Rebreather Noise Profiles and Their Effect on Mapping Accuracy

Why "Quiet" CCRs Still Corrupt Cave Maps

A JJ-CCR on a 6m deep survey pass produces a measurable acoustic event every 2.1 to 3.4 seconds as the solenoid fires to add O2, plus a continuous broadband hiss from the scrubber canister as CO2-laden gas passes through Sofnolime. The rebreather's solenoid O2 injection and scrubber cycles produce periodic acoustic events that live in the 200 Hz to 4 kHz band. That overlaps exactly with the cave-geometry signal EchoQuilt needs to separate: flow through constrictions peaks around 400 Hz to 1.2 kHz, and bedding-plane reflections come back in the 600 Hz to 3 kHz band.

The DAN reference on closed-circuit rebreathers as a "different way to dive" documents reduced bubble noise as the core CCR value proposition — and that's true relative to open-circuit. But "quieter" is not "silent." Scientific diving research on CCR reliability (PMC's deep-diving device analysis) shows CCRs remain the safest deep platform, which is exactly why survey teams use them — but the same airflow-over-valves physics that makes CCRs attractive also injects periodic signals into the cave's acoustic record.

Without profiling, these rebreather signatures get baked into the map. A scrubber harmonic at 780 Hz starts looking like a chamber resonance. A solenoid click pattern at 2.3-second intervals gets auto-classified as scalloping on the passage wall. The survey ends up with phantom geometry — features that exist only because the diver was breathing. For teams doing mapping quality control passes on deco stops, uncorrected rebreather noise is the single largest source of false-positive structure.

Stitching Around the Diver's Own Rebreather

EchoQuilt treats the rebreather as the first thing to filter out of the quilt. Before a survey series starts, each CCR on the team goes through a 90-minute calibration dive in a known open-water chamber: the unit's acoustic signature gets captured across depth bands (10m, 20m, 40m, 60m) and work-of-breathing profiles. The resulting noise mask becomes the diver's personal patch-level filter.

The math is standard underwater acoustics. The Brüel & Kjær reference on underwater acoustic measurement documents FFT techniques that split a mixed signal into frequency components; EchoQuilt runs that decomposition at 30-second intervals across each dive's audio stream. The rebreather's fingerprint — solenoid duty cycle, scrubber broadband character, exhale-check-valve tick — gets subtracted patch by patch. What's left is the cave.

Three specifics that make this actually work in field conditions:

1. Per-diver signatures, not per-model. Two JJ-CCRs off the production line behave differently after 300 hours of use. EchoQuilt stores each unit's signature by serial number, so when the survey team rotates CCRs between dives, each patch gets filtered against the specific rig that recorded it. Low-signature units like Muro CCR and other STANAG AEODP-07 compliant platforms still show measurable unit-to-unit variance.

2. Scrubber-life drift correction. As Sofnolime loads up with CO2 over a dive, the scrubber's acoustic signature shifts — the canister resonance drops by 80 to 120 Hz as absorber density changes. EchoQuilt tracks dive-elapsed time and applies a drift correction to the noise mask automatically.

3. Work-of-breathing awareness. A diver at 80m sprinting to catch the DPV tow rope has a different respiratory signature than the same diver hanging a 40-foot deco stop. The filter adjusts against heart-rate-inferred work-of-breathing, which comes from the CCR's own handset telemetry. The same workload curve also predicts the next solenoid firing within roughly 40 percent accuracy across a steady-state swim, which is enough for the EchoQuilt engine to pre-clear short windows of audio expected to contain the firing event. Pre-clearing avoids the false-positive geometry that would otherwise show up at the firing instant.

4. Cell-failure and sensor-validation noise. A failing oxygen cell triggers an audible alarm on most modern CCRs, and the divers' validation responses (cell switches, manual flushes, alarm-clear button presses) all generate brief acoustic events that look nothing like cave structure. EchoQuilt cross-references the CCR alarm log against audio and excises those seconds from geometry inference automatically when the controller event stream is available.

The quilt that comes out after this patch-level subtraction is what the cave actually sounds like — not what it sounds like with a rebreather running in front of the hydrophone. This matters because NOAA's underwater noise reference makes clear that anthropogenic acoustic sources corrupt signal-to-noise in any subsurface survey. Cave environments amplify the problem because there's nowhere for stray signal to attenuate.

