Feedback Compressor II - Gentleman's Edition (discussion)
- KVRian
- Topic Starter
- 1169 posts since 24 Feb, 2012
No, you don't need to (and shouldn't) route the output back to the side-chain. I wouldn't advice it's use for serious program compression in general, as any SC processing and/or delay can easily destroy the musical "balance" of the compressor.
I add it for special trickery, e.g. intentional pumping or other weird stuff. But I am no great fan of this idea as it's too easy to mess up everything with manual SC processing.
About the tilt filter, I have no idea how to implement a proper tilt filter efficiently. Those I saw where simple "shelvish" fakes (all with 6dB slopes!). But I'll ask Tony what he did.
A significant problem of the tilt filter idea is that most compressors don't like overshoots in the side-chain as they introduce very audible over-compression problems (as soon the side-chain signal surpasses the level of the actual content). That is, the level of the SC must be reduced accordingly. From my observations, SC processing for program material compression should be overshoot free and free of phase distortion/delay in order act musically on a really wide range of material. Every variance from this ideal directly reduces the range of use.
A wideband slope filter asks for very deeper thresholds, so any higher slope would have unreasonable effects on the user-interface responsibility and usability in general. It's always a difficult compromise.
I add it for special trickery, e.g. intentional pumping or other weird stuff. But I am no great fan of this idea as it's too easy to mess up everything with manual SC processing.
About the tilt filter, I have no idea how to implement a proper tilt filter efficiently. Those I saw where simple "shelvish" fakes (all with 6dB slopes!). But I'll ask Tony what he did.
A significant problem of the tilt filter idea is that most compressors don't like overshoots in the side-chain as they introduce very audible over-compression problems (as soon the side-chain signal surpasses the level of the actual content). That is, the level of the SC must be reduced accordingly. From my observations, SC processing for program material compression should be overshoot free and free of phase distortion/delay in order act musically on a really wide range of material. Every variance from this ideal directly reduces the range of use.
A wideband slope filter asks for very deeper thresholds, so any higher slope would have unreasonable effects on the user-interface responsibility and usability in general. It's always a difficult compromise.
Fabien from Tokyo Dawn Records
Check out my audio processors over at the Tokyo Dawn Labs!
Check out my audio processors over at the Tokyo Dawn Labs!
-
- KVRer
- 23 posts since 18 Jun, 2013
I can appreciate the need to avoid overshoot and phase distortion/delay in the SC filter (with peak detection), so I now understand why using the external SC filter is a bad idea except for tricks like ducking and noise gating. Therefore, we users are depending on the compressor to provide flexibility in the internal SC filter (while avoiding overshoot and phase distortion/delay). Hope your investigation into the head model, the tilt filter and the 3dB/Oct wideband filter produces some industry-leading flexibility.
I really like your idea of the 3dB/Oct wideband SC filter. It occurred to me that this filter (or any other) could be faked with existing compressors (feedback or feedforward) by putting the desired filter before the compressor, and then the inverse filter after the the compressor. What do you think? It is inconvenient and tricky to manually adjust and keep these two filters as inverses, but it eliminates the SC filter and any SC filter overshoot and phase distortion/delay. It doesn't matter whether the pre and post filters have overshoot and phase distortion/delay as long as the inverse relationship holds. I am using REAPER, which has 64-bit audio paths. As long as the compressor plug-in supports 64-bit audio I/O (such as the original TDR Feedback Compressor), the sound quality may be just as good as using an internal SC filter. I hope the GE edition will have 64-bit audio I/O (in 64-bit audio hosts) so that users have the flexibility to use this trick (and for better sound quality in general).
Maybe this pre+post filter idea could give you a new way to effectively implement the SC filter with complete flexibility (no worry if there is overshoot and phase distortion/delay in the pre and post filters). The user controls it like a traditional SC filter (easy to understand), but the actual SC filter is eliminated (along with artifacts caused by SC filter overshoot and phase distortion/delay). In fact, it may sound better than your existing SC filters, which do have some phase distortion. You can ensure that the pre and post filter are always maintained as inverses, so it would be as convenient as today's compressors. What do you think?
I really like your idea of the 3dB/Oct wideband SC filter. It occurred to me that this filter (or any other) could be faked with existing compressors (feedback or feedforward) by putting the desired filter before the compressor, and then the inverse filter after the the compressor. What do you think? It is inconvenient and tricky to manually adjust and keep these two filters as inverses, but it eliminates the SC filter and any SC filter overshoot and phase distortion/delay. It doesn't matter whether the pre and post filters have overshoot and phase distortion/delay as long as the inverse relationship holds. I am using REAPER, which has 64-bit audio paths. As long as the compressor plug-in supports 64-bit audio I/O (such as the original TDR Feedback Compressor), the sound quality may be just as good as using an internal SC filter. I hope the GE edition will have 64-bit audio I/O (in 64-bit audio hosts) so that users have the flexibility to use this trick (and for better sound quality in general).
Maybe this pre+post filter idea could give you a new way to effectively implement the SC filter with complete flexibility (no worry if there is overshoot and phase distortion/delay in the pre and post filters). The user controls it like a traditional SC filter (easy to understand), but the actual SC filter is eliminated (along with artifacts caused by SC filter overshoot and phase distortion/delay). In fact, it may sound better than your existing SC filters, which do have some phase distortion. You can ensure that the pre and post filter are always maintained as inverses, so it would be as convenient as today's compressors. What do you think?
- KVRian
- Topic Starter
- 1169 posts since 24 Feb, 2012
The pre/de-emphasis idea is pretty useful, but has its drawback, too. From my experience, the approach is particularly well suited for all kinds of color stuff: saturation most of all, but also dynamics. Especially for obvious color.
However, running a pre and post filter isn't "clean" at all. The SC filter approach is 100% clean, as it cannot directly affect the audio signal, the latter is 100% "protected". While the mirrored filter re-constructs the original phase, the distortion itself is still phase distorted. And it does something else as well: Classic pre/de emphasis schemes such as the ones used for tape, vinyl and broadcast (i.e. pre high freq boost, post high frequency attenuation) are known to introduce severe problems in dynamics processing. They tend to boost the unharmonic IM distortion introduced by the processing by unreasonable amounts (at very audible frequency regions).
To be clear: Such an approach in probably just the right thing for a color device (especially with saturating pre-de-emphasis filters ). But this specific compressor is built to preserve the original contrast and harmonic relation as carefully as possible, while still delivering effective dynamic range reduction. I always evaluate several approaches before I decide which one to implement.
IMHO, the best approach to frequency dependent compression (in the range of the classic SC HP filter approach) is a linear phase split, where only the upper band is compressed. Sadly, linear phase filters at such low frequencies are either cpu heavy or very difficult to build and ask for several hundred ms of additional latency. But who knows, I will maybe implement such an options in version III
Don't worry about the 32bit I/Os. They are generally absolutely fine, even for several dozen plugins in series. The parts that really matter are the recursive elements of the algorithms. These may ask for higher precisions to deliver accurate results, but the "public" I/O precision doesn't really matter (at least not at this high level of precision).
The reasoning behind is that I had to unify the development of all supported formats and versions to save time and increase the overall testing and production times and stability.
However, running a pre and post filter isn't "clean" at all. The SC filter approach is 100% clean, as it cannot directly affect the audio signal, the latter is 100% "protected". While the mirrored filter re-constructs the original phase, the distortion itself is still phase distorted. And it does something else as well: Classic pre/de emphasis schemes such as the ones used for tape, vinyl and broadcast (i.e. pre high freq boost, post high frequency attenuation) are known to introduce severe problems in dynamics processing. They tend to boost the unharmonic IM distortion introduced by the processing by unreasonable amounts (at very audible frequency regions).
To be clear: Such an approach in probably just the right thing for a color device (especially with saturating pre-de-emphasis filters ). But this specific compressor is built to preserve the original contrast and harmonic relation as carefully as possible, while still delivering effective dynamic range reduction. I always evaluate several approaches before I decide which one to implement.
IMHO, the best approach to frequency dependent compression (in the range of the classic SC HP filter approach) is a linear phase split, where only the upper band is compressed. Sadly, linear phase filters at such low frequencies are either cpu heavy or very difficult to build and ask for several hundred ms of additional latency. But who knows, I will maybe implement such an options in version III
Don't worry about the 32bit I/Os. They are generally absolutely fine, even for several dozen plugins in series. The parts that really matter are the recursive elements of the algorithms. These may ask for higher precisions to deliver accurate results, but the "public" I/O precision doesn't really matter (at least not at this high level of precision).
The reasoning behind is that I had to unify the development of all supported formats and versions to save time and increase the overall testing and production times and stability.
Fabien from Tokyo Dawn Records
Check out my audio processors over at the Tokyo Dawn Labs!
Check out my audio processors over at the Tokyo Dawn Labs!
-
- KVRer
- 23 posts since 18 Jun, 2013
The pre/post filter idea may be useful for an RMS-only controlled compressor (slow-acting), but I see your point about the distortion sounding wrong when using peak-based control.
Linear-phase filters create a pre-echo type of distortion, so they have their problems too. Linear phase may be OK in a SC filter, but my listening tests suggest that they should be avoided in the main audio path. However, you could put a delay line in the main audio path to delay match a linear-phase SC filter.
My reading on psychoacoustics research (https://en.wikipedia.org/wiki/Jens_Blauert) suggests that humans are not sensitive to phase distortion at higher frequencies. Thus, we don't hear the phase distortion of some minimum-phase filters (which include all-pole filters). It's been a number of years since I read Blauert's book, so I forget at what frequency that humans start to become insensitive to phase. Hope this bit of psychoacoustics info gives you some extra options for designing a SC filter that subjectively optimizes your goal ("preserve the original contrast and harmonic relation as carefully as possible, while still delivering effective dynamic range reduction").
Linear-phase filters create a pre-echo type of distortion, so they have their problems too. Linear phase may be OK in a SC filter, but my listening tests suggest that they should be avoided in the main audio path. However, you could put a delay line in the main audio path to delay match a linear-phase SC filter.
My reading on psychoacoustics research (https://en.wikipedia.org/wiki/Jens_Blauert) suggests that humans are not sensitive to phase distortion at higher frequencies. Thus, we don't hear the phase distortion of some minimum-phase filters (which include all-pole filters). It's been a number of years since I read Blauert's book, so I forget at what frequency that humans start to become insensitive to phase. Hope this bit of psychoacoustics info gives you some extra options for designing a SC filter that subjectively optimizes your goal ("preserve the original contrast and harmonic relation as carefully as possible, while still delivering effective dynamic range reduction").
-
- KVRAF
- 2973 posts since 10 Sep, 2003 from Karlskoga, Stockholm, Sweden
Is this out yet..?
Struggling to learn all my compressors, so one more can't hurt. Right?
Looking forward to it!
Struggling to learn all my compressors, so one more can't hurt. Right?
Looking forward to it!
- KVRian
- Topic Starter
- 1169 posts since 24 Feb, 2012
@Crackbaby: I can guarantee that the GE edition won't hurt. No release date yet.
@Blipper: Of course, I'm constantly looking for inspiration, your input is highly appreciated.
@Blipper: Of course, I'm constantly looking for inspiration, your input is highly appreciated.
Fabien from Tokyo Dawn Records
Check out my audio processors over at the Tokyo Dawn Labs!
Check out my audio processors over at the Tokyo Dawn Labs!
- Banned
- 10196 posts since 12 Mar, 2012 from the Bavarian Alps to my feet and the globe around my head
I've heard that too much feedback compression can cause headaches. Is it true?FabienTDR wrote:I can guarantee that the GE edition won't hurt.
- KVRian
- Topic Starter
- 1169 posts since 24 Feb, 2012
I've heard all good things cause headache sooner or later, women, party...Tricky-Loops wrote:I've heard that too much feedback compression can cause headaches. Is it true?FabienTDR wrote:I can guarantee that the GE edition won't hurt.
