[asterisk-dev] downsampling slinear16 to ulaw (or alaw or g729)

[Paul Albrecht]

On Mon, 2010-08-30 at 11:04 -0500, Kevin P. Fleming wrote:
> On 08/30/2010 10:53 AM, Paul Albrecht wrote:
> > On Mon, 2010-08-30 at 10:30 -0500, Kevin P. Fleming wrote:
> >> On 08/30/2010 10:16 AM, Paul Albrecht wrote:
> >>> On Mon, 2010-08-30 at 10:02 -0500, Kevin P. Fleming wrote:
> >>>> On 08/30/2010 09:39 AM, Paul Albrecht wrote:
> >>>>
> >>>>> I have a question about asterisk transcoding from wide slinear to ulaw
> >>>>> (or alaw or g729). Specifically, the result I get when I translate from
> >>>>> AST_FORMAT_SLINEAR16 to AST_FORMAT_ULAW is truncated, that is, I don't
> >>>>> get a 160 samples in the output frame. Is this a bug or should I have
> >>>>> expected the translator to truncate the result?
> >>>>
> >>>> How many samples were in the input frame? Is there a smoother involved?
> >>>>
> >>>
> >>> The slinear16 frame contains 320 samples and I'm getting 137 samples of
> >>> ulaw out which is not what I expected. I was looking for a full 160
> >>> samples of ulaw.
> >>>
> >>> I don't know if a smoother was involved. Here's how I do the
> >>> translation:
> >>>
> >>> trans = ast_translate_build_path(AST_FORMAT_ULAW,AST_FORMAT_SLINEAR16);
> >>> out = ast_translate(trans,in,0)
> >>
> >> That certainly sounds like a bug then; there's no smoother involved with
> >> that type of construction. Can you post the contents of the ast_frame
> >> structure called 'in' (but not the data)?
> >>
> > 
> > The translated frame is silent, that is, all the samples are zero so
> > that's easy. Here's the code for how I set the input frame:
> > 
> > static struct ast_frame f;
> > static unsigned char data[AST_FRIENDLY_OFFSET + 640];
> > 
> > f.data.ptr = (void *)(data + AST_FRIENDLY_OFFSET);
> > 
> > f.frametype = AST_FRAME_VOICE;
> > f.subclass = AST_FORMAT_SLINEAR16;
> > f.samples = 320;
> > f.offset = AST_FRIENDLY_OFFSET;
> > f.datalen = 640;
> > 
> > Is this enough or do you want me to write particular fields to the log
> > file or is there another way to dump the frame?
> 
> No, that's enough, and it appears to be correct. I don't see any obvious
> reason why that would be failing. To debug it further, I'd change it to
> output AST_FORMAT_SLINEAR, and see if you get 160 samples; if not,
> you've found the step that is causing the problem, and if so, you've
> still found it :-)
> 

I have already done that, but since I received that expected result, 160
samples, I figured it was likely a transcoding bug, so I posted to the
list. What else do you need? Is there a frame dump routine I can call
after I get the frame translated?

> -- 
> Kevin P. Fleming
> Digium, Inc. | Director of Software Technologies
> 445 Jan Davis Drive NW - Huntsville, AL 35806 - USA
> skype: kpfleming | jabber: kfleming at digium.com
> Check us out at www.digium.com & www.asterisk.org
> 
-- 
Paul Albrecht