0

What is the low level actual format of sound data when read from a stream in Java? For example, use the following dataline with 44.1khz sample rate, 16 bit sample depth, 2 channels, signed data, bigEndian format.

TargetDataLine tdLine = new TargetDataLine(new AudioFormat(44100,16,2,true,true));

I understand that it is sampling 44100 times a second and each sample is 16bits. What I don't understand is what the 16 bits, or each of the 16 bits, represent. Also, does each channel have its own 16bit sample?

user1565671
  • 23
  • 3
  • 1
    The 16 bits make up a number, which is the amplitude of the sound wave at the time of that discrete sample, ultimately measured in a value between -1 and 1. But why does the underlying data format matter? – bzlm Jul 31 '12 at 12:46
  • I am wanting to be able to read thru the data and pick out instances of a certain frequency. – user1565671 Jul 31 '12 at 12:56
  • 1
    Then you need to learn alot [more than the digital audio basics](http://en.wikipedia.org/wiki/Fast_Fourier_transform). :) To expand on [@IronMensan's answer](http://stackoverflow.com/a/11740529/7724), the frequencies contained in a portion of the sampled sound are the wave lengths of all the sound waves the diaphragm makes when moving. [This is computable](http://stackoverflow.com/questions/3287518/reliable-and-fast-fft-in-java), but complex, and not an exact science. – bzlm Jul 31 '12 at 13:00

2 Answers2

0

You should learn about the Digital Audio Basics (Wiki gives you a start and lots of links with further reads). After that 44.1khz sample rate, 16 bit sample depth, 2 channels, signed data, bigEndian format should immediately tell you the low level format.

In this case it means 44100 samples/sec, 16 bit signed integers representing each sample and finally endianess determines in which order the bytes of a 16bit int are put into the stream (big endian = most significant byte first).

Durandal
  • 19,919
  • 4
  • 36
  • 70
0

I'll start with your last question first, yes, each channel has its own 16-bit sample for each of the 44100 samples each second.

As for your first question, you have to know about the hardware inside of a speaker. There is a diaphragm and an electormagnet. The diaphragm is the big round part you can see if you take the cover off. When the electromagnet is charged it pulls or pushes a ferrous plate that is attached to the diaphragm, causing it to move. That movement becomes a sound.

The value of each sample is how much electricity is sent to the speaker. So when a sample is zero, the diaphragm is at rest. When it is positive it is pushed one way and when it is negative, the other way. The larger the sample, the more the diaphragm is moved.

If you graph all of the samples in your data, you would have a graph of the movement of the speaker over time.

IronMensan
  • 6,761
  • 1
  • 26
  • 35