Subject: Re: Re[2]: [linux-audio-dev] peakfiles and EDL's
From: Richard C. Burnett (burnett_AT_tality.com)
Date: Wed Feb 28 2001 - 18:34:45 EET
Well I look at the calculations this way, the only signals that are going
to cause a wave shape for the envelope are those that don't have at least
one peak represented in every peak bin, so if you are checking at 2048
samples for a peak, then using signed data, if you had a signal at 2049
samples a peak, the wavelength would be 2049 * 1/44.1k = 46.46 ms or 21
Hz (Which I don't think most audio goes to much :) ), so if you take the
suggestion of for simplicity sakes, 4000, this corresponds to 11.025 Hz as
the bottom end of where frequencies are going to cause waves in the
magnitude envelope. So in the first case each bin represents a max peak
per 47ms and in the second 90ms.
Now, as far as max/min versus absolute, If you take an absolute value you
can use a smaller bin, because the sample peak will show up twice in a
cycle, so 2048 would correspond to 11 Hz when absolutes are used.
I agree with robby in not understanding the benefit of computing min
values, The top bottom is probably going to be pretty symetric for all
purposes, so you could just mirror it.
The plan is I think to switch between the amplitude envelope and the real
wave data based on zoom level if I understand Paul properly.
Rick
On Wed, 28 Feb 2001, Robert Schrem wrote:
> Can't follow you: Why dows 2048 correspod to an audible frequency
> with 44.1 khz audio?
>
> Well I think you need only to compute one peak value for a
> sample data block: Max = abs(all samples for a block)
> I still don't understan what information we gan if we compute
> seperate min and max peak values.
>
> To display only the absolute value on screen on the y axis
> would double the resolution of the peak data display on screen
> (you would have the double number of pixels for each peak value).
>
> You don't need to display something that looks like a waveform
> if it is an evolope. Maybe the display should show a waveform
> display for close up's when you work with the real sample data
> and switch to envolope mode if you zoom out. I think this would be
> a 'honest' solution...
>
> robby
>
+------------------------+-----------------------+
| T a l i t y | +------+ |
+------------------------+ +----+-+ | |
| Richard Burnett | +-+ | |
| Senior Design Engineer +---+ +----+ |
| burnett_AT_tality.com | | |
| | | |
| Phone: 919.380.3014 | |
| Fax: 919.380.3903 | | |
+------------------------------------------------+
This archive was generated by hypermail 2b28 : Wed Feb 28 2001 - 19:04:47 EET