The Favorite 2013 Softsynth Is...
-
- Banned
- 254 posts since 7 Jun, 2008 from On this 1 world!
murnau wrote:dont you put Satin over them?Delfinoverde wrote:Best Softsynths are:
- Alchemy
- Absynth 4 or 5
- GForce ImpOSCar 2
- Rob Papen BLUE
U-he synths are not in my best of list.
I find they sound artificial, not authentic.
----------------------------------------------
Since when is SATIN a soft-Synth? 
---------------------------------------------------------------------------------
-
- KVRAF
- 5564 posts since 13 Jan, 2005 from the bottom of my heart
what? it was kind of a joke to say put satin over the u-he's synths (then they sound better).. ahhh.... i tried to beat that guy in absurdity..Delfinoverde wrote:murnau wrote:dont you put Satin over them?Delfinoverde wrote:Best Softsynths are:
- Alchemy
- Absynth 4 or 5
- GForce ImpOSCar 2
- Rob Papen BLUE
U-he synths are not in my best of list.
I find they sound artificial, not authentic.
----------------------------------------------
---------------------------------------------------------------------------------
Whoever wants music instead of noise, joy instead of pleasure, soul instead of gold, creative work instead of business, passion instead of foolery, finds no home in this trivial world of ours.
-
- KVRAF
- 12615 posts since 7 Dec, 2004
It's a matter of tweaking to get the perfect odd/even harmonic balance, drive, feedback vs. signal path saturation and then applying an awful lot of oversampling (16?) to eliminate the aliasing.Shy wrote:I mean it should be pretty easy, since we're clearly far beyond the point of "good enough", based on what these forums make people believe.
Here is some two minutes I put into it. My feedback sat was already set accurately (I just agree with the levels used in the sh- and many other synthesizers with regard to diode clipping), but I had to add a load of stage sat and a small offset to get closer to the identical harmonic levels.
What I didn't bother with was really tweaking the resonance to sit at exactly the perfect level and I think my example has the input signal level and feedback level set slightly off.
When it comes down to manually tweaking for hours to get a result that matches close enough for someone to say "okay, it matches", I think it is really ridiculous.
Shy, I don't know how you could say this doesn't come close enough and I expect that you'll understand after hearing/examining this example that it is in fact possible given enough effort to reproduce the example you posted exactly. It is very close already, even the behavior of the oscillation is nearly identical.
http://xhip.net/temp/lowpass_sat.flac
I'd have a hard time getting exactly the same result with my own sh-101!
Free plug-ins for Windows, MacOS and Linux. Xhip Synthesizer v8.0 and Xhip Effects Bundle v6.7.
The coder's credo: We believe our work is neither clever nor difficult; it is done because we thought it would be easy.
Work less; get more done.
The coder's credo: We believe our work is neither clever nor difficult; it is done because we thought it would be easy.
Work less; get more done.
-
- KVRAF
- 5139 posts since 27 Jun, 2004
That's extremely far from it in many ways, aciddose, and not in a "nuances" way. Across the sweep, the resonances are wrong, it drops too much, then rises too much, kind of a resonance stepping effect and the sample just sounds bad in comparison. I'm sure you know that if you make a long slow sweep like in the SH-101 example, it will be even much more apparent how the analog filter has a totally solid resonance with no annoying dips and peaks which at higher frequencies get a life of their own and ruin the whole sound even more, and how its "fading" from one frequency cutoff to another has a "liquid"-like quality that makes the different resonations with the oscillator source "gel" together seamlessly. You know what I'm talking about. I'm sure that in a static setting/comparison you could make it much more similar and likely even indistinguishable, but no software comes anywhere near the real thing yet.
"Music is spiritual. The music business is not." - Claudio Monteverdi
-
- KVRAF
- 12615 posts since 7 Dec, 2004
This is due to the incorrect feedback + input levels vs. signal path saturation.Shy wrote:That's extremely far from it in many ways, aciddose, and not in a "nuances" way. Across the sweep, the resonances are wrong, it drops too much, then rises too much, kind of a resonance stepping effect and the sample just sounds bad in comparison.
It's the same as hard-sync vs. soft-sync. I'm not sure if you've used soft-sync before, but the effect in the filter is that if the input level is too high it'll "push" the phase of the filter harder, causing the self-oscillation to sync up which creates the "steps". As the input signal is reduced it has less effect on filter phase and stepping is reduced.
