Hi, I am using USRP 1 for an application. I am using complex baseband representation, so i am setting bandwidth almost equal to the sampling rate. But many times i see that when i set the sample rate I get target sampling rate is different from actual sampling rate. I checked the wiki, it says 64Msps/N should be an integer. How should i decide the value of N so that the division is an integer. what is the significance of N? Secondly I have a doubt on samples/symbol. If I use complex baseband and say bpsk modulation, i set bits/symbol as one. should i set samples/symbol as 2 as there is one sample in I and one in Q. So is it default 2 samples/symbol or does it depend on anything else. Thank you.
on 2013-05-30 16:52
on 2013-05-30 17:05
On Thu, May 30, 2013 at 7:51 AM, Karan Talasila <email@example.com> wrote: > samples/symbol or does it depend on anything else. > Karan, In all discussions regarding complex baseband, a "sample" is counted as a single complex number, not separate I and Q samples (though with the USRP, I and Q are in fact sampled with separate ADCs in synchrony.) Regarding the configuration of the USRP1, the sample rate must be an integer fraction of 64Msps, and also less than or equal to 8 Msps due to the speed limitation of the USB 2.0 bus. N in this case is that integer; however, you don't need to specify this, but rather the resulting sample rate. The way the wiki page is written is probably a hold over from the (very) old, discontinued USRP blocks in GNU Radio, which did take N as a parameter. With BPSK modulation, depending on the technique you are using for matched filtering and timing recovery, you may want to set your initial receive rate to greater than one sample per symbol.
on 2013-05-31 03:58
ok..So N can be any integer. I am using USB 3.0 and it's allowing me sample rates almost close to 60MSPS without an error. I have an 8 core processor. But the surprise is even if i enter values greater than 64 MSPS it doesn't throw an error. Is that a bug as it doesn't show me any warning whatsoever saying you have gone beyond the device's maximum sampling rate. Secondly could you please explain briefly how samples per symbol depend on matched filter and timing recovery. And does it vary on modulation scheme used like say 64 QAM or so. I am of the opinion that whatever modulation scheme you use samples per symbol remains same but only bits/symbol changes. Am I right? Thank you. On Thu, May 30, 2013 at 8:33 PM, Johnathan Corgan
on 2013-05-31 04:03
On 05/30/2013 09:56 PM, Karan Talasila wrote: > ok..So N can be any integer. I am using USB 3.0 and it's allowing me > sample rates almost close to 60MSPS without an error. I have an 8 core > processor. But the surprise is even if i enter values greater than 64 > MSPS it doesn't throw an error. Is that a bug as it doesn't show me > any warning whatsoever saying you have gone beyond the device's > maximum sampling rate. You simply can't pack more than 8Msps onto the USB-2.0 interface on the USRP1. It doesn't speak USB-3.0, so your USB-3.0 hub will simply downshift. You can go to 16Msps with 8-bit samples, and that's as far as the USRP1 will go. It probably should throw an error for any sample rate that it can't actually deliver, although UHD *does* try to find the "closest rate", and it will comment on what rate it actually chose given the input rate request.