[plug] Audio Speed Control
Brad Campbell
brad at seme.com.au
Wed Oct 24 15:00:55 WST 2001
Simon Scott wrote:
>
> A couple of things
>
> a) at the samplerate you are talking about, I would triple check
> your code to ensure it does what you think its doing. I cant believe it
> would sound that bad. I have written similar code on the Commodore 64,
> pushing 8Khz, 4 bit samples through the volume register and it sounded
> great.... and I was shifting nearly an octave at times.
Ok.. I'm picky about my sound quality. It sounds decidedly sampled.
It has an edge about it that reminds me of 8 bit 8khz sampling.
Almost un-noticeable on some tracks, but do it over a nice clean picked
guitar and it jumps out a mile.
> b) Obviously you cant shift it too far, or it will sound crap no
> matter what you do.
+/- 12% Max
> c) If it is just noise, I have heard it said that if you introduce
> random noise into the sample, it can in fact sound better.... dunno if you
> wanna go this far though
Yeah.. I was looking at dithering the insert/remove points.
> Does it sound better/worse if you shift up/down?
Both..
> How are you calculating when to double/drop a sample?
At the moment, a linear counter.. I just count samples and then insert/drop at
the points required to stretch/compress the sound.
I'm thinking the noise I'm hearing is related to the constant intervals at which
I'm interfering with the waveform. Perhaps I should look at inserting/dropping
when the wave is closest to zero and using some sort of dithering algorithm to
make sure the intervals are not regular..
> At the end of the day, this sort of algorithm will only do so much
> before you have to look at resampling which, (to my limited knowledge) is
> akin to your linear interpolation but is 'non-linear', if you know what I
> mean. But even that might sound crap.
Yeah.. I'm looking at that also, but thats big cpu type calculation..
--
Brad....
/"\
Save the Forests \ / ASCII RIBBON CAMPAIGN
Burn a Greenie. X AGAINST HTML MAIL
/ \
More information about the plug
mailing list