26 Oct
2009
26 Oct
'09
2:33 a.m.
Hi everybody,
I have a question regarding amplifying a PCM frame. For each frame, I get a
float number, and multiply this number by 2, and then output frames to alsa
one by one using function snd_pcm_writei. But the wav file sounds the same
to me, nothing gets amplified. Could anyone explain this to me? I appreciate
your help!
Warmest regards,
Dripstone
26 Oct
26 Oct
5:07 a.m.
On 10/26/2009 07:33 AM, Drip Stone wrote:
Hi everybody,
I have a question regarding amplifying a PCM frame. For each frame, I
get a float number, and multiply this number by 2, and then output
frames to alsa one by one using function snd_pcm_writei. But the wav
file sounds the same to me, nothing gets amplified. Could anyone explain
this to me? I appreciate your help!
you should hear an amplification by roughly 6dB... what might happen is
this: your file already has a peak at 1.0f. after your operation, it
will peak at 2.0f, which your DAC can't handle. so the file will have to
be clamped between -1.0f and 1.0f, effectively undoing your gain
operation. just a guess.
or you might be expecting the wrong thing... times two doesn't feel
"twice as loud". then again, +6dB is quite audible in any case.
6:38 a.m.
Well if the audio was clipped, there would probably be some audible
distortion.
Perhaps if a code fragment was posted we could give a more informed
opinion.
Victor
On 26 Oct 2009, at 09:08, Jörn Nettingsmeier wrote:
On 10/26/2009 07:33 AM, Drip Stone wrote:
Hi everybody,
I have a question regarding amplifying a PCM frame. For each frame, I
get a float number, and multiply this number by 2, and then output
frames to alsa one by one using function snd_pcm_writei. But the wav
file sounds the same to me, nothing gets amplified. Could anyone
explain
this to me? I appreciate your help!
you should hear an amplification by roughly 6dB... what might happen
is
this: your file already has a peak at 1.0f. after your operation, it
will peak at 2.0f, which your DAC can't handle. so the file will
have to
be clamped between -1.0f and 1.0f, effectively undoing your gain
operation. just a guess.
or you might be expecting the wrong thing... times two doesn't feel
"twice as loud". then again, +6dB is quite audible in any case.
_______________________________________________
Linux-audio-dev mailing list
Linux-audio-dev(a)lists.linuxaudio.org
http://lists.linuxaudio.org/mailman/listinfo/linux-audio-dev
7:07 a.m.
On 10/26/2009 11:38 AM, Victor Lazzarini wrote:
Well if the audio was clipped, there would probably be
some audible
distortion.
Perhaps if a code fragment was posted we could give a more informed
opinion.
i was thinking of operating on files, where clamping can be done with
perfect fidelity. but i may have misunderstood the OP.
6:05 p.m.
Hi,
Thank you for providing valuable information. I will put some code here:
while ( 1 )
{
for ( int i = 0; i < frames; i++ )
{
float t = 0;
if ( left ) //for left channel
{
t = *((float*)&(fbuffer[pos+i*4])); //4 bytes each frame
t*=2;
//printf("%f ",t);
for ( int j = 0; j < 4; j++ )
buffer[i*8+j] = ((char*)&t)[j]; //8 bytes for 2
channels
}
rc = snd_pcm_writei ( handle, buffer, frames );
pos += frames * 4;
}
I think the value is not out of [-1.0,1.0]. BTW, what does it mean when a
value is negative?
Best,
Dripstone
2009/10/26 Jörn Nettingsmeier <nettings(a)folkwang-hochschule.de>
On 10/26/2009 11:38 AM, Victor Lazzarini wrote:
Well if the audio was clipped, there would
probably be some audible
distortion.
Perhaps if a code fragment was posted we could give a more informed
opinion.
i was thinking of operating on files, where clamping can be done with
perfect fidelity. but i may have misunderstood the OP.
_______________________________________________
Linux-audio-dev mailing list
Linux-audio-dev(a)lists.linuxaudio.org
http://lists.linuxaudio.org/mailman/listinfo/linux-audio-dev
6:39 p.m.
On Mon, 2009-10-26 at 15:05 -0700, Drip Stone wrote:
Hi,
Thank you for providing valuable information. I will put some code
here:
while ( 1 )
{
for ( int i = 0; i < frames; i++ )
{
float t = 0;
if ( left ) //for left channel
{
t = *((float*)&(fbuffer[pos+i*4])); //4 bytes each
frame
t*=2;
//printf("%f ",t);
for ( int j = 0; j < 4; j++ )
buffer[i*8+j] = ((char*)&t)[j]; //8 bytes for 2
channels
}
rc = snd_pcm_writei ( handle, buffer, frames );
pos += frames * 4;
}
I think the value is not out of [-1.0,1.0]. BTW, what does it mean
when a value is negative?
