Convolution reverb has been established among audio engineers as a popular option when it comes to realistic reverberation. While the artificial sound of digital reverb has become an aesthetic standard for many applications, and “tweak-heads” value the extensive control this method offers, convolution reverb has become the go-to tool when it comes to the realistic recreation of acoustic spaces, as well as vintage equipment like plate and spring reverbs.
But how does it work? And how close of a representation of the actual room is a sample-based convolution reverb? I decided to conduct some testing to see for myself.
How does convolution reverb work? To sample a room for convolution reverb, an impulse response (IR) is created. An IR is the recording of a short noise burst played back in an acoustic space, or through an arificial reverberator. The impulse is used to excite the reverberation evenly across the audible frequency spectrum.
download for Waves IR1
Library Info

Because I sampled some interesting and nice sounding spaces for my research, I decided to put together a little IR library. My preferred convolution reverb plug-in to use it is Waves IR1. It eliminates the direct sound from the IR, which lets you decide on the amount of direct sound that is right for your production. You can either download the library for IR1 or just the WAV files for use with your favorite convolution reverb plug-in.

Specs: 470MB (73MB as ZIP download) - WAV - 24bit - 96kHz - peak normalized to -20dBFS

donload as WAV files
Although reverb is usually thought of as a time-based effect, reverb essentially consists of information in two domains: frequency and time. The reflections that make up the reverb have a different intensity and decay at different frequencies. The recorded IR contains information in those two domains, which is all that is needed to realistically recreate acoustical spaces. However, it does not capture any other factors, like harmonic distortion, that may contribute to the characteristic sound of vintage reverberation devices and equalizers.

In practice it proves difficult to acoustically reinforce impulses, as the short, full-frequency information will cause significant distortion in most speakers. To achieve more accurate results, a sine sweep is used. Instead of exciting the reverberation of the room simultaneously at all frequencies, sine sweeps move a sine tone through the frequency spectrum over time. The sine sweep response captures exactly the same information as the IR, only in a different format. Because an IR is needed for convolution reverb, the sine sweep response then needs to be mathematically converted from the time domain to the frequency domain, using the Fast Fourier Transform. This is done on the computer, using special software, which in my case was “Fuzz Measure Pro”. The resulting IR audio file is what convolution reverb software like Waves IR1 or Avid TL Space uses to create reverberation. The IR is multiplied with every single sample of the audio signal.
In order to evaluate the accuracy of convolution reverb, I recorded short audio examples along with sine sweeps.  I then applied convolution reverb created from the sine sweeps to the dry audio examples and compared them to the versions I recorded in the actual space

I recorded various spaces around the Jacobs School of Music. I recorded in Auer Hall, Recital Hall, Ford Hall, and the MAC stage, as well as a staircase in the MAC and the plenum below the MAC auditorium. The acoustics of a space are very different in different locations throughout the room. Each IR only captures one specific combination of the positions of the speakers and microphones. Therefore the placement is crucial for the sound of the reverb. I recorded in three different positions for each of the halls. Auer Hall has banners that can be raised and lowered to mechanically change the reverb time of the hall, so I recorded a sample with the banners all the way up, half way down, and all the way down for each position in Auer Hall. I decided to place the speakers on stage, where a performer would be located, and recorded with a spaced pair of omnidirectional microphones at various distances.
Audio Examples

Dry Drums

The same Drums played back, and recorded in Auer Hall

Drums with Convolution Reverb from Auer Hall

While I encountered some issues with noise, which were largely due to limitaions of my equipment, as well as activity in the halls, the convolution reverb examples generally sounded astonishingly similar to the examples recorded in the actual hall. All the characteristics of the reverb were captured in the IR and recognizable in the convolution reverb version.

These results explain why convolution reverb is the first choice when it comes to the realistic recreation of natural spaces. Applying convolution reverb to a digital audio signal will make it sound like it was recorded in the same space, through the same microphones, in the same position as the IR was recorded. Not more and not less than that. Modification is very limited. This faithful reproduction is both, a great strength and weakness of convolution reverb.