[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Date Index][Thread Index][Author Index]

Re: OT Create your own Impulse models for freeware Convolution Reverbs



Rick Walker schrieb:
> I did a little research tonight and discovered this freeware software for
> creating *_your own_* Impulse files (digital modelling files) to use in a
> shareware or freeware Convolution Reverb.
First of all, thanks to all of you (including you, Rick), for the posts 
on the original "mics for recording drums" thread - I'll try to 
summarize and comment that thread later on.

However, as Rick pointed out, for reverbs convolution-based processors 
are an option. I'd like to give some wisdom on this topic for those who 
haven't dealt with it so far.

It definitely took some time from Kurzweil's KDFX board for the K2500 
workstation which first included the LazerVerb algorithm (also found in 
the K2600, KSP8 and Rumour) that convolution reverb became such widely 
used. Today, every plugin manufacturer offers one, there are some 
freeware/rather cheap options on the market (thanks to Rick for the S1R 
suggestion), and most DAW programs include such a plugin (have one in 
Cubase, checked for Logic and Sonar). But was is this "convolution 
reverb thing"?

Basically (and slightly simplified), the reverb patch is described by 
the frequency response of the reverb space we want to model. Now for LTI 
(linear time-invariant) systems [1], theory has it that the entire 
behaviour of the system is described by its impulse response, or by its 
complex frequency response [2], both of which are essentially the same [3].

This consequently means that the reverb will always be LTI, that means 
there is no modulation, no distortion and no saturation of any kind.

A very good application is to model a specific good-sounding room at 
sensible SPL levels, simply by recording the impulse response of that 
room [4]. But what can be done for actual rooms can also be done for 
things which model rooms - e.g. you could simply take your treasured 
Lexicon 480 effector, load a patch and record its impulse response (and 
this is actually done; in some of those free impulse response 
collections you'll find impulses from such devices).

Or, you can record the IR of, say, a grand piano or an acoustic guitar 
and use that in your convolution reverb plugin.

Or you can use something entirely different, and not record an impulse 
response at all, rather take any sound (e.g. a snare drum hit, or some 
noise bursts from a shortwave receiver) and load them into your 
convolution reverb plugin and see what happens.

However, there's a simple downside: it works perfectly if you want to 
model a specific space for which you have an impulse response available. 
If not, you'll be most of the time working more efficiently by simply 
using a high-quality algorithmic reverb and use that.

Yours,

          Rainer

[1]: A LTI (or linear time-invariant) system is a system with inputs and 
outputs where the transfer function describing the output as a function 
of the input and time (let's discuss this for only one input and output 
each) is both time-invariant and linear.
Linearity means that if you make the input louder then the output will 
become louder by the same amount but not change otherwise. Also, if you 
send two input signals at once into the system, then the output is the 
sum of the signals you've gotten if you'd sent the inputs individually. 
In other words, it does not distort. Ever.
Note that this does normally NOT mean that the system has a flat 
frequency response (which colloquially is often referred to as "linear").
Time invariance means that you always get the same result, no matter at 
what time you send something into the system. A good counter example 
would be a flanger or modulated filter (basically everything with a 
time-dependant modulation).
All ideal EQs and ideal delays are LTI.
A nice side-effect of LTI systems is that you can arrange their order 
freely, i.e. it doesn't matter if the EQ goes before or after the delay.

[2]:  This is  the frequency response which does not only have an 
amplitude over frequency ("frequency plot"), but an additional phase 
over frequency. Again theory has it (Kramers-Kronig relation) that for 
the systems we're looking at, the amplitude directly defines the phase 
and vice versa.

[3]: Fourier transform can be applied freely in LTI systems. The 
(complex) frequency response is the fourier transform of the impulse 
response and the other way round. The impulse response can theoretically 
be perceived as "what the system does if you play an infinitely short 
and infinitely high pulse into the system".

[4]. Doing that brings several problems. The first is a proper stimulus 
for the object (how do you make that infinitely short pulse?): typical 
devices are some signal pistols, bursting balloons, very specific 
thingies (somewhat larger versions of those things you use to calibrate 
measurement microphones) or on the other hand playing back a frequency 
sweep on a good speaker. The second is to record it - again, any 
limitations of the microphone, micpre or recorder will have a negative 
influence. The third is to make an impulse response file out of your 
recordings, but again, Rick already pointed out some software for that.

-- 
http://moinlabs.de
Follow me on twitter: http://twitter.com/moinlabs