Re: [Discuss-gnuradio] question about bandwith B210 - USB3

[Marcus Müller]

Hi Przemek,

I'll address the different aspects you've mentioned below:

On 01.07.2016 11:30, Przemek Lewandowski wrote:

To James Humphries 

8Gb Ram, 1TB Hdd hybrid with 8 GB SSD, linux kubuntu and about USB 3: this is mu output:

address@hidden:~$ lsusb

Bus 002 Device 002: ID 8087:8000 Intel Corp. 

Bus 002 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub

Bus 001 Device 002: ID 8087:8008 Intel Corp. 

Bus 001 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub

Bus 004 Device 005: ID 2500:0020  

Bus 004 Device 001: ID 1d6b:0003 Linux Foundation 3.0 root hub

Bus 003 Device 005: ID 8087:07dc Intel Corp. 

Bus 003 Device 004: ID 174f:1486 Syntek 

Bus 003 Device 003: ID 058f:6366 Alcor Micro Corp. Multi Flash Reader

Bus 003 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub

address@hidden:~$ lspci | grep USB

00:14.0 USB controller: Intel Corporation 8 Series/C220 Series Chipset Family USB xHCI (rev 05)

00:1a.0 USB controller: Intel Corporation 8 Series/C220 Series Chipset Family USB EHCI #2 (rev 05)

00:1d.0 USB controller: Intel Corporation 8 Series/C220 Series Chipset Family USB EHCI #1 (rev 05)

These are controllers known to work relatively well. I think that the problem really is in the CPU load your Flow graph has, probably

To Marcus D. Leech:

So, if your signal is only a few kHz wide, why do you need it to be sampled at 18Msps?  - 

Answer is: I want to transmmit also different modulation like AM, FM Narrow, FM Wide, QAM, QPSK, GFSK in other channels in the same time - for my university project. So Im optimizing each modulations to run as high as possible :) and im looking good practises for makeing better performance. And I know that every block is getting some prcessor calculations , im checking where is the limit :)

Yes, but your limit checking is done in a very inefficient manner – for example, the band pass filter (which really seems mis-configured in sampling rate, considering the "16k" in the file name, also, not sure you would actually want to use a band pass here) could very efficiently do the job of the 90-fold interpolation, and thus you could completely safe the effort that the rational resampler poses (which should contribute *very* drastically to the overall CPU load).
Also, when constructing a signal out of N evenly spaced subcarriers, it's recommendable to use the Polyphase synthesizer[2], which is *way* more efficient than having N interpolators and a summation.
So, the problem really is that you're checking /some/ limits, but it's not inherently clear they will somehow be related to the thing you need to solve later on.

 To Marcus Muller:
I can't really make much sense of your flow graph; maybe you could explain what you need to do? The parametrizations of the different blocks and the file name in the wav file source kind of contradict. I think, however, we could find something that does just what you want. And as a gut feeling: Your computer will probably be fast enough for it in the end, if we only do things that are necessary.

Answer: I know that Sampling Rate should be as low as possible and minimum of 2*fmax

This is complex baseband, so 1*f_max is sufficient (2*f_max is for the real-valued case, only). So, for the USRP side, 18 MS/s means 18 MHz of bandwidth.

but as I said I need wide bandwith to put another modulations in channels. So on my flow graph I was thinking that setting 18 MS on USRP will make 18 MHz bandwith, so my rational resampler is changing sample rate from a small stream of 16 Khz to wider band.

yes, but you seem to treat your 16kHz signal as if it was sampled at 200 kHz

This flowgraph that was sent shoud be a narrowband SSB modulation : ) inside a wideband of 18 MHz - if Im not right please correct me.

Yeah, as said, the parametrization of the band pass doesn't seem right. Also, SSB in baseband, by definition, doesn't have a spectrum you can represent using a real-valued signal, so that's wrong, too.
Remember: real-valued signals must always be symmetric in spectrum, and every real filter can only preserve that symmetry. I happen to have just held a short lecture on that topic at a amateur radio conference[1], if the theory sounds new to you.
Also: the Qt GUI graphical sinks are really your friend! Replace your USRP sink by a Qt GUI frequency sink. Set its center frequency to what you would have tuned the USRP to, and make sure it's . That's what your spectrum on the air would look like. You'll notice that it would be symmetric around the center frequency, still, because you've kept the real nature of your original signal all the way through.

Maybe an example of how one can have multiple channels in a wider bandwidth quite easily with the Polyphase Synthesizer helps you[2].

My mentor at university has got another Ettus Reasearch devices such as N - series and X - series, so maybe i will verify if it is not a USB problem.

you could try to replace the USRP sink by a "probe rate" block, connected to a "message debug" sink's "print" input. That way, you could see if the rate the flowgraph can run at is sufficiently high.

Best regards,
Marcus

[1] http://marcus.hostalia.de/sdra16.pdf . Youtube Videos are promised to appear at some point in time.
[2] https://gist.github.com/d98a47c29098f68628d3b32959b02432 Double-click in the Waterfall plot to change the frequency of the channel-selecting xlating FIR filter.

All the best Przemek

2016-07-01 9:46 GMT+02:00 Marcus Müller <address@hidden>:

Hi Przemek,

I can't really make much sense of your flow graph; maybe you could explain what you need to do? The parametrizations of the different blocks and the file name in the wav file source kind of contradict. I think, however, we could find something that does just what you want. And as a gut feeling: Your computer will probably be fast enough for it in the end, if we only do things that are necessary.

Best regards,

Marcus

On 01.07.2016 00:48, Przemek Lewandowski wrote:

Hello 

I need to generate as wide bandwith as possible. Yes I know that this is waste of energy, but beside this simple example I want to put some other channels. 

On my USRP B210 I can afford now about 18 MS.

I read on Ettus page that B200 can tx about 40 - 60 MS - depends on USB driver.

So 15 MS is 3 times slower :) Im very beginer with filters at gnu radio so I suppose that default filters can be not so good optimized and propably I will need it to fix it manually.

What do You think ???

This is my graph:
https://www.dropbox.com/s/xzwvb0dm25ubh5i/bandwith_test2.grc.png?dl=0

All The Best.

_______________________________________________
Discuss-gnuradio mailing list
address@hidden
https://lists.gnu.org/mailman/listinfo/discuss-gnuradio

_______________________________________________
Discuss-gnuradio mailing list
address@hidden
https://lists.gnu.org/mailman/listinfo/discuss-gnuradio