Diffusion, vocals, and real rooms

One of the common controls found on artificial reverberators is “diffusion.” The control can have different underlying meanings, but is usually used to tailor the echo density during the initial portion of the reverb decay. A typical description of the Diffusion parameter comes from the Lexicon PCM70 Owner’s Manual:

DIFFUSION controls the degree to which echo density increases over time. High DIFFUSION settings result in high initial buildup of echo density, and low settings cause low initial buildup. After the initial buildup (in which echo buildup is controlled by DIFFUSION), density continues to rise at a preset rate determined by the program.

Recently, I started to wonder why the Diffusion control was so common in artificial reverberators, as it has very little to do with physical reality. In a physical room, the initial echo density is determined by the lower-order reflections. A conventional model of a room will model the first order reflections as those coming from the 4 walls, the floor and ceiling (6 in all). In reality, rooms have things in them – furniture, small structural details, people, columns, etc. All of these contribute their own complex reflection patterns. The low initial echo density suggested by turning the Diffusion control down in a conventional reverb would only be achieved in a very large room with no furniture. Real acoustic spaces tend to be fairly diffuse, at least by the standards of artificial reverberators.

A bit of insider information reveals the likely answer. Most artificial reverberators use series allpass delays somewhere with their structure to generate a higher echo density. The original Schroeder 1962 AES paper suggested 2 series allpass filters, processing the output of 4 parallel comb filters. Other early reverb structures moved the allpasses to the inputs of the combs. Later work by Gerzon and Griesinger placed allpass filters inside of feedback structures, in order to create an echo density that builds with time.

Allpass delays are a convenient way of generating a high echo density relatively quickly, but they are not without their drawbacks. The term “allpass” means that, over the long term, the frequency response is flat – but over the short term, they can sound similar to comb filters. In particular, impulsive sounds processed through several allpasses in series can sound somewhat metallic. A periodic series of pulses will cause the filters to “ring out.” Unfortunately for reverb designers, a periodic series of pulses is a great description of a vocal waveform. For reverbs that rely on series allpasses for initial echo density, a setting that sounds good for drums might sound far too metallic for vocals.

This is where the Diffusion control comes in. By turning down the coefficients of several of the allpasses within the reverb structure, the metallic artifacts caused by vocal tract pulses can be reduced. This may also result in a slightly grainy echo sound on sibilants, but for the most part the sound of the reverberator on vocals will be improved. The resulting sound may not be identical to a physical room, but it will sit in a mix better than the same reverberator processing vocals with high diffusion.

Leave a Reply

Your email address will not be published. Required fields are marked *