In the DSP prior to sending to USRP I scale everything by a const 0.05, the TX gain for the USRP is 0.5 normalized. ----------------------------- Jacob Knoles
On Fri, May 11, 2018 at 11:40 AM Ian Buckley <i...@ionconcepts.com> wrote: > What TX gain have you got set? Are you sure you are operating the PA in > it’s linear region? OFDM waveforms are notorious for there PAPR > requirements. > > On May 11, 2018, at 11:23 AM, Jacob Knoles <knole...@gmail.com> wrote: > > Hey guys, > > Please pardon my ignorance, I am trying to learn everything I need for DSP > / OFDM on the fly. But I have noticed something with this OFDM gnuradio > example. If I take the example and run it directly into the usrp without > changing any variables, I get a 350 KHz wide signal out. > Now the sample rate is only 100k, the occupied carriers are -26 to 26 > (zero omitted) with pm 21 and pm 7 as pilot carriers. The fft length is > 64. > > If the bandwidth is directly related to the sampling rate how am I getting > 3x the bandwidth at low sample rates. > > For some further comparison I changed only the sample rate to a few other > values, here is what I observed: > > Sample Rate : Observed Bandwidth > 100k : 350 KHz > 50k : 350 KHz > 1M : 840 KHz > 20M : 17 MHz > > The 1M and 20M rates make sense but I don't understand what is happening > with the 100k and 50k rates. > > Thank you for the help. > ----------------------------- > Jacob Knoles > > > > On Wed, May 9, 2018 at 6:22 PM Marcus D. Leech <mle...@ripnet.com> wrote: > >> On 05/09/2018 07:57 PM, Jacob Knoles wrote: >> >> Thanks for the quick reply guys. >> >> Marcus the Re-sampling option makes perfect sense, and I believe, in >> theory, since I am writing data to a file for later use I could interpolate >> it just before writing then read it out at the usrp sample rate, right? >> >> Yes. >> >> >> Ian, very interesting suggestion. I will have to give it a try. Thanks >> for the input. And since I am doing all of the heavy processing prior to >> tx'ing I don't image this change will create too great of a burden on the >> CPU. As for reading from the file, I am just creating a small data set >> which gets loaded into memory and repeated. >> >> Thanks!! >> ----------------------------- >> Jacob Knoles >> >> >> >> On Wed, May 9, 2018 at 4:21 PM Ian Buckley via USRP-users < >> usrp-users@lists.ettus.com> wrote: >> >>> >>> > On May 9, 2018, at 4:07 PM, Marcus D. Leech via USRP-users < >>> usrp-users@lists.ettus.com> wrote: >>> > >>> > On 05/09/2018 06:53 PM, Jacob Knoles via USRP-users wrote: >>> >> Hello All, >>> >> >>> >> I am trying to generate OFDM signals of various bandwidths using the >>> X300 (UBX-160), particularly 20/40/80 and 160 MHz bandwidths. >>> >> I have used the gnuradio ofdm_tx.grc example file to generate a data >>> file which I then feed into the USRP an monitor on a spectrum analyzer. >>> >> >>> >> To quickly note, I do not care about the data being transferred, it >>> will not be received or demodulated in any way and is simply an interfering >>> signal. >>> >> >>> >> At this time I can produce a 20 MHz wide OFDM signal as well as a 100 >>> MHz wide signal (?) but the 40/80 MHz signals are rounded and look more >>> like an 802.11b signal. >>> >> >>> >> I have noted a message from the X300 that the requested sample rates >>> (40/80 MS respectively) cannot be achieved due to the 200/x ratio being odd. >>> >> >>> >> So my question is this, how do I decouple the USRP's sample rate with >>> the bandwidth of the signal I am trying to produce? >>> >> To put it another way, I produce a data file at 40 MS/s rate then run >>> it on the X300 at 100 MS/s and I get a 100 MHz wide signal instead of the >>> 40 MHz I want. >>> >> >>> >> Thanks for the help. >>> >> ----------------------------- >>> >> Jacob Knoles >>> >> >>> > You would need to interpolate it up to the desired rate. UHD has no >>> way of knowing that your samples represent data sampled at 40Msps, so when >>> you >>> > pull it out of your file at 100Msps, it will get presented as if it >>> were 100Msps data. >>> > >>> > You'll need to use some DSP code, or Gnu Radio to up-sample your >>> sample file. >>> > >>> ….or perhaps generate it off line using a non 2^n Fourier transform size >>> that targets the USRP sample rate…for example instead of 64 bins @ 40MHz, >>> 80 bins @ 50MHz, >>> With zero data in the extra outlying bins (as you would have anyway in >>> other bins). Might get interesting getting high bitrates out of a file, but >>> equally, high bitrate M:N sample rate conversion will also be tricky for CPU >>> _______________________________________________ >>> USRP-users mailing list >>> USRP-users@lists.ettus.com >>> http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com >>> >> >> >
_______________________________________________ USRP-users mailing list USRP-users@lists.ettus.com http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com
