Heart sounds, PCG audio, clicks, knock & tumor plop

Cardiophonogram in the overlay

Select PCG in Hemodynamic Overlay for a separate bottom lane with relative amplitude. It shares the pressure/ECG clock, hover/touch inspection and main pause control. Its fixed scale cannot alter the pressure axis.

Short damped oscillations depict the existing M1/T1, A2/P2, clicks, snaps and selected S3/S4/knock. Event phases use pressure-derived valve opening/closure. Positive pressure gradients gate selected valve murmurs, with delayed onset for the prolapse example. Relative amplitude is bounded and does not grade a lesion; a quiet signal does not exclude severe disease.

The synthetic frequencies are approximately 30 Hz for low-frequency gallops, 70 Hz for closure components and 105 Hz for clicks. A 600 Hz drawing lattice resolves these bursts without altering the 250 Hz pressure lattice. Sound-event histories are tied to each mechanical beat’s start and duration, so crossing a beat boundary, scrolling or revisiting a paused time cannot reuse another beat’s sound times. No audio recording, microphone response, chest transmission, prosthetic acoustic signature or comprehensive shunt-murmur model is supplied. VF, ventricular pauses and absent native flow do not receive an organized acoustic cycle.

Listen to the synthesized PCG

Select PCG, keep the overlay open and choose Listen. Adjust Volume in the same browser; headphones may make low tones easier to hear. Listening requires a local action, including on a student screen. A teacher can share the waveform and running clock but cannot enable a student’s audio or change their local volume. Hover and paused inspection select displayed samples without scrubbing or triggering sound.

The browser uses a dedicated acoustic sampler with the same pressure-derived event times, rhythm gates and relative event strengths as the visual PCG. Damped vibrations are evaluated at each audio-device PCM time. The expensive clinical envelopes use a fixed 1000 Hz cache; separately filtered acoustic noise uses a 2000 Hz grid with smooth interpolation. The drawing sampler remains a compact illustrative trace rather than a literal plot of every audible frequency. There is no microphone access, recording or audio network transfer. Nonfinite values become silence. A fixed 0.4 playback gain multiplies local volume; neither chunks nor individual beats are normalized. Acoustic shaping and headroom are original rendering choices, not fitted patient coefficients or calibrated sound-pressure levels.

Web Audio schedules short in-memory buffers no more than 120 ms ahead. A 15 ms initial scheduling lead and a check after PCM rendering prevent a slow cold render from starting a buffer in the past; an exhausted deadline re-anchors queued playback. These are engineering choices, not physiological timing constants. A rewind, clock drift above 70 ms or a changed PCG sampler cancels queued buffers and re-anchors playback to the current simulation time; a frozen monitor clock cancels queued sound instead of replaying it. This keeps rhythm and waveform context together without queuing an entire future heartbeat. Browser and audio-device latency can still create a modest audible delay; sample-exact synchrony at the listener’s ears is not guaranteed.

Main Pause cancels queued playback while keeping the local Listen choice armed, so Resume can continue it. Closing or hiding the overlay, deselecting PCG, a hidden document or leaving the page turns listening off. Future buffers are cancelled; an already playing fragment may fade for up to 6 ms to avoid an abrupt click. The audio shares the same synthetic events and murmur envelope as the PCG; it does not add measured chest transmission, stethoscope filtering, a listening position, diagnostic sensitivity or lesion severity.

Audible heart sounds and murmur texture

Normal S1 retains its short, low-frequency damped body; S2 is shorter and brighter, with a decreasing-frequency vibration. These normal profiles remain unchanged by the pathological-texture update. Gallops and clicks/snaps retain separate profiles. Acoustic onset is anchored to the unchanged modeled M1/T1 and A2/P2 event times: no burst anticipates its mechanical event. A soft attack means the acoustic peak occurs shortly after that marker. This causal audible envelope differs from the symmetric Gaussian marker drawn by the schematic PCG. Components from the preceding beat are retained while their bounded tails remain active, but pauses, VF and absent native flow remain silent.

The synthesis is inspired by published multicomponent S1 and transient S2 models and by the measured relative spectra of S1 and S2. The cited S2 chirp study used four pigs with pulmonary hypertension; it does not establish human diagnostic coefficients. The actual frequencies, decay times, attack and noise mix here are original educational sound design. At reference event width, S1/S2 retain 4 ms attacks and maximum tails of 140/110 ms; S3/S4 retain 6/180 ms and clicks 2.5/70 ms. The knock now has a separate 2.5 ms attack, 90 ms maximum tail and a main chirp decreasing from 170 toward 90 Hz. Existing event width scales these durations in compressed cycles. These are rendering parameters, not new physiological timing measurements.

Each valve murmur now combines two overlapping, independently seeded noise bands. A smooth blend changes their balance with the current modeled driving pressure, adding spectral movement within the existing clinical envelope. The synthesis follows the general idea of time-varying stochastic murmur models; the cited Debiais study used selected bioprosthetic-valve recordings and simulations, not a universal native-valve spectrum. Band limits, blend weights, pressure-response knees and grain depth here are original rendering parameters. They do not measure lesion severity or establish diagnostic frequency boundaries.

AS and PS have separate rough ejection textures, MR and TR have separate blowing textures, and MS and TS retain a lower inflow rumble. The selected acute-MR example has a lower, rougher mix than the chronic-MR example. AR keeps a brighter blowing mix, while the PR texture illustrates a lower-pressure example with a lower spectral balance; a separate high-pressure Graham Steell texture is not selected automatically. The current positive driving-pressure gates and marker times remain unchanged. Softer acoustic pressure-response knees preserve more of the rising and falling ejection contour and diastolic decay instead of flattening early at maximum murmur gain. Existing beat strength still controls relative emphasis; these acoustic changes do not alter the hemodynamics or assign a clinical grade.

Small, smoothly varying nonperiodic grain changes the murmur texture through absolute simulation time. The two bands and the grain use independent deterministic seeds, without a repeated beat-length noise loop or an added pitched oscillator. Their 97-tap Blackman-windowed filters run on a 2000 Hz noise grid; a 16-tap windowed-sinc reconstruction with 64 interpolated fractional phases supplies native PCM. Normal S1/S2, gallops, clicks, snaps, knock, rub and tumor-plop carriers retain their existing profiles. Finite filters have transition bands rather than hard acoustic cutoffs.

Pathological noise streams and the shared reconstruction table are created only when requested. Bounded 512-cell caches and finite filtering reproduce the same samples after seeking or chunk changes, without resetting noise at a beat or playback-buffer boundary. The existing six-millisecond prolapse-onset attack remains an acoustic smoothing choice. No ambient hiss or artificial room reverberation is added. Spectral and continuity checks assess software behavior; they do not establish perceived realism or clinical validity.

Noise-filter variance is scaled once from its fixed coefficients; no rendered buffer or individual beat is normalized. Fixed synthesis and playback gains preserve relative event emphasis before bounded soft headroom and the local Volume control, but peaks and perceived loudness still vary with the waveform. Speaker/headphone response and volume change perception. Recorded stethoscope responses differ substantially: these color choices do not reproduce a particular instrument, listening position, measured chest transmission or patient acoustic signature. The resulting audio remains original educational synthesis, not a patient recording or a validated auscultation model.

Supported pathological sound examples

The abnormal textures are original teaching illustrations around the existing pressure-derived sound events. The normal S1/S2 rendering remains unchanged by this murmur-quality update. Pitch and texture describe selected examples, not a recording library or a complete auscultatory examination.

Clinical murmurs vary with cause, flow and loading. Acute MR can be brief or quiet as pressures equalize; acute AR need not sound like chronic AR. Pulmonary regurgitation (PR) without pulmonary hypertension can be shorter and lower-pitched than a high-pressure Graham Steell murmur. The current renderer does not identify these etiologies from sound, and a quiet example does not exclude severe disease.