What you're describing are "nuances". It's simply a matter of getting the whole equation to be exactly correct. If you adjust the feedback level or signal gain you'll find you can get the "stepping" effect on the sh-101 no problem. Yours may be maxed out, in which case you can use the external input although that'll throw the phase off a little in some ways due to the additional amp/filters in the feedback path. It's difficult to match two sh-101 or other synthesizers due to variations in the resistors, which generally are 5%, 10% or 20%.Shy wrote:I'm sure you know that if you make a long slow sweep like in the SH-101 example, it will be even much more apparent how the analog filter has a totally solid resonance with no annoying dips and peaks which at higher frequencies get a life of their own and ruin the whole sound even more, and how its "fading" from one frequency cutoff to another has a "liquid"-like quality that makes the different resonations with the oscillator source "gel" together seamlessly. You know what I'm talking about. I'm sure that in a static setting/comparison you could make it much more similar and likely even indistinguishable, but no software comes anywhere near the real thing yet.
To get exactly the effect in the clip I posted, or the clip you posted (you can do either with either) we're talking about accuracy better than 1%.
Free plug-ins for Windows, MacOS and Linux. Xhip Synthesizer v8.0 and Xhip Effects Bundle v6.7.
The coder's credo: We believe our work is neither clever nor difficult; it is done because we thought it would be easy.
Work less; get more done.
The coder's credo: We believe our work is neither clever nor difficult; it is done because we thought it would be easy.
Work less; get more done.
-
- KVRAF
- 5139 posts since 27 Jun, 2004
You say it's "nuances" and "a matter of getting the whole equation to be exactly correct", I disagree. It sounds like everyone's "equations" are not just "not exactly correct", but also "way off", always. I'll just keep waiting for someone to make something that comes anywhere near the likeness of the real thing, but I stopped holding my breath long ago.
"Music is spiritual. The music business is not." - Claudio Monteverdi
-
- KVRAF
- 12615 posts since 7 Dec, 2004
Send me these clips:
I'll see if I can put a couple days toward it and match it perfectly.
- Self-osc 50hz, 1k, 5k, res 100%
- Ramp 130, 261, cut 1k res 0, 50, 100
- Pulse 130, 261, cut 1k res 0, 50, 100
I'll see if I can put a couple days toward it and match it perfectly.
Free plug-ins for Windows, MacOS and Linux. Xhip Synthesizer v8.0 and Xhip Effects Bundle v6.7.
The coder's credo: We believe our work is neither clever nor difficult; it is done because we thought it would be easy.
Work less; get more done.
The coder's credo: We believe our work is neither clever nor difficult; it is done because we thought it would be easy.
Work less; get more done.
-
- KVRAF
- 5139 posts since 27 Jun, 2004
Thanks but sorry, I sold it a while ago. But you have an SH-101, you could easily make a sample similar to the one I posted and it will sound essentially the same (not absolutely identical but with exactly the same important characteristics). Then you could try to match it with whatever method. If you manage to do that, forget about this silly discussion, work on refining the performance, packing it in a graphical user interface, selling it and achieving fame as the guy who made the first proper emulation of a best-selling classic analog synthesizer.
"Music is spiritual. The music business is not." - Claudio Monteverdi
-
- KVRAF
- 12615 posts since 7 Dec, 2004
Here is an example with a few variations of an "ultra-smooth" extra crunchy saturation.
http://xhip.net/temp/lowpass_sat_b.flac
It's just really super difficult to tweak it to be exactly like your example without having exact measurements from the exact synthesizer. I can get my sh-09 perfectly recreated no problem, but it has far better linearity than the IR3109 in the sh-101 and mc-202 and other synths.
(I don't own a sh-101, are you kidding? At $1500+?
I've heard lots of sh-101 examples online and I've heard vast differences from your examples, unfortunately. I doubt the sh-101 is anywhere near that stable in variations of the same batch, let alone between countries and years apart.)
I was thinking I could actually +12db the gain into the sh-09 and adjust the diode clipper to see if I can get the same amount of saturation, although the IR3109 is going to be different than the two lm13700 or four ba662 because they're simply different circuits. The saturation of the lm13700 is easy to simulate using two tanh-ish functions (the approximation of tanh is better than tanh at approximating the lm13700! lol) and I do have some IR3109 and ba662, but they're kind of "in use" and I don't really want to rip apart synthesizers that are working just fine as they are.
The idea of refining performance is not going to work. It isn't possible without oversampling and I have the most efficient decimation filter possible, 10x faster than the usual "high speed" polyphase filters and way better performance.
So much computation goes into the filter itself too that it just doesn't make sense to "optimize" anything. At best it might be possible to run stereo or quad filters using SSE but optimization usually works because you're doing more computation than you have to. With a working filter there is a certain bare minimum that can not be reduced further. A filter like this is simply a complicated thing, even at the bare metal.
http://xhip.net/temp/lowpass_sat_b.flac
It's just really super difficult to tweak it to be exactly like your example without having exact measurements from the exact synthesizer. I can get my sh-09 perfectly recreated no problem, but it has far better linearity than the IR3109 in the sh-101 and mc-202 and other synths.
(I don't own a sh-101, are you kidding? At $1500+?
I've heard lots of sh-101 examples online and I've heard vast differences from your examples, unfortunately. I doubt the sh-101 is anywhere near that stable in variations of the same batch, let alone between countries and years apart.)I was thinking I could actually +12db the gain into the sh-09 and adjust the diode clipper to see if I can get the same amount of saturation, although the IR3109 is going to be different than the two lm13700 or four ba662 because they're simply different circuits. The saturation of the lm13700 is easy to simulate using two tanh-ish functions (the approximation of tanh is better than tanh at approximating the lm13700! lol) and I do have some IR3109 and ba662, but they're kind of "in use" and I don't really want to rip apart synthesizers that are working just fine as they are.
The idea of refining performance is not going to work. It isn't possible without oversampling and I have the most efficient decimation filter possible, 10x faster than the usual "high speed" polyphase filters and way better performance.
So much computation goes into the filter itself too that it just doesn't make sense to "optimize" anything. At best it might be possible to run stereo or quad filters using SSE but optimization usually works because you're doing more computation than you have to. With a working filter there is a certain bare minimum that can not be reduced further. A filter like this is simply a complicated thing, even at the bare metal.
Last edited by aciddose on Sun Dec 15, 2013 6:48 pm, edited 1 time in total.
Free plug-ins for Windows, MacOS and Linux. Xhip Synthesizer v8.0 and Xhip Effects Bundle v6.7.
The coder's credo: We believe our work is neither clever nor difficult; it is done because we thought it would be easy.
Work less; get more done.
The coder's credo: We believe our work is neither clever nor difficult; it is done because we thought it would be easy.
Work less; get more done.
-
- KVRAF
- 5139 posts since 27 Jun, 2004
aciddose: ah, I thought I read that you had one, weird..
You know, I couldn't care less about SH-101 compared to your SH-09-based freaking fantastic sounding synth. But even with that design that you know every bit of, you can't recreate it perfectly except in some static settings or some specific dynamic settings. What about your old "match_this" sample? Even with the single pole filter, I haven't heard anything similar enough. If you can do that by now, I'd love to hear it. And in any case, it's too bad that none of that can be done in real-time...
You know, I couldn't care less about SH-101 compared to your SH-09-based freaking fantastic sounding synth. But even with that design that you know every bit of, you can't recreate it perfectly except in some static settings or some specific dynamic settings. What about your old "match_this" sample? Even with the single pole filter, I haven't heard anything similar enough. If you can do that by now, I'd love to hear it. And in any case, it's too bad that none of that can be done in real-time...
"Music is spiritual. The music business is not." - Claudio Monteverdi
-
- KVRAF
- 12615 posts since 7 Dec, 2004
I'm doing it in real time ATM, it just doesn't make that much sense dedicating a whole core to one voice.
"match this" would have been my own filter based on the KORG35 in bandpass mode. That one is ridiculously difficult to emulate because not only are there many different non-linear portions but in addition you have a ridiculously high level of harmonics from the combination (mega oversampling?) plus frequency modulation. That said, if you did want to dedicate the insane amount of processing power to it that it requires I see no reason it isn't possible, other than the fact it's so damn complicated nobody has managed to come close yet.
The IR3109 and other OTA-based filters however are not so complex. You need a few tanh-like functions to model the effect of each differential amplifier + current mirror and more importantly you only need to model one stage and the diode clipping of the feedback. Every stage is identical, so it isn't difficult to get very close very quickly. It's really very similar to the famous moog filter in that respect.
