Impulse Response
From DRC
Contents |
Impulse Responses
This is a bit unstructured. Please feel free to submit some better text.
If you are using DRC, then an Impulse Response is what you get if you convolve an accurate recording of the sound a log sine sweep through your system, with the "inverse" of the original signal. (Note, the inverse is a slight simplification)
It looks like a tiny single spike in a sea of silence when you open up Cooledit. However, if you zoom in you will see that it's an initial big spike which quickly decays to noise, but with the odd bump every few ms.
How do I view Waterfall Plots?
Have a look at ETF or PureBits for waterfall plots (or search google for others)
How do I analyse an Impulse Response?
Take another look at the original impulse response. You are kind of looking at sound arriving at the listening position through time. The big spike is the first soundwave to hit you. If you look a few ms further on then you will see another smaller spike, and so on. The smaller spikes are reflected sound off floor, ceiling and walls.
Sound travels at around 330m/s so you roughly say that's 1m every 3ms. So quite guide line would be if your rear wall is 1m behind the speaker, then the reflected sound travels an extra 2 m, and hence you will see your spike some 6ms after the main spike (use cooledit for looking at this stuff).
OK, so you can now chop your impulse response up into, say, 3 ms chunks and plot a freq response of each chunk. Put these all on one graph and you now have a waterfall plot! Equally if you want to plot the freq response of your room without room re-enforcement effects (ie see how your speakers sound in an anechoic chamber), then you just take the first few ms of the impulse (ie before the first reflection spike), and graph that. This by definition is the anechoic response. See - not so difficult huh!
Now, you can do some other quality tests as well, for example you might wonder if the ability of your system to play a completely square wave was a measure of it's quality (think instant rise time, speaker cone moves instantly to it's max travel and stays there). So we would be testing some kind of dynamics. OK, but putting such a high level signal into the speakers would melt them pretty quickly, but we can take the impulse response of this hypothetical signal, convolve with our square wave and look at the output. The distortions will be obvious (oh and it will also look like a triangle for technical reasons, search google for more info).
So we can use the impulse response to see how our system would actually play all kinds of wierd inputs, and from this we might devise some measure of quality. Isn't the impulse response useful! And as we said already you can easily add selective time filters to the signal to remove the effects of say, rear wall echo, ceiling bounce, etc.
Now the problem is that the ear considers reflections (reverb) at greater than around 10ms (must check that number) to be beneficial and interesting to listen to. But reverb with shorter time periods just causes a kind of smearing and also some frequency masking. This is the Has effect. So this explains why it's beneficial to damp those first early reflections in the typical listening room
Further Reading
See also this excellent writeup on how to analyse the Impulse Response
Impulse response analysis - How to read an impulse response and measure your room
