Despot - hard hitting ZDF compressor with antialiasing - development snapshots

VST, AU, AAX, CLAP, etc. Plugin Virtual Effects Discussion
RELATED
PRODUCTS

Post

I am having trouble moving the sliders in Ableton 9 (x86)
It's easy if you know how

Post

How exactly the problem manifests itself?

2camsr
Aliasing might appear innoticeable at first, but degrades audio in a subtle way, unlike instantly noticeable harmonic distortion and IMD. It appears normal at the first glance, but becomes indistict, weak and smeared. It was the reason for developing a compressor in the first place (first version wasn't ZDF at all and antialising was the main focus).

One not concerned by aliasing always can disable OS. It will still have somewhat lower aliasing than normal.

UPD: figured what could cause the mess with sliders with Ableton. Don't have the program itself to test, but 99% sure it is. It looks like it handles display updates improperly.

New build with temporary fix for Ableton:
http://stash.reaper.fm/v/24146/despot-1432341304.zip

Post

Most of the time, aliasing will sound terrible. Very clean leads (with low or no harmonics) will suffer the most from it. Some signals won't... have you listened for aliasing distortion via clipping white noise? :)
I run my DAW at 96khz and sometimes aliasing does appear. The effect sample rate has is more legroom for harmonics in the ultrasonic band.

Listened to the second version just now. Attack is very smooth, but the release, it seems a bit hyper. Jamming the unit to full input gain in soft knee mode, 90ms attack and 100ms release, some very large transients on the order of 15db or so would appear as the input magnitude decayed. Those transients sounded ok, but as a compressor goes, too much amplitude appeared at the output given the settings. Maybe it was wrong to use it in such extreme manner... thought it may be helpful.

Post

dekadenz wrote:UPD: figured what could cause the mess with sliders with Ableton. Don't have the program itself to test, but 99% sure it is. It looks like it handles display updates improperly.

New build with temporary fix for Ableton:
http://stash.reaper.fm/v/24146/despot-1432341304.zip
Working fine now :tu:
It's easy if you know how

Post

Glad I'm guessed correctly.

Listened to aliasing caused by extreme compression (on the edge on distortion) on different full-band material (from rock and electronics to solo and orchestral classical) in "pro" grade headphones with very good HF response. It's most noticeable on toppy instruments. You know, instant "digital top". Usually, 4x oversampling does the trick of pushing it low enough (if the compressor itself uses bandlimited or quasi-bandlimited functions). If not, even 16x doesn't help much.

Not as noticeable with 3-4 db of compression and slow release, but I prefer keeping it lower than theorethical hearing limit even if I can't hear it myself.

BTW release function it the same on hard knee and soft knee. Considered making different, less aggressive type of release as an option, but picked the most uncompromising that sounded good for the first try (will solve equations for different type of follower in future). I'm trying to make the hardest and the most relentless compressor imaginable, after all.

Soft knee wasn't meant for driving much, it already gives circa 3-4 dB GR when only hitting zero without drive. Hard knee is not 100% hard (as it will cause aliasing other way), it's more like 3 db knee.

People who deal mostly with digital processing usually used to aliasing and tend not to notice it. The ones who used to pre-digital music and clean analog hardware are very sensitive to it.

Personally, I'm trying listening to only "zero aliasing" music before making tests or no music at all.

Speaking of aliasing, it can even sound good and add "character" and useful overtones on basic signals, like sine or saw. On full-band signal it forms a solid signal dependent "grungey noise floor" (not like the harmless signal-independent noise).

P. S. To be fair, "analog" people has similar quirks too. Exactly the reason why said CLEAN analog and not just "analog". They tend to ignore harmonic distortion, even intolerable to normal people amounts, users of highly colored equipment (especially modern BS distorted tubes, unlike proper oldschool clean high headroom tube gear) and tape fans in particular.