The ba662 filter in the sh-09 is extremely easy because it's just so linear. You need a small amount of saturation in the signal path and the usual diode clipper in feedback and boom, you have the three (1 even, 2 odd) harmonics you need. They don't really drive the filter very hard either, which is why I've been meaning to modify the thing for ages to at least 4x the inputs which would get it a lot closer to the sh-101 for sure although I can't say how close without actually doing it.
If you really want to see a good IR3109 model you might try asking Urs to invest some effort there. They have all the components in place to just add it as another filter model. Actually, don't they already have one? I'd be a little surprised if not.
http://xhip.net/temp/lowpass_sat_c.flac
Slight improvement. Still needs tweaking, there are an awful lot of variables.
- Stage saturation (intensity, shape, symmetry)
- Feedback diode clipping
- Feedback level (Higher 'feedback drive' causes feedback to 'fill out gaps', eliminates the low/high level variation you mentioned)
- Input level (Higher level, lower resonance, higher sync)
It's pretty close to perfect now though. I really could struggle for hours but without those examples I mentioned it's really hit and miss.
Optimized it to get ten voices. Is this anywhere near acceptable? Ten voices? What is this, 1999? Oh wait, I was in debug. I mean 36.
"match this" would have been my own filter based on the KORG35 in bandpass mode. That one is ridiculously difficult to emulate because not only are there many different non-linear portions but in addition you have a ridiculously high level of harmonics from the combination (mega oversampling?) plus frequency modulation. That said, if you did want to dedicate the insane amount of processing power to it that it requires I see no reason it isn't possible, other than the fact it's so damn complicated nobody has managed to come close yet.
The IR3109 and other OTA-based filters however are not so complex. You need a few tanh-like functions to model the effect of each differential amplifier + current mirror and more importantly you only need to model one stage and the diode clipping of the feedback. Every stage is identical, so it isn't difficult to get very close very quickly. It's really very similar to the famous moog filter in that respect.
The ba662 filter in the sh-09 is extremely easy because it's just so linear. You need a small amount of saturation in the signal path and the usual diode clipper in feedback and boom, you have the three (1 even, 2 odd) harmonics you need. They don't really drive the filter very hard either, which is why I've been meaning to modify the thing for ages to at least 4x the inputs which would get it a lot closer to the sh-101 for sure although I can't say how close without actually doing it.
If you really want to see a good IR3109 model you might try asking Urs to invest some effort there. They have all the components in place to just add it as another filter model. Actually, don't they already have one? I'd be a little surprised if not.
http://xhip.net/temp/lowpass_sat_c.flac
Slight improvement. Still needs tweaking, there are an awful lot of variables.
- Stage saturation (intensity, shape, symmetry)
- Feedback diode clipping
- Feedback level (Higher 'feedback drive' causes feedback to 'fill out gaps', eliminates the low/high level variation you mentioned)
- Input level (Higher level, lower resonance, higher sync)
It's pretty close to perfect now though. I really could struggle for hours but without those examples I mentioned it's really hit and miss.
Optimized it to get ten voices. Is this anywhere near acceptable? Ten voices? What is this, 1999? Oh wait, I was in debug. I mean 36.
Free plug-ins for Windows, MacOS and Linux. Xhip Synthesizer v8.0 and Xhip Effects Bundle v6.7.
The coder's credo: We believe our work is neither clever nor difficult; it is done because we thought it would be easy.
Work less; get more done.
The coder's credo: We believe our work is neither clever nor difficult; it is done because we thought it would be easy.
Work less; get more done.
-
- KVRist
- 363 posts since 4 Jul, 2007
erm, what?Delfinoverde wrote:Best Softsynths are:
- Alchemy
- Absynth 4 or 5
- GForce ImpOSCar 2
- Rob Papen BLUE
U-he synths are not in my best of list.
I find they sound artificial, not authentic.
----------------------------------------------
Just a couple of points -
- apart from Imposcar, none of your favourite synths (Alchemy etc) are striving for 'authenticity' - they are deliberately 'artificial' (in a good way)
- none of them were release in 2013
and
As a long time OTB/analogue fanatic (have owned various Moogs, SE Code, Dave Smith etc), to call U-he synths 'artificial' is surprising to say the least. I for one don't mind admitting I can't tell the difference between Diva and the analogue gear it is designed to emulate.