On Fri, 2008-06-13 at 10:56 +0200, Fons Adriaensen wrote: I think, while absolutely true from an engineering aspect, there is something to be said for /dev/dsp for lowering barriers of entry into audio programming. Doing experiments like this was how I started doing it, and it would be a shame to ever remove this option completely. cheers, dave

http://www.granularsynthesis.com/index.php On Tue, 2008-06-17 at 05:34 +0530, AlgoMantra wrote:

AlgoMantra wrote: Well, it can be fun to poke around with a little bit of C code, but in the case of GS this probably won't get you very far unless you really get serious and write your own realtime synthesis engine. Which isn't easy. So why not use SuperCollider? SC is well-suited for this kind of application because of the ease with which you can allocate massive amounts of voices dynamically. All the realtime audio and control stuff that you need is already there, and you can still program everything the way you want it, rather than using some ready-made GS software. Albert -- Dr. Albert Gr"af Dept. of Music-Informatics, University of Mainz, Germany Email: Dr.Graef(a)t-online.de, ag(a)muwiinfa.geschichte.uni-mainz.de WWW: http://www.musikinformatik.uni-mainz.de/ag

AlgoMantra wrote: Well, then go for it. :) Sorry, I didn't understand that C was an absolute requirement. Yes sure, exactly my point. You can use SC from any language that has an OSC interface (like C, Python, Haskell, Q, ...), no need to program in SC language if you don't want to. -- Dr. Albert Gr"af Dept. of Music-Informatics, University of Mainz, Germany Email: Dr.Graef(a)t-online.de, ag(a)muwiinfa.geschichte.uni-mainz.de WWW: http://www.musikinformatik.uni-mainz.de/ag

On Tue, 2008-06-17 at 11:55 +0530, AlgoMantra wrote: Alrighty then. You previously had sound set up with OSS as well as the sum of four sinewaves playing 5 seconds thru your speakers. Next step is low latency and realtime control. The following program will do the same as sound.c except that this time around you will have to bang the keyboard to actually hear anything. I also added a rudimentary envelope (as well as removed an embarrissing typo in setting up frequencies.) The basic idea here is that the play() routine blocks for a short while - here 64 samples - after which you do a non-blocking call to read_input() to see if anything has happened and if so, act accordingly and then play() the next 64 samples (and so ad infinitum ...) --8<-- rt_sound.c ------------------------------------- // Compile with for example // gcc4.3 -O3 -msse -ftree-vectorize -o rt_sound rt_sound.c #include <stdio.h> #include <string.h> #include <fcntl.h> #include <sched.h> #include <unistd.h> #include <sys/ioctl.h> #include <sys/select.h> #include <linux/soundcard.h> /* We will use the keyboard for musical input */ int kbhit() { static const int STDIN = 0; struct timeval timeout; fd_set rdset; FD_ZERO(&rdset); FD_SET(STDIN, &rdset); timeout.tv_sec = 0; timeout.tv_usec = 0; return select(STDIN + 1, &rdset, NULL, NULL, &timeout); } /* Inverse of the odd factorials (n!) 1 - 9 */ #define IF_1 (1.f) #define IF_3 (1.f/6.f) #define IF_5 (1.f/120.f) #define IF_7 (1.f/5040.f) #define IF_9 (1.f/362880.f) /* ------------------------------------- * 16 bit Taylor approximation of sin() * ------------------------------------ */ inline float tsin(float x) { float x2 = x*x; float r = IF_9; r *= x2; r -= IF_7; r *= x2; r += IF_5; r *= x2; r -= IF_3; r *= x2; r += IF_1; return r * x; } #define VEC 4 #define PI 3.141592f #define RATE 44100.f typedef struct { float x[VEC]; float y[VEC]; float w[VEC]; } sound_t; void read_input(sound_t *s) { static sound_t s1 = // A preset { {0,0,0,0}, {8000,7000,6000,4000}, {220, 277,330,440} }; if(!kbhit()) return; int ch = getchar(); // unused for now ... int i; for(i = 0; i < VEC; ++i) { s->x[i] = s1.x[i]; s->y[i] = s1.y[i]; s->w[i] = 2 * tsin(s1.w[i] * PI/RATE); } } void play(short *buf,sound_t *s,int size) { int i,j; for(i = 0;i < size; ++i) { for(j=0;j < VEC; ++j) { // Waveform s->x[j] -= s->w[j] * s->y[j]; s->y[j] += s->w[j] * s->x[j]; // Envelope s->x[j] *= .99999f; } buf[i] = s->x[0] + s->x[1] + s->x[2]+ s->x[3]; } } int main(void) { int i; int out = open("/dev/dsp",O_WRONLY); // Sound i=16; // 16 bit sound ioctl(out,SOUND_PCM_WRITE_BITS,&i); i=1; // 1 channel ioctl(out,SOUND_PCM_WRITE_CHANNELS,&i); i=44100; // 44.1 KHz ioctl(out,SOUND_PCM_WRITE_RATE,&i); // Latency / jitter i = 0x00020007; ioctl(out, SNDCTL_DSP_SETFRAGMENT, &i); // Realtime struct sched_param schp; memset(&schp, 0, sizeof(schp)); schp.sched_priority = sched_get_priority_max(SCHED_FIFO)* 0.8; if (sched_setscheduler(0, SCHED_FIFO, &schp) != 0) fprintf(stderr, "\n Can't get permission to run in realtime!" "\n Try logging in as root instead?\n"); sound_t sound; memset(&sound, 0, sizeof(schp)); int running = 1; puts("\n Hit the any key for sound!"); while (running) { short int wave[64]; read_input(&sound); play(wave,&sound,64); write(out, wave, sizeof(wave)); } close(out); return 0; }

