[SDL] SDL useful for sound synthesis?

David Olofson david at olofson.net
Thu Sep 21 10:42:30 PDT 2006

On Thursday 21 September 2006 18:46, Louis Desjardins wrote:
> Hi there,
> I've been browsing through the SDL site and the examples to try to
> figure whether SDL is appropriate for my application.
> I'm basically synthesizing my own sound using an "additive
> synthesis" technique, i.e. adding individual continuous sinusoids
> that each have their own phase, frequency, and amplitude.
> I am able to compute 10 seconds of sound and then output this to the
> audio device simply via OSS. However, I need to be able to do this
> in an "on-the-fly" manner, i.e. I cant be precalculating 10 seconds
> of sound, but rather, I will be calculating a few milliseconds of
> sound samples at a time and playing these as soon as they get
> processed.

Works For Me(TM). ;-)


		(Uses a predecessor of the above.)


		DT-42 DrumToy

All of the above use the SDL audio API directl; no SDL_mixer or 

simplemixer and speaker are rather small and simple and might serve as 

DT-42 deals with synchronization of the display with audio output with 
significant latency, which is a must in such applications if you 
can't get "almost zero" latency output.

> I just have to make sure that everything is real-time and sounds
> continuous and that there are no clicks or anything between samples.

Well, there's the problem... In my experience, this works fine on Mac 
OS X out of the box, and Linux, as long as you avoid sound daemons. A 
10-20 ms should work fine; below 5 ms on a properly tuned system with 
a preemptive or low latency patched Linux kernel. That's quite 
sufficient for real time synthesis, at least if you take care to 
timestamp and implement incoming MIDI events and the like properly. 
(Maintain a fixed latency, rather than handling all events starting 
at the first sample frame of each buffer.)

Whether you use SDL, or the underlying API (OSS, ALSA or CoreAudio) 
directly doesn't matter, as the layer is rather thin (minimal 
overhead) and doesn't involve any context switches or such.

However, I've never actually seen anything like low latency on Windows 
using SDL. DirectSound seems to be able to provide acceptable 
latencies with serious drivers (though many soft synths seem to use 
various low level hacks to achieve this), and ASIO should work too 
(designed specifically for low latency audio), but AFAIK, SDL doesn't 
support ASIO, and I don't know if "pro audio" DirectSound drivers 
would actually help SDL.

Now, considering that you mention OSS, I assume you're not on Windows, 
so maybe you won't have to worry about this.

> I am totally new to audio programming so if someone could help
> direct me as to whether SDL or SDL_mixer (or maybe another library)
> would be best suited for my needs (and why?) that woudl be great!

SDL_mixer and most other libs layer over the SDL audio API, so they 
don't really offer anything you'd need if you're already doing the 
synthesis in your code.

What you need to do is adapt your code to the callback model used by 
SDL. For each buffer, generate audio for that buffer, based on the 
current state of input.

Try to keep the CPU load as even as possible across buffers. Doing a 
large window FFT or something like that every N buffers is a bad 
idea, and will cause audio drop-out long before you're anywhere near 
utilizing the full CPU power.

//David Olofson - Programmer, Composer, Open Source Advocate

.-------  http://olofson.net - Games, SDL examples  -------.
|        http://zeespace.net - 2.5D rendering engine       |
|       http://audiality.org - Music/audio engine          |
|     http://eel.olofson.net - Real time scripting         |
'--  http://www.reologica.se - Rheology instrumentation  --'

More information about the SDL mailing list