This document details a novel experimental setup for capturing light propagation at ~2 billion frames per second using time-domain multiplexing, photomultiplier tube detection, and precision nanosecond-scale pulse generation. The system reconstructs a 1280×720 pixel video from 921,600 repeated laser pulses, each capturing a single spatial position at a different temporal offset.
Key Innovation: Rather than continuous video capture, this system exploits the deterministic nature of light propagation to build a composite image from thousands of identical events, with each laser pulse capturing one pixel location at sub-nanosecond temporal resolution.
Source Material: This technical artifact is based on the experimental work documented in two YouTube videos by BetaPhoenix:
The 555 timer, ubiquitous in hobby electronics for pulse generation, becomes fundamentally unusable when attempting to generate 50-nanosecond laser pulses. This failure isn't a limitation of implementation but of physics.
The 555 timer's insurmountable bottleneck lies in its internal switching characteristics. Standard NE555 circuits exhibit rise and fall times of approximately 100 nanoseconds - twice the desired 50ns pulse width. Attempting to generate a 50ns pulse results in a waveform that resembles a single sawtooth ramp rather than a clean rectangular pulse. Stack Exchange The output never reaches a stable high state before the circuit attempts to transition back low. Additional limitations compound this problem: internal comparator propagation delays of ~16 microseconds, Texas Instruments E2E maximum practical operating frequencies around 1 MHz, Wikipedia and RC timing networks that cannot charge or discharge quickly enough at nanosecond timescales.
For reference, the minimum reliable pulse width from a 555 timer is approximately 200-500 nanoseconds - a full order of magnitude too slow for this application. Even if the circuit could theoretically trigger at the correct intervals, the slew-rate-limited output would fail to properly drive the laser driver circuit, resulting in poorly defined laser pulses with excessive jitter and inconsistent pulse energy.
The solution employs 74AC14 hex inverter with Schmitt trigger inputs from the Advanced CMOS (AC) logic family. This chip provides the speed necessary for sub-100ns pulse generation through fundamentally faster switching mechanisms.
The 74AC14 achieves 3-5 nanosecond rise and fall times Newark Electronics - approximately 20× faster than the 555 timer. Onsemi This speed comes from the Advanced CMOS fabrication process, which minimizes parasitic capacitances and optimizes charge carrier mobility. The Schmitt trigger input provides hysteresis of approximately 1.0V between positive-going and negative-going thresholds, creating sharply defined, jitter-free transitions that are essentially insensitive to temperature variations and power supply noise.
The pulse generation circuit uses a two-stage approach:
Stage 1 - Base oscillator: One 74AC14 inverter is configured as an RC oscillator to generate a square wave at the system repetition rate (~3 kHz, set by the motor/encoder system).
Stage 2 - Pulse differentiation: A capacitor-resistor differentiator (high-pass filter) converts the square wave edges into narrow pulses. The differentiator responds to rapid voltage changes (dV/dt), allowing only the rising edge transition to pass through as a brief voltage spike.
Pulse width calculation: The RC time constant determines pulse width:
τ = R × C
For 50ns pulse: R ≈ 500Ω, C ≈ 100pF
τ = 500Ω × 100pF = 50nsThe capacitor passes the high-frequency edge components while the resistor provides the discharge path. Typical component values range from 100pF to 1nF for the capacitor and 50Ω to 1kΩ for the resistor, All About Circuits with a multi-turn trim potentiometer allowing precise adjustment of pulse width to match the laser driver's optimal trigger characteristics.
The circuit gangs four 74AC14 inverters in parallel at the output stage - a configuration critical for properly driving 50Ω loads. This design decision addresses a fundamental impedance matching problem.
Current drive requirements: A single 74AC14 inverter can source or sink 24mA. Onsemi However, driving a 50Ω load (standard for oscilloscope inputs, coaxial transmission lines, and many laser driver inputs) at 5-6V requires:
I = V/R = 6V / 50Ω = 120mAParallel inverter physics: When inverters are paralleled, their output current capabilities add directly. Four inverters provide approximately 96mA combined drive current - sufficient to approach the theoretical 120mA requirement. The actual drive current slightly exceeds this due to the 6V supply (discussed below).
