Impulse Responses & Convolution Reverb: How To Sample An Acoustic Space

Those familiar with audio production probably know that there are two types of digitally synthesised reverb effect. The first, and generally most popular given its byte-sized (heh) use of computer resources is known as “algorithmic” reverb, where the incoming signal is, sample by sample, multiplied by a factor dictated algorithmically by various twiddly knob-like parameters. Such reverb types, although efficient on resources, can be less than convincing upon application, especially when applied to particularly exposed instruments or voices; unsurprising since they are merely a mathematical approximation of the kind of thing reverb should probably sound like.

 

1A twiddly knob algorithmic reverb unit.

 

The second, more sophisticated type of reverb is known as “convolution” reverb, and it is this type that is the focus of today’s brain spillings.

 

You see, convolution reverb more precisely replicates the acoustical properties of an actual real-life environment by manipulating the original recording via a method similar to the algorithmic reverb technique, but crucially different in a very specific way. This time instead of relying on a mathematically produced set of rules to determine the multiplication of the incoming signal, an “impulse response” – an actual recording of an actual real-life actual environment – is used.

 

Actually.

 

2

A convolution reverb unit.

 

“Impulse response” is just a fancy way of saying a “recording of a signal processed via some system”. In the case of convolution reverb, the system is the reverberation of a physical space. Stick a microphone in the middle of the Sydney Opera House, record the sound of a balloon popping plus the ensuing room acoustics, and, hey presto, you have yourself an impulse response. An accurate impulse response must feature all frequencies within the audible spectrum (20 Hz – 20 KHz) in order to be effective for use as convolution reverb, which is why a balloon pop makes for a fairly commonly used source, especially given the ease with which such a scenario can be set up. Generally speaking, a sudden burst of white noise of this kind contains spectral content of sufficient bandwidth to be practically useful.

 

There is however, as impulse response elitists never tire of pointing out, a problem with this method, since no two balloon pops are exactly alike, and the intensity of frequencies across the spectrum may vary wildly. A higher intensity at 500Hz than 2KHz will bias the response of the room in favour of 500Hz. Therefore, for the sake of accuracy, a frequency sweep played back through a flat response studio speaker is considered the definitive method of sampling an acoustic space. This way it is ensured that all frequencies are played at equal intensity. Of course there is then some deliberation about the quality of speaker, microphone and preamp used, however it is claimed that a fairly modest system can at least produce a very reasonable approximation.

 

I should point out at this juncture, ladies and gentlemen, that I did, in order to look clevererer than I necessarily am, try to find a more detailed explanation of the process of multiplying the input signal by the impulse response at a sample level, but my efforts merely resulted in my being sick on my desk. So forgive me if I quietly refrain from that, and instead offer that I think it might have something to do with very, very tiny squirrels.

 

I think.

 
 

Isn’t it?

 
 

Well, okay – here, I drew a picture. I think it’s something like this:

3

Make of that what you will.

 
 

Anyway, so far, so good – find a sonically interesting environment, record a frequency sweep, run it through some jiggery pokery to create a usable impulse response, load it into your favourite convolution reverb plugin – I will be using Cubase’s inbuilt REVerence – and marvel at your digital recreation of your original environment. Sounds like fun to me.

 

So, choosing two sonically interesting spaces at the University of Sussex, I put it to the test. The Meeting House is a large, circular, chapel with a domed roof, apparently designed in the midst of a fairly potent acid trip, whilst the drama studio is a much smaller, enclosed room, yet of equivalent acoustic interest.

4

The Meeting House.

5

The Drama Studio.

 
 

Within these environments I set up a Genelec 8030A for playback of the frequency sweep, a pair of AKG C451Bs in ORTF configuration on the far side of the room and then, to the bemusement of any passing visitors, recorded the signal. Then, in the same positions, I recorded a short segment of acoustic guitar for reference when assessing the authenticity of my resultant convolution reverb. For completeness, I also recorded balloons popping, so that a true comparison could be conducted.

6

7

8

 
 

So then, once all the data had been collected I could return to the lab to analyse the results. Some subtle treatment of the recorded signals was required to eliminate ambient noise where possible, always ensuring never to damage the fidelity of the actual recordings. Once the material had been analysed, treated and turned into usable impulse response files via Voxengo Deconvolver, I could load them into the REVelation Convolution Reverb unit, apply the plugin to a source signal (in this case a dry recording of the guitar riff I had played in each environment), and cross my fingers that it had worked.

 
 

Here are the results. Click to listen.

 

#1: Meeting House
— Guitar in Room / Frequency Sweep Reverb / Balloon Pop Reverb.

The results here are actually pretty good. Firstly it’s noteworthy that the frequency sweep does indeed produce better results than the balloon pop. The balloon pop has a deficit of high end information and a swelled, rather ugly middle. The frequency sweep on the other hand does a reasonably good job of replicating the guitar test recording. Both however produce a reverb tail that is fairly authentic. I am pleased with these results.

 

#2: Drama Studio
— Guitar in Room / Frequency Sweep Reverb / Balloon Pop Reverb.

A generally brighter reverb this time but the results are consistent with the previous environment. Again, the frequency sweep method has produced a far more convincing result.

 

So, all in all, that seems to have proven very successful. It is worth noting that, as well as actual physical environments, the frequency sweep method can also be used to sample hardware or software reverb FX units. By processing a frequency sweep with an interesting reverb unit and then generating an impulse response from the resulting file, a startlingly accurate clone of the original effect can be made. Here are two examples of such a practice, whereby the previous dry guitar recording has been processed first by a dedicated reverb unit, and then by an impulse response clone of that unit:

 

#3: Spring Reverb
— Original Reverb / Cloned Reverb

 

#4: FX Reverb
— Original Reverb / Cloned Reverb

 

So that’s it. Impulse response generated convolution reverb.

Lawks a lordy.

 

I’m going to eat some chicken.

Ta-ra.

 

Advertisements

About James Gasson

My name is James Gasson. I am a musician, sound engineer, artist and chief operator of Third Circle Recordings. I journey through life trying to work out what exactly is going on whilst doing my best to avoid tripping over. Some days are more successful than others. View all posts by James Gasson

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: