Is a virtual room
Finite Impulse Response (FIR) generator designed to create high quality
natural-sounding reverberberation to be processed with:
- MultiVolver™ (Classic,
VST or WCP)
- Lake
Huron™ or CP4™
- other convolvers, such as
VST-based, that can read WAV-format FIRs
|
|
|
|
Is a fast and general off-line
multi-channel reverb/mix (classic) convolver:
general design
convolves and mixes up to
8 x 8 channels in x out in any combination
allowing for up to an 8 channel
reverb/mix e.g. to 7.1
Oct 2010 two new convolvers were added to the 5th edition of
The Suite:
- MultiVolver VST™ an 8x8 VST plugin
convolver (plus separate 1x1, 1x2 and 2x1 versions)
- MultiVolver WCP™ an NxM offline convolver
Both using a common simple but flexible text file for the FIR
setup (for MATLAB, WAV or Lake SIM FIRs), an example:
[Setup]
InChannels = 1
OutChannels = 2
BlockSize = 4096
Gain = 0
[FIRs]
Format = WAV
01->01 = MVTST_L.WAV
01->02 = MVTST_R.WAV
|
|
MultiVolver™ Classic
MultiVolver WCP™
MultiVolver VST™
(in AudioMulch VST host)
|
|
- creates natural-sounding, high-density, full frequency range
32-bit FIR reverberation using virtual microphones:
- mono (omni or cardioid)
- coincident stereo (cardioids)
- AB-stereo (variable spacing,
also "wide AB")
- B-format (WXYZ)
- 5-channel (L, C, R, Ls,
Rs e.g. for natural-sounding stereo to 5-channel upmix, interactive
mic setup)
- the 5-channel implementation
allows for individual microphone selection of:
- location
- aim
- type (cardioid, super-cardioid,
hyper-cardioid, figure-of-8, special figure-of-8+ or omni-directional)
- setups can be saved and reused
- uses virtual mono or stereo sources. By creating several sets
of mono or stereo source responses up to 8 sources can be used together
with the MultiVolver™ since it will handle the relative calibration/scaling
between responses automatically even for filters created at different
times
- uses an FIR late reverberation algorithm similar to the one
used in CATT-Acoustic™:
- is rather a virtual room
impulse response creator than just a tool giving reverberation
- is based on physical principles
and creates responses with a theoretically correct, with time increasing,
reflection density avoiding the granularity in the tail created by many
Infinite Impulse Response (IIR) algorithms
- is based on the experience
from ten years of computer modeling and auralization of real rooms
- is used by Deutsche Grammophon
together with a Lake Huron™ to enhance too dry recordings
- was used together with a
Lake CP4™ to add reverberation to the music score
at the World Cup in Football (Soccer...) final at Le Stade de France,
1998 (actually CATT-VRoom, the PureVerb™ predecessor)
- creates reverb for VR audio applications (together with Lake
hardware, PureVerb™ can e.g. create all files required
to run Lake AniScape™)
- supports several samplerates:
- 44100 Hz
- 48000 Hz
- 88200 Hz
- 96000 Hz
- 16 000 Hz (for Lake AniScape™
reverb)
- has a straight-forward interface with settings (*in octave-bands
125 to 16k Hz) for:
- room size
- virtual source location
- source directivity index*
- virtual microphone location
- reverberation time*
- diffusion*
- absorption distribution
(uniform, audience or detailed*)
- reflection incidence distribution
- reflection density
- optional use of the Image Source Model for 1st order
reflections
- optionally creates responses with the direct sound excluded
and the original signal can be bypassed when mixing in the reverb
- has a large set of presets to promote realistic reverb settings
and give good starting points for further fine-tuning:
- user expandable "Generic
Halls" presets where average size and RT values can be taken from actual
well know halls
- suggests reverberation time
based on the volume and type of music (chamber music, opera, choir
etc.) using curves of optimal ratios from literature (see fig below)
- suggests a volume based on the reverberation time and type
of music (chamber music, opera, choir etc.) using curves of optimal
ratios from literature (see fig above)
- suggests listener (or virtual
microphone) location based on the reverberation radius of the selected
hall (Close, Medium, Far)
- creates CATT PLT-files (2D plans, shaded 3D
models and RT/Volume graphs) that can be viewed, printed and exported
either via the included stand-alone CATT PLT-viewer or
with CATT-Acoustic™ (from v7.2,
also the demo version)
- creates a new PureVerb eXtended (PVX) single-file,
multi-input, multi-output, multi-microphone impulse response format
that also stores a PureVerb settings-file.
A PVX-file can be loaded by File|Open
for examining the parameters used when creating the impulse responses,
or to re-calculate the same basic responses after some changes in parameters.
The PVX-format simplifies use of PureVerb responses in MultiVolver
since only one file needs to selected for a complete
set of filters. The PvxViewer
shows the contents of a PVX-file, impulse responses, octave-band
filtered impulse responses and decays.
