ZL Compressor

Have previosly written a review. Thank you so much for your detailed response! Further testing revealed some minor quirks of the plugin, so I want to address them here in case those are not intended.

• The "Vocal" style does not behave linearly as the manual says but exponentially and a bit differently from the "Classic" style (which is still interesting though).

• Reducing "Wet" signal to 50% or less gradually flattens peaks of reduction curve as if "Range" got reduced, instead of just weakening the effect without changing the curve shape of reduction.

Best regards.

P.S. I have realised negative ratio by expanding sidechain signal in another instance, so I don't mind its absence anymore! Thank you for your investigation.

Update: Lower "Ratio" flattens reduction curve as well.

1. Because of the feedback design, the *linearity* of "Vocal" won't hold during processing. A lots of parameters won't hold exactly for the feedback design, e.g., threshold, ratio, attack, release.

2. The "Wet" is directly applied on the reduction decibel values. It does not mix the input with the output. Mathematically they are equivalent but the former one is much more efficient. Therefore, you will see the changes on the reduction values. It might be a bit confusing at first but I think a lower CPU usage is worthy :).

I apologise for bothering you repeatedly (I'm enjoying the plugin overall and thank you for your response!). About the "Preparation" option I talked about before, I found an open-sourced compressor plugin with the function: Relief Compressor by yg331 (It's fixed at 0.5ms though). It sounded exactly the same as pajSoil EiS with matched settings in my test, so it should be implemented in the same or a similar way! In some cases, the current "pure" lookahead causes clicks before transient by steep reduction. I think such lookahead smoothing would be a perfect solution! I hope this helps. (I personally wouldn't mind if latency became much longer. Also, I imagine that smooth lookahead could make interesting effect when ridiculously long like 50−200ms lol).

By the way, it's silly of me talking about what I've said I didn't mind, but I've come up with an idea that it might make negative ratio work if "Curve" above 50 were ignored. I thought "Curve" 50 or below making the ratio for loud signal toward 1:1 would not be a problem in negative ratio, and the value 50 onward does not make any difference in ratio 100:1, so I guess it would be a natural behaviour to users.

I don't know if those are possible, so please dismiss them if not.

Best regards.

> It sounded exactly the same as pajSoil EiS with matched settings in my test, so it should be implemented in the same or a similar way.

According to the README and the code, the *Relief Compressor by yg331* applies a low-pass to the side-chain when lookahead is active. You can do it manually by using the side-chain equalizer in ZL Compressor.

> By the way, it's silly of me talking about what I've said I didn't mind, but I've come up with an idea that it might make negative ratio work if "Curve" above 50 were ignored. I thought "Curve" 50 or below making the ratio for loud signal toward 1:1 would not be a problem in negative ratio, and the value 50 onward does not make any difference in ratio 100:1, so I guess it would be a natural behaviour to users.

Everything comes with a cost :) If the plugin do so, it needs to do an additional *if* check per sample, which will hurt performance (more precisely, it is called branching).

Actually after some investigation, the RMS in ZL Compressor should be the most similar one, cause *Relief Compressor by yg331* uses FIR low-pass which should leak more energy into nearby, just like a RMS.

Thank you for your investigation and detailed explanation! I've just tested your suggestion (both low-pass and RMS), and yes, combination of both things works, but not as smoothly as the other plugins. Main problems are.

･ Clicks are still heard although much quieter.

･ Gain reduction is much weakened, so peaks come through.

･ Curve styles somewhat lose character depending on value of RMS Mix (I'm sure that's how RMS compression should work and it has its own charm :D).

About pajSoil, it is certain that it utilises FIR filter for lookahead, considering its behaviour, but I've also found that it also causes clicks nonetheless...

I tested Kilohearts Compactor as well, and it turns out, its lookahead function, "Attack" works differently from pajSoil: It makes no clicks or pops. It should not have something to do with FIR. It must be "fade-in of held peak" (with lookahead); its basis must be "Hold". Such "Swelling Hold" should make a nice pumping effect as itself with peaks retained! Is this one possible?

P.S. I thought attack (with or without much "Pump") could work like a fade-in of peak-hold, but it didn't. It has the same problems as RMS, failing to eliminate clicks and to attenuate peaks.

It is hard to fully attenuate peaks via attack/release. Perhaps you can try to adjust the 'Clipper' a little bit? You can also try the 'Hold' parameter, but it might not be what you want.

Thank you again for responding and even adding some of my suggestions! I'm HAPPY with their implementation! Also, the UI has become much more intuitive with the new peak meter! Although there's a bug that all the graphs are frozen for a while after the plugin window is opened, it's not a huge issue to me :D.

About your suggestion, sadly, both are indifferent to preventing click noises which I mind by far the most... I've tried all the parameters and combinations, but I'm sure there is currently no solution to that.