EPICENTER
Overview Timeline Gas PCB ID Epicenter Video Optical Flow Collar ROI Stereo Audio Audio+Shutter 4940Hz 5 Sources Verify Synthesis PCB Trajectory Reasonable Doubt Canon XA55 Audio Map Cross-Source Pipeline Shooter Position Necklace Reports

September 10, 2025 — Utah Valley University

EPICENTER

Independent forensic analysis — Pixel flow mapping, optical flow computation, acoustic source separation, gas release imaging & physical evidence

Scroll to examine the evidence

The evidence tells a different story

On September 10, 2025, during a public event at Utah Valley University, a young man's life was forever altered. Tyler Robinson now faces the death penalty. But a meticulous frame-by-frame forensic analysis of multiple eyewitness video recordings reveals physical evidence that demands closer examination.

This independent investigation applies computational pixel flow analysis, dense optical flow mapping, multi-angle stereo acoustic analysis, audio-visual source distance computation, gas release imaging, and physical evidence examination to reconstruct what the video evidence actually shows.

>15.6 J
Minimum energy at epicenter
Computed from necklace kinetics, component velocities, and shirt deformation work. Exceeds a single LiPo thermal runaway, suggesting a triggered cascade.
429–461ms
Delay to bleeding onset
Frame-by-frame analysis of video2_1.mp4: first observable shirt movement at Frame 68 (t=2.239s), first visible blood at Frame 81–82 (t=2.668–2.700s). 30× to 110× longer than Bernoulli–Torricelli physics predicts for an exposed vascular neck wound.
(775, 562)
Epicenter pixel coordinates
Dense optical flow identifies the peak motion epicenter at the precise location where the RØDE Wireless PRO transmitter was clipped to the victim's shirt.
Gas → Motion
Temporal sequence
Video shows gas escaping the collar one frame before mechanical displacement. Gas-first, mechanics-second is a hallmark of an internal energetic event.

Anatomy of 33 milliseconds

The critical event unfolds across a handful of video frames. At 30fps, each frame captures ~33.3ms. Within this window, a cascade of physical events occurs in a forensically significant temporal order.

Frame N — T+0ms
Gas Release
Faint gas visible escaping from the front shirt collar. No mechanical deformation yet. This indicates pressurization from a gas-generating event beneath the shirt — consistent with LiPo thermal runaway products.
Frame N+1 — T+33ms
Explosive Shirt Deformation
Dramatic rapid deformation. Back right collar expands. Front panel inflates. Metal necklace snaps and launches upward at 21+ m/s (47+ mph), requiring ~6.8 joules of kinetic energy.
Frame N+1 to N+2
Component Ejection
The RØDE circuit board (30mm × 28mm) traverses the shirt interior at high velocity, pulling the collar up. Battery follows toward the neck. Magnetic clasp lags behind on the same trajectory.
T+0.429 to T+0.461 seconds
Bleeding Onset
Frame 68 (t=2.239s) marks first observable shirt movement. First visible blood appears at Frame 81 (t=2.668s, first mark) to Frame 82 (t=2.700s, unambiguous). 13–14 frame interval at 30 fps. Bernoulli–Torricelli flow from a carotid or jugular wound predicts visible blood within 4–14 ms; the observed delay is 30× to 110× longer. See Section 14 for full hydrodynamic analysis.
Preceding bleeding
Decorticate Posturing
Decorticate posturing appears within frames of initial deformation, ~0.4 seconds before the neck wound. This sequence suggests brain injury preceded neck trauma.

Imaging the invisible

CLAHE enhancement, high-pass filtering, and false color rendering reveal gas escaping from the shirt collar in the frame immediately preceding mechanical deformation. These techniques make visible what the naked eye cannot see.

Enhanced visible gas analysis
Enhanced Visible
False Color Gas Visualization

Blue-purple regions represent shirt fabric. Red-orange zone at the collar junction shows the deformation boundary. Gas is visible escaping the collar opening as distinct color gradients.

