Ian:

Thank you for your detailed response. If I stick with eight bits and
increase to 50Msps, will that solve the decimation rate problem? I
realize this would result in a decimation rate of two, which would not
be ideal. I may decide to only grab 25MHz, not the full 30, to keep
the decimation rate at four.

Thank you, again.

Paul B. Huter

Ian Buckley <ianb@ionconcepts.com> wrote:


Paul,
With (the default) 16bit complex sample format the maximum sample rate
supported on USRP2 is 25Msps, simply because that is pretty much as
much data as you can get over 1G Ethernet in one direction.
50Msps is supported if you use 8bit complex samples and loose some
dynamic range.

The decimation rate must be an integer, thus 25Msps is a decimation
rate of 4 from the raw ADC converter clock of 100MHz. Also, since the
receive DSP chain cascades 2 Halfband FIR filters (each with fixed
decimation of 2) before a CIC filter (With arbitrary integer
decimation), it is preferable to always use decimation rates that
factor by 4 so that both half band filters are active which results in
the flattest passband. UHD will automatically use the half band
filters in preference to the CIC when possible with the specified
decimation.

-Ian

On 10/23/2013 07:44 PM, Paul B. Huter wrote:
> Paul B. Huter
Keep firmly in mind that your computer will need enough "grunt" to deal
with the resulting torrent of samples from the USRP.

Even if you're only recording them, you'll need a minimum of
100Mbyte/second sustained write speed to your disks.

If you're actually *doing something* with the samples, in the
mathematical sense, you'll need lots of aggregate GFLOPS to keep up.

Thanks, Marcus. I am just saving the data into a data file for later
analysis. I only plan to save about a minute of data at a time. I intend
to
start small, though, and work my way up to see if my computer can handle
things.