xoxos

murnau · Post by **murnau** » Fri Jul 26, 2013 4:01 pm

hi xoxos,

i must say as much as i like your plugins and your comments here. it's really hard to understand everything if english/american is not your native language. really hard..

ps. dunno what you can do about? i think not much, i just wanted to let you know.

theodore_whitmore · Post by **theodore_whitmore** » Fri Jul 26, 2013 4:26 pm

English IS my first language but I haven't got a clue with ninety percent of the technical details being elucidated. I'm mostly just along for the ride. The other ten percent though-a product development digest that gives me ideas to play with is welcome. Come on Xoxos, what's the secret behind the 808 cowbell?

xoxos · Post by **xoxos** » Fri Jul 26, 2013 4:51 pm

if there is a secret, it's the schmitt trigger contour. and all i've got on that is my dr110, i'm not certain that the 606/808 schmitts are the same. gimme a couple months before further discussion

i recorded waveforms from that part of the pcb, which would be really nice to see online someday.

the 606 and 808 i've been picking out the osc freqs by analysis.. hours, pages of scribbles dividing the harmonics down to identify the fundamentals. i think every 606 or 808 wav i've looked at, five of them are easy and the sixth i end up guessing at.

slap it between two buns, here's your burger. say it's perfect to every nuance, get the glossy press rolling and here's your $400 software. go speedracer.

xoxos · Post by **xoxos** » Sat Jul 27, 2013 4:18 pm

iso those missing frequencies yesterday for the dr110 (the fundamental is there but without the same presence/harmonics as the other oscs), discovered that a group of higher freqs were spaced by them

found the insight translated.

not really confident that i ought to word the product description as "perfect" or "every detail captured", for my absolutes i'd like to go with:

impeccable
precise
dishwasher safe

without a 606 or 808 to pick open, peering through the filter resonance on samples is less informative regarding the waveforms.. the dr110 was roland's last analog box, one hopes that their best methods made it in. my emulation is extended to other boxes on this ground.. last night this worked well for me, sounded nice/useful at least.

for superlatives, let's say this vst sounds exactly like the roland boxes, if they were digital and designed by me. :p

(disclaimer.. of course the fundamental criteria for a saleable emulation is to develop beyond chucking oscillators into a filter.. the 808 rimshot is a good example)

xoxos · Post by **xoxos** » Sat Jul 27, 2013 5:27 pm

gotta offer some more b/g on this here...

sure, i'm doing this to avail the popular, but also

i modded my dr110 in the 90s and rarely recorded it as other things were less fiddly to record. it got used for a few tracks i did with rappers.. dr110 mods generally expose this exotic tone, the lower oscillator count emphasises the funk of the schmitt trigger oscillators.

the tone hits you like a block of cheddar.. fragrant and substantial. it's *cheesy* but rare.. exotic. lemme dig one up -

(this project was "turn neighborhood kids in the 90s into rappers to give them confidence" - if i were smart i'd hold out and post only the good stuff by the elements, but only on cd and no drive here)
http://www.xoxos.net/temp/dr110hok.mp3

it sounds sorta "legit"/familiar and at the same time like no hat anyone else ever used. atm the emu sounds close enough to determinedly remind me of this tone.. as you can hear, it's more "frequency sparse" than the 606 or 808.

xoxos · Post by **xoxos** » Sun Jul 28, 2013 5:05 pm

i must be the luckiest developer. 4 oscs on the dr110 is a whole lot less to figure out than 6.

i'd tried XOR logic on the oscs back in the first build in june but in the wrong configuration. having probably spent 3 weeks of evenings picking through analog drum machine cymbal harmonics, it was nice to recognise a series of secondary peaks, especially the split frequency peaks.

set them up right, the generated frequencies are in the right place, without the low frequencies that produce beating. suddenly that peak that isn't harmonically related to the oscs around 1.6k is there.. and the little regular clusters of peaks under the secondary peaks in the low end.. looks right, sounds a whole lot righter.*

can't tell everyone what i did, no point in assuring everyone it's the finest thing that ever happened to their daw. i can still waffle on with...

here's my dsp "secret" for the sniffhounds..

for some kinds of oscillators, such as the ones i selected for this task, a contour can be produced without aliasing if the wavelength in samples is an integer (and, for other contours, an even integer). the disadvantage of integer length periods is of course quantised tuning, debilitatively so with higher frequencies.

otoh, to produce an antialiased waveform of this type that can be tuned continuously would take several magnitudes of cpu more. i think it's a pretty valid economisation.

a good marketer wouldn't mention this and would let customers discover that the hat pitch is quantised on their own (at least judging by hardware synths) but STILL SOUNDS UTTERLY FANTASTIC.

*as far as my rightness, which i should never mention, i can maybe give myself a 70% on the tone.. could be generous or fickle with that. eg. the noise algorithm in the dr110 is a shift register (clocked at 80k per the service manual), i recorded it off the pcb, the cycle on it is about 0.6 seconds.. standard stuff...

when i look at the noise after it's summed with the oscs and filtered, there's a peak around 12k (the highest schmitt frequency is under 1.2k). i can't think of any way to generate this procedurally from the existant structure.. and it does give a subtle but nice edge to the hats out of the box. i could stuff a peak filter in there or something extra to emulate it.. the frequency content there is rich and there's plenty of level in the highs.. i think i'm going to leave that out as it can easily be amended in post with minor equalisation, as the flat "vanilla" spectrum at the top is raw and friendly material for processing.

xoxos · Post by **xoxos** » Mon Jul 29, 2013 10:18 pm

thinking some of the stuff i wrote earlier is inaccurate

one of the great things about the dr110 is that the service manual indicates the frequency of each oscillator. all the dr110s i have samples of have some variance, 5%, a dozen, few dozen Hz.

the two highest oscillators are present, as are their harmonics. tapping the circuitboard, as said, outputs four discretely pitched oscillators per documentation. the outputs meet and are fed to two bandpass filters.

my electronics knowledge is rudimentary and finite (at least to me haha). and as others and i have observed, "there's more going on than summing four oscillators". i am hypothesising that the signal is logic gated in the chip, however that makes it to the bandpasses.

here's the gating using two oscillators, oscs change by a few Hz (one stays the same in first two). neither the pitch of the oscillator nor its harmonics are in the resulting signal:
http://www.xoxos.net/temp/cymlogic.wav

the dr110 frequencies match up nice putting two of these together, these types of harmonics are all over the analog drum boxes. doesn't sound like it from the example, the right pitches have that 606/808 ~metal dragging timbre around 12k when summed (110 is sparser).

caveat that my observations may be false..

the 606 seems to have no integer harmonic series in the output and sounds entirely logic gated. ("am i being vague? i'd prefer to make this easier for everyone *after* i've made some income..")

the 808 has the same metallic sheen while also having some observable integer harmonics. this could simply be fortuitous as there are under half a dozen meaningfully observable harmonics.. (viewing samples from four machines).

xoxos · Post by **xoxos** » Wed Jul 31, 2013 5:09 pm

how many consecutive days of cymbal synthesis now?

looks like my ears are going to survive this. singing bowl model and nighttime rides with motorcycle helmet was hard.

cron · Post by **cron** » Thu Aug 01, 2013 12:36 am

xoxos wrote:how many consecutive days of cymbal synthesis now?

looks like my ears are going to survive this. singing bowl model and nighttime rides with motorcycle helmet was hard.

Ouch!

I'm sure every other sound seems a million times better after that much cymbal. Like eating an orange after you've been surviving on bland pasta for weeks. Nobody should have to endure that much cymbal.

Still hugely looking forward to this one. Many thanks for the fascinating development log.

xoxos · Post by **xoxos** » Thu Aug 01, 2013 8:56 pm

while i "enjoy" dabbling with sensationalism, i can admit, extended percussion synthesis is not winning.. and "other sound" ... are you kidding me?

not so bad. balancing the days where 20 minute breaks are required and the days where it's not going to get done unless you put in the time listening to it.

spent yesterday extending the hat algorithm to a non-emulative model, a highlight in the process

thinking perhaps an august release. bong is not vaporware!

i rarely venture into multitimbrality and this project demands it. current thought is 4 stereo outs with assignable instruments.

if anyone wants to talk me into being strictly GM, now's the time. i'm clinging to 12 instruments, 12 keys in an octave mapping as well, though the optional cowbell makes 13 and i'm wondering what to make swappable.. not the cymbal..

i also have this thought where (what on earth am i doing spending ten days on a rimshot) i delay it for another two months to slap together a cheesy multifx unit for each group. i don't want to do it, why can't people just like me for who i am? i'm making crap up there, this is a synthedit release, people aren't going to be concerned about an fx unit.

tanabarbier · Post by **tanabarbier** » Thu Aug 01, 2013 11:07 pm

I love this thread.

xoxos · Post by **xoxos** » Fri Aug 02, 2013 5:50 pm

it reminds me of the jamms version of "dancing queen".

moving my lips so much before there's any concrete product.. living dangerously.

wrapped up the cymbals yesterday.. they have a fairly hefty cpu (5% of 1.6g cpu) not too much difference between triggering the hat or both hat and cymbal, about the same overhead. and friends, i tell ya, they're impressive!

i went with the 13x oversampling.

or at least i'm happy with the final, which is a favourable outcome, and a good place to be. and really, no one is going to be hearing them until you can move the knobs and hear the change. i don't want anyone hearing an mp3 and deciding they know all about it.

y'all.

folks.

dinty.

xoxos · Post by **xoxos** » Fri Aug 02, 2013 10:09 pm

"actually" (as evinced by the presence of this journal) there are gratifying elements to this development. as it's multitimbral, when i finish a sound engine the next step isn't automatically gui and manual. i occasionally get to go back to parts i "finished" a month ago and have had time to reflect and distance myself from.

i'm having "fun" (let's keep today a "some normal words in quotes" day) with parameter economisation. i'm shooting for 3 to 8 a voice. today i added the clap highpass filter. this intentful limiting of the user interface i see as an imposition on the user, and turns facilitation into an "art" inspired by "grace". how much ability shall i grant you today? maybe, not so much.

i am thinking that a fair amount of the percussion vst paying for public aren't going to handle stoq or friction too well. and i prefer to keep my high expectations of the user rather than insult them. but, this time, YOU GET THREE KNOBS ONLY. et c.

so i get to engage in artfulness. i've given you the right three knobs (well either that or i've egregiously debilitated the functionality).

on the technical side, given the relatively static expectations of percussion sounds, a "new for me" was turning the inverse of the logarithmic highpass coefficient into gain, so that eg. very highpassed hats are as loud as barely highpassed. logarithmic scaling of this for oversampling, and you can do stuff like pow(n, 0.7) to make it less extreme (pulls the coefficient closer to 1.0).

xoxos · Post by **xoxos** » Fri Aug 02, 2013 10:12 pm

i've given you the right pow(n, rightness)

cron · Post by **cron** » Sat Aug 03, 2013 12:03 am

xoxos wrote:i'm having "fun" (let's keep today a "some normal words in quotes" day) with parameter economisation. i'm shooting for 3 to 8 a voice. today i added the clap highpass filter. this intentful limiting of the user interface i see as an imposition on the user, and turns facilitation into an "art" inspired by "grace". how much ability shall i grant you today? maybe, not so much.

i am thinking that a fair amount of the percussion vst paying for public aren't going to handle stoq or friction too well. and i prefer to keep my high expectations of the user rather than insult them. but, this time, YOU GET THREE KNOBS ONLY. et c.

so i get to engage in artfulness. i've given you the right three knobs (well either that or i've egregiously debilitated the functionality).

One of the things I like about your plugs is that they tend to give you enough rope to hang yourself with. You wanna go nuclear and fry your CPU to a crisp for some ultra-complex silence? Go ahead!

'Perhaps' the fewer controls angle will be an interesting direction to take. On the other hand I surmise that a drum machine will be a little more understandable to the uninitiated than say, and imaginary demon vocal tract (despite the excellent and informative documentation that comes with said tract and other contraptions). Elder Thing I've ran entirely from modified presets. However, restricting your drum machine to parameters that sound 'nice' or 'authentic' is indeed just as much an artistic decision as a technical (or perhaps pragmatic) one.
---
Oh yeah, sorry to go off topic but just wanted to say that Radian is f**king brilliant. Missed it first time around and I can't leave it alone since discovering it over the past week. Some algorithmically generated rhythmic sequences pitch-locked to whatever scale, going into the 'alg' end of Amoeba, going into a big old delay with some tasteful distortion along the way pretty much equals a complete track. Outstanding.

xoxos

KVR Audio

bong vst