Re: [Discuss-gnuradio] usrp A/D scaling question

[Brian Padalino]

On Wed, Jul 8, 2009 at 11:54 AM, John DuBois<address@hidden> wrote:
> I am puzzled by a very basic issue relating to the usrp1 output. Using GRC,
> I connect the usrp output through a complex to mag conversion to a scope
> sink and tune to, say, 1GHz. The daughter board is a DBS_RX.
>
> Using a cw signal generator, I tune for a maximum usrp output and then raise
> the usrp input  up to about -45 dBm. As the input increases, I see rising
> output values (on the scope display) from just above zero to a maximum of
> about 16400.
>
> The puzzle is this: if the usrp A/D converter is 12 bits for I and Q
> channels and I am doing a complex to magnitude conversion, why don't I see
> 4095*1.414 as the maximum output number?
>
> The maximum output I do see seems like 14 bits, but without any 1.414
> factor.
>
> What am I missing ??

Decimation and Interpolation using CIC filters causes bit-growth which
is only marginally compensated for using bit-shifts.

Moreover, CORDIC mixing causes bit-growth in the output vector as well.

Hopefully exploring those avenues solves your puzzle.

Good luck.

Brian