FabFilter User Forum

Weird Pro-MB expansion behavior in single band mode

Hi,

When using Pro-MB as downwards expander with only a single band covering the entire spectrum, the attack transient seems to get boosted even though range is negative and band level is 0 dB.

The problem is 100% reproducible with a simply decaying sine input. Threshold, ratio or knee seem to have no effect. I used attack 1%, release 20%, lookahead 20 ms. Reducing lookahead reduces the boost but only turning the lookahead function completely off seems to remove it.

Is this intentional? Can I avoid it while keeping lookahead?

Antti

We're going to have a look at this to see if we can reproduce it!

Cheers,

Frederik (FabFilter)

The simplest way I came up with:
1. Generate some silence followed by a 1 kHz sine at -3 dB (level doesn't actually matter as long as it's above threshold).
2. Open Pro-MB with default settings.
3. Create a band that cover the whole spectrum.
4. Set mode to expand, range to -30 dB, knee to 0 dB, lookahead to 20 ms. Leave other settings at default.

Checking the result in any waveform editor shows that the first 100 - 200 ms gets boosted by about 1 dB.

Antti

Hi Antti,

I've investigated this now, and the behavior is indeed as expected. Without getting too technical, what you see is a side effect of the internal lookahead implementation we chose to use for Pro-MB.

If you're looking for more precise full band limiting, I recommend that you use Pro-L 2 of course :-)

Floris (FabFilter)

Hi,

Thanks for looking into this. Although I'm not sure how Pro-L 2 would help with expansion...

My use case is reducing the boominess of drum samples by using Pro-MB as a clean multiband expander with low ratio. The reason I noticed the issue was that the attack parts of the samples ended up getting clipped due to the gain increase. Testing revealed it to behave the same in single band mode and hence the issue report.

Downwards expansion (the range is negative, afterall) actually increasing the level just seems completely illogical and more or less an outright bug (in algorithm if not in code).

Antti
Reply to this topic Go to the forum topic list