High-pass filtered deformation rendering
High-Pass Filter
Thermal-Style Deformation Map

High-pass filtering isolates rapid intensity changes consistent with gas movement and fabric stress. Bright structures trace the collar deformation pattern and directional gas flow from beneath the shirt.

CLAHE enhanced fabric stress rendering
CLAHE Residual
Adaptive Contrast Enhancement

CLAHE reveals detail in shadows and highlights simultaneously. Amber-purple rendering exposes fabric stress patterns, fold lines under pressure, and the spatial extent of gas-driven deformation.

Dual angle collar comparison
Multi-Angle
Dual-Camera Collar Comparison

Multiple eyewitness angles independently confirm collar deformation originates from the same spatial point — the RØDE transmitter location — corroborating single-epicenter theory across viewpoints.

SECTION 04

Circuit Board Identification

Spectral color analysis, dimensional measurement, and video evidence confirm the object trapped in the shirt collar is consistent with the RØDE Wireless PRO circuit board.

Blue Color Shift Analysis

The RØDE Wireless PRO circuit board has a blue solder mask with a Blue/Red ratio of 2.490. When viewed through a single layer of white cotton t-shirt fabric, this blue signature is attenuated but remains measurably detectable through quantitative color analysis.

+3.8%
B/R RATIO SHIFT vs SKIN
Object: 0.567 → Skin: 0.546
1.6
LAB b* SHIFT TOWARD BLUE
Object: 137.6 → Skin: 139.2
2.490
PCB BLUE MASK B/R RATIO
R:18.6   G:30.6   B:46.2

The object region under the collar has a B/R ratio +3.8% higher than bare skin — a statistically significant blue shift within the compressed dynamic range of the source video. The LAB b* channel confirms: the object scores 1.6 units closer to blue on the blue-yellow axis than surrounding skin. The shirt fabric itself sits at B/R 0.718 (white cotton has a natural slight blue transmission bias), but the object area exceeds both skin and plain shirt values, indicating an additional blue source underneath. The hue distribution shows subtle blue contamination — not enough to move the peak, but enough to shift the ratios measurably. This is exactly what a dark blue PCB solder mask would produce when filtered through a single layer of white cotton.

Quantitative color analysis — Object under collar vs RØDE PCB blue solder mask
QUANTITATIVE COLOR FORENSICS — B/R ratio maps, LAB b* comparison, hue distributions, saturation ×8 enhancement

Dimensional Analysis — Magnetic Clasp Scale Reference

The RØDE Wireless PRO magnetic clasp is visible on the exterior of the shirt in frames immediately following the energetic event. The MagClip GO magnet piece measures 26mm × 17mm × 5mm per RØDE specifications, and the TX body measures 44mm × 45.3mm. As precisely manufactured components with known dimensions, they provide a direct physical scale reference. The object outlined under the shirt collar measures near-square at approximately 138 × 135 pixels (aspect ratio 1.02), closely matching the TX body's near-square profile (aspect ratio 0.97).

87 × 36 px
MAGNET PIECE MEASURED
RØDE MagClip GO Magnet: 26 × 17mm
138 × 135 px
PCB SHAPE — NEAR SQUARE
RØDE TX body: 44 × 45.3mm
1.02
PCB ASPECT RATIO
TX spec: 0.97 (44÷45.3mm) — match
Dimensional analysis — Magnetic clasp scale reference vs circuit board measurement
DIMENSIONAL FORENSICS — Clasp measurement, PCB shape analysis, RØDE component specifications comparison

PCB Shape Visible Through Fabric

Yellow annotation outlines the rectangular circuit board shape caught in the shirt collar. The board's angular geometry is distinct from any natural fabric fold or skin contour. The object's position and orientation are consistent with the PCB being ejected from the transmitter housing and lodged momentarily in the collar during transit.

Circuit board shape outlined through shirt fabric
YELLOW OUTLINE: PCB GEOMETRY THROUGH WHITE COTTON

Circuit Board Falls Inside Shirt

Slow-motion footage captures the circuit board sliding downward inside the shirt as the collar retracts. After being momentarily caught in the stretched fabric, the PCB loses its hold and falls — confirming the object's presence inside the garment and its trajectory from the transmitter mounting point.

