[ALSA - driver 0001151]: Audigy sampling rate seems off

[bugtrack@alsa-project.org]

The following issue has been ASSIGNED.
======================================================================
<https://bugtrack.alsa-project.org/alsa-bug/view.php?id=1151> 
======================================================================
Reported By:                felix
Assigned To:                jcdutton
======================================================================
Project:                    ALSA - driver
Issue ID:                   1151
Category:                   PCI - emu10k1
Reproducibility:            always
Severity:                   minor
Priority:                   normal
Status:                     assigned
Distribution:               Suse 9.1
Kernel Version:             2.6.5-7.147-default
======================================================================
Date Submitted:             06-02-2005 23:07 CEST
Last Modified:              07-18-2006 14:45 CEST
======================================================================
Summary:                    Audigy sampling rate seems off
Description: 
I'm doing an experiment that involves generating a tone, sending it out
the
sound card and into a (passive) circuit, and recording from the circuit
back
into line in at the same time.

When I play a sample: mono, 8kHz sample rate, containing a synthetic 1kHz
sin signal, the recording I get is of a signal 1000.12Hz. If I send 100Hz,
I get back 100.012Hz. I use 'snd' to do a 256K-point FFT across a 10
second sample just to be sure. I've also verified this using FFTW and
correlation-based programs of my own.

I've also verified that the reference 1kHz and 100Hz signals are exact.

As a sanity check, I tried the same experiment on the built-in sound card
on that same PC. The signals come back correct.

Note the seeming coincidence that the frequency of the signal coming back
is 1/8000 higher than the signal output.

I've never seen anything like it. Not that I've been looking before...
======================================================================

----------------------------------------------------------------------
 sst4 - 06-03-05 00:29 
----------------------------------------------------------------------
If it's rather off by 1/8192 (8192 is 2 to 13th power), then it'd look very
much programmatic.

Just an observation - I know nothing about "audigy".

----------------------------------------------------------------------
 perex - 06-03-05 10:59 
----------------------------------------------------------------------
Note that the routine which calculates "pitch" for playback is in
alsa-kernel/pci/emu10k1/emupcm.c - emu10k1_calc_pitch_target(). It's quite
simple, so you may do experiments with it.

Also, upgrade to latest ALSA 1.0.9, there are some cleanups for the sample
cache, and it might be also related.

----------------------------------------------------------------------
 felix - 06-03-05 17:45 
----------------------------------------------------------------------
to sst4, re. powers of 2: Good point. I also tried 32kHz sampling, and the
problem follows! ie the frequency error drops to ~ 1/32000. I *wish* I
could tell whether it was 32k or 32K, but my measurement tools can't give
me enough significant digits :-( I can, however, try more sampling rates
to see whether the problem follows 1/sample rate or jumps between powers
of 2...

re. audigy: my vague. /proc/asound/cards says:
0 [Audigy         ]: Audigy - Sound Blaster Audigy
              Sound Blaster Audigy (rev.3, serial:0x531102) at 0xa000, irq
11

to perex: Thanks for the code pointers. Will also definitely upgrade to
1.0.9.



----------------------------------------------------------------------
 felix - 06-04-05 03:42 
----------------------------------------------------------------------
More results: Upgraded kernel to 2.6.11.11, installed 1.09a drivers outside
the kernel and all 1.09 libraries and tools. The problem remains.

However...
I ran a spectrum of sampling rates through emu10k1_calc_pitch_target() and
found that from 6kHz through 96kHz, at intervals of 3kHz, pitch_target
ends up being divisible by 1024. And as it turns out, the frequency of a
tone at these sampling rates is exactly correct. Perhaps this will serve
as a clue to someone familiar with the driver and hardware. For all I
know, this is common to all emu10k1 cards.

And yes, 44.1kHz displays the error, whereas 48kHz and 96kHz do not.

----------------------------------------------------------------------
 felix - 06-04-05 03:58 
----------------------------------------------------------------------
Two more exciting data, to see whether it is possible to isolate the
problem to playback or recording:

1. synthesize and aplay a 1kHz tone at 8kHz sampling rate,
   simultaneously record at 12kHz sampling rate
   => Recorded tone is shifted up by ~ 1/8k

2. synthesize and aplay a 1kHz tone at 12kHz sampling rate,
   simultaneously record at 8kHz sampling rate
   => recorded tone is EXACTLY 1000Hz

Issue History
Date Modified  Username       Field                    Change              
======================================================================
06-02-05 23:07 felix          New Issue                                    
06-02-05 23:07 felix          Distribution              => Suse 9.1        
06-02-05 23:07 felix          Kernel Version            => 2.6.5-7.147-default
06-03-05 00:29 sst4           Note Added: 0004879                          
06-03-05 10:59 perex          Note Added: 0004883                          
06-03-05 17:44 felix          Note Added: 0004884                          
06-03-05 17:45 felix          Note Edited: 0004884                         
06-04-05 03:42 felix          Note Added: 0004885                          
06-04-05 03:58 felix          Note Added: 0004886                          
07-18-06 14:45 jcdutton       Status                   new => assigned     
07-18-06 14:45 jcdutton       Assigned To               => jcdutton        
======================================================================



-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV