freenode/#sicl - IRC Chatlog
Search
6:02:29
beach
I haven't done this calculation for some time, so please help me check it: Let's say I want to use mainly additive synthesis to synthesize a symphony orchestra. The sample rate is CD quality. Let's approximate that with 50kHz. So I have 20μs to compute 1 sample.
6:02:40
beach
Also, let's say that an instrument requires 100 simultaneous sine waves and that a single sample of a single sine wave requires 2 multiplications and 2 additions taking perhaps 10ns. That means a single instrument takes 1μs per sample, so I can generate 20 simultaneous instruments with a single core.
6:02:41
beach
Now if I have 8 cores or so, that should be plenty for an entire orchestra. Did I do this right?
6:04:18
beach
It is possible that some instruments require more than 100 simultaneous sine waves, but on the other hand, one should be able to inspect the total collection of sine waves and apply some psychoacoustic result and eliminate masked oscillators.
6:05:22
beach
So, if this calculation is correct, a single modern processor should be able to generate an entire orchestra.
6:05:48
no-defun-allowed
Although...that seems like a very uncommon and unusual approach for synthesis, from what I know about computer music.
6:07:17
no-defun-allowed
Sure, but from a practical/production point of view I've hardly used it, except for trivial cases of additive synthesis like electric organs.
6:08:38
no-defun-allowed
I have two physical (though digital) synthesisers: one is a "subtractive" synthesiser which produces two saw waves or square waves and filters (usually) high frequencies out, and the other is a sampler with a similar optional filter.
6:09:38
no-defun-allowed
Additionally, "frequency" (which is actually phase-) modulation is sometimes used, notably by synths like the DX7 and early sound cards like the SoundBlaster.
6:10:24
beach
Ah, I think we are talking different things here. My idea is to analyze existing instruments into something like a sum of percussion sounds and a bunch of sine waves. Then, it will be easy to modify those sounds with respect to duration, volume, pitch, etc.
6:10:31
no-defun-allowed
You are right in using it as a performance benchmark though, but I'm not sure about how applicable it is as a pitch for CLOSOS as a sound production operating system.
6:11:57
beach
Right now I am not thinking CLOSOS in particular. I am thinking about how an instrument such as the Kurtzweil that Jordan Rudess is using in his demos could be created.
6:12:29
beach
It probably used to be the case that it had to have a large number of dedicated signal processors.