A method loosely based on several established techniques for recording acoustic music and ambient sound, including:
- Optimal Stereo Signal* by Jürg Jecklin.
- Bilateral Ambisonics** by Z. Ben-Hur, D. Alon, O. Berebi, R. Mehra, and B. Rafaely.
- OCCO*** popularized by Robert Sandy “Boojum” Noyes, Jim Norman and Tony Faulkner.
- STAAG, by sound engineer and researcher Jamie Tagg.
Description: Two Ambisonics subarrays are centered over each side of a 35cm diameter round baffle, and stereo sets of microphones are placed beside them. OVJ is short for Omni Variable Jecklin.
How each pair works:
- The pair of Ambisonics microphones (capsules 5 to12) separated by 36cm deliver Jecklin-affected* bilateral** signals.
If using FOAs, the arrays are physically rotated 45° so the leftmost and rightmost capsule are opposite to each other. Resulting signals are correspondingly rotated once converted to Ambisonics B-Format. We can be call this Modified for Symmetry First-Order Ambisonics (MFS-FOA).
Even though in Ambisonics theory this should not be necessary, in practice the capsules probably don't exhibit textbook-like polar responses, therefore a more symmetrical layout results in more symmetrical results.
- The inner non-Ambisonics pair (1, 2) can adhere to OSSv2: omnidirectional capsules, also separated by ∽36cm, as close as physically possible to the Ambisonics sub-arrays, to facilitate W-signal replacement:
- Since calculations by Oren Levy and empirical testing have shown an SRA (Stereo Recording Angle) of 82° for Jecklin OSS, distances and angles for other stereo sets within this array default to that same SRA as a starting point. The outer pair (3, 4) is therefore separated by ∽78cm at 0°. This pair can be used for Jecklin effect mitigation if needed, while providing other spatial cues. Using APEs for omni capsules or cardioids can provide something original OSS lacks: front-back differentiation:
- The array allows for unobtrusive hyper-realistic location recording from a single stand with abundant versatility when mixing.
- With time-coherent processing of signals from the Ambisonics arrays and the other pairs, they can all be used in a mix. As in OCCO, mixing can cause comb filtering, which can require mitigations such as predominance by >9dB.
- The geometry allows for extra bass forward gain from the sum of all capsules on each side, LPF applied to that sum avoids comb filtering.
- Hyper-realistic multichannel results are available with modest processing. Suggested for Dolby Atmos within Apple Logic Pro:
- Audio Brewers Transcoder for Ambisonics A to B format.
- Audio Brewers AB Rotator for compensating MSF FOAs in B-format.
- W-channel replacement with high-pedigree omnis (1, 2) closest to the subarrays.
- Audio Brewers Advanced Decoder from B format to beam-formed signals corresponding to discrete Atmos channels.
- Panning each Ambisonics subarray ∽50° and 20% or more to each side delivers realistic immersion. Too little panning and the image can become blurred, too much panning and you can get a "hole in the middle". It's also a good idea to attenuate the mid front from each side by at least 3dB because they will be summed.
- As in STAAG, backfacing beams from Ambisonics allow for less dependance on artificial processing and confer a more realistic sense of the space where the recording took place.
More findings:
- The “Jecklin effect” survives fold-down from Atmos to stereo.
- The experience on headphones is similar to binaural, to an extent that a more natural result is possible by turning off frontal binaural rendering in the Dolby Atmos workflow, something which unfortunately does not seem possible with the Apple binaural renderer at this time.
- Excellent mono-compatibility occurs because a significant portion of the signal originates from Ambisonics, unlike traditional binaural which falls apart when summed.
- If less channels are available or more portability is required over quality and options, the outer and/or inner pairs can be sacrificed and both mono, stereo and spatial signals can be extracted from the Ambisonics pairs.
- Baffle absorption is critical for vocals and similar sounds in which comb filtering due to reflections can be very noticeable. Basotect is a winner for absorption on each side.
- Baffle core rigidity seems critical for LF performance, probably because flexing turns it into a passive radiator which passes on LF. Acrylic and carbon fiber work well.
Notes:
Tests took place using:
- 2x Rode / Soundfield NT-SF1 Ambisonics microphones,
- 2x Schoeps MK 5 microphones, usually in omnidirectional mode,
- 2x Earthworks QTC30, somteimes with custom 40mm APEs (Acoustic Pressure Equalizers), or 2x Immersive Soundscapes Earsight Thumb microphones.
- 35cm round Jecklin disk baffles made out of varying materials, including perforated metal, polyester fiber, Basotect, carbon fiber and acrylic.
* The “Jecklin effect” is the result of the Optimum Stereo Signal, an acoustic process accomplished with a round vertical baffle between matched capsules. OSS delivers spatial cues from varying intensity and frequency response depending on incidence angles, as well as time precedence caused by inter-capsule distance. Unlike dummy-head or current software-based binaural, this method is highly compatible with stereo loudspeakers.
** OVJ is not by-the-book Bilateral Ambisonics, but a simplified process with Ambisonics arrays for each, where spatial cues from the Jecklin effect are presented to the listener somewhat separately for each ear, similar to the way OSS works.
*** OVJ inherits from traditional OCCO the concept of the single-axis horizontally spaced, vertically coincident multi-microphone phased array, but not the practice of setting the inner pair to a different SRA than the outer pair.
The OVJ name is inspired by Christian Amonson and his outside-in nomenclature.
SRA prediction diagrams come from the Neumann Recording Tools app.
All original content on this page by @ignace / Ignacio Rodríguez de Rementería is open source licensed under CC BY-NC-SA 4.0 https://creativecommons.org/licenses/by-nc-sa/4.0/Many thanks to the following friends and teachers for their help in alphabetical order: Benjamin Maas, Bradford Richards, Brian Peters, Christian Amonson, Darío Bustos, Eduardo Monteiro, Eric Weber, Francisco González, E., Hudson Fair, Ian Wood, Jack Reynolds, Jake Purches, Joel Rec, John Cone, Julian J. Ludwig, Lucas Guitink, Manfred Schmidt, Mario Vila, Oren Levy, Paal Rasmussen, Paul Fee, Stefan Kießling and many others from the excellent CMLR group on Facebook, and especially to Alejandro Cabrera, Carlos Fernández, Diego Rodríguez B., Don Booth, Esteban Zabala I., Gricelda Duarte, Helmuth Reichel-Silva, Iván Órdenes, Jaime Valbuena M., Joaquín Luppi, Jorge Montesi, Jorge Sacaan M., Juan P. Quezada, J. Alberto Palacios, Julio Figueroa M., Leonard Moskowitz, Mauricio Landeros, Pablo Saavedra, Ricardo Henríquez, Romualdo Castro, Sebastián Errázuriz, Xavier León and last but not least my lovely wife Patricia Reichel for putting up with me during more than a year of research, design and testing.