From the original SynthMaker forum (2004–2014), with its members’ names and dates.
datsound ·
here is hopped P-F stream, thanks to the info from MBenjamin and Oddson and all it is about 1/10th efficiency
i am very chuffed because i do maths like a monkey. i get the figures and multiplications that people give me, wire them about 100 different ways until one combination works. it took me a while.
exonerate ·
Nice one 🙂
It surprises me that it is so much faster.
By the way you may as well leave the hop off, it makes no real difference. Hop are only beneficial on large code segments that take up a fair bit of CPU. There's no benefit to hopping a multiply! in fact it may increase CPU usage because the hop has a bit of overhead it's self.
Also the float "c" is not needed you can write directly to the output "f = a * b;"
Cheers 😉
Exo
datsound ·
Oh right i get it. ty.
here is a second version, much more precise (on 48khz) at the mid and top end and this time it is some of the bass notes that are off by .02%, if i had an asio card i would try to figure out a totally equivalent version.
here is the figure that makes it more precise. depending on sr, i dont know. here it is at 48khz.
aliasant ·
Heres mine 🙂
Slightly faster and more precise. This one is hoped at 4 but larger hops can be used an it still sounds great.
datsound ·
oh thankyou aliasant. very amazing. it is even good for FM which my one wouldnt do:)
EDIT - i am having issues with both new versions of p.f ... i cant import alaiasant 1 very easity, it crashes, although i did manage to import it and it only half crashed, synthmaker overloaded on something.
with my one i managed to import it and put it in the synth, and then it crashed after a while and every time i reloaded and pressed a button it crashed so i have to re import it and try again to see what the bug is. edit- it seems to work okay, good idea is to close and reload SM sometimes when it bugs for no obvious reason.
MichaelBenjamin ·
heres the one modded with mhz exellent mod of shifrins stream power, gives exact the same results as the stock module and runs ~75%faster at samplerate.
edit: i was wrong, the last two places are sometimes a tiny bit different. i wonder if this comes from the 1/x precalculation. i guess its better to do the sample rate inversion with a real div and try and if not, then also the 1/12 if one needs super exact precision. i think it doesnt matter much since the actual difference is for example 466.62271 hz vs 466.6228 hz, which would be a total phase difference if played together with the length of, if my calculations are correct (1/(9*10^-5))/60= 185.185185 minutes, or ~3hours.
aliasant ·
datsound wrote: oh thankyou oddson. very amazing. it is even good for FM which my one wouldnt do:)
EDIT - i am having issues with both new versions of p.f ... i cant import oddsons one very easity, it crashes, although i did manage to import it and it only half crashed, and the the synced modules and the sound output were all bugged.
with my one i managed to import it and put it in the synth, and then it crashed after a while and every time i reloaded and pressed a button it crashed so i have to re import it and try again to see what the bug is.
Not sure what Oddson has to do with my module but heres a clue.
It uses a wavetable and the table is pretty large. I suggest you remove the after load prim inside the purgable module before saving it to he toolbox. Or even remove the pugable module al together or atleast make sure the loop cant start. The loop is only needed needed when deciding on how high the pitch should go. I did set it to 150 for this one and that should be high enough for most. You might want to keep the module for the future but click the trigger input on the last Float Array S&H to clear it.
aliasant ·
Just tested the Shifrin version. Mine uses about half of that one but it isnt as exact. Something I kind of like. Te slight differences in freq might make some nice music 🙂 Mine migt be a bt tricker to use though. You need to keep an eye on that huge loop it creates for the array. Saving a schematic with that loop in without clearing the last s&h inside the purgable module will add about 5Mb to the osm. Also clear it before dragging it into the toolbox. And of course after dragging it in from the toolbox the loop eeds a trigger to recalculate. But once you have it the way you want it you can just delete that module al together. The array wil still be saved inside the Waveread Module.
datsound ·
sweet thanks for the info. So if you had an lfo going into the array, it would make almost infinite values? so you would have to quantize to like 512 / 1024 values per octave then it would probably not be audible? i dunno. it seems really good i will have a go at some point tomorrow. i confused you with oddson for some reason, i think it is from the neurosis of having my 1tb hdd enclosure die from a wind electricity interruption. thanks aliasant
infuzion ·
Thanks guys! There is a very slight difference between SM's P-F, but that may be rounding difference. Happens alot when I ASM a set of SM's primitives.
Late Christmas Present: Faster (by 1 cycle, 15% on Intel C2D) & sexier Merry Christmas 🙂
MegaHurtz ·
Props mates! I was yust working on a project seeking the minimum amount of math for this, so now I know. Thanks MB and Infuzion!
infuzion ·
MegaHurtz wrote: Props mates! I was just working on a project seeking the minimum amount of math for this, so now I know. Thanks MB and Infuzion!
Welcome 😇 You do have to test alot using this. I tried my version out in 2 places a OSM I was hired to optimize. The first module I put it in, it actually increased CPU% using SM's CPU readout by alot from the stock SM version. The 2nd place I replaced the SM version worked at reducing overall CPU% test test test!
Wrote something here? To have a post removed or your name hidden, write to [email protected].
SynthMaker