Critical performance benefits:
Engineering note: Good supply decoupling is mandatory when paralleling inverters. The simultaneous switching of multiple gates creates significant transient current demands. Multiple decoupling capacitors (0.1μF ceramic placed immediately adjacent to the IC pins, plus 10-47μF bulk capacitance) prevent supply rail collapse during switching transitions.
The circuit operates at 6V rather than the standard 5V logic level, extracting maximum performance from the 74AC14 within its absolute maximum rating of 6.5V.
Rationale for 6V operation:
The 74AC14 datasheet specifies operation from 2V to 6V, making 6V a safe choice with 0.5V margin below the absolute maximum.
The complete system comprises five major subsystems that must synchronize with sub-nanosecond precision across hundreds of thousands of laser pulses.
A single 16V wall-wart power supply feeds the entire system through cascaded voltage regulators:
This architecture isolates the sensitive high-speed logic from the noisy laser driver switching currents while minimizing the number of power adapters required.
A DC motor drives a rotating scanning mirror that sweeps the laser beam across the scene. An optical encoder (likely reflective or transmissive optical interrupter) mounted on the motor shaft generates a pulse train synchronized to mirror position.
Encoder signal processing: The encoder's raw output feeds a 555 timer configured as a monostable multivibrator via an optocoupler. The optocoupler provides electrical isolation, preventing motor noise from corrupting the sensitive timing circuits. Each encoder pulse triggers the 555, which generates a clean, debounced pulse that initiates the next laser flash. Electronics TutorialsInstructables
Repetition rate calculation: At approximately 3000 Hz repetition rate and 921,600 total pixels:
Total acquisition time = 921,600 pixels / 3000 Hz = 307 seconds ≈ 5 minutesEach motor revolution captures one horizontal line of the image (1280 pixels), requiring 720 revolutions for the complete frame. The encoder must provide 1280 pulses per revolution, either through a high-resolution optical encoder disk or through electronic division of a higher-frequency carrier.
Timing sequence (see state diagram below):
Distinction between signal types (critical for understanding):
The most elegant aspect of this system is the ~200-foot coaxial cable spool that provides a precisely calibrated time delay for oscilloscope triggering.
The fundamental problem: The oscilloscope must trigger at precisely the moment the laser fires, not when the PMT detects light. However, the PMT signal arrives microseconds after the laser pulse (due to light travel time through the scene), and triggering on the PMT signal itself would create circular causality - you can't trigger on data you're trying to measure.
The solution: The laser driver includes a feedback output that generates a timing pulse exactly coincident with the laser emission. This signal travels through ~200 feet of coaxial cable before reaching the oscilloscope trigger input.
Delay calculation:
Speed of light in coax ≈ 0.66c (typical for polyethylene dielectric)
= 0.66 × 3×10^8 m/s = 2×10^8 m/s
For 200 feet (61 meters):
Delay = 61m / (2×10^8 m/s) = 305 nanosecondsThis ~300ns delay allows the following sequence:
By carefully trimming the cable length and adjusting the oscilloscope trigger delay settings, the operator positions the PMT signal in the optimal portion of the oscilloscope's capture window.
Why sub-nanosecond accuracy matters: At 2 billion fps, each frame represents 0.5 nanoseconds of elapsed time. Timing jitter of even 1ns would blur two frames together, destroying temporal resolution. The coaxial delay system provides stable, repeatable timing because the cable's propagation delay is constant and temperature-stable (varying by only ~100 ppm/°C for quality coax).
The photomultiplier tube (PMT) serves as the single-pixel detector, converting individual photons arriving from the scene into measurable electrical pulses. Wikipedia
PMTs operate at high voltage (1000-2000V) to create sufficient electron multiplication gain. Wikipedia Each incident photon ejects an electron from the photocathode; this electron is accelerated through a series of dynodes, each ejecting multiple secondary electrons, creating a cascade that amplifies the signal by factors of 10^6 or more. WikipediaLaser Focus World
Critical warm-up period: PMTs require 10-30 minutes of warm-up time after power application to achieve stable gain. During warm-up, thermionic emission (dark current) decreases as the photocathode reaches thermal equilibrium, Laser Focus World and the dynode chain stabilizes. Attempting to acquire data before warm-up completion results in drift artifacts and inconsistent pixel intensities.