Tactical Filtering for Survey-Grade CCR Work

Teams pushing JJ, rEvo, AP Inspiration, or Muro units on survey-grade cave work need three additional layers beyond the baseline noise mask:

Pre-expedition calibration discipline. Every unit on the team gets a chamber calibration inside 30 days of the expedition, and every major component swap (head, canister, cells) triggers a recalibration. EchoQuilt tracks calibration age on each patch and lowers the patch's survey grade if the mask is older than 30 days. The tolerance tightens for publishable work.

Solenoid-firing schedule awareness. Solenoid firing is not random — it follows the diver's metabolic rate and the setpoint differential. Divers coordinating with scooter-assisted mapping higher speeds passes should know that accelerating from rest on a DPV can trigger a solenoid cluster within 8 seconds — EchoQuilt flags those clusters for exclusion from geometry inference.

Cross-domain pattern reuse. Mine-rescue teams filtering changing ventilation signatures face the same math in a different domain. The approach in acoustic interpretation for bratticing changes mirrors how CCR divers isolate solenoid events from passage geometry. Survey teams crossing both domains — coastal-resource cave rescue teams especially — use the same filter library.

Diluent-flush event isolation. Manual diluent flushes during a dive — to clear a stuck cell reading or to prove the loop on a buddy check — produce a brief, characteristic broadband event that bears no relation to cave geometry. Without isolation, a single diluent flush near a constriction can produce a spurious 4-meter feature in the quilt. EchoQuilt cross-references the CCR controller's event log against the audio stream and tags every flush window with a "diver action" marker so the geometry inference engine ignores those seconds. Teams running JJ-CCR controllers with logging firmware get this for free; teams on older Inspirations or unlogged units have to flag the events manually in the post-dive review.

Buddy-pair signature collision. When two CCR divers swim within 3 meters of each other on a survey pass — typical for a primary and secondary surveyor working a passage — both rebreathers contribute solenoid clicks to both audio streams. The two unit signatures cross-contaminate, and the EchoQuilt engine can confuse one diver's solenoid for a real cave event in the other diver's quilt. The fix is to capture each unit's controller event timestamps and use them to label which click came from which rebreather. Teams running NACD-style buddy-pair survey protocols already coordinate hand signals at constrictions; adding an event-stream sync at each major waypoint takes seconds and pays off in the post-dive quilt.

Bailout cylinder cross-talk. A bailout regulator hung in front of the EchoQuilt cluster on the diver's chest can produce intermittent first-stage tick noises as ambient pressure changes during depth shifts. The ticks are small but rhythmic enough to be mistaken for biological signal. Mounting bailouts on the diver's left hip, with the first stage angled away from the receiver cluster, eliminates the contamination. Sidemount divers on Yucatán cenote surveys have refined this rigging detail across hundreds of dives.

What's left on the quilt after three layers of filtering is a map you can stake publication on.

Filter Your CCR Out of the Map

Survey teams running CCRs on WKPP-style DPV tows, Ressel-style deep sump pushes, or GUE project dives know the rebreather is both life support and acoustic source. EchoQuilt's per-unit calibration treats your JJ-CCR or rEvo as a known artifact to remove, not noise to tolerate. Join the Waitlist for Cave Diving Survey Teams and bring your chamber-calibration data — we're prioritizing teams with three or more CCRs in rotation for the first filter-library build.

Share your unit inventory (JJ-CCR, rEvo, AP Inspiration, Muro), each unit's current calibration date, your typical setpoint and dive profile, your bailout staging convention (left-hip versus chest), and your federation affiliation (NSS-CDS, GUE, QRSS, NACD); we will scope a per-unit acoustic fingerprint baseline against the EchoQuilt filter library, set up the controller-event-stream sync template for buddy-pair signature de-collision, prepare the diluent-flush event marker workflow for unlogged Inspiration units, and configure the 30-day calibration-age tracker that lowers patch grades when masks expire. Priority cohort access goes to teams running three or more CCRs through active publication-grade survey campaigns.