Digital 'scope & spectrum analyzer Re: [Open-graphics] PCIe know-how?

[Dieter]

> > > If the granularity of our control of the signal is 330Mhz,
> >
> > Ouch, that's a bit of a limitation.  I don't suppose the two heads
> > could be interlaced together somehow to get 660 Mhz?
> 
> Ok, what we can probably do is use a fixed 330MHz and then, in the
> amplifier, delay green by 1/3 and blue by 2/3 of a clock period
> relative to red and then just add them together.  That gives us
> effectively 990MHz.  Good enough?  Actually, it's not quite that
> simple, but since everything is a smooth curve, I think we can fake
> it.
> 
> > (Still low, but
> > better than 330.)  And we lose half of that to Nyquest?
> 
> That loss is elsewhere in the problem, I think.  That is, we would get
> the same loss whether we used analog or digital to encode the signal.

I was thinking of 330 MHz on the digital side.  If it is actually the
analog side then never mind about Nyquest.  A 990 MHz sine wave should
be enough for what I have in mind.  I'd also be happy with a 990 MHz
spectrum analyzer.  Your mileage may vary.

> > > can we
> > > encode all of the information in the TV signal?  Would the steps
> > > between digital levels (1024 of them) be too noisy?  Could we fix that
> > > with a low-pass filter?
> >
> > Does 1024 levels imply a S/N of 30.103 dB?   If so, that would be plenty
> > for ATSC, but IIRC a bit low for good quality NTSC.  I haven't seen numbers
> > for PAL, SECAM, or DVB-T.
> 
> Tying the three channels together effectively triples that.

It isn't obvious to me how interleaving red, green, and blue increases the S/N.

I googled a bit, and found S/N numbers higher than I recall.  :-(
Looks like we'd want 56-60 dB S/N.

On the other hand, those numbers are likely referring to thermal noise,
where this is quantization levels.  A bit apples vs oranges?