Operating mode: For this application, the PMT likely operates in analog current mode rather than photon-counting mode. Each laser pulse reflects thousands of photons from the scene, and the PMT output current is proportional to instantaneous photon flux.
The PMT is spatially fixed - it doesn't move. Instead, the rotating mirror scans the laser beam across the scene while simultaneously scanning the reflected light across the PMT's photocathode. Each mirror position corresponds to one pixel location.
Angular encoder registration: An angular encoder (either optical or magnetic) mounted on the motor shaft provides precise mirror position feedback. The acquisition computer reads:
Pixel intensity mapping: The oscilloscope samples the PMT signal at a specific time delay after the laser pulse. By adjusting this delay across multiple experimental runs, the system captures different temporal "slices" of the light propagation. A complete "video" requires:
Total measurements = 1280 × 720 pixels × N temporal samples
For N=30 time samples: 27,648,000 individual laser pulses
At 3 kHz: 9,216 seconds = 2.56 hours acquisition timeThe oscilloscope serves as both the timing arbiter and data acquisition system, requiring careful configuration to avoid false triggers and timing artifacts.
The problem with triggering on PMT data: If the oscilloscope triggered directly on the PMT signal, it would face an ambiguity - is this signal due to the current laser pulse or a reflection from a previous pulse? Additionally, the PMT signal amplitude varies with scene reflectivity, making threshold-based triggering unreliable.
The feedback coax solution: The laser driver's feedback output provides a clean, constant-amplitude timing reference that is perfectly synchronized to laser emission but electrically isolated from the PMT signal path. This feedback signal:
A trim potentiometer in the pulse generator circuit allows fine-tuning of the pulse amplitude to precisely match the 74AC14 Schmitt trigger thresholds. The Schmitt trigger has two threshold voltages:
The trim pot adjusts the pulse peak voltage to:
Adjustment procedure: While monitoring the oscilloscope trigger signal on one channel and the laser driver output on another, the operator adjusts the trim pot to achieve:
The following state diagram illustrates the complete experimental sequence:
┌─────────────┐
│ SYSTEM │
│ POWER ON │
└──────┬──────┘
│
▼
┌─────────────┐
│ PMT │
│ WARM-UP │
│ (10-30min) │
└──────┬──────┘
│
▼
┌─────────────┐
┌─────────┤ MOTOR │◄────────┐
│ │ RUNNING │ │
│ └──────┬──────┘ │
│ │ │
│ ▼ │
│ ┌─────────────┐ │
│ │ ENCODER │ │
│ │ GENERATES │ │
│ │ PULSE │ │
│ └──────┬──────┘ │
│ │ │
│ ▼ │
│ ┌─────────────┐ │
│ │ 555 TIMER │ │
│ │ TRIGGERS │ │
│ │(via opto) │ │
│ └──────┬──────┘ │
│ │ │
│ ▼ │
│ ┌─────────────┐ │
│ │ 74AC14 │ │
│ │ GENERATES │ │
│ │ 50ns PULSE │ │
│ └──────┬──────┘ │
│ │ │
│ ▼ │
│ ┌─────────────┐ │
│ │ LASER │ │
│ │ DRIVER │ │
│ │ FIRES │ │
│ └──┬───────┬──┘ │
│ │ │ │
│ │ └─────┐ │
│ │ │ │
│ ▼ ▼ │
│ ┌──────────────┐ ┌─────────────┐
│ │ LASER │ │ FEEDBACK │
│ │ PULSE │ │ SIGNAL │
│ │ PROPAGATES │ │ TO COAX │
│ └──────┬───────┘ └──────┬──────┘
│ │ │
│ ▼ │ (~300ns delay)
│ ┌──────────────┐ │
│ │ LIGHT │ │
│ │ REFLECTS │ │
│ │ FROM SCENE │ │
│ └──────┬───────┘ │
│ │ │
│ ▼ │
│ ┌──────────────┐ │
│ │ PMT │ │
│ │ DETECTS │ │
│ │ PHOTONS │ │
│ └──────┬───────┘ │
│ │ │
│ ▼ ▼
│ ┌─────────────────────────┐
│ │ OSCILLOSCOPE │
│ │ - Trigger on coax │
│ │ - Sample PMT output │
│ │ - Record amplitude │
│ └──────┬──────────────────┘
│ │
│ ▼
│ ┌─────────────┐
│ │ COMPUTER │
│ │ STORES │
│ │ PIXEL DATA │
│ └──────┬──────┘
│ │
│ │ ┌──────────────┐
└───────────┴────┤ COMPLETE │
│ ONE PIXEL │
│ RETURN TO │
│ NEXT ENCODER│
│ PULSE │
└──────────────┘Key timing parameters:
| Event | Time (relative to laser fire) | Precision required |
|---|---|---|
| Laser driver receives 50ns pulse | T₀ | ±0.5ns |
| Laser emission begins | T₀ + 5ns | ±1ns |
| Light travels 15m through scene | T₀ + 50ns | N/A (physics) |
| PMT detects photons | T₀ + 50-100ns | ±2ns (detector response) |
| Feedback signal starts down coax | T₀ + 2ns | ±0.5ns |
| Feedback signal arrives at oscilloscope | T₀ + 305ns | ±0.5ns (cable stability) |
| Oscilloscope triggers | T₀ + 305ns | ±1ns (trigger jitter) |
| Oscilloscope samples PMT | T₀ + 50-100ns | ±0.2ns (timebase accuracy) |
Critical insight: The oscilloscope trigger occurs after the PMT signal has already arrived and been captured in the oscilloscope's analog memory. Modern digital oscilloscopes continuously buffer incoming signals; the trigger event simply tells the oscilloscope which portion of its buffer to save. This "post-trigger" capability is essential for capturing events that occur before the trigger signal arrives.
The final "video" is not a continuous recording but a composite reconstruction from hundreds of thousands of individual laser pulses, each capturing one spatial pixel at one temporal offset.
For each pixel (x, y):
For each temporal frame:
Total acquisition time (for 30-frame video):
30 frames × 921,600 pixels/frame × 0.33ms/pixel = 9,145 seconds ≈ 2.5 hoursThis technique requires deterministic, repeatable events:
For filming light propagation through a static scene (milk-filled water tank, optical components on table), these requirements are readily met. For dynamic scenes, this technique fails - you cannot film a person walking using this method.
Raw oscilloscope data requires processing before visualization:
Pulse generator:
Laser driver:
Photomultiplier tube:
Coaxial cable:
Oscilloscope:
Motor and encoder:
Unstable oscilloscope triggering:
Inconsistent laser pulses:
PMT signal too noisy:
Spatial registration errors:
This experimental system demonstrates that ultra-high-speed imaging is achievable without million-dollar commercial cameras by exploiting three key principles:
Time-domain multiplexing eliminates bandwidth requirements: Rather than capturing all pixels simultaneously (requiring terahertz-bandwidth detectors and electronics), this system captures one pixel at a time, with each measurement requiring only megahertz-bandwidth components. The 2 billion fps temporal resolution emerges from precision timing, not detector speed.
Nanosecond pulse generation is achievable with commodity logic ICs: The 74AC14 Schmitt trigger, costing under $1, provides rise times adequate for sub-10ns pulse generation when properly implemented. Parallel ganging overcomes drive current limitations, while the RC differentiator shapes pulses without requiring specialized components.
Coaxial cable delay provides picosecond-stable timing reference: A $20 spool of coax cable becomes a precision delay line with <100ps jitter, enabling oscilloscope triggering that would otherwise require atomic clocks and complex synchronization electronics.
For researchers seeking to reproduce this work, the essential requirements are patience (2.5-hour acquisition times), precision (sub-nanosecond timing alignment), and deterministic events (static scenes or perfectly repeatable phenomena). Within these constraints, the system enables visualization of light-speed phenomena that remain invisible to human perception and conventional cameras alike - all from parts available at electronics distributors for under $500.
Complete bill of materials estimate: $450 (excluding oscilloscope and computer)
The true cost lies not in hardware but in the weeks of iterative debugging required to achieve sub-nanosecond timing stability across a system built from hobby electronics components - making this documentation essential for future experimenters seeking to reproduce these results without reinventing every troubleshooting step.
Source Videos:
Note to Readers: While this document is based on the experimental work documented in the BetaPhoenix YouTube videos, specific implementation details should be verified against the original video content, as certain specifications represent reasonable technical inferences from the described system architecture rather than direct transcriptions.