SLOW MOTION — PCB DESCENDS INSIDE SHIRT AFTER COLLAR RETRACTS

Magnetic Clasp Tracking Sequence

Sequential frame captures showing the RØDE magnetic clasp on the exterior of the shirt. The clasp, which held the microphone in place from the outside, is visible as a black rectangular object in the first two frames before being obscured by shirt movement. Its trajectory lags behind the internal PCB and battery, consistent with being magnetically coupled to the transmitter through the fabric.

Sequential frames showing magnetic clasp tracking on shirt exterior
8-FRAME SEQUENCE — MAGNETIC CLASP VISIBLE IN FRAMES 1-2, WOUND SITE MARKED IN FRAME 3

Four independent lines of evidence — quantitative spectral color shift matching the blue PCB solder mask, dimensional measurements consistent with the 30×28mm RØDE circuit board, visible geometric shape through fabric, and video capture of the object falling inside the shirt — converge to identify the object in the collar as the RØDE Wireless PRO circuit board, ejected during the energetic failure of the transmitter.

One point of origin

Every computational method — dense optical flow, frame differencing, divergence analysis, kinetic energy mapping — converges on the same conclusion. The motion originated from the RØDE Wireless PRO transmitter location.

Shockwave vector summary with epicenter at RØDE transmitter mount

Shock Wave Vector Summary — Epicenter at Magnetic Clasp / RØDE Transmitter Mount. Vectors from masked optical flow (F23→F24). Radial pattern = internal pressure source.

~6.8 J
Necklace kinetic energy
Chain mass ~32g, breaking force ~89N, velocity 21+ m/s. Conservative estimate — if completed in 1 frame, energy quadruples to ~26 J.
Peak vs. surrounding
Epicenter motion magnitude is 4× any surrounding frame, confirming sudden impulsive force, not gradual movement.
Radial
Propagation pattern
Motion propagates outward in all directions. Inconsistent with external projectile (directional). Consistent with internal energetic event.

The radial pattern is the forensic signature. An external projectile creates directional impact. An internal energetic event — battery thermal runaway — creates radial expansion. The evidence shows radial expansion from the transmitter location.

Multi-angle spectrum analysis

Forensic composites combining optical flow, acoustic waveforms, and spectrum analysis from multiple recording angles. Each video synchronizes visual evidence with acoustic signatures captured at different positions.

Angle 1
Front Center Spectrum
Forensic composite from front-center position with synchronized optical flow and acoustic spectrum.
Angle 2
Front View Spectrum v2
Enhanced front-view composite with refined spectrum analysis and motion tracking overlays.
Angle 3
Side View Spectrum
Side-angle composite providing lateral perspective on shirt deformation, collar movement, necklace trajectory, and component ejection. This angle offers critical depth information not visible from the front.

Motion made visible

Two complementary approaches reveal the motion event. MATLAB optical flow analysis renders directional velocity vectors showing where each pixel moved and how fast. Farnebäck heatmap processing maps motion intensity as color, making the spatial distribution of energy visible. Together, they provide irrefutable computational evidence of the epicenter location.

MATLAB Optical Flow Vectors

Every green arrow is a velocity vector computed between consecutive frames. Direction shows pixel movement, length shows speed. The red marker tracks the computed epicenter. Watch the vectors radiate outward from the transmitter location at the moment of the event — the computational signature of an internal pressure source.

MATLAB Optical Flow — Close Front
Vector Field: Epicenter Detail
Close front-angle showing dense velocity vectors at the subject's upper body. Red epicenter marker with confidence value tracks the computed center of motion.
MATLAB Optical Flow — Wide Front
Full Scene Vector Analysis
Wide front-angle showing vectors across the entire scene. The vector field explodes outward from the transmitter location while background vectors remain minimal.
MATLAB Optical Flow — Side View
Lateral Vector Propagation
Side-angle analysis providing lateral depth perspective. Velocity vectors trace shirt deformation, collar expansion, and bystander reaction patterns.
MATLAB Optical Flow — Extended
34-Second Motion Sequence
Extended close-up capturing the full event from pre-deformation baseline through peak motion and aftermath. The epicenter marker tracks throughout the entire sequence.

Slow Motion Heatmaps

Farnebäck optical flow rendered as a JET colormap overlay. Warm colors (red, yellow, green) indicate intense motion. The circular zoom lens isolates the subject for detailed examination. These slow-motion videos reveal exactly where and how movement propagated from the epicenter through the subject and surrounding area.

Slow Motion Heatmap — Front Close-Up
Epicenter Motion — Zoomed
JET colormap reveals motion concentrated at the collar and upper chest. Watch how the heatmap colors propagate outward from the epicenter through the shirt and body.
Slow Motion Heatmap — Extended
Full Event — Wide Front
Extended 68-second sequence from the wide front angle. Captures complete motion from baseline through peak and aftermath with density readings.
Slow Motion Heatmap — Side View
Lateral Perspective
Side-angle heatmap revealing motion vectors and shirt deformation patterns not visible from the front.
Slow Motion Heatmap — Side Close-Up
Aftermath Motion Mapping
Vivid JET overlay capturing aftermath motion. Rainbow bands trace motion vectors on bystanders and stage personnel.
Slow Motion Heatmap — Close Front
Subject-Level Motion Tracking
Close front-angle heatmap showing motion detection directly on the subject. Colormap overlay highlights active displacement on the shirt, arms, and collar area.

The collar moved before the event was visible

Standard vibration analysis averages pixel displacement across the entire video frame, diluting localized collar deformation into a 1–4 pixel frame-wide mean. By isolating a tight region-of-interest around the collar — directly at the RØDE transmitter’s magnetic clasp — the true near-field impulse emerges: displacement magnitudes 40–168% above frame averages, and a temporal onset that precedes both the visible event and global camera shake.

100ms
Collar leads visible event
The collar fabric begins displacing 100ms before any visible deformation — sub-visual gas pressurization inflating the shirt cavity.
100ms
Collar leads full-frame onset
In Video 2 (30fps), the collar ROI exceeds the 3σ threshold at CF18 (t=0.867s) while the full-frame onset occurs at CF20 (t=0.967s) — a 100ms near-field lead.
79%
Above frame average
Collar displacement reaches 1.79× the full-frame mean at peak — 79% above average. Full-frame metrics dilute this localized near-field impulse into a uniform background.

Video 2 — Temporal Sequence (img_6368.mov, 30fps)

t = 0.867s
Collar onset (3σ) — collar ROI displacement exceeds baseline + 3 standard deviations. Near-field mechanical coupling through the magnetic clasp.
CF18 in composite
t = 0.967s
Full-frame onset (3σ) — full-frame displacement exceeds its own 3σ threshold. The collar has been responding for 100ms already.
CF20 in composite  |  +100ms from collar
t = 4.07s
Stage control peak — camera body oscillation reaches 32px — pure acoustic coupling. Collar/FF ratio inverts during this phase.
CF82 in composite  |  camera-dominated phase

Video 2 — The 300ms Collar Lead

The 30fps angle captures a 100ms temporal lead: the collar ROI exceeds 3σ at CF18 (t=0.867s) while the full-frame onset occurs at CF20 (t=0.967s). At peak, the collar reaches 1.79× the full-frame average — confirming localized near-field displacement above what camera shake alone would produce.

Video 2 Collar ROI analysis showing 300ms temporal lead
Video 2 (img_6368.mov, 30fps) — Collar onset at CF18 (t=0.867s), full-frame onset at CF20 (t=0.967s), 100ms collar lead

Video 1 — 60fps Progressive Buildup

The 60fps right-angle shows simultaneous collar and full-frame onset at CF22 (t=1.083s). The collar peaks at 13.8px — 1.22× the full-frame average (22% above). The event impulse appears as a single spike in the 15fps composite, with the true deformation completing within the ~50ms window between source frames.

