IR's and the difference......

[hermitage59]

In the process of learning about convolution and the structure of IR's, a question springs to mind.

As I understand it, a convolution IR comes from the 'tail' of an initial loud noise or percussive effect,i.e. starter pistol, etc. (Please correct my ignorance if this isn't the case.)

So my question is:

Does an IR built with a 'percussive' note from a brass instrument (for example) differ dramatically, aurally, from that of a pistol? My instinct tells me it should be different, but i'm wondering that's really the case. And would an IR built from a musical instrument vary in some degree of 'warmth and humanity', as opposed to a pistol?

Alex.

[Dietz]

In principle, you can take any signal for use in a convolution-process. The idea with IR-based "sampling reverbs", though, comes from the idea to have an acoustically "true" impression from a real room. For this reason you would need a well-defined signal source - ideally an impuls which is has equal loudness in any frequency range at a given moment.

In the Real World this kind of impulses are actually impossible to produce; hand-claps, balloon-pops and pistol shots are only a vague aproximation, and signals played through loud-speakers are limited due to its dynamic response and frequency range.

For professional IR-production, mathematically generated noise bursts or (even better) sweeps are used, which allow to compute an "idealized" impulse response from the recorded room, due to the well-known parameters of the initial signal.

So - yes, "an IR built with a 'percussive' note from a brass instrument (for example) differ dramatically, aurally, from that of a pistol"! [:)]

... but I have to admit that I don't really understand what you're asking for with this sentence:

@Another User said:

And would an IR built from a musical instrument vary in some degree of 'warmth and humanity', as opposed to a pistol?

[Nick Batzdorf]

I think the question is because he's unclear on the concept of impulse responses. That's understandable, because this is all very confusing. My understanding of convolution is still very crude.

Alex, an impulse response is theoretically all frequencies at one instant - which as Dietz says isn't possible to produce. That's the reason to use swept sine waves and then a program that reads the sweeps and figures out what they'd look like if they were compressed to a single point in time.

In other words, a recording of a horn is not an impulse response, it's a sample of a horn's effect on a room. That might be a useful effect, and it might well give a nice warm color, but it's not the normal way to sample a room to make a realistic reverb.

[hermitage59]

@Nick Batzdorf said:

I think the question is because he's unclear on the concept of impulse responses. That's understandable, because this is all very confusing. My understanding of convolution is still very crude.

Alex, an impulse response is theoretically all frequencies at one instant - which as Dietz says isn't possible to produce. That's the reason to use swept sine waves and then a program that reads the sweeps and figures out what they'd look like if they were compressed to a single point in time.

In other words, a recording of a horn is not an impulse response, it's a sample of a horn's effect on a room. That might be a useful effect, and it might well give a nice warm color, but it's not the normal way to sample a room to make a realistic reverb.

Exactly Nick. I'm unclear about the concept, and don't have the right terminology i guess to ask the question in an understandable fashion.

It's an interesting idea though to sample the 'tail' of a brass note in a room. Could this be done in such a way, as to be considered an 'IR' of sorts?
It might be useful to have a small collection of these, for each section in the orchestra.

Thanks for the information gentlemen.

Regards,

Alex.

[Dietz]

@Nick Batzdorf said:

[...] an impulse response is theoretically all frequencies at one instant [...]

Actually, what you describe here is the _impuls_ itself, the impuls that generates a response, so to say.

[Dietz]

@hermitage59 said:

[...]
It's an interesting idea though to sample the 'tail' of a brass note in a room. Could this be done in such a way, as to be considered an 'IR' of sorts?
It might be useful to have a small collection of these, for each section in the orchestra.
[...]

This is more or less just an "ordinary" reverb-trail (or a collection of them). As I tried to outline above: The basic feature of an "impuls" in our sense of the word is its known set of parameters, so we can replace the impulse by _any_ signal on behalf of convolution and (ideally) get the proper acoustic reaction of the room --- as if the new signal was recorded on the same place as the impuls.

... the same is true for any other linear (electro-)acoustic system, of course. That's why you can sample the behaviour of EQ's and the like, as long as there are no time-variant aspects to take into account. IOW: sampling synthetic reverbs will often not catch them completely, due to (time-variant) chorusing and similar stuff being part of their algorithms.

HTH,

[MarkOfTheStoat]

The audioease Altiverb 6 demo QT video explains the basics of convolution reverbs... with diagrams. [[;)]]

http://audioease.com/Pages/mp3%20Altiverb%20Tests/Altiverb6DemoEnglish.html

[Nick Batzdorf]

@Another User said:

Actually, what you describe here is the _impuls_ itself, the impuls that generates a response, so to say.

Yes, sorry. [H] [[;)]]


This actually points to an aside: you can create some interesting special effects with convolution processors by sampling all kinds of things. At King Idiot's suggestion I once tried sampling a whole mix, for example.

...not that I've ever found a use for that. [:)]