live sound fm with double square mixer and txo

[Martin Czech]

Now , while things are running wild with heterodyne ideas, I can speak
about my recent ideas on frequency shift without too much embarassment...


First of all, let me say that I like very much what J.H. did in base-band.

I too think that the phase difference method is the only reasonable way
to get rid of one of the sidebands. You have to modulate to get a double
sideband, and there is no reasonable way to mix this down into base band
or near base band for better filtering, because the Lower Side Band will
mess the Upper Side Band then. One could choose a modulation frequency
that is quite low (10 KHz), the LSB may appear "folded", this is not harmfull.
But then carrier suppression is an issue...

So the plan is to modulate to 60kHz, with a fixed (quartz derived)
carrier. I plan to use 1496 for this, they are made for rectangular
carriers, that's why the linearisation is left away (1495). It is easy
to set up two 90-Deg phase shiftet 60kHz square waves by logic and
dividers etc.

Now comes a Phase Difference Network, as described by Mr. Hutchins in E.N.
(No, I will not disclose that here). I think the most I can hope for is
-60dB sideband rejection, and I'm not shure if this is possible at all.

You can easily compute that the phase error needs to be about 0.1 Deg!
This is hard to do over temperature and aging. OTOH 10% caps can
be used (100ppm/Deg tempco should be), the tolerance has to be trimmed
out via one resistor in the usall phase shift circuit we all know.

The plan could be to have the network for 10Hz to 40kHz bandwith,
the resulting PDN will be 16th order or so.

Why? Well, I don't want to have low frequency or higher frequency
components to mess my signal. So I want to filter at 20 Hz and 20kHz
and I need some transitional band for the filters to drop.
Really, it could be 40Hz-17khz audio, so I could save something there.

Ok, the audio signal is now 90 Deg shiftet, and square wave modulated.
The hope is that the nonlinearities in the circuit will not mess up the
first harmonic sidebands with higher harmonics (rest of 2nd @ 120kHz and
strong 3rd @ 180kHz).

Since this modulator is pretty fixed (quartz - divider) the modulator
speakthrough (estimated -60 - -70dB) could be fixed by adding an
appropriate inverse and slighly phase shiftet signal, I only want to
cancel the 1st harmonic, so the waveform is not so important, i hope I
can use a RC-shifter, resistive adder thing for canceling.

If all fails there could be two holes in the front cover of the shifter
with two trimmers behind for each multiplier, in order to cancel the
carrier after a few sessions.

The add/substract stages for sideband separation could also suffer from
phase shift, high bandwidth opamps, and an opamp even in the noninverting
path may cure that (equal delays).

So, with some good fortune I would have USB and LSB 60dB seperated,
and the carrier down to -80dB or so.

Now comes the mixer to baseband. This needn't be very cute, because the
new carrier is 40-80kHz, so it will not mess up with the audio band. OTOH
if the resulting signal is to be processed by ADC (recording) it is
better not to have so much hf dirt... If it is rectangular, a polarity
switcher would do, but the 1496 are quite cheap , -50 - -60dB could be
achieved without trimming...

The plan is to use a pll derived square for mixing into baseband.
A divide by 10000 would give about 8 Hz resolution and so on, and this is
where the trouble really starts.  In order to have a very low increment,
the divider has to be more then 16bit...  As a result, the pll input
frequency is very low (slow servo loop) or the divider input frequency
has to be very high.

I don't think that a cascaded 74AC163 divide by n will run faster then
10-20MHz. And AC logic and such frequencies are quite some fun,
I`ll need a double sided board, at least a ground plane is a must.

A remedy for missing pll resolution could be to use two nominal equal
quartz crystals, one for the up mixer and one for the down.  One of
them could be "pulled" then with a varactor diode.  A voltage controlled
crystal osc (TXO, VXO). This is nothing new and described in Horrowitz/Hill.

Gotcha... no the crystals have to be a factor 2 away from each other
to avoid locking effects. Another problem...

I have no clue about the performance.

Even if it may turn out that the device does not work as intended
(carrier supression, intermodulation, side band suppression), it will
be a very good exercise to do it.

Ah, at the end the signals have to be cleaned by low-pass filters,
of course.  Note that the variable carrier does not appear in audio band
if the fixed carrier is suppressed good enough.

The hope is to get away without any compander/gate circuits.
The sideband rests will be there, of course, but no audio in means no sidebands...

Ok, this proposed device could do fixed frequency shifts.

Two programmable points of variable frequency could do sweeps
via pll lag.

Crazy. Yes, indeed. There are so many unknown things...
But I don't think that it is impossible or will fail alltogether.


Is there any clever idea to get finer frequency resolution,
or let's say frequency ratios without very high clock frequencys?

m.c.