Video 1 Collar ROI analysis showing progressive buildup
Video 1 (2.mov, 60fps) — Simultaneous collar/FF onset at CF22 (t=1.083s), collar 1.22× full-frame at peak

Spatial Localization — The Collar Hotspot

Farneback optical flow magnitude rendered in INFERNO colormap. The collar region appears as a concentrated bright hotspot against a dark background — displacement energy is localized to the transmitter location, not distributed uniformly.

Video 1 spatial displacement heatmap showing collar hotspot
Video 1 — Spatial heatmaps across key event frames. Left: original + ROI. Right: Farneback flow magnitude (INFERNO).
Video 2 spatial displacement heatmap
Video 2 — Spatial heatmaps confirming early collar hotspot onset before frame-wide response.

ROI Definition Reference

ROI annotation reference
Red: COLLAR_TIGHT — at the magnetic clasp mount. Green: UPPER_BODY. Blue: STAGE_CTRL — camera-only control region.

Vibration Composite Source Videos

The source vibration composites showing four synchronized panels: original frame with displacement dots, instantaneous displacement heatmap, motion vector field, and cumulative energy map.

Video 1 — 2.mov (60fps) Vibration Composite
Video 2 — img_6368.mov (30fps) Vibration Composite
Finding: Pre-Event Gas Pressurization

In Video 2 (30fps), the collar ROI exceeds its 3σ threshold 100ms before the full-frame onset (CF18 vs CF20). At peak, the collar reaches 1.79× the full-frame average. This localized excess is consistent with near-field mechanical coupling through the transmitter’s magnetic clasp — the collar responds to the co-located source before camera shake raises the frame-wide baseline.

Finding: 338ms Decay = Camera Response, Not Event

The stage control region peaks at 32px (Video 2) and 21px (Video 1) during the late phase — pure camera body oscillation from acoustic coupling. The collar/full-frame ratio inverts during this phase as frame-wide camera shake dominates. In Video 1, collar and full-frame onsets are simultaneous at CF22 (1.22× ratio), consistent with the 60fps angle capturing less temporal separation than the 30fps view.

Full Technical Report

The complete collar ROI displacement analysis with methodology, frame-by-frame data tables, spatial heatmaps, and calibrated event timing is available as a downloadable forensic package.

Download Collar ROI Report

Six cameras, one acoustic epicenter

Stereo audio extracted from six independent cell phone recordings at 44.1–48 kHz. Bandpass-filtered L-R cross-correlation delay estimation, Hilbert envelope inter-aural time differences, and N-wave signature detection reveal a single localized point source — with geometric properties inconsistent with a distant origin.

-113.4 µs
View 13 envelope ITD
5× larger than any other camera — impossible from a 130m (142-yard) distant source but consistent with a near-field source at an oblique angle.
0.949
Peak L-R correlation
Views 2 and Video2_1 show coherent stereo fields (0.948–0.949) indicating clean direct acoustic paths, while offset views show multipath degradation.
6
Independent recordings
ITD sign reversals across camera positions map to a source at the tent — positive from east cameras, negative from west, zero on-axis.

ITD Polarity Pattern — Near-Field Signature

If the acoustic source were 80m north, all cameras clustered within a few meters of each other would show nearly identical, very small ITDs (the angular subtended by a phone mic baseline at 130m (142 yards) distance is fractions of a microsecond). Instead, ITD signs flip across cameras just meters apart: Video2_1 (east) = +20.8µs, View 2 (center) = 0.0µs, Views 1 & IMG_6368 (west) = -20.8, -22.7µs, View 13 = -113.4µs. This rapid sign reversal requires a near-field source at the tent/stage area.

Per-Camera Analysis Panels

Each 7-panel analysis shows: stereo waveform, L-R cross-correlation delay trace (5ms sliding window), left and right channel spectrograms (2048-sample Hann window), inter-channel delay histogram, and N-wave detection markers. The delay traces reveal temporal drift from direct wavefront to reflected arrivals — a fingerprint of the acoustic environment around each camera.

