USRP RF output frequency is much faster than intended

I build a flow graph like an image in the following link:

After then, I connect the USRP and an oscilloscope and see output, which

As you see, both are square waves, but the frequency, the above is 1k
the below is 7.8 kHz.
If I set a frequency of signal source to 2k, the below says 16.6 kHz.
course, there are some deviation)

How could I understand this?
If the oscilloscope shows a frequency slower than the GNU radio,
I think that is acceptable since there are latency and processing time

But the output is 7.8 times faster than what I inteded.
The only thing I can guess is… 7.8 is rounded up to 8, and 8 indicates
‘8’ bits.
But it seems a bit unnatural and weird…

Could anyone give me a hint?

If you think it is more appropriate to be posted on USRP-users, I’ll
it there later. :slight_smile:


Hi Jeon,

there’s a bit wrong with this:
32kHz is not a valid sampling rate of any USRP that can take basicTX,
LFTX or UBX daughterboards, which is what you’d need to produce a
baseband signal. Your console window should show a warning saying which
sampling rate was actually used. We can deduct which rate was used:
32kHz * 7.8 ~= 250kHz.

Best regards,

Hi Jeon,

I stared at this over my morning coffee. I can’t quite explain why
getting 7.8kHz output on a scope, but there’s some other considerations.

  1. Your UHD block has a center frequency of 0Hz. When you run your
    flowgraph do you get messages from UHD? You should see something about
    being able to set a center frequency of 0Hz and using x Hz instead.

  2. 32 kHz is not a good sampling rate for USRPs. You should see
    similar to #1 about setting the clock. The signal you are seeing is
    probably a result of generating a signal with some expected sampling
    frequency and looking at with another. If you compare the actual USRP
    rate to 32kHz and do some math I wouldn’t be surprised if it comes out
    be a factor of 8 different.

  3. Imagine an even simpler case with just a sine wave. You have a sine
    of 1kHz at complex baseband going in to an upconverter (imagine this as
    multiplying your signal by an oscillator at your center frequency). What
    would you expect on the scope? If you don’t know, then do the math; it’s
    fairly simple trig.

Good luck,

Dear all, Mleech, Marcus and Nathan

Thanks for all your advices.

I’ve set the samp_rate to 250k and it works perfectly.


I’ve checked your answer on USRP users mailing list a few minutes ago.
Some settings were set wrong, so the mail was not delivered to me.
Sorry for late.


LFTX and LFRX can use center frequnecy of 0 Hz (maybe…) which are
what I
am using now.


2015-06-25 22:58 GMT+09:00 West, Nathan [email protected]: