Question

I’m looking into writing a audio syntesizer in Java, and was wondering if anybody has any advice or good resources for writing such a program. I’m looking for info on generating raw sound waves, how to output them into a usable form (playing over speakers), as well as general theory on the topic. Thanks guys.

Answer 1

1. This problem is basically about mapping functions to arrays of numbers. A language that supports first-class functions would come in really handy here.

2. Check out http://www.harmony-central.com/Computer/Programming and http://www.developer.com/java/other/article.php/3071021 for some Java-related info.

3. If you don’t know the basic concepts of encoding sound data, then read http://en.wikipedia.org/wiki/Sampling_rate

4. The canonical WAVE format is very simple, see http://www.lightlink.com/tjweber/StripWav/Canon.html. A header (first 44 bytes) + the wave-data. You don’t need any library to implement that.

In C/C++, the corresponding data structure would look something like this:

typedef struct _WAVstruct {     char headertag[4];     unsigned int remnantlength;     char fileid[4];      char fmtchunktag[4];     unsigned int fmtlength;     unsigned short fmttag;     unsigned short channels;     unsigned int samplerate;     unsigned int bypse;     unsigned short ba;     unsigned short bipsa;      char datatag[4];     unsigned int datalength;      void* data; //<--- that's where the raw sound-data goes }* WAVstruct; 

I’m not sure about Java. I guess you’ll have to substitute ‘struct’ with ‘class’ and ‘void* data’ with ‘char[] data’ or ‘short[] data’ or ‘int[] data’, corresponding to the number of bits per sample, as defined in the field bipsa.

To fill it with data, you would use something like that in C/C++:

int data2WAVstruct(unsigned short channels, unsigned short bipsa, unsigned int samplerate, unsigned int datalength, void* data, WAVstruct result) {     result->headertag[0] = 'R';     result->headertag[1] = 'I';     result->headertag[2] = 'F';     result->headertag[3] = 'F';     result->remnantlength = 44 + datalength - 8;     result->fileid[0] = 'W';     result->fileid[1] = 'A';     result->fileid[2] = 'V';     result->fileid[3] = 'E';      result->fmtchunktag[0] = 'f';     result->fmtchunktag[1] = 'm';      result->fmtchunktag[2] = 't';     result->fmtchunktag[3] = ' ';     result->fmtlength = 0x00000010;     result->fmttag = 1;     result->channels = channels;     result->samplerate = samplerate;     result->bipsa = bipsa;     result->ba = channels*bipsa / 8;     result->bypse = samplerate*result->ba;      result->datatag[0] = 'd';     result->datatag[1] = 'a';     result->datatag[2] = 't';     result->datatag[3] = 'a';     result->datalength = datalength;      result->data = data; // <--- that's were the data comes in      return 0; // an error code, not implemented, yet ...; in Java: return result } 

Again, I’m not sure about Java but the conversion should be straightforward if you convert the void-pointer to an array corresponding to the bitrate.

Then simply write the entire structure to a file to get a playable wave file.