Hey Guys, thanks so much for all the input. Now I need to process/experiment with it all and get back to you. Jens!!! The code you're sending is most helpful. We were able to play a few nice loops today in class using pure sine synthesis in C, and I of course suggested they start working on a keyboard-input thingie like a toy Casio. Of course, they are 18-20 years old, just out of high school. So you cannot imagine how timely this is... Frank: I don't like the PD patchwork interface much, but while studying PD and Puckette's bible I found your footils most helpful. Why I don't like the PD interface, um.....have you played with VVVV? If you're into visual synthesis, try it, and you'll see why its much more intuitive. http://vvvv.org/ -- for windows only :( My own plans are far more nefarious....um, I dare not spell them out so soon. warm regards, ------- -.- 1/f ))) --. ------- ... http://www.algomantra.com

Jens M Andreasen &lt;jens.andreasen(a)comhem.se&gt; writes: There was a very nice curses based midi keyboard for OSS, called midikb or something like that. It solved this problem very neatly by allowing you to switch between a few different modes. IIRC one was to send note off when a new key was pressed (ie like a monophonic synth), one was to always play each note for a set time, and last I remember was to press key a second time for note off. I wonder what happened to that application. It used to be in Debian. I used it all the time. Often these modes of playback made it easier to play some parts than a real midi keyboard (or vkeybd). On the other hand of course at other times it was annoying that it could not function more like a real midi keyboard. I think/hope I have the source somewhere. Even did some modifications to it to send specific control change event easily for some hardware synth I used at the time. -- /Pelle

Hallo, AlgoMantra hat gesagt: // AlgoMantra wrote: I didn't mean to recommend Pd as a general language (this time), but as a kind of library that takes away many tedious tasks like GUI writing or audio interfacing and lets you concentrate on the dsp side in your C code, while still being able to quickly run it. Of course you could also use LADSPA for this. Yep, and not open source. As I don't run Windows anymore, I cannot try it. Personally I like the Pd interface, but I'm also very used to it. Ciao -- Frank Barknecht _ ______footils.org__

On Wed, 2008-06-18 at 07:52 +0530, AlgoMantra wrote: welcome to the real world. or at least, the linux version thereof. or maybe just the academic version ... there are no steve jobs in this corner of the universe - just the smorgasborg of technology to play mix-n-match with.