Return-path: Received: from mail-wi0-f180.google.com ([209.85.212.180]:61491 "EHLO mail-wi0-f180.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755755AbaGPI4X convert rfc822-to-8bit (ORCPT ); Wed, 16 Jul 2014 04:56:23 -0400 Received: by mail-wi0-f180.google.com with SMTP id n3so880132wiv.13 for ; Wed, 16 Jul 2014 01:56:18 -0700 (PDT) MIME-Version: 1.0 In-Reply-To: <201407161038.10130.sw@simonwunderlich.de> References: <1405441966-32681-1-git-send-email-sw@simonwunderlich.de> <4111404.UJkIUxkzQm@bentobox> <201407161038.10130.sw@simonwunderlich.de> Date: Wed, 16 Jul 2014 10:56:16 +0200 Message-ID: (sfid-20140716_105643_829035_F958E1A9) Subject: Re: [RFC 0/2] ath10k spectral scan support From: Michal Kazior To: Simon Wunderlich Cc: Sven Eckelmann , "ath10k@lists.infradead.org" , linux-wireless , "Giori, Kathy" , Mathias Kretschmer Content-Type: text/plain; charset=UTF-8 Sender: linux-wireless-owner@vger.kernel.org List-ID: On 16 July 2014 10:38, Simon Wunderlich wrote: > >> On Wednesday 16 July 2014 08:24:58 Michal Kazior wrote: >> > > * So far, we always get 64 bins of fft data in the samples, even if we >> > > >> > > are in HT40 or VHT80 mode. That doesn't seem right, as there should >> > > be more bins for wider channels (e.g. 128 or 256 bins). Is there any >> > > parameter to change in the configuration to get the output from >> > > multiple channels / higher bandwidth? >> > >> > Hmm.. I suspect this may be because you use scan command. Scanning >> > uses channel list which is programmed with legacy modes a and g >> > (implying 20Mhz). Perhaps that's the limiting factor now. Using other >> > (wider) modes breaks scan from what I remember but it might be worth >> > checking out if it impacts fft bin sizes. >> >> I did the HT80 and HT40 stuff on a single channel in the spectral >> "background" mode. They reports from the firmware (10.1) still only had 64 >> bins. The HT80 stuff wasn't tested with the "main" firmware because it >> refused to switch to HT80 mode as AP/Sta. > > Just to clarify this a little more, we tried AP (running hostapd) as well a > STA in 40 and 80 MHz modes. We always got 64 bins returned. It seems the number of bins is 2^ (fft_size - bin_scale) = 2^(7-1) = 64. Changing fft_size to 9 yields 256 bins. Perhaps fft reports are (implicitly) relative to the mode the hw is in at a given time and if you want a reasonable bin resolution for, e.g. VHT80 you need to increase fft_size? Michał