SynthMakerinfuzion ·
file 2048 array WaveTables Ramp & Squares in low (-3 & down) seem to have a high-pitched squeal. 256 WT don't have this sound. I didn't really notice on my laptop speakers, but on 2 headphones boy is it there. verifiableby others
1.0.9 WT issues in low octives
From the original SynthMaker forum (2004–2014), with its members’ names and dates.
sambean ·
Sounds OK to my cloth ears.. and the FFT looks good too..
I'd say there's a problem in the 256 table size though- If you set the octave to +1 you can see aliasing in the frequency domain, perhaps there's a problem with the 256's table offset?
Sam
infuzion ·
sambean wrote: I'd say there's a problem in the 256 table size though- If you set the octave to +1 you can see aliasing in the frequency domain, perhaps there's a problem with the 256's table offset?
Yea I see it, but I don't hear a problem, but I don't have a good sound system to test.
I'm wondering if the 2048 issue is from aliasing, like synths aren't supposed to go that low?
xmidiman ·
I've noticed this too, well even before this update. But I was really hoping with the new sizes in the wavetables the low end was going to rock, but I'm disapointed now that there is not too much improvement on the low end at all. I actually testing a saw @ 256,1024,2048,4096,8192 and all of them sounded exactly the same, except the 256 which doesnt sound very good.... Basically like water farts 😊 I checked the specturm analysis, and it looks like there was something poping up around 7khz and also some unstable other "things" poping around... I saw this post awhile back but never saw a responce from the devs if this indeed was a bug or are we doing something wrong? I've used all new modules in building new wavetables as well to ensure no errors.
Xavier
malc ·
Don't worry this hasn't been forgotton - still looking into it. Thanks for your patience.
infuzion ·
malc wrote: Don't worry this hasn't been forgotton - still looking into it. Thanks for your patience.
That's cool; I imagine you're working on the FL edition, family, etc.
Acrobat ·
I'm working on wavetable module @ 2048 waves but I still didn't hear anything wrong... 😕
infuzion ·
malc wrote: Don't worry this hasn't been forgotton - still looking into it. Thanks for your patience.
I'm not the only one who notices this: [old link] cheers 🙂
xmidiman ·
Any good news on this issue? Does anyone need examples?
infuzion ·
xmidiman wrote: Any good news on this issue? Does anyone need examples?
A clear example or 2 is always fun 🙂 I wonder if something in the developer toolkit could be a workaround? [old link]
kirkagur ·
infuzion wrote: xmidiman wrote: Any good news on this issue? Does anyone need examples?
A clear example or 2 is always fun 🙂 I wonder if something in the developer toolkit could be a workaround? [old link]
the wave table osc in the developer toolkit has the same problems.
the problem is in the wave table module ,it`s not generating enough harmonics in the low octave tables .
xmidiman ·
infuzion wrote: xmidiman wrote: Any good news on this issue? Does anyone need examples?
A clear example or 2 is always fun 🙂 I wonder if something in the developer toolkit could be a workaround? [old link]
Here you go! It's been a long time since I last asked about this one, and you just can not do some nice bass lines with these "water farts"... On a side note, I've done tests long ago doing doing all the multiples I believe up to 8192 and it did not make any difference in quality.
I attached a spectrum view so you can see exactly what we have been hearing. The harmonics are just not there... As kirkagur stated, I was excited to see someone fixed the wavetable in that developer toolkit, but actually I didn't understand what it did better becuase it does the exact same thing as far as quality...
Also another example (XO_SEPARATION_01a.mp3) is the synth I've long been working on, which I'm using low notes on the supersaws and pad, however it's lacking those harmonics to make it sound full.
infuzion ·
xmidiman wrote: Here you go! It's been a long time since I last asked about this one, and you just can not do some nice bass lines with these "water farts"...
Yep, SM sounds like water farts to me! 😝 Dang, how I wish we had a SDK, so we could just make our own OSCs!
exonerate ·
I suspect this may be fixed in the next release, the issue is not enough tables being generated for the low notes at large array sizes. As stefan Found.
So malc know exactly what the issue is, so I'd assume it's already been fixed and in the next beta. 😀
Cheers 😉
Exo
infuzion ·
A test OSM with a button or 2 that makes the fart sound would help.
xmidiman ·
infuzion wrote: A test OSM with a button or 2 that makes the fart sound would help.
