One cycle of an 808 tuned to E1 lasts 24.3 milliseconds, almost five times longer than the shortest release Krum-B will let you dial in.
That gap is where the thump comes from. Set a recovery time shorter than the note's own cycle and the compressor stops managing the envelope of the sound and starts redrawing the wave itself.
You have probably met the standard fix already. Slow the release past 100 ms, keep the attack off the floor, and the low end tidies up. That advice is correct, and it is also where most explanations stop.
So one cycle arrives at the detector as two peaks. On that E1 note the control voltage gets pushed twice every 24.3 ms, which lays a ripple near 82 Hz across the gain reduction.
That ripple is not level control. It is a second signal multiplied into your audio at a frequency related to the note, and ripple left sitting after the rectifier causes distortion that typically dwarfs anything the surrounding circuitry adds on its own.
Averaging detectors soften this, because they measure across a window instead of chasing peaks, which is why RMS-style detection has always been the recommendation for compressors expected to handle low frequencies. The trade-off is that an averaging detector is slower to notice a real transient, so the click at the front of an 808 can slip past before any gain reduction arrives.
Now look at the release range, which runs from 5 ms out to 1000 ms. At the fast end you are recovering gain about five times per cycle on a low E, so a release faster than the note itself is not a tone decision. It is a distortion setting.
That fix only works when there is content above the filter for the detector to read. Split a bass part at 150 Hz and the low band holds nothing but material below 150 Hz.
Put a detector high-pass at 100 Hz on a band like that and you have thrown away most of what the detector had to measure. The compressor is now reading the quiet edge of its own input.
So the sub band has no filter escape route. The only levers left are the time constants and the way the detector averages, which is exactly why sub-bass compression behaves worse than the same settings on a full-range bass.
Its attack averages 10 ms. Two time constants, one signal, no user control over either.
That is what adaptive means in practice, and it is measurable rather than decorative. A modern sub-release algorithm is chasing the same behavior with different parts, holding the low band down long enough to skip the cycle-tracking problem while still letting go before the next note lands.
That is the argument for giving the low band its own timing. A low band with its own adaptive release saves you from picking one recovery value for a part whose cycle length changes every time the note does.
Because the cycle length really does move. E1 gives you 24.3 ms, C2 at 65.41 Hz drops to 15.3 ms, and a C1 at 32.70 Hz stretches out to 30.6 ms.
A useful starting point is the note length rather than a fixed number of milliseconds. At 140 BPM an eighth note runs about 214 ms, so a recovery time in that region lets the low band settle between hits without chewing on individual cycles.
Your meter will tell you when you have it wrong. Gain reduction that flutters or vibrates is the detector tracking individual cycles, while gain reduction that swings once per note and settles is the detector doing the job you wanted.
That gap is where the thump comes from. Set a recovery time shorter than the note's own cycle and the compressor stops managing the envelope of the sound and starts redrawing the wave itself.
You have probably met the standard fix already. Slow the release past 100 ms, keep the attack off the floor, and the low end tidies up. That advice is correct, and it is also where most explanations stop.
The detector sees twice as many events as you do
Before any gain reduction happens, the compressor rectifies the signal to measure it. Full-wave rectification is the normal choice, and it folds the negative half of the wave up onto the positive half.So one cycle arrives at the detector as two peaks. On that E1 note the control voltage gets pushed twice every 24.3 ms, which lays a ripple near 82 Hz across the gain reduction.
That ripple is not level control. It is a second signal multiplied into your audio at a frequency related to the note, and ripple left sitting after the rectifier causes distortion that typically dwarfs anything the surrounding circuitry adds on its own.
Averaging detectors soften this, because they measure across a window instead of chasing peaks, which is why RMS-style detection has always been the recommendation for compressors expected to handle low frequencies. The trade-off is that an averaging detector is slower to notice a real transient, so the click at the front of an 808 can slip past before any gain reduction arrives.
Now look at the release range, which runs from 5 ms out to 1000 ms. At the fast end you are recovering gain about five times per cycle on a low E, so a release faster than the note itself is not a tone decision. It is a distortion setting.
The sidechain high-pass trick cannot save a sub band
On a full-band compressor the usual answer to low-frequency triggering is a detector filter. Roll the sidechain off around 100 to 120 Hz on bass-heavy material and the compressor stops lunging at energy it should leave alone. Hardware versions offered fixed positions near 30, 60, 105, 125, and 185 Hz.That fix only works when there is content above the filter for the detector to read. Split a bass part at 150 Hz and the low band holds nothing but material below 150 Hz.
Put a detector high-pass at 100 Hz on a band like that and you have thrown away most of what the detector had to measure. The compressor is now reading the quiet edge of its own input.
So the sub band has no filter escape route. The only levers left are the time constants and the way the detector averages, which is exactly why sub-bass compression behaves worse than the same settings on a full-range bass.
Adaptive release times are an old fix, not a new one
Program-dependent recovery has been shipping since 1965. The Teletronix LA-2A releases roughly 50 percent of its gain reduction in about 60 ms, then takes anywhere from half a second to five seconds for the rest, depending on what the program material did beforehand.Its attack averages 10 ms. Two time constants, one signal, no user control over either.
That is what adaptive means in practice, and it is measurable rather than decorative. A modern sub-release algorithm is chasing the same behavior with different parts, holding the low band down long enough to skip the cycle-tracking problem while still letting go before the next note lands.
That is the argument for giving the low band its own timing. A low band with its own adaptive release saves you from picking one recovery value for a part whose cycle length changes every time the note does.
Because the cycle length really does move. E1 gives you 24.3 ms, C2 at 65.41 Hz drops to 15.3 ms, and a C1 at 32.70 Hz stretches out to 30.6 ms.
A useful starting point is the note length rather than a fixed number of milliseconds. At 140 BPM an eighth note runs about 214 ms, so a recovery time in that region lets the low band settle between hits without chewing on individual cycles.
Your meter will tell you when you have it wrong. Gain reduction that flutters or vibrates is the detector tracking individual cycles, while gain reduction that swings once per note and settles is the detector doing the job you wanted.
