Loss-compensated filters?

angstrom · 24 Jan 2006

sounds good to me, I've thought abut this before but never tried it.

Of course a lopass with resonance peaking will probably raise the volume, necesitating volume attenuation - but that will still work.

I guess it depends on how 'perceived loudness' compares to metered loudness. Metered 0db of gain at 100hz is a lot more tolerable than 0db of 5000hz. Perhaps a frequency scale might be needed

mindplay · 24 Jan 2006

the envelope detectors would of course have to be "hearing compensated", yes - I think this is commonly done by using something called "fletcher-munson compensation", although I don't know how this is done.

angstrom · 24 Jan 2006

there are a variety of curves to deal with the sensitivities in hearing. lots of research done for dithering down to 16bit recordings, where they use Fletcher Munsen, but also others such as this one.

that's from Bob Katz's site, an article about dither extracted from his book.

stefan · 24 Jan 2006

mindplay wrote: Just an idea that turned up in my head today: loss-compensated filters...

Filters take stuff away - which means that the overall volume of the signal is lowered. Sometimes, this can pose a challenge - like in house and techno music, if you want to cut away all the bass for a moment to build up some tension, you have to automate the filter cutoff as well as the output volume to compensate for the loss of volume.

What about simply putting an envelope follower on the unfiltered signal, and putting one on the filter output - the factor between these two envelope levels is the amount of loss, and you can then multiply by the reciprocal to bring the volume up to whatever it was before it went through the filter.

Is this a good idea, do you think? Why/not? Has anyone tried it?

Hm. Basically a compressor with a twist. I like the idea, could be interesting to explore it with other effects than filters as well.

You've already described how to do it though, and there's a ready-to-use envelope follower in the toolbox you could use to try this out.

infuzion · 24 Jan 2006

mindplay wrote: Filters take stuff away - which means that the overall volume of the signal is lowered. Sometimes, this can pose a challenge - like in house and techno music, if you want to cut away all the bass for a moment to build up some tension, you have to automate the filter cutoff as well as the output volume to compensate for the loss of volume.

hmmm, when I kill the EQ while DJing on a sound system, you don't really worry about the volume loss; there is usually a 3-way sound system, so you're just cutting out 1 set of speakers. People's ears will adjust to the loss, focusing on what is available.

In post-production, any cutting out of instruments will be compinsated by a multi-band compressor.

Sometimes (esp in pop/rock), a producer will "ride the channels" & boost certain insturments to make up for a silent insturment, & cut as more instuments are added. I thought this meathod is best simulted automaticly by a side-chained compressor the the side-chain connected to an EQ'd signal (or the signal that you're bringing in/out)

mindplay · 25 Jan 2006

angstrom wrote: that's from Bob Katz's site, an article about dither extracted from his book.

Personally, I'm not inclined to paying any regard to anything written by Bob Katz - I bought his book "Mastering Audio", which was utterly useless and filled with some of the most obvious and straight out dumb advice I've ever heard 😉