Despot - hard hitting ZDF compressor with antialiasing - development snapshots
- KVRian
- 1091 posts since 8 Feb, 2012 from South - Africa
Is the envelope follower in a feedback compressor arrangement? Otherwise ZDF would not make any/much difference actually, except maybe if you want a better diode curve approximation, but in the feedforward compressor case it is just easier to solve it explicitly with the Lambert_W function, if for practical purposes you only use the upper branch Lambert_W0 it is pretty doable.
Cool project! Looking forward to using it once all the kinks are ironed out.
Cheers
Andrew
Cool project! Looking forward to using it once all the kinks are ironed out.
Cheers
Andrew
-
- KVRist
- Topic Starter
- 40 posts since 16 May, 2015
It has feedforward topology. When doing this, wanted to test the concept, so picked something that's easier to solve, besides more "modern" and aggressive compression. Thought it won't make much difference, but was clearly wrong. It's subtle, but it's present. Compare with JSFX version if interested. It has different functions, but will be sufficient anyway.
If really interested, I'll add non-ZDF mode in the next snapshot.
Once I will finish this project, looking forward for more advanced tasks, like feedback compression and higher order filters. They will most likely require numerical methods and number crunching.
If really interested, I'll add non-ZDF mode in the next snapshot.
Once I will finish this project, looking forward for more advanced tasks, like feedback compression and higher order filters. They will most likely require numerical methods and number crunching.
- KVRian
- 1091 posts since 8 Feb, 2012 from South - Africa
Would be fun, but not compulsory. Btw, did you solve this analytically? The furthest I came with envelope followers was with step invariant transform + simple waveshaping. But I've done some ZDF stuff for synth vcfs and phasers.dekadenz wrote: If really interested, I'll add non-ZDF mode in the next snapshot.
Feedback-comp seems insane, andy(cytomic) is the only one I know of who has done it that way, but there surely must be some others.dekadenz wrote: Once I will finish this project, looking forward for more advanced tasks, like feedback compression and higher order filters. They will most likely require numerical methods and number crunching.
Cheers
Andrew
-
- KVRist
- Topic Starter
- 40 posts since 16 May, 2015
New build was released.
Now with adjustable attack and release and any sample rates support. Make sure you're tried low-aliasing zero attack.
By the way, solving the general case for arbitrary attack and release was really hardcore. One typo got into huge equations and it took long to figure out what was wrong. Finally fixed, enjoy.
Future version will have less limited attack/release and nonlinear sliders for easy dialing. Without nonliear I could not make the scale too wide without making it inсonvinient.
Now with adjustable attack and release and any sample rates support. Make sure you're tried low-aliasing zero attack.
Indeed. True TPT/ZDF and no trickery. I think I'll make separate non-ZDF version in order to avoid adding extra option and confusing users. Developed ZDF envelope follower not for disabling it.Ichad.c wrote: Would be fun, but not compulsory. Btw, did you solve this analytically? The furthest I came with envelope followers was with step invariant transform + simple waveshaping. But I've done some ZDF stuff for synth vcfs and phasers.
By the way, solving the general case for arbitrary attack and release was really hardcore. One typo got into huge equations and it took long to figure out what was wrong. Finally fixed, enjoy.
Future version will have less limited attack/release and nonlinear sliders for easy dialing. Without nonliear I could not make the scale too wide without making it inсonvinient.
-
penguinfromdeep penguinfromdeep https://www.kvraudio.com/forum/memberlist.php?mode=viewprofile&u=193898
- KVRAF
- 1993 posts since 18 Nov, 2008
I think ZDF should become stable in any plug-in where it's possible
circuit modeling and 0-dfb filters are cool
-
- KVRist
- Topic Starter
- 40 posts since 16 May, 2015
Regarding aliasing: did a sine test on zero attack and 100 ms release on 176.4 kHz.
Not pristine like older non-ZDF version (because of the simplified functions), but I think it's acceptable for early dev version. Many hyped and expensive plugins has much, much worse aliasing in fact, even with oversampling.
With current situation, it's far more viable to develop GUI for peanuts, strap some 60's digital filters and sell for 20$.
A plugin with ZDF should cost 100$ minimum, not 20.
Not pristine like older non-ZDF version (because of the simplified functions), but I think it's acceptable for early dev version. Many hyped and expensive plugins has much, much worse aliasing in fact, even with oversampling.
Considering the complexity of development and completely dampened market, it's impossible. Only DSP aficionados doing someting new without money involved.penguinfromdeep wrote:I think ZDF should become stable in any plug-in where it's possible
With current situation, it's far more viable to develop GUI for peanuts, strap some 60's digital filters and sell for 20$.
A plugin with ZDF should cost 100$ minimum, not 20.
-
- Banned
- 12368 posts since 30 Apr, 2002 from i might peeramid
nah ZDF is still $200 material.
they should also be standardised every place you don't understand them, but are certainly willing to pay for them.
they should also be standardised every place you don't understand them, but are certainly willing to pay for them.
you come and go, you come and go. amitabha neither a follower nor a leader be tagore "where roads are made i lose my way" where there is certainty, consideration is absent.
-
- KVRist
- Topic Starter
- 40 posts since 16 May, 2015
Before I'll release the next build with nonlinear attack-release scale and two knees - soft and hard, I want some advice from potential users.
Is such a table is ok?
Is such a table is ok?
Code: Select all
0 - 0.01 ms - 0.001 ms step
0.01 - 0.1 ms - 0.01
0.1 - 1 ms - 0.05
1 - 10 ms - 0.5
10 - 100 ms - 5
100 - 1000 ms - 10
1000 - 3000 ms - 100
3000 - 6000 ms - 500
- KVRAF
- 3303 posts since 6 Jul, 2012 from Sick-cily
Have You updated?
-
- KVRist
- Topic Starter
- 40 posts since 16 May, 2015
No, waiting for for some replies. Not sure if this attack-release scale is ideal and covers everything. Any ideas?
-
- KVRist
- Topic Starter
- 40 posts since 16 May, 2015
So, no suggestions. Fine, left it to my tastes.
Now, almost everything is in place, only oversampling is left for the future build. It will be HQ linear phase oversampling with user selectable ratios from 2x and up, no quality/latency compromises, which means the latency will be pretty high with oversampling enabled (0 without). Not an issue for latency-compensated hosts, but not suitable for real-time live use with oversampling.
On the bright side, linear phase and latency compensation means it can be used for parallel compression and won't cause any phase issues.
A new build is out.
1432326119
- stepped nonlinear attack/release scale
- hard and soft knee
Notes: quieter signal in soft knee mode is not a bug. This is 100% soft knee, it begins compressing very early (even from zero signal), so the drive should be backed off if you want less compression.
http://stash.reaper.fm/v/24139/despot-1432326119.zip
Now, almost everything is in place, only oversampling is left for the future build. It will be HQ linear phase oversampling with user selectable ratios from 2x and up, no quality/latency compromises, which means the latency will be pretty high with oversampling enabled (0 without). Not an issue for latency-compensated hosts, but not suitable for real-time live use with oversampling.
On the bright side, linear phase and latency compensation means it can be used for parallel compression and won't cause any phase issues.
A new build is out.
1432326119
- stepped nonlinear attack/release scale
- hard and soft knee
Notes: quieter signal in soft knee mode is not a bug. This is 100% soft knee, it begins compressing very early (even from zero signal), so the drive should be backed off if you want less compression.
http://stash.reaper.fm/v/24139/despot-1432326119.zip
-
- KVRAF
- 7420 posts since 17 Feb, 2005
Compressors can get technical, maybe try posting the question in the DSP subforum.
I wouldn't worry so much about aliasing, most material with a broad frequency spectrum will mask most if not all of the sound of aliasing, if it's at a low enough level. You should also consider CPU load, having large delays is sure to slow things down.
I wouldn't worry so much about aliasing, most material with a broad frequency spectrum will mask most if not all of the sound of aliasing, if it's at a low enough level. You should also consider CPU load, having large delays is sure to slow things down.