Re: CVS commit - 0.95.07

On 5 May 2007, at 12:28, Jan Depner wrote:

> On Sat, 2007-05-05 at 12:12 +0100, Steve Harris wrote:
>> On 4 May 2007, at 18:47, Jan Depner wrote:
>>
>>> On Fri, 2007-05-04 at 12:24 -0500, Jan Depner wrote:
>>>> OK, I've got the meters set up for RMS on output and I've got the
>>>> function calls set up in intrim.c to handle it.  If someone can
>>>> figure
>>>> out how to compute RMS using Steve's routines -
>>>> http://www.ecs.soton.ac.uk/~swh/rms/ we can get this thing
>>>> working.  I
>>>> am assuming that the calls would need to go in the following  
>>>> loop in
>>>> process.c :
>>>>
>>>>     for (pos = 0; pos < nframes; pos++) {
>>>> 	for (port = 0; port < nchannels; port++) {
>>>> 	    const float oa = fabs(out[port][pos]);
>>>>
>>>> 	    if (oa > lim_peak[LIM_PEAK_OUT]) {
>>>> 		lim_peak[LIM_PEAK_OUT] = oa;
>>>> 	    }
>>>> 	    if (oa > out_peak[port]) {
>>>> 		out_peak[port] = oa;
>>>> 	    }
>>>> 	}
>>>>     }
>>>>
>>>> but I'm probably wrong ;-)
>>>>
>>>
>>>     Just for fun I implemented the "naive" RMS in the above loop
>>> and it
>>> looks purty (as we say down here in the South).  It's probably not
>>> what
>>> we're looking for but it does prove that I've got the meters and
>>> callbacks set up properly ;-)
>>
>> Great!
>>
>> To hook up the code, you just need to initialise the RMS object with
>> a call to rms_new(), then call rms_run_buffer() on each audio buffer,
>> and the float that you get back is the RMS signal level.
>>
>
>     I just put this into process_signal :
>
>     for (port = 0 ; port < nchannels ; port++)
>       {
>         rms *r;
>         r = rms_new (sample_rate, 0.1);
>
>         rms_peak[port] = rms_run_buffer (r, out[port], nframes);
>
>         rms_free (r);
>       }
>
> rms_peak is a misnomer but I was just duping the out_peak code and it
> was easier ;-)  The RMS meters come up with a value of about -20.0 and
> they don't look quite right to me so I probably did something  
> wrong.  I
> can commit the changes and then you can take a look to see what's  
> wrong
> if you have time.

Unfortunately I'm off to Canada tomorrow, and I'm a bit busy before  
then. One immediate problem though is the rms_new(): it needs to be  
called once per channel, and the struct needs to be held onto between  
calls. You could make them static in the processing function and call  
_new if its NULL. eg.

static rms **r = NULL;

if (!r) {
	r = malloc(nchannels * sizeof(rms *));
	for (port = 0 ; port < nchannels ; port++) {
		r[port] = rms_new (sample_rate, 0.1);
	}
}

for (port = 0 ; port < nchannels ; port++) {
...
}

the rms_free() can be skipped.

Could also be done in the process initialisation function, if there  
is one.

- Steve

-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/