World: r3wp

How can you test if the sound port is already open?

Never mind just used an "if error? try [" statement..

ICarii, I just quickly tried your synth2.r. It plays the song here 
on Linux, and matches the score. I wonder, though, how you are going 
to scale the program. Rebol's sound port is currently underpowered, 
and Rebol isn't fast at number crunching which is necessary for audio 
work. I don't like the idea of waiting for hours for a full song 
to render.. I would be looking for an audio processing system to 
hook into rebol.

im going to port it to R3 tomorrow and rely on Call to play sounds

but yes it is horribly slow.

Well, I've been fiddling with Tiny C Compiler - maybe it can eventually 
be used for fast crunching.

R3 build-song speed (without optimization) plus write to disk is 
1810ms vs R2 of 2862ms. So even without optimizing in R3 there is 
a large difference in performance.

added an interpolate function for dual channel sound and works fine 
:)  still slow on initial build though :(
interpolate: func [series1 series2 chunksize /local ts1 ts2 c][
	ts1: copy series1
	ts2: copy series2
	c: copy #{}
	forall ts1 [
		insert tail c copy/part ts1 chunksize
		insert tail c copy/part at ts2 index? ts1 chunksize
		take/part ts1 chunksize - 1
	]
	c
]

hmm that works with the take/part in the forall but: 		loop chunksize 
- 1 [next ts1]
may be safer?

ahh.. i should have used forskip .. never used that before

interpolate: func [series1 series2 chunksize /local ts1 ts2 c][
	ts1: copy series1
	ts2: copy series2
	c: copy #{}
	forskip ts1 chunksize [
		insert tail c copy/part ts1 chunksize
		insert tail c copy/part at ts2 index? ts1 chunksize
	]
]

Try with :

interpolate: func [series1 series2 chunksize /local ts1 ts2 c][
	ts1: copy series1
	ts2: copy series2
	c: copy #{}
	forskip ts1 chunksize [
		insert/part tail c ts1 chunksize
		insert/part tail c at ts2 index? ts1 chunksize
	]
]


Why do you copy the 2 series? It's not required here, you're not 
modifying them.

You should also pre-allocate the 'c buffer : c: make binary! <max-size>.

ahh thx - re copy - that was an artifact from the first iteration 
of the function where i was using take in a loop.

next on the list of todo's is to separate top and bottom notes and 
timings and add in rests.

Just tried synth2.r, nice work by the way.

:)  its getting there slowly - i really need to look at ADSR envelopes 
so 16bit samples dont sound so horrid :)

ICarii, have a look at my PM-101 http://www.rebol.org/view-script.r?script=pm-101.r

I use 8-point free envelope there (ADSR can be easily emulated using 
just 3 points), maybe it will be of some use for you

will do :)  ive been looking at using piano waveforms but they require 
too much number crunching for rebol :(

I was compiling the math functions to C and the result was about 
5-10x faster (not optimalized), but I cannot find that code right 
now.

pm-101 reminds me of an aircraft console ;)   once it stops scaring 
me ill look more closely :)

hehe :)

i was thinking of prebuilding and saving envelope shapes and modifying 
the tone like a mask

I was also working on tracker-like program (not published). I had 
a simple format for notes there that was similar to sequencers - 
I was using tuple! for timing information, it may be better than 
having two blocks, one for notes and one for timing.

so it looked like 1.1.1.1 C1 0.0.1.0 1.1.2.1 D1 0.0.1.0 etc

I think it was used in my demo pitch.green for Rebol demo contest

i intend to fully reproduce sheet music so the format is still a 
little in teh air at the moment

the next version of synth is using a top/bottom stave block with 
separate timings

Ah, OK

i am also trying to make the R3 conversion as simple as possible 
by keeping most of it generic

its taking me a little longer than i thought as im having to learn 
acoustic physics :P

Yes it's better to have the generator totally separated from the 
UI.

yes - i have an R3 synth version now that can build the wav files 
in the same manner as R2

nice

I should have a drumsynth for R3 somewhere, I'll try to find it.

heh :)

wait ... for R3? can that do sound now?

No, just save the sound as WAV or AIFF

ahh ok

it uses vector! for storing sound data

i thought vector was broken? - ive been avoiding it since i heard 
that

Well vector has its limitations but if you try hard enough, it's 
usable :)

noticed an Endian bug in bitrates of 16+ when saving to wav files 
- will fix this in next release.

ok - im trying to calc the optimal intersection point where 3 different 
frequency sine waves create the highest complex wave -
w1: freq 440.0Hz  w2: freq 530.0Hz w3: freq 720.0Hz
i was using: sine (90 * w2 / w3) + sine (90 * w1 / w3) + 1.0

this was under the assumption that the greatest cumulative distance 
from 0 would be at the 90degree mark from the highest freq wave - 
unfortunately this is not the case :(

does anyone have a nice (or not too nasty) formula that will find 
the optimum value? - i will post this in maths too

Do you mean you are trying to find the repeating period of the three 
sine waves, once they are combined ?
I think that means finding the lowest common denominator.

i am trying to find the maximum amplitude of the complex wave

Ok, different problem than I thought..

sorry - wasnt very clear in my original post

So you want to calculate it in advance.

yes - it is a huge time save when i go to do loudness balancing