[hpsdr] Perseus is a reality

[John Miles]

> > That does make sense; I hadn't really thought about the fact that it's
> > down_sampling_ as well as downconverting.  The gain is more
> like 4-5 bits in
> > that example (10*log(downsampling factor) dB, rather than 20), but still
> > very much worthwhile.
>
> Why 10*log ? Aren't we talking about the _voltage_ that can be
> represented with 16 or 24 bits ?

A "bit" is a voltage ratio, but process gain is expressed as a power ratio,
hence the 10*log(N) expression.  Each bit of resolution doubles your voltage
range, so it's worth 6 dB in "power" space (numerically, 20*log(2) ~= +6
dB).  Once you have the process gain in dB using the 10*log(downsampling
factor) function, you can turn that into an estimate of equivalent bits by
dividing by 6.

To some extent, we're in a state of sin when we use dB to talk about voltage
ratios, but "everybody else does it!"

-- john, KE5FX



 1192065518.0