Recording - October 2009 - 64

It turns out that removing room reverb is a very hard thing to do. There are some things you can do that will sort of deal with very reverberant tracks. You can use a gate or a “downward expander” with the decay time set very fast, so that it chops off the end of the note as soon as it starts to decay. The effect is very artificial, and it can be pretty annoying on vocals, but if it’s just one element in a dense mix, the snapping of the gate cutting off can be hidden a little. Sometimes you might want to add just a little actual reverb to the track after doing that. Sometimes you might want to just pass it off as a Phil Collins kind of drum sound. But sometimes it just doesn’t work and the only thing you can do is retrack. Back in the Seventies there was a huge obsession with eliminating any leakage between tracks and totally removing all room sound. This made for an easy time editing tracks and punching in and out bad notes, but it also made for very artificial sound quality and a complete absence of real Guest editorial by Scott Dorsey Dereverberation stereo imaging, something the audio community is still recovering from thirty years later. So, what can you do? One thing that people have been proposing for years is the notion of deconvolution. If you measure the impulse response of a system, you can generate an inverse function and then apply it to normalize the impulse response. This has two problems in the real world: First of all the impulse response is hard to accurately measure, and secondly it changes somewhat as things move around in the room—even things like the drummer’s arms! Reverb isn’t just a great a cappella group (see the photo)— it’s a problem that every engineer has to deal with sooner or later. Wouldn’t it be cool to be able to just make it go away? Scott Dorsey considers the problem of removing the room A decade ago I heard a Grateful Dead audience tape, where someone had made a very loud sound in the hall, then used that loud sound to determine the impulse response of the hall using Matlab, then applied an inverse function to the file. The effect was very strange it clearly made for a more dry sound, but it was also very weird and artificial sounding, and it progressively got more artificial as the concert went on. There have since then been various attempts at systems that dynamically estimate out the impulse response of the room. Some of these are actually good enough for communications-grade work, like teleconferencing. Keisuke Kinoshita and some other folks from NTT in Japan have come up with an algorithm that models late reflections in the room and removes them. It makes no attempt to deal with the close-in reflections, and the only goal is improved intelligibility, but it does actually do that. So, we have come some distance in the past thirty years, but not far enough so that we have a tool that is actually useful for audio production work at this point. And while I am constantly amazed at what the digital signal processing guys keep coming up with, I don’t see this as being a problem that will ever be solved to a degree where it’s satisfactory for production work. So, for better or worse, the solution is to track everything so that it sounds good in the first place. And the great thing about the modern digital world is that if you’re not sure if you really want a distant mic or a close one, you can record both because there’s no shortage of tracks. So in that way, there is something of a solution, brought to the problem by digital technology Just add chops. Scott Dorsey loves this stuff. Write to him at dorsey@recordingmag.com. You’d think that since it’s so easy these days to generate artificial reverberation that it would be easy to remove it too. Who out there hasn’t at one time got the idea that if you invert the output of a digital reverb you could use it to remove reverb from tracks that are too wet? Sadly it doesn’t work that way. Finding a way to remove room sound from tracks has been a sort of Holy grail for the digital signal processing people, since back in the 1970s. Not a month goes by that I don’t hear from someone or other who has some ingenious new device that they claim is going to remove room sound and give you a nice dry track. But you’ll notice that, like the fabled 100 MPG carburetor, nobody actually sells such a thing. 64 RECORDING OCTOBER 2009

Recording - October 2009

Table of Contents for the Digital Edition of Recording - October 2009