Documentation

 

 

1. Overview and History

 

Convolution Reverb is a technique that has gained a lot of attention over the last few years and has now firmly situated itself as one of the most powerful tools for artificial reverberation. Exciting any resonating system (like a room) with an impulse, one can record the resulting response of that system, yielding what is known as an Impulse Response. Through the mathematical process of convolution, this Impulse Response can then be imposed on an arbitrary sound source to create the realistic impression of this source being played throuh (or in) the system the Impulse Response (henceforth referred to as „IR“) was recorded in. While Convolution Reverb for stereo applications is now widely used, the possibility of multi-channel convolution remains relatively obscure outside of certain applications like VR gaming. This contribution aims to fill this gap by offering a comprehensively sampled room under creative commons, free to use for artistic projects of any kind.

 

The St. Gertrud Church in Cologne is one of the most remarkable specimen of contemporary sacral architecture after the second world war in Germany. Designed by renown architect Gottfried Böhm and constructed from 1962 – 1965 in the style of Brutalism. Its structure is composed of concrete, employing geometric patterns in its interior design. Of special importance are the four folds of the roof construction, which give both the optic as well as the acoustic room impression additional spaciousness. It features an underground crypta that is connected to the main hall through a wide, open staircase, as well as a chorus/organ chamber that is comparatively low in reverb. All of these features create a unique sonic space, with a noticeable church-like impression but nonetheless a distinct character. Although rarely used for services, St. Gertrud remains a consecrated church. It hosts regular exhibitions, concerts and other cultural events.

T he interior of St. Gertrud

 

2. Recording and processing Method

 

For generating the Impulse Responses, the exponential sweep technique was employed. Instead of a single impulse covering the entire frequency range, a sinewave sweeps through the audible range over the course of 60 seconds, as this method is more reliable and resistant against possible external noise. This was recorded through a variety of microphones in different set-ups. Time constraints were a concern so not every position has been recorded in all possible combinations.

 

The IRs were recorded with a Sennheiser Ambeo microphone for Ambisonics, a Sennheiser binaural head equipped with two OKM II as well as two pair of Schoeps CMC 621. The Schoeps have been used either in an AB quad stereophonic set-up, regular AB Stereo or as super-wide pseudo-stereo (as indicated in chapter 5). Some recording positions have been recorded in true stereo, although most use a monophonic sound source.

 

For the purpose of portraying the full range of this unique sonic space, we first focused on the positions that would be closest to a regular listeners or visitors experience, recording these with all microphones and in true stereo. We then moved on to the more experimental positions, like positioning the speaker in the crypt or the microphones in the organ chamber or the confession chairs. As these were logistically more difficult to realize, they were recorded with fewer of microphones.

 

All IRs were recorded first as sweeps. To generate a usable impulse response, the recorded sweep (that includes the rooms reverberation) has to be deconvolved with the original, clean sweep. The free tool fscape offers this possibility, as well as very useful batch processing capabilities. The resulting file was cut and further cleaned with Izotope RX. This yields an IR file that can be used with any convolution reverb plugin.

 

3. Suggested Workflow for implementing Ambisonics Impulse Responses

 

Ambisonics Impulse Responses are a relatively new area and not a lot of attention has been given to them outside of some VR applications and gaming. Wwise offers a library for implementation in this context and the french manufacturer Noise Makers has released Ambi Verb HD, a commercial plugin coming with a library of 1st-order Ambisonics Impulse Responses. However, neither of these solutions are open source or freely available so we set out to find an alternative way of utilizing Ambisonics IR. It should be noted that any multichannel convolution is a highly CPU-intensive undertaking so efficiency plays a huge role.

 

The Ambisonics in the attached folder are supplied in A-Format, as they have been recorded. For implementing this method however, they need to be converted into B-Format, which can be done with the free Sennheiser Ambeo plugin. These B-Format IRs then need to be loaded into the HO-SIRR (Higher-Order Spatial Impulse Response Rendering) tool by Leo McCormack et. al. In this tool, the desired speaker set-up needs to be entered in the „Output Loudspeaker RIR“ section. Rendering this yields one IR file for each loudspeaker based on its distribition in the space and the original room impulse response. These can then easily be appliad to each corresponding ambisonics channel in any multi-channel capable tool. We used Max MSP for this purpose, but it should be possible to achieve similar results in PD or any comparable software. Although the fact the IRs were recorded in first order Ambisonics is limiting the precision of localization, the resulting room impression is highly realistic and, in our opinion, warrants further creative exploration.

 

4. Software and Equipment used

 

Speakers

• RCF TT25-A

 

Microphones

• Schoeps CMC621

• Ambisonics AMBEO VR

• OKM II

• Sennheiser binaural head

 

Recording Equipment

• Aspen Media Andiamo.MC

   

Software

• Ableton Live 10 (www.ableton.com)

• Fscape (https://www.sciss.de/fscape/)

• Izotope RX 8

• HO-SIRR (https://leomccormack.github.io/sparta-site/docs/plugins/hosirr/)

• Max MSP 8

 

 

 

6 . Floor Plan of Recording Set-Ups - See Download

​

7. Credits and Acknowledgement

 

This project has been made possible through the support of the Folkwang University of the Arts and the Institute for Computer Music, specifically Jorge Vallejo and Michael Edwards. It has furthermore been made possibly by the support of the St. Gertrud Kirche + Kultur and Peter Otten, who allowed us to record there. This project has been realized as part of the Folkwang University’s coursework. Thank you to the course participants Leon Focker, Theo Voerste, Simon Bahr and Christian Duecker, who have helped with recording and set-up. Thank you to Hans Holger Rutz, Leo McCormack and countless others that are providing invaluable tools for free.