Rotating machinery sings in tones. A healthy induction motor produces a fundamental frequency tied to shaft speed and harmonics stacked at integer multiples. Bearings add their own signature when races develop pits or cages loosen. Mounting hardware rattles when bolts fatigue. Trained vibration analysts use accelerometers on bearing housings and years of catalogued fault frequencies. Sonarish offers something humbler and still useful: record a spectrum when the machine is known good, store it as a baseline on your phone, and later compare new recordings to see which frequency bins grew louder — all without uploading audio to a cloud service that promises magical diagnosis.

The approach is acoustic fingerprinting rather than full physics simulation. You are not solving Newton's equations for a drum assembly. You are asking whether today's sound contains more energy at eighty-seven hertz than the baseline captured eighteen months ago when the washing machine still spun smoothly. That question is answerable with a microphone, an FFT, and honest diff math.

A repeatable workflow in the field

Consistency beats clever algorithms. Place the phone at the same standoff distance — half a meter to one meter is typical — on the same surface if possible. A rubber mat on a metal panel reduces extraneous vibration paths. Record thirty to sixty seconds of steady-state operation after the machine has warmed up, not during startup transients when currents surge. Tag the capture as baseline with a human-readable machine name.

Months later, repeat the geometry as closely as you can. Sonarish computes the short-time Fourier transform of the new capture, converts magnitudes to decibels, and subtracts the baseline spectrum bin by bin. The diff view highlights sustained rises above about six decibels. Transient clicks from solenoids or relays can be suppressed with an optional median filter so one-off events do not paint false alarms across the spectrum.

Reading a diff without pretending to be Oracle

Suppose a front-loading washer spins at six hundred RPM during the extract cycle. Six hundred revolutions per minute equals ten revolutions per second, so you expect strong energy near ten hertz and its harmonics. After a year and a half of service, diff view might show plus twelve decibels near eighty-seven hertz and nearby multiples. That frequency may correspond to a bearing cage calculation for the drum diameter — or it may reflect a loose counterweight. Sonarish does not auto-label the fault. It shows the change clearly enough that you can search bearing frequency formulas, call a technician, or decide the noise is still acceptable.

That restraint is intentional. False confidence destroys trust faster than a raw graph. We would rather be the instrument you bring to the basement than the app that confidently mislabels a belt squeal as catastrophic bearing failure.

Limits you should respect

Microphone saturation ruins everything. Pressing the phone against a vibrating metal panel drives the analog front end into clipping. The clipped waveform looks louder in time domain but its spectrum lies — harmonics appear that were not physically present. Background room noise changes between sessions when HVAC seasons shift or doors open. Different phone models have different frequency responses; compare on the same device when tracking a machine over years.

Signal processing foundations live in our FFT for mortals post. Level calibration and A-weighting context appear in A-weighting and decibels explained. If you want to see raw spectra before diffing, Phyzix exposes a live microphone analyzer suitable for classroom demos.

Where this fits in the uranashel stack

Baseline comparison is the feature that turned Sonarish from a party trick decibel meter into something I keep installed next to the fuse panel. It shares FFT parameters with the live spectrum view, uses the same approximate SPL calibration, and stores sessions in the app sandbox described in our on-device architecture. No account, no upload, no model trained on your washing machine humming at odd hours. When the diff stays flat for months, you have evidence the repair held. When it creeps upward, you have a timestamped graph to show a technician instead of vague words about " sounding rougher lately."