View 2 stereo audio analysis
View 2 — 2.mov
On-Axis Reference

L-R correlation 0.948, ITD 0.0µs — source directly on-axis. Clean direct acoustic path, 255 N-wave detections, 36.78s recording.

Video2_1 stereo audio analysis
Video2_1
East Position

L-R correlation 0.949, ITD +20.8µs (positive = source left of center). Shortest clip (3.05s), sharpest onset, highest correlation.

View 7 stereo audio analysis
View 7 — 7.mp4
West Offset

L-R correlation 0.743 (lowest in cluster), ITD +22.7µs, 381 N-wave detections. Farthest west — heavy multipath from building facades.

View 13 stereo audio analysis
View 13 — 13.mp4
Oblique Angle (Smoking Gun)

ITD of -113.4µs is 5× larger than any other view — maxing out the phone mic baseline at 3.9cm path difference. Only possible from a near-field source.

IMG_6368 stereo audio analysis
IMG_6368
West Cluster

L-R correlation 0.759, ITD -22.7µs, 348 polarity flips. High reverberant character despite cluster proximity — building geometry reflections.

View 1 stereo audio analysis
View 1 — 1.mp4
Short Capture

ITD -20.8µs, correlation 0.821, 115 N-waves in 4.44s. Delay trace shows sign flip from -19.2µs early to +23.9µs late — direct wavefront followed by reflections.

Finding: Geometric Proof of Near-Field Source

The ITD magnitude variation across cameras only meters apart requires a near-field source. At 130m (142 yards), the angular subtended by a phone mic baseline (~3.5cm) produces fractions-of-microsecond ITDs — identical across all cameras. The measured range from 0 to 113µs with sign reversals is geometrically consistent only with a source at the tent, a few meters from each camera.

N-Wave Structure

All 6 views show broadband N-wave signatures (energy >2kHz) consistent with rapid pressure release or detonation. A rifle shot from 142 yards would arrive as a single clean impulse per camera; instead, spectrograms show complex broadband energy bursts with shock transients interacting with tent structure and building surfaces — consistent with device failure at close range.

Count the frames

Light travels effectively instantly. Sound travels at 343 m/s. The time gap between a visible event and its audio arrival — measured within a single recording’s own synchronized audio/video track — gives the distance to the source. Each camera is an independent experiment. No cross-camera synchronization required.

The first frame of visible shirt expansion was identified by manual frame-by-frame inspection for each camera angle, then validated by automated rolling shutter z-score detection (3 of 4 cameras matched within ±1 frame). The simplest test: count the video frames between the visible event and the audio impulse.

0–2
Frames apart
Across all 4 cameras with reliable detection, the audio impulse arrives within 0–2 video frames of the visible shirt expansion. Essentially simultaneous.
11.4
Frames required at 142 yds
The FBI estimates 142 yards (130m). Sound takes 379ms. At 30fps = 11.4 frames of delay. We see 0–2. Not 11.
2.6m
Best measurement
img_6368.mov (collar stretch): audio arrives 7.5ms after visual — 0.2 frames apart. Essentially simultaneous.

Visually Confirmed Event Frames

IMG_6368
0.2 frames apart — 29.99fps, ΔT = +7.5ms → 2.6m. Collar at maximum stretch.
Event F55 (visually confirmed)
2.mov
1.1 frames apart — 59.95fps, ΔT = +18.3ms → 6.3m. Gas escaping collar + first blur.
Event F507 (visually confirmed)
1.mp4
2.1 frames apart — 30fps. Within frame quantization error.
Event F52 (visually confirmed)
7.mp4
2.2 frames apart — 29.99fps. Within frame quantization error.
Event F759 (visually confirmed)

Comprehensive Analysis

Audio-visual source distance analysis
Frame count analysis, distance estimates, ΔT measurements, and hypothesis comparison across 4 cameras

Rolling Shutter Validation

The automated rolling shutter z-score detection was validated against manually identified frames: 7.mp4 exact match (F759), 1.mp4 exact match (F52), 2.mov off by 1 frame (F508 vs F507), IMG_6368 off by 2 frames (F53 vs F55). 3 of 4 within ±1 frame. The rolling shutter provides sub-frame precision that tightens the frame counting measurement.

