Ambi Effect Response Trace

s3g Ambi Effect Response Trace 64 captures a bank of short directional responses from an incoming 1OA through 7OA ACN/SN3D field: one convolution kernel for every pickup in the selected Listener body. The source, captured spatial differences, and result remain ambisonic. The response can stay local, cross the field, diffuse through neighboring regions, or roam around the Listener.

Workflow

1. Place Response Trace on an ambisonic track and select the field order.
2. Follow it with an ambisonic decoder. Additional Ambisonic effects may be placed before the decoder when their supported order and channel count match the field.
3. Leave Body at AUTO and reduce Mix or Response gain.
4. Set Time, play the sound that should become the kernel, and press Capture. Capture uses the next incoming passage.
5. The completed response remains bypassed. Press Apply only when ready, then use Apply / Bypass to A/B without deleting it.
6. Use Tone, Response gain, topology, and the directional mask to place the convolved field.

Capturing a Response

Capture first decodes the HOA field through the selected Listener model, then records every virtual pickup for 0.05 to 1.5 s. Each pickup becomes its own kernel, so a response arriving strongly from one region is not flattened into a shared mono trace. The kernels are DC-filtered and faded at their tails, then scaled together against bounded peak, energy, and absolute-sum budgets. This linked scaling preserves their relative directional levels.

Starting a new capture fades the old response to bypass. After recording, response conditioning and frequency-domain kernel preparation are spread across successive audio blocks instead of creating one large processing spike; the Listener display reports the complete preparation phase. Completion remains bypassed unless Apply was explicitly requested during preparation, in which case that request waits for the bank and then fades in. Changing Order or Body during an active capture cancels it because either change alters the pickup analysis.

An arbitrary captured passage is a creative convolution kernel, not automatically a measurement of an acoustic system. To capture a recognizable physical response, feed the plugin a prepared impulse or a response recovered from a known excitation. Short percussive, noisy, and decaying sounds are often the easiest creative starting points.

The captured body and all of its directional kernels are embedded in CLAP project state together with the controls, applied or bypassed state, and camera view. No external media file is required when the project is reopened. Clear first fades an audible response to silence and then deletes it; Apply / Bypass retains the response.

Signal Flow

1. Decode the incoming HOA field onto the selected Listener body. During capture, store one response kernel per pickup.
2. Give the complete decoded body one linked excitation budget so a hot field cannot drive different regions into unrelated gain reduction.
3. Convolve each Listener pickup with its corresponding directional kernel using uniform partitioned convolution.
4. Apply Response gain and Tone, then contain an excessive linked wet return before soft bounding.
5. Apply Local / Cross / Diffuse / Roaming topology and the directional wet mask.
6. Re-encode the processed correction and combine it with a latency-aligned copy of the original HOA field.

The plugin reports 1024 samples of host latency. Its dry path is delayed by the same amount, so Mix remains aligned with partitioned convolution. The reported tail is the captured response length plus that fixed latency while the response is applied.

Listener Body and Topology

AUTO uses Icosa12 for 1OA and 2OA, Dodeca20 for 3OA, and Sphere24 for 4OA through 7OA. You can also select Tetra4, Cube8, Icosa12, Dodeca20, or Sphere24 explicitly. The body selected at capture time defines the stored kernel bank. If Order or Body is changed later, the response briefly fades out, stored kernels are assigned to the new body by nearest direction and rebuilt when necessary, then the response fades back in. The visual Listener shows input energy, response activity, and mask attenuation.

• LOCAL returns convolution where the field excited it.
• CROSS returns each response from the opposing Listener region.
• DIFFUSE shares response energy through neighboring regions.
• ROAMING moves the response relationship around the body at Roam Rate.

Controls

Level Safety

Kernel conditioning limits several kinds of captured energy instead of normalizing peak alone. Its gain is linked across all stored kernels so it does not erase the directional level pattern. During playback, a linked decoded-excitation governor targets a 0.25 Listener peak with immediate reduction and a 250 ms recovery. A second linked governor targets approximately 0.50 after Response gain, before soft bounding and an emergency wet slew limit. These stages prevent a strong kernel or fast control move from turning into a crackle while preserving one shared spatial gain relationship.

After Mix and the smoothed Output control, one linked soft-knee -1 dBFS guard protects the complete HOA output with immediate reduction and a 300 ms recovery. It remains active while the response is bypassed, so positive Output gain cannot expose an unguarded dry path. GOV reports whichever linked excitation or wet-return governor is working harder; SAFE reports the final output guard. Persistent activity means the safeguards are audibly shaping the result, so lower Response, Mix, or Output for a more transparent sound.

REAPER Routing

Set the track and plugin pins to the complete channel count for the selected order—4, 9, 16, 25, 36, 49, or 64—with one-to-one channel mapping. Ensure plug-in delay compensation is enabled so the reported 1024-sample latency remains aligned with other tracks. Host project state and title-band LOAD/SAVE presets include the directional kernel bank, controls, and camera view; selecting INIT restores parameter defaults without clearing the captured response.