Pathological sound limits

HOCM can change the selected S2 splitting and S4 markers, but an HOCM outflow murmur is not synthesized. Apart from the explicit TOF RV-outflow example, general shunt-flow murmurs and continuous ductal murmurs remain outside the acoustic coverage, as do prosthetic signatures and a complete set of right-sided gallops or opening snaps. The presence of a hemodynamic condition does not imply that every expected clinical sound is rendered.

No listening site, radiation, patient chest transmission or calibrated sound-pressure level is supplied. The referenced qualitative descriptions and recorded spectral studies guide teaching choices; neither those references nor software checks validate this synthetic audio for diagnosis, lesion grading or auscultation proficiency testing.

Inflammatory rub and unrepaired TOF

The acute and transient-constrictive pericarditis examples add a representative scratchy rub, separate from valve murmurs and the constrictive knock. It can have atrial-systolic, ventricular-systolic and early-filling components; AF omits the organized atrial component. Resolved inflammation removes this component, without changing unrelated normal sound profiles. A real rub is variable and not obligatory; this is original synthetic teaching audio, not a recording or a diagnostic sensitivity model.

Unrepaired TOF adds an illustrative RV outflow ejection texture and a quieter pulmonic closure component. Its large unrestrictive VSD does not receive a separate high-gradient murmur. Low pulmonary flow can reduce a real outflow murmur despite substantial obstruction; loudness is not a severity or cyanosis scale. General congenital murmur coverage and repaired TOF acoustic signatures remain outside the model.

S1 and S2 splitting

S1 separates mitral and tricuspid closure by a short reference interval, widened in the selected RBBB example. S2 follows A2/P2 timing with ordinary inspiratory widening, wider splitting in selected RBBB/PS examples, and reversed/paradoxical ordering with delayed LV ejection in selected left-sided conduction/outflow states.

Large ASD has a fixed-splitting example; PAPVR alone does not inherit it. Pulmonary hypertension uses an accentuated P2 and a selected narrow split rather than assuming fixed splitting in every patient. Positive-pressure ventilation does not blindly apply spontaneous inspiratory rules.

Opening snap and clicks

The opening-snap delay uses LA pressure and LV systolic pressure: 0.15 − 0.004 × (LA pressure − 3) + 0.0002 × (LV systolic pressure − 120), bounded to 0.045–0.14 seconds. This is a synthetic timing relation, not a valve-area equation. Transmitted external pressure is excluded from its filling interpretation.

Ejection clicks track the relevant pressure-dependent opening event. The MVP click shifts with loading/prolapse factor rather than grade alone. Very short cardiac cycles compress some timing to fit available diastole; the tooltip identifies that limitation.

Gallops and knock

A congested selected myocardial/regurgitant phenotype can add S3. S4 requires an organized atrial contraction and a compatible stiff-ventricle phenotype, so AF does not receive a normal S4. Marker presence and amplitude are teaching assumptions, not diagnostic sensitivity or specificity.

The constrictive knock is associated with abrupt arrest of early ventricular filling after the mitral valve has opened. Tyberg and colleagues recorded a brighter early sound 90–120 ms after A2 in a small constriction series, coinciding with a premature filling plateau; some patients had no knock. That interval guides a selected example, not a universal clock or a substitute for checking the current filling event.

The implemented knock targets 105 ms after modeled A2, stays after modeled mitral opening, and is compressed or omitted when the remaining diastole cannot contain it. Its marker and audible onset use the same per-beat event. The target is not a filling-arrest time inferred from a measured ventricular volume curve.

Tumor plop and mobile inflow obstruction

The selected left atrial myxoma adds a separate damped, low-pitched plop when diastolic mass motion is checked. Its event follows the shared mobility/obstruction state rather than a fixed mitral-stenosis opening-snap delay. The repeating pattern is an authored demonstration; real positional mobility and acoustic pitch vary, and a plop is not obligatory.

The associated inflow texture follows positive atrial–ventricular driving pressure and the selected obstruction. Neither its loudness nor the plop establishes mass size, histology or embolic risk. Low plop, brighter knock and short opening snap are deliberately distinguishable synthetic profiles, without a validated auscultation classifier.