Every microcontroller, radio, and wristwatch you own contains a tiny slab of quartz that vibrates millions of times per second. Quartz is piezoelectric: squeeze it and it generates a voltage; apply a voltage and it deforms. Cut a wafer to a precise thickness and it has a mechanical resonance so sharp and stable that it beats every RC or LC oscillator by orders of magnitude.

The crystal behaves electrically like a series RLC circuit (motional inductance, capacitance, and resistance) in parallel with a shunt capacitance from the electrodes. This gives it two close resonances: a series resonance f_s where impedance is minimum, and a parallel (antiresonance) f_p a few kHz higher where impedance is maximum. Oscillator circuits exploit one or the other.

The most common circuit is the Pierce oscillator: a single inverting CMOS gate with a feedback resistor biases the amp into its linear region, and two load capacitors (typically 12-22 pF each) to ground pull the crystal to its rated frequency. This is the circuit hiding inside literally every AVR, STM32, and PIC's HSE pins.

Load capacitance matters. A crystal is spec'd at a load capacitance (say, C_L = 18 pF). The two external caps appear in series to the crystal, plus stray board capacitance (~3-5 pF). Rule of thumb:

• C₁ = C₂ = 2 × (C_L − C_stray)
• For C_L = 18 pF, C_stray = 5 pF → caps ≈ 26 pF (use 27 pF standard)

Get this wrong and your 16.000 MHz crystal runs at 16.0003 MHz — enough to break USB enumeration or drift a GPS receiver.

Frequency stability is where quartz shines. A cheap watch crystal (32.768 kHz, chosen because 2¹⁵ divisions give exactly 1 Hz) holds ±20 ppm — about 10 seconds per week. A temperature-compensated crystal oscillator (TCXO) hits ±0.5 ppm. An oven-controlled version (OCXO), which holds the crystal at 75°C in a tiny heated chamber, reaches ±0.01 ppm — one second per three years. Cell towers and GPS receivers use these.

Real-world gotcha: AT-cut crystals (the standard cut) have a cubic temperature curve with a turnover point around 25°C. That's why an unheated crystal drifts fastest at the temperature extremes and is most stable at room temp. It's also why "the clock is fast in winter" is a real complaint in unheated equipment closets.

Aging is real too: crystals drift ~1-5 ppm in their first year as internal stresses relax, then slow dramatically. Precision timing labs "burn in" oscillators for months before deploying them.