

In addition to the original two tones I get the whole series of 
harmonics of them with the second harmonic starting at just 6dB down 
from the fundamental. And there are also strong IMD products, but these 
are far away from the two tones! In fact the second order IMD product is 
exactly as strong as the second harmonic - but it falls on the 
difference frequency between the two tones, that is, in the audio range! 
  My two tones, distorted this way, create a strong IMD product on 1kHz. 
But that's no problem in a practical amp, since audio is of course 
stripped off!

Within the range of interest, the strongest IMD products are the 3rd 
order ones, and interestingly they are very much weaker than the 3rd 
harmonic: While the 3rd harmonic is only 20dB down, the 3rd IMD products 
are way over 60dB down!

Then I started adding nonlinearity to the remaining half wave, thus 
starting to distort the envelope. Even a small amount of envelope 
distortion drives the close-in IMD way up, like to -20dB for a pretty 
modest amount of nonlinearity.

So, the self-generated answer to my question is: Severely distorting the 
waveform of a modulated RF signal does indeed created IMD, but the IMD 
products falling within the bandwidth of the amplifier are weak enough 
to ignore. The harmonics instead are so strong that they absolutely need 
filtering. All strong IMD products fall on audio frequencies, far 
outside the response of the amp.

This, folks, is what makes it possible to run single-ended class AB tube 
amps without messing up the spectrum!

When designing amplifiers for ham radio, we need to keep the envelope 
response highly linear, and we should also keep the phase response 
linear. I don't know yet HOW linear we need to keep the phase 
response... But the actual waveform of the RF signal at the active 
devices doesn't matter, at least not by itself. Anything goes. We just 
have to turn it back to a clean sine wave before it reaches the antenna, 
by means of low pass filters, tank circuits, or whatever, so that we 
don't put out the strong harmonics of a distorted waveform.

Thus I see no fundamental problem in using switchmode amplifiers for 
linear amplification. That said, there sure are a lot of practical 
problems, but they can be solved! A switching amplifier, pulse-width 
modulated or supply-modulated, with envelope detectors before and after 
it, a comparator and a high gain baseband error amplifier dynamically 
controlling the gain, should produce high efficiency along with high 
envelope linearity. If in addition the MOSFET capacitances are low 
enough so that their voltage-induced variability results in low-enough 
phase modulation, we should be all set.

It's a bit funny to do this analysis nowadays. It surely has been done 
by others many decades ago, and likely even a century or two ago on a 
more theoretical basis! But ham radio is a lot about self-instruction, 
isn't it?

Manfred


========================
Visit my hobby homepage!
http://ludens.cl
========================