The floor beneath every voice measurement: thermal inertia, a demodulating diaphragm, and the information limit on jitter

A ceiling claim means nothing without a floor. Acoustic voice measurement reports differences of hundredths of a percentage point and has never stated the limit below which no measurement of any kind can go. This paper states it, in three connected quantities. THE ACOUSTIC FLOOR. Air in thermal equilibrium never stops drumming on a diaphragm, and the fluctuation-dissipation theorem makes this obligatory: a medium that absorbs energy must also fluctuate. Integrated over the audible band at room temperature the floor is approximately -23 dB SPL. A sound-treated booth sits 43.5 dB above it, an ordinary clinical room 70.4 dB, and the crew cabin of a spacecraft 86.0 dB. Nothing in physics is constraining anybody. THE DIAPHRAGM IS A DEMODULATOR. For a mylar film over a 20 micrometre air gap the areal heat capacity is 8.13 J m^-2 K^-1 and the gap conductance 1300 W m^-2 K^-1, giving a thermal time constant of 6.26 ms and a thermal corner at 25.4 Hz. The acoustic carrier therefore never reaches the thermal channel -- suppressed by 31.9 dB at 1 kHz and 43.9 dB at 4 kHz -- while the envelope passes intact. The correct axis for the picture is modulation rate, not frequency, and on that axis the sources of a recording separate: steady machinery near DC, breathing at 0.2-0.5 Hz, syllables at 3-8 Hz, handling and vibration above 10 Hz. THE INFORMATION FLOOR ON JITTER. From the Cramer-Rao bound on time-delay estimation, with the root-mean-square bandwidth measured on the specimens themselves (234 Hz on clean sustained vowels), the smallest resolvable mean jitter in a three-second recording at 100 Hz is 0.014 percentage points at 30 dB signal-to-noise ratio and 0.044 at 20 dB. The measured repeatability of a single clinical recording is 2.47 percentage points -- between fifty-six and one hundred and seventy-six times larger. The gap is in the estimator and in the room, not in the air. WHAT THIS IS NOT. It is not an instrument. Under the most generous possible assumption -- that the diaphragm absorbs every watt of incident acoustic power -- a voice at 70 dB SPL raises its temperature by 7.5 nanokelvin over one thermal time constant, and the real figure is smaller still. A picture that cannot be turned into an instrument is labelled as such here rather than left to look like one. AN ERROR OF OUR OWN, in the text rather than a footnote. Our first derivation of the acoustic floor was wrong by twenty-eight decibels: the expression used is correct for a regime a microphone diaphragm is not in. A book proposal from this project also stated the clinical-practice-to-floor gap as 43 dB; that figure is the booth-to-floor gap, and clinical practice sits 70 dB above the floor. The script that produces every table is included and runs in under a second.