in the meantime: ssb!

[Martin Czech]

At work I have to run very long compile times and I can't
do anything else in between. So time for some little experiments
in order to entertain the Sun Ultra 10 which has finally a (crappy)
soundcard.

The question with ssb was: how much unwanted side band
supression will *I* need? 

I did some simulation. In order to avoid trouble, I've created test
sinusoids, or groups of them. It is easy to shift them upwards
and downwards, no need for Hilbert transformers etc.

The most critical case is a single sinusiod, quite high pitch.  And this
shifted upwards several octaves.  In this case the ear will recognize
carrier feedthrough and the unwanted sideband with ease. The headphones
are a bit crappy walkman style, so certainly I have to have another
listening, but so far I can say:


-20dB : forget it
-40dB : well, hmm, *perhaps* tolerable
-60dB : good, i can't hear the wrong sideband moving into
        the other direction.
        
Note that this is about wide band sweeps, not 1Hz displacement or so
for phasing. Phasing people should be happe with -30 .. -40dB.

The more partials I use, the more beating I get between them,
and the harder (really amazing) it is to hear the unwanted sideband,
even if only -20dB!

So, for shifting white noise, -20dB might be good enough.
Is this an option for random signals? Would it be a counterpart
to the T. Henry fm proposal? I.e. take some narrow band noise
and shift it down to allmost 0?

I'm very happy that I did not start any hardware yet, a listening test
in advance is very usefull.
And also a second reason: polyphase. Since Mr. Hutchins read all Hilbert 
transformer  suggestions and wrote them all down in E.N. the active
phase shifter is in and the passive polyphase is out.

The passive network needs lots of components, it is not cheap.
OTOH the are some pros:

-low absolute value requirements
-only pole matching (i.e. a diy bridge meter could be used, and a simple DMM)
-it offers 0, 90, 180 and 270 DEG outputs at the same time, thus the name
 "polyphase"
-little error with cap temp coeff.
-no trouble when publishing schematics
-anti ripping (Doepfer), because untolerable cost for mass production ;->
-perhaps a module of it's own, though I don't know what to do
 with four shiftet outputs, may be stereo panorama things?
-it is different from other proposals, I like to be different ;->
 
Now, what is this ployphase property good for? I came across this recently,
the HAM people really use all four shifts together with four modulators!
Sounds like stupid waste of money. No.
If these are mixed in the right way, carrier and sideband suppression
is further optimised. Sounds promising.

The good thing is if I have invested a lot of time and money into the network
and if it works, it is allways usefull, regardless of ssb method.
It is also possible to do a baseband approach a la Haible,
because the Tietze-Schenk sin/cos generator will deliver all
needed phase angles as well.

I hope that I get a little bit forward, after the listening tests.
A ssb effect is (together with vocoder) the effect I always wanted to
have and that no other machine I have can do. And it is not such a *big*
project like a vocoder is.

The next weeks will have a 200km+ cycling marathon each (Oetztal, Alpen
Brevet, Alpirsbach, Altdorf, Riderman), so soldering anything will be
unlikely.  But I have written some very usefull little software tools,
for listening test as well as for angle error vs. component matching
studies as well.

m.c.