# Interactive COFFEE Analog Speech Synthesizer

> Long before microchips, 7 analog circuits synthesized a voice saying "COFFEE." Explore the interactive circuits in your browser.

- **Canonical URL:** https://coffee.jasonshim.net/
- **Exhibit Provenance:** Engineered by Philips Research Laboratories (NatLab) and the Institute for Perception Research (IPO) in Eindhoven, Netherlands.
- **Technical Architecture:** Self-contained, zero-dependency Web Audio API + interactive SVG single-file instrument (`index.html`) designed for multi-decade archival longevity, classroom STEM education, and offline museum kiosk deployment.

<!-- Creator Attribution (uncomment when ready):
- **Digital Restoration By:** Jason Shim (https://jasonshim.net)
-->

## How the Analog Speech Synthesizer Works

Long before microprocessors or digital audio recordings existed, this machine synthesized a human voice saying **"coffee"** (`C–O–FF–EE`) in real time using the **acoustic source-filter model of speech production**:

1. **Unvoiced Air Turbulence (White Noise Generator):** Simulates the burst and hiss of air through the mouth for unvoiced consonants (`C` and `FF`).
2. **Voiced Vocal-Cord Buzz (Sawtooth Oscillator on Board 5):** Simulates the vibration of human vocal cords (`85 Hz` Bass to `240 Hz` Treble, with a `1900 Hz` low-pass glottal roll-off) for voiced vowels (`O` and `EE`).
3. **Analog Formant Filter Boards (Boards 1–4):** Shape the raw noise and buzz into four distinct English phonemes sequenced by an analog capacitor clock (Board 6):
   - **Board 1 (`C`, `/k/` unvoiced velar plosive, `70 ms`):** White noise routed through dual parallel bandpass filters (`1450 Hz, Q=4.0` and `2600 Hz, Q=3.5`).
   - **Board 2 (`O`, `/ɒ/` voiced back vowel, `195 ms`):** Sawtooth buzz routed through three parallel formant bandpass filters (`F1 = 570 Hz, Q=5.5`, `F2 = 840 Hz, Q=6.5`, `F3 = 2410 Hz, Q=8.0`).
   - **Board 3 (`FF`, `/f/` unvoiced labiodental fricative, `115 ms`):** White noise routed through a high-pass filter (`2800 Hz, Q=0.8`).
   - **Board 4 (`EE`, `/iː/` voiced front vowel + prosodic glide, `355 ms`):** Sawtooth buzz routed through three parallel formant bandpass filters (`F1 = 290 Hz, Q=4.8`, `F2 = 2250 Hz, Q=8.5`, `F3 = 2950 Hz, Q=9.5`).
4. **Prosodic Inflection Contour (Board 7 & Left Rotary Knob):** Bends the oscillator frequency across the final `EE` vowel, sweeping upward (`+75%`) to ask a polite question (*"Coffee?"*) or dropping sharply (`-85%`) for an exclamation (*"Coffee!"*).
5. **Why did the Coffee machine speak with a Dutch accent?** Because the fixed analog resistor-capacitor and inductor formant filters were tuned by Dutch acoustic researchers at Philips NatLab in Eindhoven, the synthesized English word *"coffee"* carries an authentic Dutch vowel coloration.

## Interactive Features at https://coffee.jasonshim.net/

- **5-Step Interactive Guided Tour:** Built-in curatorial walkthrough with live **"Show me"** audio/visual demonstrations for each circuit stage and control.
- **Live 7-Board Circuit Window:** Watch red indicator lamps chase across the 4 phoneme boards (`C`, `O`, `FF`, `EE`) and lower oscillator/sequencer boards (`5`, `6`, `7`) in real time.
- **Rotary Knobs & Ergonomic Touch Dock:** Adjust **Inflection** (`EXCLAMATION!` / `STATEMENT` / `QUESTION?`) and **Vocal Cord Pitch** (`85 Hz` Bass to `240 Hz` Treble), or toggle hands-free **Auto-Run Mode** (repeats every `2.9 seconds`).