Ouch, you're doing a byte-by-byte copy of a 32bit float?? Please, just
use the correct data types.
Floating point audio data is usually signed, so -1.0 and 1.0 are the
farthest extents, and 0.0 would be the default no-signal state.
Peter.
7 p.m.
Nothing in the posted code fragment suggests that neither in nor out is
remotely related to floats. Where do you get yor in-data from? Why is
fbuffer[] a bytearray (if that is what it is?) How was ALSA opened?
/j
On Mon, 2009-10-26 at 15:05 -0700, Drip Stone wrote:
while ( 1 )
{
for ( int i = 0; i < frames; i++ )
{
float t = 0;
if ( left ) //for left channel
{
t = *((float*)&(fbuffer[pos+i*4])); //4 bytes each
frame
t*=2;
//printf("%f ",t);
for ( int j = 0; j < 4; j++ )
buffer[i*8+j] = ((char*)&t)[j]; //8 bytes for 2
channels
}
rc = snd_pcm_writei ( handle, buffer, frames );
pos += frames * 4;
}
I think the value is not out of [-1.0,1.0]. BTW, what does it mean
when a value is negative?
Best,
Dripstone
2009/10/26 Jörn Nettingsmeier <nettings(a)folkwang-hochschule.de>
On 10/26/2009 11:38 AM, Victor Lazzarini wrote:
Well if the audio was clipped, there would
probably be some
audible
distortion.
Perhaps if a code fragment was posted we could give a more
informed
opinion.
i was thinking of operating on files, where clamping can be
done with
perfect fidelity. but i may have misunderstood the OP.
_______________________________________________
Linux-audio-dev mailing list
Linux-audio-dev(a)lists.linuxaudio.org
http://lists.linuxaudio.org/mailman/listinfo/linux-audio-dev
_______________________________________________
Linux-audio-dev mailing list
Linux-audio-dev(a)lists.linuxaudio.org
http://lists.linuxaudio.org/mailman/listinfo/linux-audio-dev
8:50 p.m.
Hello again,
I obtain data for fbuffer in following way:
char *fbuffer = ( char * ) malloc ( 10240000 ); //10M bytes
long length = fread ( fbuffer, 1, 10240000, file );
I also set alsa in float mode:
snd_pcm_hw_params_set_format ( handle, params,
SND_PCM_FORMAT_FLOAT_LE );
I pasted all the code at http://pastebin.com/m7fdf0957.
Thank you,
Dripstone
On Mon, Oct 26, 2009 at 3:59 PM, Jens M Andreasen
<jens.andreasen(a)comhem.se>wrote;wrote:
Nothing in the posted code fragment suggests that
neither in nor out is
remotely related to floats. Where do you get yor in-data from? Why is
fbuffer[] a bytearray (if that is what it is?) How was ALSA opened?
/j
On Mon, 2009-10-26 at 15:05 -0700, Drip Stone wrote:
while ( 1 )
{
for ( int i = 0; i < frames; i++ )
{
float t = 0;
if ( left ) //for left channel
{
t = *((float*)&(fbuffer[pos+i*4])); //4 bytes each
frame
t*=2;
//printf("%f ",t);
for ( int j = 0; j < 4; j++ )
buffer[i*8+j] = ((char*)&t)[j]; //8 bytes for 2
channels
}
rc = snd_pcm_writei ( handle, buffer, frames );
pos += frames * 4;
}
I think the value is not out of [-1.0,1.0]. BTW, what does it mean
when a value is negative?
Best,
Dripstone
2009/10/26 Jörn Nettingsmeier <nettings(a)folkwang-hochschule.de>
On 10/26/2009 11:38 AM, Victor Lazzarini wrote:
Well if the audio was clipped, there would
probably be some
audible
distortion.
Perhaps if a code fragment was posted we could give a more
informed
opinion.
i was thinking of operating on files, where clamping can be
done with
perfect fidelity. but i may have misunderstood the OP.
