2010-10-06

Verb Musings

This morning, my friend Sudara complained that he was loving Eos too much. (Hard to fault him, it's my absolute favourite reverb).

The conversation rambled towards convolution reverbs. I'm fairly ignorant about them beyond the layman's description that we're given to make us layman stop asking questions. This ignorance allows me to formulate all kinds of crazy bullshit.

See, we talked about recording sounds in small reverberant spaces like bathrooms. It occured to me that people like the sound of stuff in their bathrooms more because of the resonances of the room (the way certain frequencies are boosted) than any reverb.

That's a subjective statement for sure...  but bear with me.

Which led me to wonder, does convolution reverb capture the EQ/resonance of a room and is it able to impart it to the affected signal.

My initial guess was no.  But thinking more on it, the IR inherently has the EQ of the impulse (how can it not?) so unless the convolution black magic loses it somehow, the affected signal must have at least some of this EQ curve applied.

I can certainly imagine an experiment that could test this (at least subjectively) but it's so much easier to ask the Google of Mountain View.

And having asked, I learn that there are several methods of gathering impulse responses.  Some involve a sine sweep over the audible frequency spectrum and others involve some kind of digital noise burst.  So if the IRs you have in your convolution reverb are made with sine sweeps,  the implication is that every audible frequency is covered and would be applied to whatever you were running through it.

Sort of revealing my earlier assertion as the naive bullshit it was.

Which raises questions about the kinds of IRs I have kicking around...

And thus we learn.

/me is off to play with impulse responses

CC licensed photo by flickr user: Pulpolux !!!

2 comments:

seancostello said...

Convolution can fail to capture the high frequency response of a large space, due to the thermal currents resulting in time-varying reflection patterns. This time-variation turns into averaging in convolution, which turns the high end into mush.

However, the low end resonances are captured to a high end of accuracy in convolution. And a smaller room (i.e. those suitable for drums) can be captured quite well with an impulse response. It is only when you get up to very large spaces, such as halls, that the time variation has an audible effect on the high frequency response.

Mind you, I prefer modulated reverbs (I wrote the Eos algorithms). But convolution can work great for all sorts of reverbs, especially smaller rooms.

MMI said...

Wow. I'm used to having a reasonable expertise in OS internals and random programming arcana. It's really neat and weird to be a "dangerous" n00b in audio.

Thanks Sean (and yes, I knew you're the Eos dude which is why I bought shimmer without hearing it; to say I'm happy would be an understatement).