Ok I made something for you, hopfully you will at least be able to press the key on your keyboard as I did not make it automated. 😈
I think looking at the analyzer, the 512 WT looks like it doesn't mess with the harmonics as much as +1024 does.(Although I do see it making some kinda notch in the band) I was always using high value WT(2048), but maybe 512 is best for right now until this is fixed?
exonerate wrote: I suspect this may be fixed in the next release, the issue is not enough tables being generated for the low notes at large array sizes. As stefan Found.So malc know exactly what the issue is, so I'd assume it's already been fixed and in the next beta.
I hope your right but I will assume it isn't fixed because there has already been 3 releases last year that never addressed this issue. I think there is about ~3 or more threads that I've seen about these water farts. 😉 2 of them I posted a please/beg for this to be addressed. Malc replied, but I think it needs another booty bump. 😕
infuzion ·
Thanks for the file! Yeap, crappy. Darn, so much for my bass-synth! 😠
rl ·
how about mixing waves together? a hi-rez wave for lower notes and a bandlimited one for higher notes?
xmidiman ·
rl wrote: how about mixing waves together? a hi-rez wave for lower notes and a bandlimited one for higher notes?
Yeah I was going to suggest that, but I was thinking of having some kinda switch network by midi note, not mixing. So if midi note is less than "40" use this osc(raw), if greater than use the wave table etc. The raw saw at low frequencies will still alias, but I think most ears will probably not notice? 😮 I really don't know how well this will work when your working in a poly stream though... 😕 Also I dont know how you could do this with more complex waveform vs basic shapes you can make with code.
exonerate ·
xmidiman wrote: rl wrote: how about mixing waves together? a hi-rez wave for lower notes and a bandlimited one for higher notes?
Yeah I was going to suggest that, but I was thinking of having some kinda switch network by midi note, not mixing. So if midi note is less than "40" use this osc(raw), if greater than use the wave table etc. The raw saw at low frequencies will still alias, but I think most ears will probably not notice? 😮 I really don't know how well this will work when your working in a poly stream though... 😕 Also I dont know how you could do this with more complex waveform vs basic shapes you can make with code.
I experimented with that approach quite a while back, I used the pitch output on the midi to poly and a bit of code, It didn't find it usable because going from the non bandlimited to band limited there was a very distinct change in sound of the osc, which didn't sound good, you press say C and then C# sounds very different in texture.
Actually mixing them would probably sound better, but would use more CPU.
For more complex shapes you'd have to use the float array to poly I guess for the non bandlimited part, but that would probably sound very crappy.
Cheers 😉
Exo
xmidiman ·
exonerate wrote: I experimented with that approach quite a while back, I used the pitch output on the midi to poly and a bit of code, It didn't find it usable because going from the non bandlimited to band limited there was a very distinct change in sound of the osc, which didn't sound good, you press say C and then C# sounds very different in texture.
Yeah, I know what you mean, kinda like early sample librarys that would only sample a few octaves...~Oh... thats a good key! 😀
exonerate wrote: Actually mixing them would probably sound better, but would use more CPU.
Thinking about it, the main reason I think this might not work(sound good), is because your lower harmonics would be at a higher level when mixed with the bandlimited osc thus changing the sound even more.(I guess it'd be kinda like a low shelf eq the way I imagine it because you are adding 2 things that already share some common harmonics) You'd also have to use perfect copies of each other or your going to get some funkyness.
BTW I made a quick cheesy osc switcher from what I was talking about... Probably someone could take this idea and make a nice one out of this, or I'm sure there is a better way you guru's could come up with. Going from F to E doesnt sound too bad other than the levels, I think most people wouldnt tell the difference when its got all that filter/fx etc on it. 😇
exonerate ·
This is what I would do for the osc switching...
Your right though it does sound ok, apart from the slight volume difference.
Cheers 😉
Exo
xmidiman ·
Hey thats pretty nice and simple! Might sound silly but you could probably use that "keyboard splitter" for a keyboard split for different synth parts too!(bass/lead) With some good specturm analysis, time, and ears, you could find the best split points based on quality/alaising. I picked 41 early because the graph showed it loosing harmonics on the WT version right below that note. Thanks Exo!
Tom7777 ·
-
exonerate ·
Tom7777 wrote: Unless im doing something wrong it seems the split is not working -
Shouldn't it play noise also? Or am i mistaking it for something else? I seems like it should split the frequency so it will play one on high notes and another on lower?
The splitter requires a 0-1 normalized frequency to work.
Cheers 😉
Exo
Tom7777 ·
-
infuzion ·
exonerate wrote: I suspect this may be fixed in the next release, the issue is not enough tables being generated for the low notes at large array sizes. As stefan Found. So malc knows exactly what the issue is, so I'd assume it's already been fixed and in the next beta.
I have a 1.1.3 pre-beta 0.7.1, & it still has the poor quality issues as described here. Perhaps this is one of the things delaying the next release. I'm glad it is not just me. 😉 I do need this to be fixed with in the next month or 2 please.
xmidiman ·
infuzion wrote: I have a 1.1.3 pre-beta 0.7.1, & it still has the poor quality issues as described here. Perhaps this is one of the things delaying the next release. I'm glad it is not just me. 😉 I do need this to be fixed with in the next month or 2 please.
How many samples are you using? I used SynthMaker 1.1.3 beta 0.7.2 for testing and it really depeneded on how many samples you were using up to about 2048, then things have less audiable differences... You can somewhat see it better with these pictures I've attached. Acutally the F2 ones, are just showing some issues if your using ~1024 or below samples.
Overall these damn things sound excellent @ 2048 and up... So you could say they are fixed, other than at low samples you still see this strange notch in the freq band. I am willing to say the quality & alaising testing showed it was better than even Sytrus's saw.(Which was the best I had seen in the past from other testing)
Exporting from original to beta showed the dll increase: before 2,461,696 bytes after with WT fix(just exporting in the beta): 5,956,359 bytes) So about 3Mb increase?(hehe I wonder why 😈 ) Edit: the dll i was refering to was a project I'm working on
infuzion ·
xmidiman wrote: infuzion wrote: I have a 1.1.3 pre-beta 0.7.1, & it still has the poor quality issues as described here....
How many samples are you using? I used SynthMaker 1.1.3 beta 0.7.2 for testing... Overall these damn things sound excellent @ 2048 and up... So you could say they are fixed, other than at low samples you still see this strange notch in the freq band.
2048 sample wavetable; maybe I'm not used to a note played at C0 sounding so bright.
xmidiman ·
infuzion wrote: 2048 sample wavetable; maybe I'm not used to a note played at C0 sounding so bright.
So I am guessing it is working correctly for you now? 😕 You don't have to go that far down, but it's sometimes interesting to hear the quality. On a side note: Does anyone know how we can use the wavetables that are created first, for loading for future use? Rather than it generating the wavetables every single time you load the VST/OSM?(Or maybe I have this thought out wrong?) Seems I could save about 6 seconds if I could do this. Perhaps I should just put this ? in the other thread about loading speeds...
infuzion ·
xmidiman wrote: infuzion wrote: On a side note: Does anyone know how we can use the wavetables that are created first, for loading for future use?
Are you asking for Float Array, Float Array Sample & Hold, Text?
xmidiman ·
infuzion wrote: Are you asking for Float Array, Float Array Sample & Hold, Text?
Nope, I am refering to the wavetables being generated everytime the VST/OSM loads. At least I was assuming that this was happing? edit: I currently use a Float Arry --> wavetable --> wavetable read(osc)
I have been doing some experiments with some switching system that would "turn on" the whole wavetable network from start to finish, using the after load trigger... All things dont really work correctly even tho it speeds up loading.(Some wavetable osc never get their tables etc)
It takes me about 10 secs exactly to load my project which contains 56 wavetable osc/Float Arrays etc... Just looking for a quicker way about this. 😉 Note that I have a Intel Q9450, so anyone else loading this is going to need a smoke break. 🙂
exonerate ·
xmidiman wrote: infuzion wrote: 2048 sample wavetable; maybe I'm not used to a note played at C0 sounding so bright.
So I am guessing it is working correctly for you now? 😕 You don't have to go that far down, but it's sometimes interesting to hear the quality. On a side note: Does anyone know how we can use the wavetables that are created first, for loading for future use? Rather than it generating the wavetables every single time you load the VST/OSM?(Or maybe I have this thought out wrong?) Seems I could save about 6 seconds if I could do this. Perhaps I should just put this ? in the other thread about loading speeds...
The wavetables are created automatically at start up, there's currently no way round that. I've had the same idea though and made a request HERE.
Bump that thread if you like. 😉
Cheers 😉
Exo
xmidiman ·
exonerate wrote: Bump that thread if you like.
Woohoo! Exactly what I had in mind and I added some 🙂 Thanks! 😀
Wrote something here? To have a post removed or your name hidden, write to [email protected].