KVR Audio

Music Engineer · Post by **Music Engineer** » Wed Feb 17, 2010 9:37 pm

antto wrote:Robin: btw, my sequencer uses a very compact pattern storage, a whole pattern of mine wastes only 33 bytes, and there are additional non-303 stuff (extras) like the step length per step

step-length per step - interesting idea. i currently have only a macro steplength parameter (for all steps at once). perhaps, one could even have both by just defining them as step-length multipliers.

the format of my patterns is binary, each on/off value is a single bit, it was a bitwise exercise for me ;P~ quite fun

my main concern currently is not the most economic way of storing patterns but the protocol for manipulating patterns through the VST interface (via setParameter). for my own derivative-synth (which won't rely on that VST-way of passing information from GUI to the core), i'll probably go with xml-based storage as i do in all my plugins. although, the storage format does not really have anything to do with the manipulation protocol. however, life is too short to juggle with single bits

antto · Post by **antto** » Thu Feb 18, 2010 12:07 am

not sure i understand all that..
i have my steps in structures, both in storage and manipulation format
which is very handy

struct step { mbyte x,y; };
// this holds the two bytes that store all data for 1 step

struct step2
{
bool gate,accent,slide;
mbyte note,length;
};

i convert between "step" and "step2" structures, i work with "step2" and store the result as "step"
the functions for converting between these are ugly (macros, bitwise crap..)

yeah, i have step length per each step, it's range is from 1/8 to 8/8
it's not something big, but it can make some swingish/funkyish feel to some patterns
it has no effect on slided notes, since the length there is always 8/8 (full)

asseca · Post by **asseca** » Thu Feb 18, 2010 7:24 am

Robin from www.rs-met.com wrote:
Download: Open303-g.7z
link did not work for me

Sorry about that, link fixed ... download open303-g.rar

Robin from www.rs-met.com wrote:
asseca wrote: apparently it is by implementing a notes/steps grid consisting of 13 notes x 16 steps, resulting in the use of 208 parameters. Each grid "point" allows for "off","on","Transpose 1+","Transpose 1-","Glide","Accent". By using 16 values for each parameter all combinations are possible ...
mmm...so would you suggest to use VST-parameters to encode the sequencer state? mmm - maybe 16 parameters for the pitches of the 16 steps (where the pitch is encoded as 1.f/127.f - perhaps with zero indicating 'off' state)...and another 16 paramters to encode the the 4 possible combinations of glide and accent (as 0.f, 0.25.f, 0.5f, and 1.0f)? that would make fo 32 additional parameters to include in the vst-dll. not really what i would call a clean design but perhaps a workable solution.

Yes it is a bit primitive, but mGUI can directly fill the matrix from a 13 by 16 matrix. 208 parameters is only 832 bytes which should not be a problem. Each grid point parameter can have 13 values: (which is basically a knob-strip)

[edit]updated, now when clicking inside the grid, a yellow focus-rectangle will appear, using the cursor keys you can move to grid-points and change the grid-point value with SPACE, num +, num - [/edit]

mztk · Post by **mztk** » Thu Feb 18, 2010 9:24 pm

(besides the first)

wdf. really? it misses the first note?
@#????

how can that be? there must be some kind of
explanation for that. what if you copy it twice?
drops another note?
if you copied it 16 x, no notes?

1/length per step is a mc202 feature(ok tho')
2/you could combine original display with userfriendly
display with flip panels+2way function/display

Q: how do you avoid patterns forming with pseudo-randoms?
(Q:what's the problem with dotnetfx?)

robin: a/b'd acidevil 1+2, and, hey it sounds nice. 2 is more
organic, and has a bit of growl/wapp-but could go further-but the
decay doesn't seem to work at all. the slide time is wrong/too long?
makes the tuning sound weird.
accent attk on v1 is fun. would be good with (lfo). all round..well, they're
quite different really.

everyone read this?
http://www.geuggis.com/tb-303/art06.htm

rv0 · Post by **rv0** » Fri Feb 19, 2010 1:06 pm

mztk wrote:(besides the first)

wdf. really? it misses the first note?
@#????

