Re: Combining WAV files
Kat,
1. Could you ramble on a little more, relative to how to take a sample of
speech in a wav file and *reduce* the pitch some, while still keeping
approximately the same overall duration of the sample?
2. Since wav file wouldn't have negative numbers, would zero-crossing be
half-way between biggest & smallest number in wav file, or something else
more complex?
3. Just curious: when you say "a 75 watt PWM 1Mhz signal", since I think
people can't hear 1Mhz, I'm guessing "PWM" refers to *modulating* a 1Mhz
signal with an audio signal?
Dan
----- Original Message -----
From: "Kat" <gertie at PELL.NET>
To: "EUforum" <EUforum at topica.com>
Sent: Monday, September 17, 2001 12:02 PM
Subject: RE: Combining WAV files
>
> On 17 Sep 2001, at 14:08, Bernie Ryan wrote:
>
> >
> > Chris Bensler wrote:
> > > Hi all,
> > >
> > > I already know how to read and write wav files. But I would like
to be
> > > able to convert the frequency of the file.
> > > Any ideas?
> > >
> > > For example. Convert from 1300 to 44100
> > >
> > > Also, I am using Exotica, and it allows frequencies in the negative
> > > range.
> > > How would I calculate that?
> > >
> >
> > Chris:
> > You are asking to increase the number of samples 1300;
> > to 44100 samples. There is no way to increase the number
> > of samples that you already have. Maybe you are not explaining
> > exactly what you are trying to accomplish.
>
> Besides, 44100hz sampling is too close to Nyquist to be comfy. You'd still
> be using a brickwall filter at 20K to avoid aliases, with will make for
phase
> funnies in waveforms that you can't accurately reproduce anyhow. My
> suggestion, if you are striving for perfection, is to sample the original
analog
> at 100Khz *or more*, this will at least give you 5 samples at 20khz, you
> *still* won't be able to reproduce it properly, but it will sound better
on hi-end
> equipment. The phase accuracy will be significantly improved, but your
> mileage may vary per the performance of your person ears.
>
> I once drove tweeters with a 75 watt PWM 1Mhz signal, 50 times 20khz, or
> 25 times Nyquist at audio. Other than the sustained plasma arc on the
voice
> coil at that power level, the sound was incredible. I was swinging 100v in
2ns,
> radiated harmonics disrupted every radio in the building. With reduced
> voltage to the drivers, the arc stopped. Below 50khz, the sound quality
> suffered. Some people could not hear the sound difference until the PWM
> signal freq was lowered to 19khz,, proving that sound is subjective, and
> killing my funding into higher quality sounds. But anyhow, if you are
looking
> for higher accuracy, you must go back to the original analog. Your 2nd
best
> bet is to interpolate between samples you have now, and look for vlf IM
> products between the sampling frequency and the resulting audio, and
> modify the interpolation results. This will only work on sustained notes,
and it
> won't be original quality, but closer. For instance, you could get better
upper
> harmonics on guitar solos, but not cymbal crashes. Good luck.
>
> If you are looking to change the playback pitch of the wav file, disregard
all
> the above, locate zero crossings of the waveform, and "packetize" all the
> zero-to-zero sections of the waveform. Then double the packets to get
double
> the pitch,
>
> Original:
> 1,2,3,4,5,
> new:
> 1,1,2,2,3,3,4,4,5,5
>
> and in the new wav, drop every other sample to play back at double pitch
at
> the same sample rate. Very lossy, isn't it? But at least one commercial
> product used this technique. They used CCD "bucket brigade" devices,
> sampling and storing the analog directly, it was very noisey for audio
(but
> worked fine for video time shifting, which is only 8bits of resolution, i
was
> involved in reverse engineering the TI CCD drivers for better reliablity
in video
> work in 1980 or so). It works in reverse too, you can play back at 2x
speed,
> and drop every other packet, and stretch the remaining packets to fill the
> timespace of the two orig packets, and the resulting audio will be the
correct
> speed, and roughly the correct pitch, but tends to be "choppy" sounding.
> Every conversion throws away information. Fine tune your process to sync
> the output pitch to whatever multiple of the input pitch you want.
>
> Kat,
> rambling again.
>
|
Not Categorized, Please Help
|
|
