From: Fredrik Lingvall Subject: Re: Aliasing on ALi5451 Date: Tue, 27 May 2003 15:09:32 +0200 Sender: alsa-devel-admin@lists.sourceforge.net Message-ID: <3ED3638C.7090006@signal.uu.se> References: <3ECF2AC9.8070503@signal.uu.se> <1053786705.876.11.camel@localhost> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <1053786705.876.11.camel@localhost> Errors-To: alsa-devel-admin@lists.sourceforge.net List-Help: List-Post: List-Subscribe: , List-Unsubscribe: , List-Archive: To: manuel.jander@embedded.cl Cc: alsa-devel@lists.sourceforge.net List-Id: alsa-devel@alsa-project.org Manuel Jander wrote: >As far as i know, using a fixed rate codec implies a fixed >antialiasing filter too. So there is no need to adjust a >antialiasing filter. > Yes your right the sample rate converters should do the job. > I think thats the really reason why >most audio card use AC97 codec with a samplerate converter >instead of a programable codec, since the a varying antialiasing >filter semes more costly to me than a samplerate coverter. > >While implementing the Aureal alsa driver, i found several situations >where bad samplerate converter parameters caused strong aliasing. >I would try to find the bug in the samplerate converter setup >if any. >On the other hand, there are some audio cards like the CMedia 8378 >that have nearly useless ADC's in their codec, because of too low >samplerates. If the codec is a AC97, it should be fine, because >they usually work at 48KHz, so there could only be problems with >white noise (bad chip+board or preamp design). > Yes the codec is (should be) AC97 compatible. The sample rate converters are controlled by two registers (2Ch and 32h) one for the DAC:s and one for the ADC:s. The values of these registers for the supported sample rates are: 8000 1F40 11050 2B11 16000 3B80 22050 5622 32000 7D00 44100 AC44 48000 BB80 I have tried to find the code in the ALSA sources where these values are set without success (any help appriciated). My idea is to set the sampling rate to a fixed value, for example AC44 (44.1 kHz), recompile the alsa drivers, and check if the problem dissapears. Fredrik > >Best regards, > >Manuel Jander. > >On Sat, 2003-05-24 at 04:18, Fredrik Lingvall wrote: > > >>Hi! >> >>I've tried to do loadspeaker identification using the ALi5451 chipset >>in my laptop. I saw a strong drop in the spectrum above ~2kHz when >>a ran the ADC:s at 44100 (I used a high quality microphone and a mic >>pre-amp). I then wired the audio output directly to the input with >>the same result - strong attenuation above 2kHz (I used wideband >>white noise as identication signal). Then I generated a chirp signal >>(5 sec sine sweep from 0-24kHz using Matlab) at 48kHz sampling rate >>which I recorded using Audacity at 8, 44.1, and 48 kHz. There was >>clearly freqeuency folding (aliasing) for all sampling freqs except 48 kHz >>(see the time-freq plot at: >>http://www.signal.uu.se/Staff/fl/chirp_test.jpg). >> >> From the specs of the cs4299 chip (used in ALi 5451) it allways runs >>at 48 kHz and uses sample rate converters for the other sampling freqs. >>However, the anti-aliasing filters (if there are any) do not seem to adapt >>when changing sampling freq. I also checked >> >>cat /proc/asound/5451/pcm0p/sub0/hw_params >> >>and >> >>cat /proc/asound/5451/pcm0c/sub0/hw_params >> >>when playing/recording and the sampling freqs. was correct. >>Is this a driver issue or is it just a bad sound chip? >> >>Regards >> >>Fredrik >> >> >> >> >>------------------------------------------------------- >>This SF.net email is sponsored by: ObjectStore. >>If flattening out C++ or Java code to make your application fit in a >>relational database is painful, don't do it! Check out ObjectStore. >>Now part of Progress Software. http://www.objectstore.net/sourceforge >>_______________________________________________ >>Alsa-devel mailing list >>Alsa-devel@lists.sourceforge.net >>https://lists.sourceforge.net/lists/listinfo/alsa-devel >> >> >> > > > > ------------------------------------------------------- This SF.net email is sponsored by: ObjectStore. If flattening out C++ or Java code to make your application fit in a relational database is painful, don't do it! Check out ObjectStore. Now part of Progress Software. http://www.objectstore.net/sourceforge