Archival DSP 07

The Restoration Lab

Take public-domain early recordings and toggle between Original, De-click, Noise reduction, EQ compensation, and Hum removal. Expose the philosophical tradeoff: Which part of an old recording is the signal—and which part belongs to its history?

DSP Archival LaboratoryThe Signal vs. History Tradeoff

The Restoration Lab

Epistemic Question“Which part is signal—and which is history?”

Toggleable DSP Modules:

De-Click & De-CrackleACTIVE

Detects non-musical impulsive transients (<2 ms duration) caused by dust particles, micro-scratches, and broken shellac groove walls, interpolating across the damaged samples.

Tradeoff: Aggressive settings confuse sharp musical attacks (snare hits, harpsichord plucks, vocal plosives /t/, /k/) with scratches, dulling the natural transient punch.

Spectral Subtraction Noise ReductionBYPASS

Measures a noise profile from the groove lead-in silence, dividing the frequency spectrum into FFT bins and subtracting that estimated noise floor from the entire recording.

Tradeoff: Causes 'musical noise' or 'watery phase chirping' when musical harmonics sit near the threshold; strips high-frequency acoustic presence and room air.

Mains Hum Notch Filter (50/60 Hz + Harmonics)BYPASS

Applies high-Q recursive digital notch filters centered precisely on 50 Hz (Europe) or 60 Hz (US) and their harmonic overtones (100/120 Hz, 150/180 Hz).

Tradeoff: Notches out musical bass frequencies: any acoustic bass fundamental falling exactly on 60 Hz (near B1) is permanently excised from the performance.

Acoustic Horn Inverse EQ CurveACTIVE

Applies inverse equalization curves to flatten the severe mechanical resonance peaks (typically a massive 12 dB honk around 1.5–2.5 kHz) caused by the physical recording funnel.

Tradeoff: Amplifies frequencies where the recording horn had zero sensitivity, drastically elevating background noise in the low bass and extreme highs.

Stroboscopic Speed & Pitch StabilizationACTIVE

Analyzes mains hum frequencies or motor eccentricities to calculate the exact mechanical playback speed, correcting pitch drift caused by hand-cranked spring motors.

Tradeoff: Assumes historical performers were tuned to modern standard pitch; can inadvertently alter intended expressive pitch standards of the period.

Source Medium: Single-sided 10-inch Shellac disc (Victor 81025)

Enrico Caruso – 'Vesti la giubba' (1904)

Recorded February 1, 1904

Physical Acoustic Pathology:

Dense abrasive shellac surface roar, severe acoustic horn resonance peak around 1.8 kHz, zero bass below 160 Hz.

Target Acoustic Feature:

Preserving Caruso's legendary vocal formants without introducing digital watery artifacts.

Curatorial DSP Treatment:

Gentle de-click only; light 6 dB dynamic broad-spectrum noise expansion; preserve the natural shellac floor as historic substrate.

The Curatorial Restorer's Dilemma

“The shellac surface noise is not merely garbage to be scrubbed away; it is the physical watermark of 1904 Camden, New Jersey. Strip every trace of noise, and Caruso's voice loses its corporeal presence and singing overtone ring.”