Open303 - open source 303 emulation project - collaborators wanted

[mistertoast]

rv0, I think the lowest note must be more like 88.2Hz rather than 88.2kHz. 96kHz must be the format of the wav that kunn would like to see.

[rv0]

i meant sample rate of the wav, not note frequentie lol

anyhow.. on second thought, I might not be able to record it tonight..

My main computer died yesterday and the one I'm typing from now doesn't have firewire (for my audio interface).
Still not sure what went wrong with my system so in the worst case, no samples till next week :'(

[mistertoast]

i meant sample rate of the wav, not note frequentie lol

anyhow.. on second thought, I might not be able to record it tonight..

My main computer died yesterday and the one I'm typing from now doesn't have firewire (for my audio interface).
Still not sure what went wrong with my system so in the worst case, no samples till next week :'(

OK. I was assuming a typo. 96k is a common samplerate even though it's not a multiple of 44100. Not sure why.

[antto]

it's a multiple of 48

[mistertoast]

it's a multiple of 48

Yeah, and to those of us who still think in hexadecimal, it's $17700.

[kunn]

will record tonight..
so the lowest note possible, 88.2kHz (or do you really need 96?).

88.2kHz will do fine as long as it is bigger than ~2*28kHz. guess we'll have to wait them for a while.

[Music Engineer]

Any other samples needed? It's easy if I can do a whole bunch at once.

yeah! i'm currently looking into the question how slide and accent interact when they are used at the same time. so i would need a pattern that contains slides with all 4 possible combinations: normal->normal, normal->accent, accent->normal, accent->accent

so - something like:

Code: Select all

TM PM UDAS   WTF
G  c  s      normal->normal
O  d
-
-
G  c  s      normal->accent
O  d  a
-
-
G  c  as     accent->normal
O  d  
-
-
G  c  as     accent->accent
O  d  a
-
-

would probably be perfect...i hope i have grokked that 303 notation - just to be sure, here's what i mean graphically:



as for knob settings:
C: max, R: max, E: max, D: max, A: max

here, i think we need 88.2 kHz too since the filter will go way above 22.05 kHz with these settings.

...maybe for D and E also some lower setting could be more suitable - not sure without having seen the results. what i actually want to see is, if and how the envelope-generators switch their time constants and maybe offsets and such.

thanks in advance

edit: oh, i just see, that in my picture i have actually used c# instead of d - it doesn't really matter. the notes should just be somewhat close together such that the pitch-slide doesn't create too much mess in the spectrogram
...actually, it could be even the same notes (all c), if that does not trigger some special condition - does it?

[antto]

rv0: i want another audio too, need to catch up the amplitude levels of both waveforms

a pattern of 4 steps will do just fine:
PM: c(D) C(U) c(DA) C(UA)
TM: G G G G

init:

Code: Select all

Decay = MAX
Accent = MAX
Cutoff = MAX
EnvMod = min
Resonance = min
Tune = <center> // not so important
Tempo = 120 // don't need to be exact

now when recording:

Code: Select all

let it loop the pattern a few times with the initial knob settings
Resonance > MAX
<wait>
EnvMod > MAX
<wait>
<end!>

now the tricky part:
set the volume so that it doesn't clip (test it with Reso=Min)
i need this test with both waveforms, the volume should be the same, so i can see if the square wave is louder or quieter than the sawtooth
sampling rate: 88.2KHz would be good, tho, 44.1 should work too

thanks! ;]

[antto]

Robin: the accent is more interesting when envmod is low (or fully zero)

rv0: i got some questions about programing, i saw a video on youtube, where a guy shows how to program the pattern from FatBoySlim - everybody needs a 303

he first entered the pitches, then TimeMode
well, from what i see, you can do that even in my beta version
tho, there were no "ties" in that pattern really

[Music Engineer]

Robin: the accent is more interesting when envmod is low (or fully zero)

mmhh...generally i agree, but here i'm mostly insterested in seeing the time-constants switch. this will presumably be easier to see when envmod is high. on the other hand - the increased modulation depth on accents will probably be easier to spot when envmod is low. so maybe we can have the two variants?

important, however, is the decay=max thing - this way, the difference in the time-constant between accented and non-accented notes will be most pronounced.

