If you make an FFT block for each rtl-sdr source you can then interleave them all together and take your pick using the 'M in N' block. At the edges of each rtl-sdr frequency it will probably get quite messy as the clocks are not synchronised.
An example of the use of the FFT and 'M in N' blocks is in my scanoo_rx app which will give you a head start. The 'M in N' block is used to pick out the channel from the output of the FFT block:
> Playing around with GRC for a few months now, sorry if this question is
> from a dummy.
> I've a bunch of rtl dongles laying around, and wonder if it's possible to
> combine more then one to a single baseband signal.
> The goal is to create a super cheap full HF band web reciever like or for
> ' websdr'
> I managed to create multiple channels, where the 2nd dongle is tuned to an
> offset, equal to the samplerate of the first. (for example 2 x 2 Msps = 4
> Msps baseband)
> Is such a thing possible with GRC?
A couple of things, on top of what Mike said:
- HF bands go lower than the frequency of the dongles. Keep that in mind.
- As Mike said, clocks aren't sync'd. So, if you're using data from
multiple dongles, that'll most likely be corrupt.
- To make things worse, they drift differently too. So, little chance in
- If all you want is an (inaccurate) "spectrum analyzer" for more than 2
MHz, then this is actually not a bad idea. If you have N+1 dongles, you
can use N of them to monitor N*2MHz of spectrum, and the other one to
receive at frequencies you're interested in.
How about the PFB synthesizer detailed by TR? I'd think all the dongles would have to be phase locked and references tones (or comb) used to calibrate the channels. If your signals are not crossing channels then maybe you could get away without that.
All you channels will not only need to be phase-coherent, but also have
zero phase offset. That's the only way you can reasonably simulate
having much wider bandwidth by "gluing together" a bunch of smaller
My own experiments with the RTLSDR gear is that achieving phase
coherence across significant bandwidth appears to not be possible, even
with a common, high-quality master clock.