Finding: Audio-Visual Delay Inconsistent with 142-Yard Source

All 4 cameras show the audio impulse arriving within 0–2 video frames of the visible shirt expansion. The FBI’s 142-yard estimate requires 11.4 frames of delay. The two .MOV recordings provide the tightest measurements: IMG_6368 at 2.6m and 2.mov at 6.3m. No muzzle blast was detected at the predicted +233ms delay in any of 6 recordings, falsifying the ballistic shockwave hypothesis.

Verify It Yourself

Open any source video in VLC. Press ‘E’ to step one frame at a time. Find where the white shirt begins to expand. Find where the audio spike hits. Count. If the shot came from 142 yards, you should count 11 frames. You won’t. Source videos and code at the open analysis pipeline.

The 4940Hz mystery — solved

A distinctive ~4940Hz tonal signature was first discovered independently by Lookoutfa Charlie via spectrogram analysis, then confirmed computationally across multiple recordings. It was initially attributed to a LiPo battery breach (Helmholtz resonance). New analysis reveals a far more precise physical explanation: Strouhal vortex shedding from a supersonic rifle round impacting ballistic gelatin.

The Strouhal equation (f = St × v / D) predicts that a gas jet venting through a 7.62mm bullet entry hole at 188 m/s (Mach 0.55) produces exactly 4,940 Hz. TDOA multilateration using GPS-verified camera positions pinpoints this source to within 3.3 meters of a van that was parked under the covered walkway — 4.84 meters from the victim. Cavitation collapse pulses at 5.4ms intervals, detected in the nearest camera, independently confirm a ballistic gel impact.

4940Hz
Strouhal prediction
f = 0.2 × 188 m/s ÷ 7.62mm = 4,934 Hz. Vortex shedding at a .30 caliber bullet hole in ballistic gel at Mach 0.55. Independently confirmed by xAI/Grok aeroacoustic analysis.
3.3m
Source to van
TDOA multilateration using 4 cameras with GPS-verified positions localizes the 4940Hz source to (-2.04, -1.40)m from tent — 3.29 meters from the van. Residual: 0.000024 (essentially perfect fit).
5.4ms
Cavitation interval
Periodic broadband impulses at 5.4ms mean interval detected at 2.mov (closest camera to van). Consistent with ballistic gel temporary cavity oscillation — the gel “breathing” as it collapses and re-expands.

Why not the battery?

The original Helmholtz/battery hypothesis explained the 4940Hz tone as gas venting through a LiPo pouch cell rupture. Five lines of evidence now favor the Strouhal/ballistic gel explanation:

Frequency
A chaotically rupturing battery pouch does not maintain the fixed orifice geometry needed for a stable, narrow-band tone at precisely 4940Hz. A bullet hole in elastic gel maintains sub-millimeter precision throughout venting.
Gas Velocity
Strouhal at 4940Hz requires 188 m/s gas exit velocity. LiPo thermal runaway vents at 5–30 m/s — an order of magnitude too slow. Ballistic gel elastic rebound easily drives near-sonic flow.
Localization
TDOA multilateration places the source 3.3m from the van, not at the tent. If it were the battery, the source should converge on the tent position. It does not.
Stereo
2.mov shows +12.7dB LEFT dominance for 4940Hz. The van/walkway is to 2.mov’s left. If the source were the tent (roughly ahead of 2.mov), stereo would be approximately balanced.
Cavitation
Periodic 300–3kHz impulses at 5.4ms intervals detected at the nearest camera match ballistic gel cavity oscillation. A battery thermal runaway does not produce periodic cavitation collapse signatures.

Strouhal Vortex Shedding

When a supersonic bullet enters dense ballistic gel, the elastic rebound forces trapped gas through the narrow entry hole at high velocity. Vortex shedding at the hole lip produces a tonal signature governed by the Strouhal relation: f = St × v / D, where St ≈ 0.2, v = gas exit velocity, and D = hole diameter. For a 7.62mm (.30 cal) entry hole at 188 m/s, this yields exactly 4,940 Hz.

