See (partial) answers below:
Kind people,
I apologize for this posting being roughly half off-topic, but I think you probably have the expertise I am seeking.
I have a couple of old 1960's novice rigs in my basement. I am wondering if I could modify one of them to transmit PSK31 by using transformerless series AM to apply the (1-cos)/2 shaping function (sort of an inside-out Hann window) to a class C amplifier driven by a modern DDS VFO which has the capability to generate the phase shifts. If you are not familiar with PSK31, I will follow my questions with a short description of how it works. I realize that class A series modulation is horribly inefficient, but I am only looking for something like 20 watts PEP.
My questions are
1) Would this work [Yes. People have used separate amplitude and phase modulation of a class C or a class E amplifier to a create a linear amplifier]
2) I only need downward modulation so it seems to me that I only need to increase the plate voltage enough to overcome the resistance of the modulating device when it is quiescent. Is that correct? [No comment]
3) Could I do this with a mosfet as the modulator? I am just getting back into ham radio after a 40 year hiatus, so I have 40 years of semi-conductor ignorance to overcome.[There are recent threads on the board regarding transformer-less, pulse duration modulation (PDM) using solid state devices]
4) Can I get close to 100% downward modulation this way? [Yes... but there may be some increased non-linearity as you approach 100% downward modulation].
5) Am I better off on the plate side or the cathode side of the tube or does it matter? [No comment]
6) Can you suggest some additional reading I might do? [See the recent thread regarding: "How much peak power can a class C 6146B put out?"]
7) What gotcha's have I overlooked? [If the linearity of the amplitude modulation is as critical as you imply... to get the necessary degree of cancellation of frequency domain components... then you may want to try using modern pre-distortion techniques. Some of the new software defined transceivers (for example the latest ANAN products) employ pre-distortion to achieve significant linearity improvements. The key is to observe those aspects of a sample of the RF output signal whose levels represent the degree of presence of distortion products, and then use some algorithm to automatically adjust the pre-distortion parameters to minimize the observed levels of those representations of distortion products. You could also employ pre-distortion in the SSB/up-conversion method that you mention below.]
If you are interested, here is a brief description of the generation of PSK31 signals.
As you probably know, PSK31 is a digital mode in which binary digits are sent at a rate of 31.25 bits per second, so that each bit lasts 31msec. A 1 bit consists of 31 msec of raw carrier. A 0 bit starts out like a 1 bit, but half way through, at 15.5 msec, the phase is shifted by 180 degree. Thus a string of 0's would be a carrier multiplied by a square wave of frequencxy 15.625 cps (oops, Hz). This signal would have large bandwidth. The bandwidth is controlled by multiplying the signal by (1 - cos(2pi*31.25t)/2, and doing the phase shift when the amplitude is zero.
Thus, a string of 0's is the product of a cosine term at the carrier frequency, a term (1-cos)/2 at 31.5 Hz, and a square wave at half of 31.5 Hz. In the frequency domain, a lot of cancellation takes place as these three signals are convolved, and you end up with the carrier frequency suppressed and two dominant sideband frequencies at carrier+15.625 and carrier-15.625 giving a bandwidth of 31.25 Hz.
Usually these signals are generated using a computer sound card. The card creates an audio signal somewhere between 300 and 2500 cps. For a zero that is all you need. For a 1, the card applies the inside-out Hann window and flips the phase halfway through the bit. This audio signal is then fed into an SSB trasmitter which effectively heterodynes the signal up into the RF world and suppresses the lower sideband.
The problem with this approach is that the least little whiff of non-linearity will cause the bandwidth to become excessive. One is, after all, basically presenting the transmitter with a pair of tones and running one's own two tone test. I am wondering whether the AM scheme suggested above would be more forgiving.
Anyway, thanks for listening to my rambling.
73
Paul
