Proof Audio Analyzer

Open Proof and ten instruments run on whatever input you have: RTA, oscilloscope, spectrogram, XY scope, signal generator, SPL meter, level meter, LUFS meter, dosimeter and a transfer function. No account, no trial clock. That half is free and stays free. MEASURING IS THE PAID HALF, AND THE POINT OF IT IS THAT THE RESULT LEAVES A four-second log sweep, separated into the fundamental and its harmonics by Farina's method. What you get back is a file, not a screenshot. Records are plain readable text with a content hash, re-hashed on open rather than trusted: an edited record says so and is refused rather than drawn. Anyone can open and verify one free, paid or not. Export the record, or on the Mac as text or CSV, into REW, a spreadsheet or a driver tool, with the floor beside every value. So a result can be handed to a client or a colleague, and they can check it themselves. WHY THE NUMBERS ARE WORTH HANDING OVER Before it measures your device, Proof measures your interface: its own noise and distortion, frequency by frequency. Every reading is then held against that floor. Clear of it, you get a number. Near it, a range, never a plus-or-minus: a floor can only push a reading up. At or under it, an inequality with no digits: "quieter than this instrument can hear". Significant figures are derived, not chosen: four figures means the instrument is entitled to four. LIVE • RTA: fractional-octave spectrum, bars or line, A/C/Z weighting, averaging, peak hold • Oscilloscope: triggered, with source, slope and level on the graticule • Spectrogram: a colour map monotonic in lightness, so brighter is louder • Signal generator: sine, sweep, pink and white • XY scope: Lissajous and correlation. One channel in this version • SPL meter: A, C or Z weighting, fast or slow, with Leq and max • Level meter: peak and RMS in dBFS • LUFS meter: momentary, short-term and integrated, ITU-R BS.1770-4, gated to EBU R128 • Dosimeter: dose and projected 8-hour TWA, 85 dBA, 3 dB exchange • Transfer function: magnitude, phase and coherence; the trace stops where coherence fails rather than thinning MEASURE • Frequency response, bounded by the measured noise spectrum rather than the distortion floor, because a level is not a ratio • Harmonic spectrum, H2 to H5, each point bounded by the floor at its own frequency • THD against frequency, with the floor for a sum of orders raised the way a power sum requires • THD against level: a sweep per rung, unbounded because a floor from one drive is wrong at both ends of a ladder • The impulse response, and from an acoustic capture its energy-time curve, RT60 per octave band and cumulative spectral decay • Impedance against frequency, through a sense resistor whose value you measure • Microphone calibration from a miniDSP or REW file, corrected at each harmonic's own frequency, with the capsule serial and file hash in the record • A trust tier (Unverified, Floor measured or Calibrated), computed on every read, never stored, so a file cannot claim one it has not earned • A library that browses, verifies, compares and exports THE LIMITS, NAMED • All nine lenses compute. ETC, RT60 and waterfall want an acoustic capture, meaning a microphone with its calibration file loaded; impedance wants a sense-resistor jig. Each says which, and why • SPL reads dBFS until you calibrate against a 94 or 114 dB reference; the dosimeter shows nothing until it does • On a rig with two uncorrected clocks Proof does not report phase or arrival time, and says so • The signal generator writes into whatever you have connected. It is the only tool here that can do damage Proof connects to the App Store for your purchase and to nothing else. No account, no analytics, no crash reporting. Nothing you measure leaves your device unless you export it. Records go to a Proof folder in Documents, reachable from Files. Mac, iPad and iPhone are one app: buy once, use all three. A swept measurement wants an interface you can loop back.