Re: [Discuss-gnuradio] Quadrature Sampling

[Matt Ettus]

Quoting Prateek Dayal <address@hidden>:

> I have a question involving sampling and bandwidth. In case I sample
> say a TV signal using USRP, I can sample the signal and store it as 8
> M samples/s real and imag data. There will be no aliasing.

Correct.

> Now if in a gnuradio application, I load this real and imaginary part
> and then do a complex_to_float, then shouldn't the sampling rate
> requirement go higher. I mean if originally the signal was 7 Mhz, it
> was fine to quadrature sample it at 8 Mhz, but after the
> complex_to_float, shouldn't this signal require a 14 Mhz sampling rate
> (ie., an upconversion).

This is correct.  8 MS/s complex means you have 16 million independent
measurements per second, allowing you to see 8 MHz of BW w/o aliasing.  If you
want real samples, you should interpolate by 2, upconvert to 1/2 the old
sampling rate (4 MHz), and drop the Q part.  There are some tricks where you
can do this much simpler than an actual filtering and mixing operation.

On the other hand, why would you want to do this?  IQ samples are the easiest
way to do most operations.

Matt