[USRP-users] Not getting white noise in a USRP N210

Marcus Müller marcus.mueller at ettus.com
Thu Sep 4 18:01:20 EDT 2014


Hi Harold,
Thanks! Let me comment on this.

On 04.09.2014 23:02, Harold Daniel Moreno Urbina wrote:
> Sure Marcus,
>
> ----------------------------------------------------
> uhd_fft.grc
> ----------------------------------------------------
>
> center frequency 2.45 GHz: *http://imgur.com/mmsdO0v
> <http://imgur.com/mmsdO0v>*
Not too sure what to make of this, but I'll come back to it later
>
> center frequency 2.42 GHz: *http://imgur.com/K526vsc
> <http://imgur.com/K526vsc>*
This is the classical LO leakage: Incredibly sharp, centered at the RF
frequency.
>
> ----------------------------------------------------
> Custom USPR2FFT
> ----------------------------------------------------
>
> center frequency 2.45 GHz: *http://imgur.com/3Ukz1F8
> <http://imgur.com/3Ukz1F8>*
>
> center frequency 1.80 GHz: *http://imgur.com/CkZsFyN
> <http://imgur.com/CkZsFyN>*
>
> center frequency 900 MHz: *http://imgur.com/MfVe34m
> <http://imgur.com/MfVe34m>*
>
> center frequency 300 MHz (out of SBX used operational band):
> *http://imgur.com/zSlv6PD
> <http://imgur.com/zSlv6PD>*
These look pretty clean to me, nicely located LO leakage. I don't know
what exactly happened when you set the frequency to 300MHz, but UHD
won't let you do that, because it's physically impossible, so it usually
prints a warning saying what it tuned to instead.
>
> ----------------------------------------------------
> Tried with other frequencies and sample rate
> I looked to have a center frequency not a factor
> of sample rate
> ----------------------------------------------------
>
> Center freqency 452 MHz, Sample Rate 9 MSps: http://imgur.com/oLgXmqN
>
> Center freqency 2759 MHz, Sample Rate 9 MSps: http://imgur.com/edysiBb
On the N210, sampling rates must be integer fractions of the 100MHz
master clock rate.
Most probably, UHD prints a warning telling you that it used an adjacent
possible rate -- 10MHz or 9.0909..MHz or so.

But that aside, that's very plausible.
So tuning with a N210 happens in two steps:
1. The driver looks up a frequency that the LO synthesizer can
physically generate, and configures it accordingly,
2. The FPGA eliminates the remaining frequency offset by multiplication
with exp(2*pi*j*delta_f*n) digitally.

So, for some frequencies, the LO is right in the middle where you see it
in all of your screenshots, for others, that can not be the case.
Actually, the synth of the SBX is quite flexible and you most probably
won't be able to distinguish between the target frequency hitting a
physically possible LO frequency, and the target frequency being right
in the middle between two possible LO frequencies at your spectral
resolution.

So, this explains why the LO can be completely shifted out of your 10MHz
bandwidth: just manually specify that you want to tune with an |offset|
> bandwidth/2, and the LO will be out of band.

Comparing the 452 and 2759 MHZ spectra reveals that they look really
really much alike (which is really comforting to me) but that 2.7GHz
spectrum is about 4dB lower -- which doesn't surprise much, considering
that nothing analog really works exactly frequency-independent. So this
is actually a sign for the USRP being a nice platform, and the SBX being
a sensitive analog frontend with a wideband input with only little
distortion.
> ----------------------------------------------------
> I used a second USRP N210
> ----------------------------------------------------
>
> Center freqency 2759 MHz, Sample Rate 9 MSps: http://imgur.com/pwFFCPB
>
> Center freqency 2760 MHz, Sample Rate 9 MSps: http://imgur.com/NrBbugz
>
> Center freqency 2760 MHz, Sample Rate 10 MSps: http://imgur.com/2NTmi2f
>
> Center freqency 2760 MHz, Sample Rate 10 MSps: http://imgur.com/CZ4oHOF
>
>
> I can see that when I use 10 MSps the signals is much more flat than the
> signals captured with 9 or 11 MSps.
That could be the CIC filter's rolloff that you can only see with odd
decimations (10MS/p is 1/10 of the 100MHz master clock rate, 9.091MS/s
is 1/11).
Maybe that's again where Ian could throw in his experience

