[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Discuss-gnuradio] RFNoC FFT Size

From: Marcus D. Leech
Subject: Re: [Discuss-gnuradio] RFNoC FFT Size
Date: Thu, 03 Mar 2016 11:47:26 -0500
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:31.0) Gecko/20100101 Thunderbird/31.7.0

On 03/03/2016 11:10 AM, Simon Olvhammar wrote:
Ok thanks, so it's a limit in the RFNoC software and not related to Ettus FPGA hardware or something like that?
I dont see that limit in regular GNURadio and 8192 channels works fine, isnt the packet size also max 9000 there?
Martin/Jonathan could comment on FPGA limits--I don't think anything larger than 2048 is actually implemented, but I don't see
  what one couldn't internally compute a larger FFT in the FPGA, being careful to consider resources.

In regular GNU Radio, FFTs are being computed on the host PC, there's no packet-size limitation.  The limit in RFNOC is due to the
  way data are packetized between the X310 and the host PC.  Since there's no current mechanism for splitting FFTs across multiple
  VITA frames, you're limited to what you can do in a single ethernet-layer frame.

On Thu, Mar 3, 2016 at 3:45 PM, Marcus D. Leech <address@hidden> wrote:
On 03/03/2016 05:21 AM, Simon Olvhammar wrote:

I'm wondering what it would take to modify the current FFT RFNoC block to be able to compute a 8192 channel FFT. Someone that can guide me in the right direction? I'm a quite experienced programmer, however not in HDL.

Best regards

The main current stumbling block to large FFTs in RFNOC is that the FFT output has to be able to fit into a single transport packet, and
  that is limited to 9000 bytes, which pretty-much limits you to 2048-bin FFTs.  There will be work in the future to allow FFT outputs in
  RFNOC to span multiple packets.

Discuss-gnuradio mailing list

reply via email to

[Prev in Thread] Current Thread [Next in Thread]