ASCII Text Export Data

Hi there,

I've recently discovered the ability to export music as ASCII text data. It's a neat feature and I'd like to play with the values somehow. The problem is that I don't understand how the data is formatted.

If I were to guess, I'd guess that I'm seeing left and right channel information and that the values are intensity. I can see that 0,0 is silence, but aside from that I'm completely lost.

Does anyone know how this information is actually outputted? Is there a set value for notes, pitch, tempo, or any other specific aural data? What determines when a value is recorded? Is each line a millisecond?

I'm hoping to use the information to do some mathematical equations, so I'd like to know exactly what each entry means.


Thanks!

The ASCII text file format represents PCM audio data, just like wave files, but the sample values are in text form.
Each line of values represents digital sample(s) in time--one value on a line for mono, two for stereo.
The options for this filetype include whether the data should be normalized or not, and whether to include
a format header.  When normalized, the sample values range from -1 to 1.  When unnormalized, they
range from -32768 to 32767.  The format header, when present, includes lines at the beginning of the
file indicating the number of samples in the file, the bit depth, the number of channels, the sample rate,
and whether the data is normalized or not.  The values are strictly sample values; no values for pitch, tempo, notes, etc.

The resolution of the data is very good:  approximately 22-23 bits of floating point precision.

This filetype is useful for having a computer program read audio data as ASCII text when you don't have a
wave file parser.  Files of this filetype are significantly larger than corresponding wave files.

Hello younglove,

Thanks for your reply. That explains quite a bit about what I'm looking at.

Since the values are strictly numeric, what is telling audition to make various noises? If I have a trumpet, piano, and guitar all playing the same song, what would change in the data to illustrate the differences between the instruments? I'm afraid I don't know anything about PCM audio data or wave parsers. I am most certainly *not* an audio guy.

Try these links:

http://en.wikipedia.org/wiki/Pulse-code_modulation

http://en.wikipedia.org/wiki/Digital_audio

http://en.wikipedia.org/wiki/Midi

Nice to see Mr. Younglove lurking about here -- and thanks for an interesting read.

Thanks again, Younglove. I'll give those a read and see if I can get my equations to work.

...When unnormalized, they range from -32768 to 32767.

Can there be decimals, or is the bit depth 16 only?

Bert

[B]

Can there be decimals, or is the bit depth 16 only?

Binary Digit

Rather says it all... digits are whole numbers!

It's only the normalised values that would have the equivalent of decimal places. Normalised values for a 24-bit sample would also extend from -1 to 1, but the amount of decimal places needed would be greater, because in a 24-bit sample the integer values have to go from -8388608 to 8388607.

The numbers are relatively easy to arrive at - 2^16 = 65536, so dividing that by 2 (+ve and -ve) gives the 32767/8 value, and 2^24 etc gives the 24-bit result similarly.

The values are not integers (example -4104.632 (unnormalized) or -0.1252634 (normalized)).  Subtracting the text values from an original wave float gives a difference of about -166 dB, i.e., very little loss.

The values are not integers (example -4104.632 (unnormalized) or -0.1252634 (normalized)).  Subtracting the text values from an original wave float gives a difference of about -166 dB, i.e., very little loss.

I'm sure that this is true for the text values, but not for Bits. Bits are, by definition, what the acronym says they are.

The values are not integers (example -4104.632 (unnormalized)...

Thanks,
That was just what I wanted to know, and the obvious reason for my question.
Bert