Greetings,
Marcus


PS: regarding the first picture: yes, there are more peaks than the LO,
but they seem to be so random and the current PSD seems so uncorrelated
that I daresay that I cannot say anything.
> 2014-09-04 14:33 GMT-06:00 Marcus Müller <usrp-users at lists.ettus.com>:
>
>>  About the pictures: Would you mind uploading them somewhere else? It
>> really takes ages for me to see a single one, and it blocks parallely
>> waiting for multiples, so I must admit I only saw the first two before I
>> gave up. Maybe imgur.com?
>>
>> Greetings,
>> Marcus
>>
>> On 04.09.2014 22:12, Ian Buckley via USRP-users wrote:
>>
>> Good point, I answered that very quickly without looking at the downloads. Do you understand Marcus's point about direct conversion architectures? It was perhaps more appropriate.
>>
>>
>>
>> On Sep 4, 2014, at 1:04 PM, Harold Daniel Moreno Urbina <haroldmk at gmail.com> <haroldmk at gmail.com> wrote:
>>
>>
>>  But neither 2.45 GHz nor 2.42 GHz are factor of 100 MHz. The other frequencies used are (1.8 GHz, 900 and 300 MHz).
>>
>>
>> 2014-09-04 13:53 GMT-06:00 Ian Buckley <ianb at ionconcepts.com> <ianb at ionconcepts.com>:
>> Very likely it's a harmonic of the 100MHz clock used extensively in the N210.
>> Try tuning to a frequency thats not a factor of 100MHz and see if it is still there.
>>
>> On Sep 4, 2014, at 12:24 PM, Harold Daniel Moreno Urbina via USRP-users <usrp-users at lists.ettus.com> <usrp-users at lists.ettus.com> wrote:
>>
>>
>>  Hello,
>>
>> When using USRP N210 without an antenna I would expect to receive white noise, right?. But consistently I am getting signal at center frequency well over 20 dB compared to other frequencies and 30 or dB over the signals near the edge.
>>
>> Anyone knows what is happening here? Is something wrong in the USRP or in GNU Radio? or me? Thanks.
>>
>>
>> I am using USRP N210, GNU Radio 3.7.2.1 (version of last friday) over Ubuntu 14.04.
>>
>> I uploaded some screenshots of the FFT of the received signal, always 10Msps and getting about 5 minutes of signals. I used the example included in GNU Radio uhd_fft.grc and a custom flowgraph icluding only a USRP source and a FFT sink.
>>
>> ----------------------------------------------------
>> uhd_fft.grc
>> ----------------------------------------------------
>>
>> center frequency 2.45 GHz: https://mega.co.nz/#!WFE3UTRZ!AO04usGbl9TBa5-mCUPz7FYW2IabY5nbo5zZqSWwaM4
>>
>> center frequency 2.42 GHz: https://mega.co.nz/#!Hc0UzBJA!fSfgScdXPJ_phKdIWUKcNseonlKQt3m64yvYMR0bDhs
>>
>> ----------------------------------------------------
>> Custom USPR2FFT
>> ----------------------------------------------------
>>
>> center frequency 2.45 GHz: https://mega.co.nz/#!Hd9RyApS!FzsY5yxr5e6dJaAAO7Ww2O4BFefyCk6MDPN6ffcOs_g
>>
>> center frequency 1.80 GHz: https://mega.co.nz/#!LQcTkDjb!NQ7rIe1PDgRFCp547_U3mhxxxzsvQilPV-esTo7nHa4
>>
>> center frequency 900 MHz: https://mega.co.nz/#!eZ1TFbSY!nkt-n1mlC54aFfDwvqoJIJd7-a7gMB2GBLLhyclxxcs
>>
>> center frequency 300 MHz (out of SBX used operational band): https://mega.co.nz/#!mJ8EESwR!lI_0r4RvvTncxwwMnpEVAvrwsVhcHzLX7H6C7TA4Cj0
>> _______________________________________________
>> USRP-users mailing listUSRP-users at lists.ettus.comhttp://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com
>>
>>
>>
>> _______________________________________________
>> USRP-users mailing listUSRP-users at lists.ettus.comhttp://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com
>>
>>
>>
>> _______________________________________________
>> USRP-users mailing list
>> USRP-users at lists.ettus.com
>> http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com
>>
>>





More information about the USRP-users mailing list