_______________________________________________
Linux-audio-dev mailing list
Linux-audio-dev(a)lists.linuxaudio.org
http://lists.linuxaudio.org/mailman/listinfo/linux-audio-dev
_______________________________________________
Linux-audio-dev mailing list
Linux-audio-dev(a)lists.linuxaudio.org
http://lists.linuxaudio.org/mailman/listinfo/linux-audio-dev
8:53 p.m.
and test .wav file is at
http://gamma.cs.unc.edu/SOUNDC/Assignments/clapping!_2.wav
On Mon, Oct 26, 2009 at 5:50 PM, Drip Stone
<dripstone.blendercn(a)gmail.com>wrote;wrote:
Hello again,
I obtain data for fbuffer in following way:
char *fbuffer = ( char * ) malloc ( 10240000 ); //10M bytes
long length = fread ( fbuffer, 1, 10240000, file );
I also set alsa in float mode:
snd_pcm_hw_params_set_format ( handle, params,
SND_PCM_FORMAT_FLOAT_LE );
I pasted all the code at http://pastebin.com/m7fdf0957.
Thank you,
Dripstone
On Mon, Oct 26, 2009 at 3:59 PM, Jens M Andreasen <
jens.andreasen(a)comhem.se> wrote:
Nothing in the posted code fragment suggests that
neither in nor out is
remotely related to floats. Where do you get yor in-data from? Why is
fbuffer[] a bytearray (if that is what it is?) How was ALSA opened?
/j
On Mon, 2009-10-26 at 15:05 -0700, Drip Stone wrote:
while ( 1 )
{
for ( int i = 0; i < frames; i++ )
{
float t = 0;
if ( left ) //for left channel
{
t = *((float*)&(fbuffer[pos+i*4])); //4 bytes each
frame
t*=2;
//printf("%f ",t);
for ( int j = 0; j < 4; j++ )
buffer[i*8+j] = ((char*)&t)[j]; //8 bytes for 2
channels
}
rc = snd_pcm_writei ( handle, buffer, frames );
pos += frames * 4;
}
I think the value is not out of [-1.0,1.0]. BTW, what does it mean
when a value is negative?
Best,
Dripstone
2009/10/26 Jörn Nettingsmeier <nettings(a)folkwang-hochschule.de>
On 10/26/2009 11:38 AM, Victor Lazzarini wrote:
Well if the audio was clipped, there would
probably be some
audible
distortion.
Perhaps if a code fragment was posted we could give a more
informed
opinion.
i was thinking of operating on files, where clamping can be
done with
perfect fidelity. but i may have misunderstood the OP.
_______________________________________________
Linux-audio-dev mailing list
Linux-audio-dev(a)lists.linuxaudio.org
http://lists.linuxaudio.org/mailman/listinfo/linux-audio-dev
_______________________________________________
Linux-audio-dev mailing list
Linux-audio-dev(a)lists.linuxaudio.org
http://lists.linuxaudio.org/mailman/listinfo/linux-audio-dev
27 Oct
27 Oct
4:47 a.m.
On Mon, 2009-10-26 at 17:53 -0700, Drip Stone wrote:
I pasted all the code at http://pastebin.com/m7fdf0957.
OK! Aside from the C++ style of declaring iterators, I just made this
change in (or around) line 126 to raise left channel by 6 dB:
-8<------------------------------------------
//left
int i;
for (i = 0; i < frames; i++ )
{
float t = 0;
int j;
if ( left )
{
t = *((float*)&(fbuffer[pos+i*4]));
t *= 2.0f; // <-- Works like a charm?
...
---------------------------8<---------------------
What was the problem now again?
7:49 a.m.
Thank you for the reply. Finally I figure out the problem. The code is OK,
but I thought it should be louder. So when I was multiplying t by 2, I could
not hear the difference from my laptop speaker. But after multiplying PCM
data by 20, I heard a significant difference.
Best,
Dripsotne
On Tue, Oct 27, 2009 at 1:44 AM, Jens M Andreasen
<jens.andreasen(a)comhem.se>wrote;wrote:
On Mon, 2009-10-26 at 17:53 -0700, Drip Stone wrote:
I pasted all the code at http://pastebin.com/m7fdf0957.
OK! Aside from the C++ style of declaring iterators, I just made this
change in (or around) line 126 to raise left channel by 6 dB:
-8<------------------------------------------
//left
int i;
for (i = 0; i < frames; i++ )
{
float t = 0;
int j;
if ( left )
{
t = *((float*)&(fbuffer[pos+i*4]));
t *= 2.0f; // <-- Works like a charm?
...
---------------------------8<---------------------
What was the problem now again?
5329
days inactive
5330
days old
linux-audio-dev@lists.linuxaudio.org
10 comments
5 participants
participants (5)
-
Drip Stone -
Jens M Andreasen -
Jörn Nettingsmeier -
Peter Nelson -
Victor Lazzarini