[antto]

not sure if i get it but..

for the filter envelope, the envelope time changes pretty simple - immediately
i use the SE built-in ADSR module (only the Decay of it) and i just change the value instantly, it matches perfectly
if you use this RC decay thing, you can still do the same, just precalculate the shortest time (for decay=0) and switch to it whenevere there is accent on the step (your RC will fall faster)

but if you increase the Accent knob, things get complicated
what you'll see on the sonogram (i'm talking about the curves, the resonance frequency) is that the normal notes after an accented note get some "tail" - start off a little higher and then calm down
also, the curve on the accented note itself won't look like a normal short-decay note

if you add Envmod to that.. well, not sure ;]
this is where i've messed up - when i slide a normal to an accented note
i think most (software) emulations have ignored this case, but there is such a thing in the TB-303, tho, i don't really have good samples showing it

i've modeled my "gimmick" circuit (which is not an emulation as i said) to get the same (well, close to) effect to the real thing, but that is when there are accents on normal notes (or slided ones)
the circuit uses some peakfollowers based on the filter envelope, so if i change my AccDcy (Decay time on accented notes) the gimmick sounds false (i guess this is what happens with a DevilFishMod too)

when the first note is a normal one, and is slided, but the second one has accent, it gets complicated again
since the gimmick uses the envelope, but this time, it doesn't start on this note, but has started before it (on the previous one)
so the envelope might be high (if the Decay knob is at MAX, and tempo is high) .. so when the accented note comes in, it'll change the envelope time to short, and the effect would be similar to a single accented note

if the Decay knob is low, or enough time has passed that the envelope is dead (or almost dead) - then, there shouldn't be much of an "accent" audiable (no wOw)

if the envelope is somewhere in the middle, the "wOw" will have smaller amplitude (i mean.. on the sonogram)
i got all of these cases working kinda close with my peakfollower approach
except for when envmod is increased.. my gimmick sounds silly and not 303-ish (this is because i didn't make an exact emulation of the whole envelope+gimmick circuits..)

tho i think i can tweak it more

-----------
btw, i was searching for a faster approximitation for my oscillator's 1Volt/Octave response..
.. my sequencer outputs the pitch in this format, it's simple
the osc however, has to convert it to frequency, to calculate it's Omega
so the formula is something like 440.0 * pow(2.0,(Pitch - 5.0))
where Pitch is normaly 0 to 10, and 5 is A4 key

i entered some samples of this into Curve Expert, it found a perfect match for it:

Modified Power: y = 13.75 * pow(2.0,x)
doh, one subtraction eliminated
i currently use a damn lookup table for this (linear interpolation), but wouldn't that simple multiplication+pow() be faster?!

[kunn]

btw, i was searching for a faster approximitation for my oscillator's 1Volt/Octave response..
.. my sequencer outputs the pitch in this format, it's simple
the osc however, has to convert it to frequency, to calculate it's Omega
so the formula is something like 440.0 * pow(2.0,(Pitch - 5.0))
where Pitch is normaly 0 to 10, and 5 is A4 key

i entered some samples of this into Curve Expert, it found a perfect match for it:

Modified Power: y = 13.75 * pow(2.0,x)
doh, one subtraction eliminated
i currently use a damn lookup table for this (linear interpolation), but wouldn't that simple multiplication+pow() be faster?!

i use a pre-calculated table for all midi notes. in a tb-303 u need only 3 octaves. and the frequency:

f = 440*pow(2, (midiNote - 69)/12)

in implementation i use something like this:

f = noteFreq[midiNote]

that means sequencer outputs integers which are indexes of an array.

[mistertoast]

Yeah I do what kunn does.

[rv0]

in a tb-303 u need only 3 octaves.

not true.
there's an extra octave because you can transpose a full pattern +1 octave up either in realtime or programmed in track mode.

listen to my audio examples i posted earlier, I really show off the full range with it.

rv0: i got some questions about programing, i saw a video on youtube, where a guy shows how to program the pattern from FatBoySlim - everybody needs a 303
he first entered the pitches, then TimeMode
well, from what i see, you can do that even in my beta version
tho, there were no "ties" in that pattern really

now, what is your question about it..?



THE GOOD NEWS: computer finally fixed.. Will records all samples real soon.

[kunn]

not true.
there's an extra octave because you can transpose a full pattern +1 octave up either in realtime or programmed in track mode.

that's a simple case of multiplying the frequency by 2 or if you can transpose in semitones then yeah adding another octave is simpler.