[sc-users] PlayBuf Interpolation

[Jesper Elén]

I am very interested in these interpolation algorithms.
Any chance they will end up in the  distro?

Jesper

6 feb 2008 kl. 13.16 skrev kernel:

> On 6 Feb 2008, at 08:30, Wouter Snoei wrote:
>
>> Could you elaborate a bit more on the b-spline algorithm? In what way
>> does it sound dull? How cpu intensive is it compared to the current
>> hermite interpolation?
>
> the b-spline uses 6 points.  to my ears at least it isn't just that it
> lessens a lot of the aliasing noise that makes it sound dull - it just
> seems to lose something.  i have heard of people using a pre-emphasis
> filter for such things though.  CPU-wise the increase is minimal (on
> my intel iMac 2.33).  another option is to oversample when using such
> interpolators but that could get complex.
>
>> Sinc based interpolation would be very useful to have for audio
>> anyway I think (indeed for offline, but also for realtime on a fast
>> machine). I would like it very much if there would be an
>> interpolation arg in PlayBuf, just like in BufRd. If the options you
>> name here would become available through that argument as well that
>> would be great!
>
> the sinc interpolator sounds really nice and you can easily specify
> the number of taps it uses.  but the CPU usage is high.  it uses 1
> sin() and 1 cos() operation per tap per sample (8 taps minimum for
> good quality).
>
> anyway, i think it would be relatively easy to bundle these up into
> PlayBuf and provide an argument for the interpolation method (and
> number of taps for sync).  although, running the server at say 96KHz
> does a pretty good job of oversampling PlayBuf anyway.
>
> kernel
>
>
>>
>>
>> cheers,
>> Wouter
>>
>> Op 6-feb-2008, om 1:58 heeft kernel het volgende geschreven:
>>
>>> hi all,
>>>
>>> I've been experimenting with different interpolation algorithms with
>>> the aim of improving the audio quality of PlayBuf (aliasing mostly).
>>> so far I have two algorithms that improve the sound:
>>>
>>> 1. a b-spline algorithm that reduces aliasing but sounds a little
>>> dull
>>> but uses very little CPU and 2. a sinc based algorithm that uses a
>>> LOT
>>> of CPU but sounds really nice and can provide different quality
>>> settings (useful for offline rendering perhaps?)
>>>
>>> just wondering if anyone else has any ideas here and how useful
>>> people
>>> may find thiese.  i'm very un-tempted to release any code yet  
>>> though.
>>>
>>> thanks,
>>> kernel
>>> _______________________________________________
>>> sc-users mailing list
>>> sc-users at create.ucsb.edu
>>> http://lists.create.ucsb.edu/mailman/listinfo/sc-users
>>
>> Wouter Snoei
>>
>> info at woutersnoei.nl
>> http://www.woutersnoei.nl
>>
>> _______________________________________________
>> sc-users mailing list
>> sc-users at create.ucsb.edu
>> http://lists.create.ucsb.edu/mailman/listinfo/sc-users
>
> _______________________________________________
> sc-users mailing list
> sc-users at create.ucsb.edu
> http://lists.create.ucsb.edu/mailman/listinfo/sc-users