how can that be? there must be some kind of
explanation for that. what if you copy it twice?
drops another note?
if you copied it 16 x, no notes?

its been a long time but most likely has something to do with with unused or double used adress pins on the memory IC, and the way the memory is mapped/assigned

It could go a bit like this:
it keeps the pointer to the first note, but as you cleared the pattern and there's no time mode nor pointers to any other notes, as soon as you change pattern and go back to the destination it will retain the pointers from the source pattern, appart from the first pointer.

or something like that.. it's been a few years since i've learned how memory chips worked

I should get myself some logic probes and check out what's happening, afaik the service manual doesn't go into detail.

the memory thing gets totally crazy in track mode, you can chain patterns and do stuff with them, but as soon as you make your track too long or anything it totally f*cks up everything. there's 7 tracks you can store and most of them vary in available 'length', this is documented in the manual but can get very confusing once you start chaining tracks, thus writing from track to track without setting an end point.. Esp. if you make a mistake and it screws up an hour of manual track mode input

(yes, I've made some insane creations in track mode)

Wow, now I made it sound like it isn't any fun at all

Most acid doesn't use track mode so...

Dexrow synths · Post by **Dexrow synths** » Sat Feb 20, 2010 1:34 pm

I have been separating the graphics from the graphics engine. If you can send the parameters as text spreadsheet style I have the graphics for it. I have been testing it on csound and some programming languages. I also have some nearly finished drum machine graphics...

Music Engineer · Post by **Music Engineer** » Sat Feb 20, 2010 7:20 pm

Dexrow synths wrote:I have been separating the graphics from the graphics engine. If you can send the parameters as text spreadsheet style I have the graphics for it. I have been testing it on csound and some programming languages. I also have some nearly finished drum machine graphics...

i don't really understand that offer - what do mean by separating graphics from the graphics engine?

antto · Post by **antto** » Sat Feb 20, 2010 8:26 pm

meanwhile, i'm experimenting with nonlinearities..
inspired by a bunch of articles about diodes 'n' stuff..
i'm definately not doing wierd math to derive digital simulation of circuits or components
..but what i did so far is something different, interesting, and to me, it sounds very pleasing.. less digital, more analogish
i'm still tweaking and altering a bunch of algorithms, but i can upload an audio demo these days, if you wanna hear it

basicaly, it is inspired by a statement that diodes are not just plain-shapers, but have something like "memory"
i didn't read the whole article, and didn't paid a lot of attention (so i might have got it all wrong) .. and i didn't understand 80% of what i read..
but the whole thing started to look to me like a little LP-flavour inside the nonlinearity, so here's what i did:

i made a little clipper algo, it's perfectly (or maybe relatively) linear between -0.5 and +0.5
beyond that, it softly limits the amplitude
now, the interesting thing is that there are 2 1-pole LP filters, one for the negative and the other for the positive halfwaves
the result is pretty cool, when the LPFs react slower - most of the signal "leaks" at first..
now this might look like some kind of foldback distortion, yes..
..but not quite.. since the "folding" is frequency dependant now
at least this is how i understand it, oh well, it sounds very good on a 303
when the gain is very high, and the 303 filter is low, it adds nice little edges all around, sounds familiar ;]
but as the 303 filter freq increases, it doesn't distort it that much, because it "leaks" ..

the algo itself is kinda big for just a replacement of atan() (for example)
i'm testing it as a distortion on my synth, but i am about to put it in the filter now, i have a feeling it'll be sick ;]

Dexrow synths · Post by **Dexrow synths** » Sat Feb 20, 2010 8:34 pm

typo graphics from the audio engine.. sorry that way one graphic program can be used with several audio engines.

http://pypi.python.org/pypi/colin1911/.99%20version%203

antto · Post by **antto** » Mon Feb 22, 2010 4:06 am

my previous Filter envelope (RC) curve shaper was too sharp
i used a sample from dinsync.info, sonogram, extract datapoints, convert to exponential input (RC) and here's the new shaper:

Open303 - open source 303 emulation project - collaborators wanted