- optionally directly creates
WAV-file format and includes a file-format conversion tool that converts
Lake SIM-files to 16-, 24- or 32-bit PCM integer
WAVs for use with other convolvers. Optionally cretes WAVs
in WAVE-EX format
|
sample PureVerb™
screen-shots |
|
|
sample 5-channel responses |
|
- examples of uses (all of these can be done in one step with
auto-scaling):
- mono
mono reverb (1 x 1)
- mono
stereo reverb/up-mix (1 x 2)
- mono
B-format reverb/upmix (1 x 4)
- mono
5-channel reverb/up-mix (1 x 5)
- stereo
stereo reverb/re-mix (2 x 2)
- stereo
B-format reverb/up-mix (2 x 4)
- 8-channel
5-channel reverb/mix (8 x 5)
- cross-talk cancellation
of a binaurally recorded stereo file (2x2, CATT-Acoustic™ >=
v7 can create cross-talk cancellation filters)
- 5-channel
binaural down-mix (5x2 or 6x2 for 5.1)*)
- n-channel Ambisonic
binaural down-mix (e.g. 4x2, 6x2 or 8x2)*)
- B-format
Ambisonic decode for up to 8 loudspeakers
*) CATT-Acoustic™ >= v7 can create
binaural room filters for various listening room configurations
- each of the 8 inputs can have a gain to balance the mix
- Classic: each of the 8 inputs can be any of:
- a mono WAV
- the left of a stereo WAV
- the right of a stereo WAV
- stereo WAV (left/right mix)
- WCP:
each of the inputs can be any of:
- a mono WAV
- a selected channel of a multi-channel WAV
- Classic: the output channels are either one mono WAV
per channel or stereo WAVs assigned to successive pairs of outputs.
WCP: output channels are either
one mono WAV per channel or multi.channel WAV
- automatically handles calibration/scaling/overflow:
- separate mono
5-channel filters for each source location can be created
in PureVerb™ and they will be relative calibrated
when the processing is made, no wild trial runs required to check for
overflow
- estimates filter gains and
adjusts so that the output WAVs will not overflow. If it still happens in
some odd cases (e.g. if the input WAVs have very low margins), it will be
discovered and a safe second pass will be used
- gives a min "margin" value
on the output WAV-files (similar to that of pro DATs and similar).
If the resulting margin is too high, an automatic remake can be performed
- automatically handles delays from sources to microphones (if
PureVerb™ or CATT-Acoustic™ responses are
used, for imported responses initial delays should be included as initial
zeros)
- utilizes Lake's 32-bit integer SIM-file format
or PureVerb PVX for for the FIR
filters:
- allows off-line tests to
be made and subsequent processing using the same filters in a Lake Huron™ processor
(or the other way around)
- each set of FIR-filters can
have a base-name so that fast and safe filter changes can be made
(requires only one name selection)
- can as well use measured
FIRs via file conversion to SIM-format
- PVX load dialog:
- uses/creates 16-,24- or 32-bit PCM WAVs, optionally in WAVE-EX
format.
- uses no temporary hard disk space at all, can thus convolve
any length WAVs (as long as the resulting WAVs can be fit on the output
disk)
- performs floating point convolution of long filters typically
at 8x speed on an old 500 MHz PIII PC. Sample processing times with
88,000 tap FIRs (i.e. 2 sec reverb at 44.1 kHz) and 8
minutes of music:
|
Type of processing
|
Number of
convolutions
|
Processing
time ; speed
|
|
mono
mono reverb
|
1
(1x1)
|
1
min ; 8x
|
|
mono
stereo reverb
|
2
(1x2)
|
2
min ; 4x
|
|
stereo
stereo reverb/re-mix
|
4
(2x2)
|
4
min ; 2x
|
|
mono
5-ch reverb/up-mix
|
5
(1x5)
|
5
min ; 1.6x
|
|
stereo
5-ch reverb/up-mix
|
10
(2x5)
|
10
min ; 0.8x
|
|
maximum
8 x 8 matrix
|
64
(8x8)
|
64
min ; 0.125x
|
- real-time 2x5 zero- or low-latency processing in hardware at
44.1 or 48 kHz , as outlined above, requires:
- a Lake Huron™
with a minimum of 3 DSP-boards (12 Motorola DSPs)
giving a latency of 160 ms (2 ms latency requires 5 DSP-boards, 20 DSPs)
- or (for 2x4 only
actually):
- two
Sony DRE-S777: 2x(11+11)=44 Sony
custom DSPs total (assuming the required second DSP
boards also contains 11 DSPs)
- two
Yamaha SREV1: 2x32=64 Yamaha custom
DSPs total
- has an integrated Ambisonic decode filter creator:
- optional user-supplied shelf-
and inverse speaker filters
- energy and velocity vector
graphs
- save/load rigs or complete
filter setups
- includes a WAV-file player with single, A/B and play-list options
|
|
sample MultiVolver
VST/WCP™ screen-shots |
|
|
sample MultiVolver™
classic screen-shots |
|
|
signal flow graphs, principles |
|
|
Ambisonic decoder details |
|
Comprehensive audio demonstration/tutorial 
PureVerb™
MultiVolver™