Post

dekadenz wrote: P. S. To be fair, "analog" people has similar quirks too. Exactly the reason why said CLEAN analog and not just "analog". They tend to ignore harmonic distortion, even intolerable to normal people amounts, users of highly colored equipment (especially modern BS distorted tubes, unlike proper oldschool clean high headroom tube gear) and tape fans in particular.
Most people forget that tape machines alias too. :wink:

Post

Processor works fine, especially on drums\percussions but probably camsr can learn us something about this kind of processors... ;-)

Post

dekadenz wrote: Speaking of aliasing, it can even sound good and add "character" and useful overtones on basic signals, like sine or saw. On full-band signal it forms a solid signal dependent "grungey noise floor" (not like the harmless signal-independent noise).
It's a type of noise, better than white thermal noise IMO, because it is signal dependant, and it does not reduce dynamic range.

Post

Turello wrote:Processor works fine, especially on drums\percussions but probably camsr can learn us something about this kind of processors... ;-)
What do you mean? I've been experimenting with ZDF for a compressor, but it's not exactly conventional.

Post

I talk about your great plugins! ;-)

Post

camsr wrote: It's a type of noise, better than white thermal noise IMO, because it is signal dependant, and it does not reduce dynamic range.
Noise just reduces dynamic range, aliasing messes with the signal. Signal-dependent stuff should be MUCH lower than independent in order to be considered "harmless". Anything not entirely VST will have some amount of noise (clean analog has circa -90-100 dB floor). I'll take -90 dB noise over -90 dB aliasing any time of the year.

By the way my compressor does not have any thermal noise as well. Hard digital silence down to double precision resolution, so the user should add noise by oneself if needed.
Ichad.c wrote: Most people forget that tape machines alias too. :wink:
Isn't it just some form of intermodulation distortion with bias signal and not aliasing?

Post

Turello wrote:I talk about your great plugins! ;-)
Oh thank you. I know I mentioned a compressor to you somewhere on the forum, it will probably have to wait... developing plugins is a real PITA, not to mention the "hardcore mathematics". It doesn't help the cause that I have some presumptions about how a plugin should work. Thanks for the compliment.

Dekadenz
I would prefer no aliasing as well. But it's just not going to happen here. Sampling is plauged with aliasing. But you are being too hard on yourself and your compressor if you think those results are bad in any way. The difference is just not significant, almost any signal passed through your compressor will sound unaliased, and that's your goal. Aliasing will be there, the sound of aliasing won't.

Post

@camsr: Well, I really appreciated your weapons, I still work with your clippers and If you have created something new, please, let me know (also in PM)... Actually I works on Mastering Service and I'm glad to use something (brand\brave) new (breed), not only the same choices... Job said: "Think Different"... I think it's true cause my clients like my masters!

@dekadenz: Unfortunately I don't know mathematics, algorithms etc etc... But my advice is: don't stop working and don't stop believing on it... The plugin has ample margins for growth\expansion... ;-)

Post

2camsr
Don't think the results are bad in any way. Virtually all aliasing is below -140 dB on 176.4 and will be even lower with higher oversampling, which won't be heard in any way (though it would be worse on higher frequencies). Just compared to non-ZDF version that has pristine clean tests on the same settings. It shows that there's always some room for improvement, but this will require numerical methods (current version uses simpler functions that can be solved analytically). This numerical crunching will be less CPU friendly add will add some distortion on its own. Maybe it's better idea to invest in faster oversampling instead (compressor responds very well to oversampling unlike some that will alias no matter what the sample rate is).

There is enough compressors without any form of antialiasing or with just oversampling. Don't know, maybe someone will use it on anything non-sampled or at least samples not from nearest trap library. And I don't want to dissapoint in that case.

Even samples triggered from vintage sampler with lots of aliasing could benefit from alias-free compression, not every aliasing sounds the same, and these devices often has a certain "character" that will be spoiled by another aliasing on top.

Post

well, certainly every envelope follower can benefit from no audible resonance artifacts in modulation :)

one of my favourite compressors was in "nation of millions" - those sounds that are like squeaks. used to joke and say terminator x had a big box of gerbils and every 4 bars he'd take one out and do something with it.

but i wonder if anybody ever thinks about the squeaks and what he was doing it. some problem people parsing even things that very are simplest. then take advantage, and keep stupid, for the next opportunist advantage.

think, he said, and he also got in fights.
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.

Post Reply

Return to “Effects”