FabFilter User Forum
Pro Q3 Brickwall not filtering low frequencies
Q3 set to 30hz low cut brickwall on a kick drum.
Placed an analyzer over the master (using a Q3 but issue tracks with any analyzer) and there is a ton of low end information still come through. The Brickwall slop doesn't seem to actually work until your above 200hz and even then its more of a 48-72db slope, not brick.
The brick wall works fine when set to a high cut but terrible on low frequencies.
Am I missing something? I know its not a true brick and more like 100db slope but far more information is going out the master than that.
Please replicate on your end before disscusing EQ slopes. This seems like a major bug.
Darren — Feb 16, 2021
Which DAW is in use? IIRC there's a bug in Cubase which leads to the plugins output being mixed with used in mono, so essentially you get a 50%dry/wet result.
I've sent white noise through it and set a 30hz brickwall lowcut in ProQ3. It removes noticably much of the low end and is steeper than the 96db/o filter. Plugindoctor confirms this.
I noticed that unlike with the other filter types, cascading the brickwall does not change the result anymore which leads to the conclusion it already has removed all it could.
Keep in mind that anaylzers (especially based on FFT) have bad low frequency resolution unless they're using pretty high block sizes. I used proq3 (moderate cut), vps scope (steep cut at 30hz) and bluecataudio freqanalyst (at highest precision also a steep cut) and got 3 different results :D
Yeah, I'm going to go with analysers here as well, especially at lower frequencies they're not all that accurate. Ie, a regular analyser will never show the steep roll-off on low frequencies you would expect to see.
Guess my issue is with the visual representation of the slope on the Q3 - Its not actually doing what it is showing.
A brick wall setting shows zero slope so one would assume that the frequencies behind the line would not be present on your master. There shouldn't be any low end information for an anlizyzier to misinterpret.
But that's not how analyzers work... You can only analyze a certain size of discrete audio data - not an infinite big block (means you have a limit on resolution of the lower frequencies) and since it's discrete you have an upper bound too of nyqusit and a step size of frequencies (mean you cannot analyzer what lies between those bins or better said each bin covers a certain *frequency range* not a single frequency). Additionally to make that block suitable for anaylzing there's windowing needed which will further more "distort" the result.
Btw I just put the 30hz brickwall on some audio, bounced some minutes of it, loaded it in adobe audition, set the highest spectral size available and it shows almost a perfect brickwall cut. Lowering the spectral size creates more of the smeared result/roll off. Those super high sizes are not suitable for realtime graphics, not only do they take a good amount of CPU but they'd also add latency to a point where they are not useful anymore.
So ProQ is doing what it's showing with regards to the curve, but analyzer have a limit...