[opus] CELT CFFT size configuration

[Jean-Marc Valin]

On 13/04/17 05:04 AM, Diego Alejandro Parra Guzman wrote:
> I'm not completely sure about how the kiss_fft works for sizes which are
> not power of two, as occur in the case of celt which calculate the mdct
> using a fft sizes of 480, 240, 120, and 60, I guess that internally the
> input data is padded with zeros in order to compute efficiently the mdct
> using the Cooley and Tukey algorithm.

No, the FFT is *not* padded. The Cooley-Tukey algorithm can be
implemented for radices other than 2. The kiss-fft-derived code in Opus
implements radices 2, 3, and 5. You *cannot* use padding here.

	Jean-Marc

> If I'm correct. Means that technically,  I could compute the fft using
> the padding strategy and adjust the output such as kissfft done, someone
> could confirm if that is possible.
> 
> thanks
> cheers
> 
> On Wed, Apr 12, 2017 at 2:27 PM, Diego Alejandro Parra Guzman
> <daparrag at correo.udistrital.edu.co
> <mailto:daparrag at correo.udistrital.edu.co>> wrote:
> 
>     Dear all,
> 
>     Sorry for this simple and maybe stupid question, I'm working in the
>     implementation of opus for ARMv7e microcontroller using a library
>     CMSIS/DSP used to calculate the CFFT and MDCT based on the DCT-IV.
>     my question is:
> 
> 
>     In the implementation of Celt  you have used a fixed length CFFT
>     equal to 1920, I want to know if there is some issues which can
>     appear if a change that configuration for a cfft of size 2048, some
>     of you could give me some support about that?, I ask because the
>     library which I'm using doesn't support 1920 size.
> 
>     please let me know, I you need I could be more specific.
> 
>     thanks you so much.
> 
>     -- 
>     Diego Alejandro Parra Guzmán
>     Estudiante de ingeniería electrónica
>     Universidad distrital FJC
> 
> 
> 
> 
> -- 
> Diego Alejandro Parra Guzmán
> Estudiante de ingeniería electrónica
> Universidad distrital FJC
> 
> 
> _______________________________________________
> opus mailing list
> opus at xiph.org
> http://lists.xiph.org/mailman/listinfo/opus
>