[USRP-users] XCVR2450 measurements
jmiggal at gmail.com
Wed Nov 17 10:25:59 EST 2010
Thanks for your comments. Some answers bellow.
I tested both transmitter and receiver path of XCVR2450 on a USRP2 under
>> Matlab r2010b. Here some comments about details I found and I don´t
>> TX path: (I transmit a 100KHz single tone modulated at 5.5GHz) to a
>> spectrum analyzer. The expected spectrum is two components at 5.4999GHz and
>> 1) Besides the expected spectrum, I can see a powerful (20dB less that the
>> modulated signal) component at 5.5GHz (Leakage from the MAXIM Local
> What is the scale of the samples you are sending to the USRP sink? If you
> increase the scale, does the delta in power between your expected stuff and
> the LO also increase? It may be you aren't exercising enough of the DAC
The scale is one (1). When I set a lower value (0.5/0.1/0.01) the desired
components start to decrease while the centre component (5.5GHz) remains at
-30dB (i.e. the DAC doesn't use all its bits). If I set a bigger value,
neither the desired components nor undesired one change their power value.
However my spectrum becomes awfull with a lot of poweful harmonics at
(centre frequency+ N*2*Fsingle tone) where N=1, 2,3....
Different sample frequencies don't change this behaviour.
I assume 1 is the optimum value. However I get a -30dBm undesired component
at the carrier frequency. (While the desided ones are -6dBm with 0 XCVR2450
>> 2) I see how after 20 or 30 seconds the signal spectrum suddenly (a fast
>> change) becomes very degraded (a lot of noise in my whole band of interest
>> appears and the signal is very degraded. The only good news is that the
>> component at 5.5GHz becomes smaller)
> Does the behavior here have any correlation with your sample rate /
> interpolation? What is your sample rate?
YES. You are right. I generate signal in simulink at (1e8/512) Fs and the
set an interpolation of 512 to reach the 100MS/s expected by the ADC. After
about 2 minutes my signal (still a modulated single tone) becomes awfull
(seen with a spectrum analyzer. This means that noise raises up to the level
of the undesired carrier frequency (-30 dBm).
This behaviour is not affected by the amplitude of the digital signal.
However it is dependent on the Fs/interpolation of my USRP.
At 1e8/512 Fs and 512 interpolation the signal spectrum degrades and never
comes back (improve)
At 1e8/128Fs and 128 interpolation the signal spectrum degrades and from
time to time there is a sweep in which the goos spectrum comes back!
At 1e8/92Fs and 92 interpolation the signal spectrum degrades and comes back
every other sweep of the sprectrum analyzer.
However Simulink real time operation is needed to work with the USRP and as
long as we do more Simulink process we cannot achieve real time !! Why I
cannot transmit a good modulated tone at 512 interpolation??
>> 3) From the setting I specify on my USRP TX software I can see about 8KHz
>> frequency offset in the spectrum analyzer received signal. I connected a
>> signal generator (a good one) to the same spectrum analyzer and the offset
>> was 0.2 KHz at the same frequency (5.5GHz). Furthermore there is an
>> instability over time. (The signal is moving over time around 0.2KHz)
> I think this is expected. Oscillator only good to ~20 ppm or so?
4) Last comment. When setting a gain value (10dB )in the XCVR2450 (still in
TX mode) I can see in my spectrum analyzer the same harmonics than see the
same harmonics that I see when increasin the level of my digital signal in
simulink. My feeling is that you make use of the whole range of the DAC you
canot set gain in the XCVR2450. Otherwise the daughterboard is saturated. Am
> Can anybody explain some of these issues?
>> Many thanks,
>> USRP-users mailing list
>> USRP-users at lists.ettus.com
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the USRP-users