> Personally, I now think that it was a mistake to use this weird
> parameterization. [...]
>
> Perhaps when it comes time to put together the first SA
> corrigendum, we should consider proposing to remove this
> parameterization in favor of a simpler one.
I think it would be great if the corrigendum clarified the
parameterization -- specifically:
"At each frequency f given as a parameter, the reverberation
time (RT60) at that frequency is given by the corresponding
r parameter"
Does a post-filtering scheme suffice to implement this? Right
now, sfront does this:
|-------------| |-------------|
|four parallel| | low pass |
|---| comb filters|---| filter for |--------
| | for r0 | | f0 | |
| |-------------| |-------------| |
| |
| |
| |
|----------| | |-------------| |-------------| |
---|2 allpass | | |four parallel| | low pass | |-----|
|filters in|------| comb filters|---| filter for |------| + |--->out
|series | | | for r1 | | f1 | |-----|
------------ | |-------------| |-------------| |
| |
| |
| |
| |
| |-------------| |-------------| |
| |four parallel| | low pass | |
|--| comb filters|---| filter for |---------
| for r2 | | f2 |
|-------------| |-------------|
For the initial implementation, I made the assumption that
if you rename the f's so that f0 < f1 < f2, then r2 > r1 > r0,
which is why lowpass filters suffice. But replacing the lowpass
filters with more complex ones can remove this restriction. Does
this sort of post-filtering meet the standard, or is implementing
a reverb algorithm that has this parameterization implicitly
in its structure necessary?
--jl
-------------------------------------------------------------------------
John Lazzaro -- Research Specialist -- CS Division -- EECS -- UC Berkeley
lazzaro [at] cs [dot] berkeley [dot] edu www.cs.berkeley.edu/~lazzaro
-------------------------------------------------------------------------
This archive was generated by hypermail 2b29 : Mon Jan 28 2002 - 11:46:37 EST