Building the Erbe-Verb: Extending the Feedback Delay Network Reverb for Modular Synthesizer Use Tom Erbe UC San Diego Department of Music tre@ucsd.edu ABSTRACT The paper describes the design of a Eurorack Synthesizer reverb module that extends the feedback-delay network re- verb in several ways. The extensions are oriented around the needs of live electronic music: 1) the creation of ab- stract resonant spaces which may or may not have a resem- blance to acoustic spaces, 2) quick modulation of all pa- rameters – modeless morphing between all states of the re- verb. In this short paper I will open up the design of the Erbe-Verb, show the techniques used, and how they com- bine to make a very flexible performance reverb processor. 1. INTRODUCTION AND DESIGN CON- SIDERATIONS The Erbe-Verb was designed for use in the Eurorack synthe- sizer system, a modular synthesizer format created by Doepfer in 1995 [1]. The goals for this processor were to create a reverb that would integrate into the ecosystem of a patchable synthesizer, both functionally and sonically. Most reverb processors are used as the last effects device, the room that one places an instrument in. Although the Erbe-Verb supports this use, I wanted it to be more fully integrated into a synthesizer voice. For this reason, all pa- rameters are fully modulatable, designed to be controlled by common control-rate signal generators such as envelope generators, low frequency oscillators and envelope follow- ers. Each note, moment, or gesture can have it’s own reso- nant character. In addition, all time-related parameters (modulation speed, pre delay time) can to be clocked to ex- ternal pulses. And finally, when possible, the parameters can be modulated at the audio-rate to allow audio effects not usually associated with reverb like FM sidebands on the internal delay or absorption filters. By allowing all parame- ters to be freely interconnected and modulated by any other synthesizer module, the synthesist is given flexibility to create new ways to use and incorporate the reverb in a larger patch. Although the sonic goal of the Erbe-Verb is to emphasize synthetic or unusual resonant spaces, its basic characteristics are the same as any other reverberation device or perform- ance hall. It provides sustained sound after the source is silent. It shapes the timbre, possibly smoothing attacks, add- ing distinct echoes or diffusion. Finally, the reverb gives a strong sense of location, even when the resonant space is completely unrealistic. To achieve these basic goals, I stud- ied many previous designs, incorporating elements of each. With this solid foundation, it gave me freedom to experi- ment with extensions to the design. 2. BASIC DESIGN The Erbe-Verb is based on a 4-delay feedback delay net- work reverb (FDN) as proposed by both Michael Gerzon [2], and Puckette and Stautner [3]. This model was chosen after investigating several topologies: the early experiments of Schroeder, the multi-tap feedback delay line of Christo- pher Moore’s Ursa-Major Spacestation [4], and David Gre- isinger’s figure-8 loop [5]. The feedback delay network was chosen for several charac- teristics. First, with a unitary feedback matrix it is a lossless gain design. This gives it the ability to sustain indefinitely when the feedback gain is 1.0, and also decay exponentially when the gain is below 1.0. This ability to sustain indefi- nitely is of importance to frozen or infinite reverb. Second, it allows the use of several short delays to quickly build up echo density. Finally, early reflections can be easily in- serted as multiple taps in the delay network. This prototype FDN makes for a very compact and efficient reverb that can be easily modulated and modified. 2.1 Added Diffusion - All Pass Filters Percussive signals require even higher initial echo density than the FDN can provide. For this reason, allpass filters are placed before each delay in the FDN. The gain on the all- pass filters can be varied from 0.0 to 0.8 so that the density of early reflections can be increased. It should be noted that although allpass filters have a flat frequency response over time, the tail will ring. For this reason, the gain is kept low enough to avoid audible ringing, and the relative delay times of each allpass are kept mutually prime so that there are no common resonances. The delay time for each allpass is short enough so that individual echoes are not heard, they are kept Copyright: © 2015 Tom Erbe. This is an open-access article distributed under the terms of the Creative Commons Attribution License 3.0 Unported, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. ICMC 2015 – Sept. 25 - Oct. 1, 2015 – CEMI, University of North Texas – 262 –