# USRP + GNU Radio as 1 GHz bandwidth spectrum analyzer

I have maybe not trivial but some interesting question. I need to create 802.11ac channels spectrum analyzer with USRP.

• Hardware: USRP B210
• PC: Intel Core i3 i3210m 16 GB RAM.

According to documentation from: https://www.ettus.com/content/files/b200-b210_spec_sheet.pdf USRP has 56 MHz bandwidth, so FFT analysis greater than 56 MHz did it (by my opinion), by some multiple snapshots with 56 MHz.

During to experiments FFT with 1 GHz bandwidth refresh 1 per 3 seconds, so it can not be used for real time monitoring in that condition.

Of course it takes some processing time, but can somebody tell, it is possible to speedup FFT analysis on big bandwidth: 500 MHz, 1 GHz, at least to 1 refresh per 1 second, or it is not possible due to processing limits.

As my opinion, maybe need to investigate verilog code, that uploaded to USRP FPGA before first start, maybe it can be increased reference frequency with frequency multiplier, or not?

Marcus Müller, thanks for reply. I don't know how it can be enough detailed architecture explanation, but try to explain. user34533 = WhiteBear. Sorry i use stack exchange first, and didnt setup correct login from stack overflow. In future i will put posts from my regular account.

So: USRP connected to the laptop via USB cable. In laptop installed Linux Xubuntu and GNU Radio Companion software. GRC interact with USRP via USRP source abstraction block, that powered by official UHD (USRP hardware driver), provided by Ettus Research also as open software. FFT block - standard GNU Radio block, and did by laptop. Bandwidth that displayed on FFT directly related to sample rate of USRP, so maybe i didnt correct inform You. So when it setup to large values (greater that 400 MHz) FFT became continuously slow and slow refresh the spectrum.

Flowcharts and settings of blocks You can find on the images, ready to provide additional information. Thanks a lot for reply, hope, You can help

• an FFT doesn't have a bandwidth; it's just a base transform. Can you elaborate on how the system you're trying to works, architecturally? – Marcus Müller Mar 18 '18 at 18:22
• The USRP cannot sample at 1G. Your flowgraph shouldn't even compile. – MBaz Mar 18 '18 at 21:56
• @MBaz that's not checked at compile-time, as you might not know which sampling rates are supported at that time. – Marcus Müller Mar 19 '18 at 9:21
• @MarcusMüller You're right... it should fail at run time. – MBaz Mar 19 '18 at 12:55