Bilateral Ambisonics For Realistic Binaural Recording And Playback

    • June 26, 2023 at 5:54 pm #5862
      Len Moskowitz
      Participant

        Bilateral Ambisonics For Realistic Binaural Recording And Playback

        The Bilateral Ambisonics recording technique is unique in that it can realistically reproduce how we hear with our two ears, even as we rotate and nod our heads to locate sound sources in space, left and right, front and rear, above and below. It preserves all of the spatial information that we hear out in the world, with the precision and accuracy we’re used to hearing in our lives. Recorded with a pair of excellent higher-order ambisonic microphones, and played back using a low-latency headtracker, personal HRTF and calibrated/corrected headphones, the result is as close to real-life experience as I’ve ever had.

        Here’s a link to download a second-order B-format recording of Angelica Women’s Chamber Choir, recorded with the Bilateral Ambisonics technique, at Grace Church in White Plains, New York:

        https://u.pcloud.link/publink/show?code=XZJqPOVZzu41J3Whfdfr1xPA1jF76jMGRLa7

        Plogue Bidule Bilateral Ambisonics LayoutThe left microphone is the first 9 channels. The right microphone is the last 9.

        Here’s how to play it back correctly, so that you too can have this experience:

        I listened to the B-format file with the Plogue Bidule layout that’s pictured below. [I will be glad to provide a copy of this layout on request.]

        Two Core Sound OctoMics, spaced apart by 14 cm (my ITD), were recorded to a total of 16 tracks, 8 for each OctoMic. The two raw 8-track A-format recordings were encoded to two second-order B-format streams (one for the left side microphone, and one for the right) using two VVOctoEncode encoder plugins and the individual calibration files for each of the OctoMics. A small 3-degree fixed yaw offset was added to each of the B-format streams to allow the soundfields to slightly overlap, so there would be a continuity between the two soundfields.

        Two instances of the SPARTA AmbiBIN plugin, modified by Hector Centeno to eliminate their ITD contributions, then decoded the B-format for binaural playback. The right side passed through a gain control to correct for a gain offset between the two OctoMics. (One of the OctoMics was the first one ever manufactured, and it has a lower output than current OctoMics.)

        An Mbientlab MMR headtracker was mounted on top of a Focal Elear headphone set. The MMR headtracker was controlled by Hector Centeno’s NetMMR2OSC app. NetMMR2OSC produced two OSC data streams that the two AmbiBIN plugins used for dynamic yaw/pitch/roll.

        The Focal Elear headphone was sent to Sonarworks for measurement and calibration. Sonarworks created a frequency response correction for the Elear that was used with the Sonarworks Reference 4 playback correction app to provide a flat playback frequency response.

        The two AmbiBIN plugins read and used my personal HRTF file in SOFA format. The personal HRTF was created by the VisiSonics HRTF Creator app, using the camera on my Samsung S20 FE smartphone. (VisiSonics HRTF Creator is not yet available to the public.)

        And here’s a fixed-head binaural decode of that recording, using SPARTA AmbiBIN’s default HRIR set and with the headtracking disabled. Again, two instances of the modified SPARTA AmbiBIN were used. (The default HRIR sounds a bit too bright to me.)

        https://u.pcloud.link/publink/show?code=XZM1POVZNRiIgx17vu5ykvPuDhhR7F78j2fy

        And here’s an ORTF decode (with 14 cm spacing instead of 17):

        https://u.pcloud.link/publink/show?code=XZpnPOVZOAF4upvDWmFsYKVOQdQjOF1SHHl7

        Thanks to Marie Caruso, Angelica’s Music Director, for permission to record and post this recording of Ruggiero Leoncavallo’s Mattinata.

        Thanks to Zamir Ben-Hur and his associates for the creation of the Bilateral Ambisonics process.

        Thanks to Leo McCormack for the SPARTA AmbiBIN plugin and his guidance how to modify it for Bilateral Ambisonics.

        https://leomccormack.github.io/sparta-site/

        Thanks to Hector Centeno for creating the NetMMROSC interface app for the Mbientlab MMR headtracker, and for modifying AmbiBIN to remove its ITD contribution.

        Thanks to Jack Reynolds for his suggestion to slightly cross the two soundfields to maintain spatial continuity.

    You must be logged in to reply to this topic.