experimental scope - poking the x0xb0x

[guest]

so the envmod parameter in my equation
can actually go negative
which is what gives that effect

q10 and q11 are biased up to around 3.2v
and envmod can go from 1.8v to 10v at full up
and from 3v to 4.3v at full down

the accent doesnt do this
it always decays back down to 3.2v or so

[antto]

i took my multimeter, measured the middle pin of the cutoff pot in my x0x
..when sequencer is not running (and the envelope has decayed out long ago)
when the knob is 0 - there's 4.57V
when it's 1 - 11.85 (i think that's my 12V supply voltage)

so in that formula i guess i should set A=7.28 and B=4.57

then i tryied to measure the middle pin on the envmod pot
when 0 it gives 2.7V
when 1 it gives 0.03V
but the envelope itself is "discharged" .. i'm not sure how can i measure the peak voltage at the envmod pot (aka when the envelope is peaking)

[guest]

youll need to connect up to a scope
to measure the peak values

also
since the effective zero point in that equation is up at a higher voltage
the gains are actually a bit different
the current through r73 and tm3
is the cutoff pot voltage minus the q11 base voltage
the q11 base voltage is at about 3.8v

[antto]

i'll try with brute force now
i "parametrized" your equation and my old "approximation" and i'll write a small app which will compare the output of both functions, iteratively, comparing the error
for the coefficients, i'll try random numbers, the algorithm will record the coeffs which give smaller error
i don't expect this to magically find the best coeffs right away, but at best it will get me some small enough error which i'll plot and examine.. then retune the randomizers on each coeff manually and it'll eventually get better and better

[guest]

you could probably get a close approximation
by setting accent to 0
and shorting out r61
and setting envmod to 0

then measure the cutoff frequency of the x0x
with cutoff at 0 and at 1
these will give A and B
as the e^() term should stay constant
and reflect coefficents of 0 for C and D

you can also assume E is 0 and see what you get
its just a linear offset
and is probaby pretty close to 0 anyways

you can then keep cutoff constant
and set accent to 1
and see if that gives you D
and then repeat on envmod for C

[antto]

so i left the app struggle for an hour (but i watched it, and made changes)
it appears that whatever the value of A is, B should be approximately A*0.125
E almost always made no sense (on each record, E had very different values)
C usually sticks to 0.86
and yeah, right now i don't take the accent into account, i leave it 0
i wanna get the normal envelope ranges at first, then i'll manually figure out how to put accent in there

[antto]

i can't seem to make this good enough..
maybe i'm doing it wrong
i *did* modify the equation in order to feed envmod value and envelope value sepparately.. maybe that's wrong

*EDIT*
wait, i had a bug in one place.. now this begins to behave in a way..
now i'm adding my previous approximation of the accent into the equation, guess what happens when i push the envmod .. it appears the exp() is doing some of the voodoo
i'll have to work on it more but it looks promising now

[antto]

OMG
it was just a matter of removing my "hacks" from the C13 filter
now i change it's frequency on discharge (aka exactly as if it's a peakfollower)
and it perfectly matches the x0x

guest, THANK YOU
imma post pictures and sounds after i get back from work

[guest]

awesome
im excited to see the pictures

[antto]

http://www.box.com/s/e643b6021d543422f0d6 (mp3)

first picture is with envmod=0
second picture is the more problematic case with envmod increased
as you can see, it perfectly matches the behaviour

[guest]

thats great
now im beginning to wonder how compactly
your algorithms can run

in other words
how small of a processor will they run on
can we make a hardware version of your software version

[antto]

if the sequencer is on a sepparate processor (such as atmega2560 as per the cpu mod)
but my synth relies on floating point math, so a chip with floating point math will be needed
40 leds and a 4x20 LCD

i recently implemented midi sync (with convertion to dinsync) in it too
so it is pretty much like a software x0xb0x

in overall, it works as the x0xb0x
my sequencer outputs 3 signals to the "analog" side
it also handles midi/buttons/leds/eeprom

[antto]

so i think it's pretty possible
and i WANT one ;P~

[guest]

the arm cortex m4 has floating point
and some friends of mine are making
an arduino compatible development board for it
it runs at 72mhz and is a 32bit machine
so maybe that will work

im just imagining this super small and inexpensive x0x
a pocket acid machine

[antto]

but it needs all the knobs and buttons and leds
to be precise..
Venom has 24 buttons (exactly positioned as on the 303) and all of them have a LED (unlike the 303) plus the 16 "bank" leds as on the x0xb0x
tempo and volume pots, bank (16 positions) and mode (12 positions) switches, and the 6 "performance" knobs
ehm, here's how it looks now (it will change a bit, and the LCD is not shown here)

so i think it shouldn't be smaller than the actual 303