New Molot version with new limiter and larger GUI is almost ready but now some words about compression [или немного про компрессию]

In this post I want to show how the Molot compressor works and show visual difference between sigma and alpha modes, harder and softer knees and changes between 0.2.130 and 0.2.121 which I want to revert in the next release.

 В этой заметке будет показано, как работает компрессор “Молот”, будут показаны визуальные отличия между режимами “сигма” и “альфа”, мягким и грубым переходом и отличия между версиями 0.2.130 и 0.2.121, которые я собираюсь вернуть к старому варианту в следующей версии.

First get such kind of test impulse signal. 5 kHz of -3 dB for 1 sec and -12 dB for 1 sec.

Next apply Molot compressor with Alpha mode, -12 dB of threshold and attack and release set to maximum values. We can get something like this:

Now change the mode to Sigma. Now we can get this:

See the difference! In Sigma mode the attack of signal is soft and like a “S” letter. It means the most of attack will come through.

Now let’s setup more extreme settings. -40 dB of threshold and minimum ratio of 1:1.1. Look at this for alpha and sigma modes:

Now I think you got it. Let’s go deeper. How does the compressor work? First it builds the signal envelope. For alpha mode it uses very small attack and release independent of what you have set with knobs.

As you can see the attack is so small so envelop starts to “jump” on signal’s wave. In sigma mode the attack and release of initial envelope is much greater and depends of compressor’s attack and release. So we can get very soft envelope curve:

You can see its attack and release on picture. Next the signal envelope is converted to gain reduction by checking threshold and knee. For alpha mode we have this kind of gain reduction:

You can see it started at 0 dB then falls down to -6 dB and “jumps” on wave but then… What the…! Why gain reduction doesn’t reduce but amplify signal for 1 dB? Ho-ho-ho! This is the “feature” of harder knee. I’ll show it a big later in details. If you set knee knob to softer value the gain reduction curve looks like this:

So we have -1 dB of gain reduction for -12 dB signal due to the softness of knee. Now let’s see the same in sigma mode for harder knee:

And for softer knee:

And finally the compressor’s attack and release are applied to gain reduction curve to take something like this in alpha mode:

Hmm, the impulse at the beginning of signal is not expected. Seems to be a bug and should be fixed… In sigma mode see S-like transients:

And now the same for softer knee in alpha and sigma modes:

Boring? I don’t think so. Now about sound changes between 0.2.121 and 0.2.130. The idea was to make compressor sound a bit more transparent so the release has become a bit shorter for small signal values. There’s no visible difference in curves above between versions but for -40 dB and 1:1.1 threshold there’s some differences.

In old version with -40 dB of threshold there’s almost no visible difference between soft and hard knees and in new version in soft knee mode too: 

But in new version in hard knee the curve looks like this:

See that? The compressor does some amplification for low-level signals. A bit unexpected for me so making these pictures was a good idea.

Now try to process some real signal. An ordinal kick taken from real drum record:

Now try to use default “kick” preset for processing it (got sigma mode). The envelope looks like this:

Of course default “kick” preset got sidechain filtering turned on. If turned off the envelope looks like this:

And just for example I’ll show this envelope in alpha mode:

If you remember the next step is to get gain reduction. See those curves:

And finally the curves with attack and release applied:

After those processings final kick (in sigma mode) got this image:

See that nasty peak? Now you can imagine why I added peak limiter. Apply it to make peak a bit smaller!

But if you want to use the compressor in alpha mode for such a kick you don’t need peak limiter because there’s no peak at all:

I think you’re tired. But I’m almost completed. Do you remember that strange signal level increase when harder knee was used? See this diagram with compressor knees (for 4:1 ratio):

They set for thresholds of -10 dB, -20 dB and -40 dB. In harder knee (left curves) there’s a small signal level up just near the thresholds. Trust me it sounds cool! 🙂

And finally about new version’s sound differences related with release curve. For extreme settings such as -40 dB of threshold and 10:1 or limiting mode the curves look like that:

But for new version they look very strange. In such a way:

How ugly it is! And this is one of the reasons for I returned release curve back. And next version will be old-sounded. I’ll release it in couple of days!

