Un­cov­er­ing the truth be­hind high-pass fil­ter­ing

Future Music - - FEATURE -

From day one of mu­sic pro­duc­tion school, you’re in­structed to ap­ply fil­ters to non-bass el­e­ments in a mix, in order to re­move clash­ing low-end fre­quen­cies and there­fore re­claim mix head­room. Yet when ap­ply­ing a high-pass fil­ter, a sig­nal’s peak level can ac­tu­ally in­crease! Why ex­actly is this? And does this mean it’s bad to ap­ply in­di­vid­ual fil­ters through­out a mix? An­drew Sim­per, coding ex­pert and fil­ter guru, fills us in.

“This is a good point, and is def­i­nitely some­thing to keep in mind. One draw­back of us­ing high-pass fil­ters is that, al­though fre­quen­cies are cleaned up, you end up hav­ing a more ‘spiky’ sig­nal, which eats up some of your dy­namic range – but this is where non­lin­ear fil­ters can help, as they keep these peaks down.

“It’s prob­a­bly eas­i­est to con­sider a square wave as the worst-case ex­am­ple. If you flick back to our open­ing di­a­gram (on p27), you can clearly see that the high-pass peak level is around twice that of the in­put sig­nal! A lot of the time, a high-pass fil­ter is used to clean up ex­tra­ne­ous low-fre­quency noise from a record­ing – such as un­wanted bump­ing the body of the gui­tar, rum­ble from wind, or some other con­tent be­low that of the notes be­ing played. In this case, there isn’t any­thing fur­ther to do, be­cause the spikes won’t hap­pen.

“How­ever, if you’re cut­ting into the fun­da­men­tal fre­quency of the in­stru­ment with the fil­ter, and this is re­quired for the mix, then this will start boost­ing the peak level of the sud­den changes. So, back to the worst case sit­u­a­tion of the square wave: usu­ally this isn’t go­ing to chew up much dy­namic range be­cause the per­ceived loud­ness of a square wave will mean the vol­ume of the track will be so low that it won’t cause any is­sues. Also. the place­ment of the in­di­vid­ual spikes will not align with any other tran­sients, so these won’t add to­gether to push us over 0dBFS.

“Yet if you do need to keep these sorts of peaks down, the only op­tion is to do some non-lin­ear pro­cess­ing to tame them, ei­ther from a fast-act­ing compressor/lim­iter or a non­lin­ear low-pass fil­ter. This will add dis­tor­tion of some kind. In our ex­am­ple of the square wave, this isn’t an is­sue, since the en­tire square wave is one big dis­torted shape any­way, so adding more dis­tor­tion won’t sig­nif­i­cantly al­ter the ex­ist­ing har­mon­ics; it will just sub­tly change the am­pli­tude of a few of them – though if we’re do­ing dig­i­tal pro­cess­ing, we do still need to worry about alias­ing.

“If your sound isn’t as bright, you won’t have these spikes jump­ing out, but you may still have large tran­sients on the at­tack por­tion of the in­stru­ment. The good news here is that large tran­sients al­ready add a big splash of fre­quency to the mix, so dis­tort­ing them a bit won’t make too much dif­fer­ence, as long as there isn’t too much har­monic con­tent rid­ing on top of the tran­sient – which there won’t be for most mono­phonic sources.”

So surely we can min­imise the im­pact of fil­ter­ing in­di­vid­ual sig­nals by in­stead group­ing them to­gether and fil­ter­ing the col­lec­tive sig­nals in one go? Per­haps not, as An­drew ex­plains: “If there are mul­ti­ple sig­nals added to­gether, then we could pos­si­bly be adding dif­fer­ent sets of har­mon­ics to­gether, or hav­ing har­monic con­tent rid­ing on top of a large tran­sient. If we high-pass this summed sig­nal, and then use non-lin­ear­i­ties to tame the peaks, we will end up with lots of in­har­monic dis­tor­tion. This will con­trib­ute noise to the mix, even with all-ana­logue pro­cess­ing, so it should be kept to a min­i­mum.”

An­drew’s so­lu­tion? “I rec­om­mend do­ing any high-pass fil­ter­ing as early as pos­si­ble, then us­ing a fast compressor or lim­iter to tame the peaks of those sig­nals that you need to be loud. With this work­flow, you have a tidy sig­nal that will fit into the mix with min­i­mal pro­cess­ing re­quired down the line – per­haps some gen­tle com­pres­sion or EQ de­pend­ing on the con­text. If you have two over­lap­ping sig­nals, and you don’t want to lose the low fre­quen­cies from ei­ther of them, then the best thing to use is sidechain com­pres­sion, so only one of them is be­ing heard at a time. This pre­serves the dy­namic range and keeps things from get­ting muddy, and is al­ways some­thing to keep in mind if high-pass fil­ter­ing isn’t work­ing.”

If your sound isn’t as bright, you won’t have these spikes jump­ing out

Newspapers in English

Newspapers from Australia

© PressReader. All rights reserved.