Fabien from Tokyo Dawn Records
Check out my audio processors over at the Tokyo Dawn Labs!
Check out my audio processors over at the Tokyo Dawn Labs!
-
- KVRist
- 186 posts since 31 Jul, 2013
Wow, what a terrific compressor. I love what it does to the bass on a mix, love the clarity, love the lack of phony analogue-ism. This is really a great piece of work--I'm looking forward to learning how to make better use of it. I will gladly pay for a gentleman's edition when it comes out, just to acknowledge the high quality of this plugin
-
Hermetech Mastering Hermetech Mastering https://www.kvraudio.com/forum/memberlist.php?mode=viewprofile&u=7418
- KVRAF
- 1619 posts since 30 May, 2003 from Milan, Italy
- KVRian
- Topic Starter
- 1169 posts since 24 Feb, 2012
It grew over the months, no worries!
Vlad is currently building a powerful linear phase crossover for frequency selective compression and I'm still experimenting with improvements in the compression DSP.
Additionally, we're almost ready with our new plug-in "bed". Visually, it's some sort of toolbar, but it adds tons of great features such as Undo/Redo, User Presets, A/B, copy/paste, neat installers and several stability improvements and "historical" bugs fixes. It took us a long time, but it's worth it. We'll never have to do that work again. Of course, this affects all our previous plugins, too!
Really sorry for all the delays, I had a very busy year.
Vlad is currently building a powerful linear phase crossover for frequency selective compression and I'm still experimenting with improvements in the compression DSP.
Additionally, we're almost ready with our new plug-in "bed". Visually, it's some sort of toolbar, but it adds tons of great features such as Undo/Redo, User Presets, A/B, copy/paste, neat installers and several stability improvements and "historical" bugs fixes. It took us a long time, but it's worth it. We'll never have to do that work again. Of course, this affects all our previous plugins, too!
Really sorry for all the delays, I had a very busy year.
Fabien from Tokyo Dawn Records
Check out my audio processors over at the Tokyo Dawn Labs!
Check out my audio processors over at the Tokyo Dawn Labs!
- KVRian
- 648 posts since 6 Nov, 2011 from The Netherlands
Take all the time you need, it's ready when it's ready. And you will get our money, no questions asked.FabienTDR wrote:It grew over the months, no worries!
Vlad is currently building a powerful linear phase crossover for frequency selective compression and I'm still experimenting with improvements in the compression DSP.
Additionally, we're almost ready with our new plug-in "bed". Visually, it's some sort of toolbar, but it adds tons of great features such as Undo/Redo, User Presets, A/B, copy/paste, neat installers and several stability improvements and "historical" bugs fixes. It took us a long time, but it's worth it. We'll never have to do that work again. Of course, this affects all our previous plugins, too!
Really sorry for all the delays, I had a very busy year.
Untold Stories Vol.1 - 64 Arturia MiniFreak presets
Analog History - 84 Behringer DeepMind 6/12/12D presets
Earth & Stars - 139 Free Patches for SuperMassive
Website
Analog History - 84 Behringer DeepMind 6/12/12D presets
Earth & Stars - 139 Free Patches for SuperMassive
Website
- KVRian
- Topic Starter
- 1169 posts since 24 Feb, 2012
Thanks man!
Many of you requested frequency-selective options, here's a UI idea we're working on...
Many of you requested frequency-selective options, here's a UI idea we're working on...
Fabien from Tokyo Dawn Records
Check out my audio processors over at the Tokyo Dawn Labs!
Check out my audio processors over at the Tokyo Dawn Labs!
-
Hermetech Mastering Hermetech Mastering https://www.kvraudio.com/forum/memberlist.php?mode=viewprofile&u=7418
- KVRAF
- 1619 posts since 30 May, 2003 from Milan, Italy
Thanks for the update Fabien! That's all great to know! Make sure you get some rest over Yuletide!