Also I implemented new intelligent soft limiter with program-depended attack and release. It’s not ideal but much more better than old one. See you!

8 responses to “New Molot version with new limiter and larger GUI is almost ready but now some words about compression [или немного про компрессию]

  1. Alteregoxxx 2011/02/11 at 19:11

    Really interesting Vladg!…the last graphic shows a so called “negative ratio”, a feature of some DBX compressors…from what i remember a negative ratio is simply something like 4 : -1 (4 to minus 1)…for signals 4 dB over the threshold the processed sounds come out with a value equal to [ threshold – 1 dB]
    It would be interesting to illustrate and explain whether and how Molot manages the attack and release in a program-dependent manner…on this topic, on UAD blog there is a very simple and interesting artcile:


    the article talks about the frequency-dependant compression also (that many people confuse with the simple side-chain eq tweaking…)…sort of “multiband compressor behaviour” found on vintage compressor like LA-2A…
    I think 33609 and Fairchild 670 are, in a more or less manner, program-dependant…from my little experience on electronic circuits i think all analog compressor are more or less program-dependant…
    And this, for me, is what distinguishes a great analog emulation from a discrete emulation…
    What’s you opinion about, Molot?
    Ciao from Italy!

    • vladgsound 2011/02/12 at 22:19

      Thanks for information. I meant for program dependent limiter the release time is dynamic. For short bursts it’s short but for long limiting period it getting longer. For example Fairchild in positions 5 & 6. New limiter in Molot has attack of .2 ms and release from 10 ms to current compressor value. For frequency-dependent ratio I don’t have opinion how can be this useful for now. I think the main problems in most software compressors is that they sound very dirty when they compressing. I started Molot developing not to simulate hardware but to make software compressor as clean as possible. I did all tests on drum tracks that’s because it has such specific sound.

      • Alteregoxxx 2011/02/13 at 14:18

        Vladg i think i stumbled upon a Molot 0.2.135… i was checking the behaviour of the new molot limiter to se the type of time-dependance used for the release curve and…surprise!! there is no time-dependance at all !! for a short burst, duration 3 ms, the release is about 10 ms, like you said, but the recover time ramains the same even for a “steady state” signal of 3 sec duration! the burst and the steady state signal are at a frequency of 1Khz. I mentioned a “possible bug” because the strange fatc is that one time only it has worked as you’ve planned it should work!!! in other words with the same signal stimulus ONE TIME ONLY i’ve seen clearly the limiter time dependance, 10 ms for the burst and about 800 ms for the steady state (i set 800 ms on the release time…)…i repeated the test scores of times: nothing!! no time-dependant release…i’m on Reaper 3.75…hope this can help..

  2. Alteregoxxx 2011/02/13 at 14:23

    sorry…i forgot to add a little wordin my first sentence..the phrase should be “Vladg i think i stumbled upon a Molot 0.2.135 BUG <— (missed BUG in the sentence) 😀

    • vladgsound 2011/02/13 at 19:27

      You’re great! This is really a bug. I started with new limiter by curves but next I continued to debug “by hear” and suddenly dynamic release disappeared. My last tests were by rock mix for which the fast release is more useful so I missed it.

      Thanks again. I’ll fix it soon and replace binary.

      • vladgsound 2011/02/13 at 21:56

        I replaced binaries. I didn’t new version release but fixed version for “0.2.135f” in the corner of plugin screen.

  3. Dax Liniere 2011/03/08 at 14:31

    Hey Vlad,
    (Mr Vladimir G… I presume?)

    I would like to talk to you about some plugin stuff, but can’t find your email address.
    Will you please contact me on the email address provided?

    All the best,

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: