posted 05 October 2004 03:44            

1. Can I always assume that the actual size of the frame is the number of partials recorded in the SPC file plus four?
2. How to calculate the standard frame length? I tried:
length (nsec) = NumberOfSamplesInAFrame / SampleRate * 1000000000
length * NumberOfFrames seems different from the time shown in the Sound Browser.

I have some more undocumented kind of questions. Please answer them if they are not secrets

posted 05 October 2004 11:22            

2. The number of sample frames divided by the frame size (power of two from answer 1) is the number of spectral frames. That number times the frame duration should be the duration of the resynthesis. The frame duration should be set to indicate how much time a sprectral frame represents.

For your additional questions:

1. The MARKER chunk is used to indicate loop points for the spectrum editor. The format of the chunk is identical to an AIFF MARKER chunk. (For correctness, the marker should point to a multiple of the frame size and not to an arbitrary sample index.)

2. The INST chunk is the same as the AIFF INST chunk and provides information about the base pitch, pitch range and velocity range that the file should be used on.

posted 06 October 2004 04:06            

What I wanted to know in question 1 was why the number of partials in a frame recorded in a SPC file seemed four less than the number chosen on creation. For example, in 128 partial SPC files, 124 is recorded as the number of partials. Does that mean that in such a frame, only 124 partials are valid? On the other hand, when I write an SPC file, the data I can pack in a 128 partial frame is only 124 partials?

In question 2, suppose I have an original sample file (AIFF) and its frequency-domain analysis, how can I set the frame time so that their durations be the same? If I calculate the duration of the original file and divide it by the number of frames, then it might be possble, but the dependency is awkward.

Well, I'm not just hacking SPC files. I was trying to create a Sound in which !KeyDown triggers an SPC file fed to the Filter Bank Element, which carves the input sound. So far, SPC files can be created only by analizing and modifying the existing sound. But in this case I need the way to make them from the scratch....

posted 06 October 2004 09:41            

Yes, for harmonic spectra (the spectra with 'h' after the name), partials 1 through (nbrPartials - 4) are valid (for 's' spectral, ALL partials are valid). However, you can create your own files with the full 128 and Kyma will read them correctly (for both 'h' and 's').

>In question 2, suppose I have an original sample file (AIFF) and its frequency-domain >analysis, how can I set the frame time so that their durations be the same? If I calculate the >duration of the original file and divide it by the number of frames

That sounds like it should work.

>Well, I'm not just hacking SPC files. I was trying to create a Sound in which !KeyDown >triggers an SPC file fed to the Filter Bank Element, which carves the input sound. So far, SPC >files can be created only by analizing and modifying the existing sound. But in this case I >need the way to make them from the scratch....

Do you want to create a spectrum without analyzing an existing Sound? You could try using SyntheticSpectrumFromArray as your spectral source.