I guess full fast convolution (as opposed to partitioned) is still
best when processing offline (it will be faster), as the delay
can be compensated easily, so single-frame PVX files should
still be useful.

Victor

----- Original Message ----- 
From: "Richard Dobson" 
To: 
Sent: Thursday, January 24, 2008 10:43 AM
Subject: [Csnd] Re: Re: pvs oscillator collection? (potentially "hello 
Victor" again..)


> Tim Mortimer wrote:
>> Sorry, try again:
>>
>> My question is:
>>
>> if im simply wishing to fft analyse a single "oscillation" of a wave 
>> (that
>> is for convenience say 2048 samples long...) can i perform a valid &
>> accurate fft analysis using 2048 bins that is 1 frame in length, & save 
>> this
>> single frame as a .pvx file to use?
>>
>
> Sorry, mostly lurking as head full of other stuff at the moment. I would 
> tend to regard that as a plain FFT - somehwat meaningless to convert it to 
> full pvoc amp/freq as freq is the phase difference from one frame to the 
> next, and by definition you only have the one frame.
>
>
> The PVOCEX (.pvx) file format does not support a literal single frame - it 
> is based on CARL pvoc where there is always a zero (empty) first frame 
> (which I guess kind of gives you a starting amp/freq basis). There is a 
> potential advantage to this, though I haven't published it (yet) - to use 
> a .pvx file with just one frame to represent an single-frame impulse 
> respsonse (in raw COMPLEX format) for convolution purposes. I have a 
> private build of my "pvocex2" program that does just this. Now that we 
> have partitioned convolution this single-frame method is obsolete (the 
> latency is of the duration of the frame, which will typically be a very 
> large one), unless some other virtue can be identified for it.  In any 
> case, were single-frame pvx files to become relevant to users, we should 
> really extend PVOCEX to define it formally (as a new format type).
>
> So you would need to write 2 frames to the file, the empty first one, and 
> then your second one. But I would question whether this was the best use 
> of a pvx file per se.
>
> Note that you can create aliasing in a pvoc frame simply by writing an 
> excessive frequency value in the topmost bins. In the general case, of 
> course, if the aliasing is already present in  the time-domain signal 
> (such as a raw square wave) it will show up very vividly in the FFT.
>
>
>
>
> Richard Dobson
>
>
>
> Send bugs reports to this list.
> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe 
> csound"