🚀 go-pugleaf

In article <1993Jan23.151043.23034@impmh.uucp>, cm@impmh.uucp (Colin Manning) 
writes:

>According the the ms documentation, they accept data in 44.1Khz samples,
>20.05Kh and 10.025Khz. ADPCM is sampled at different rate.
>
>Anyone any idea how easy/difficult it is to convert audio data sampled at
>one rate to another ?

Now that you ask: Yes, it is difficult... :-)

If the only difference was the sampling rate, plain linear interpolation
would probably be satisfactory - and anyone who can do simple multiplication
and division in c could do the job.

But while Windows uses a simple, linear PCM format, ADPCM is quite a 
different thing. The "A", Adaptive, means that the step size is not
constant, but depends on the slope of the curve. The "D", Differential,
means that the encoded sample value is not a direct value of the signal 
level but the difference from the previous value. These two factors
make it quite a job for one inexperienced with digital signal processing.

Sure, to experienced DSP people, the job is "fairly simple" (at least
in theory), but I suspect that you ask because you do not know a
sufficient amount of the theory.

A while ago, I picked up the CCITT recommendation G.722, hoping to write
a G.722-to-Windows-sound. Sorry, my math background was much to weak.
Yours may be better than mine, though! (G.722 is a 7 kHz/48-56-64 kbit
encoding. G.721 is a 3.4 kHz/32 kbit encoding; they are significantly
different.)

So, rather than trying to code it yourself, try to find a library to
buy. If you want anything close to real time conversion, look for
hardware. After all, the chips to do it will be going into millions of
telephones as G.722/G.721 becomes commonplace, so they ought not be
too expensive. Maybe they still are, today.