A van was observed parked under the covered walkway at GPS coordinates 40°16’38.92”N, 111°42’50.60”W — just 4.84 meters from the tent. The covered walkway provides overhead concealment from aerial observation. The van’s rear hatch was open in temporal proximity to the event.

TDOA multilateration — 4940Hz source localized 3.3m from van
TDOA multilateration using GPS-verified camera positions. Yellow pentagon: 4940Hz source at (-2.04, -1.40)m, 3.3m from van. Green circle: Mach cone. Red star: muzzle blast direction.
Cavitation collapse pulses in ballistic gel
Cavitation bubble collapse impulses detected at 5.4ms intervals in the 300–3kHz band. Strongest at 2.mov (closest to van). Pattern matches ballistic gel temporary cavity oscillation.
Updated Finding: 4940Hz Source is Ballistic Gel Impact

The 4940Hz tone is consistent with Strouhal vortex shedding from a .30 caliber bullet hole in 20% ballistic gelatin at Mach 0.55 gas velocity. TDOA multilateration localizes the source 3.29m from a van parked under the covered walkway. Periodic cavitation collapse pulses at 5.4ms intervals confirm the gel impact mechanism. The Helmholtz/battery hypothesis remains a possible secondary contributor but cannot explain the TDOA localization offset, the stereo directional pattern, or the cavitation signatures. The full supplemental report is available in the downloads section.

Transparency Note

This analysis supersedes the previous Helmholtz/battery attribution of the 4940Hz signature published on this site. Science demands following the evidence wherever it leads. The TDOA localization, Strouhal aeroacoustic physics, and cavitation detection collectively point to a ballistic gel impact in a nearby vehicle as the more precise explanation. The original spectral discovery by Lookoutfa Charlie remains valid — only the source attribution has been refined.

Five signatures, five sources

Beyond the 4940Hz tone, computational analysis of the six cell phone recordings has identified four additional acoustic signatures, each originating from a distinct spatial location. Together, they form a five-component acoustic fingerprint that is inconsistent with a single-point event.

t ≈ 0ms
Mach Cone (500–8kHz)
Supersonic shockwave sweeping the courtyard as the bullet passes. Sub-millisecond rise time, broadband N-wave signature. Localizes toward the rooftop shooter position 127.4m (139.3 yards) away.
t ≈ 30ms
RØDE Detonation (200–6kHz)
Chemical explosion from the transmitter at the tent. Detected at 5 of 6 cameras with slow rise time (2–12ms) — the acoustic signature of a thermal/chemical event, not a shockwave. +16.2dB energy spike at the nearest camera (13.mp4, 4.2m).
50–133Hz
Chest Cavity Resonance (20–200Hz)
The victim’s thorax resonating after the RØDE shockwave impact. Peaks at 53Hz, 80Hz, and 133Hz match published chest wall structural modes and Helmholtz lung cavity resonance. 100–200ms decay — confirmed by independent biomechanical analysis (xAI/Grok).
+233ms
Muzzle Blast (30–500Hz)
Propellant gas expansion from the rooftop, arriving after the crack-to-blast delay. Confirmed across all 6 angles. TDOA localizes toward the shooter position at 100m along the tent-to-shooter bearing.
4940Hz
Ballistic Gel Tone (4.7–5.2kHz)
Strouhal vortex shedding from gas venting through 7.62mm bullet hole in ballistic gel. TDOA source: 3.3m from van. +12.7dB stereo asymmetry at nearest camera. Cavitation collapse pulses at 5.4ms intervals confirm gel impact.
Critical Finding: Multi-Source Coordinated Event

Five distinct, temporally separated, spectrally distinct acoustic signatures — each localizing to a different spatial origin via independent TDOA analysis — are inconsistent with a single-point acoustic event. A device malfunction at the tent cannot produce a Mach cone from a distant rooftop, a muzzle blast arriving 233ms later from 127 meters away, or a 4940Hz tone originating 3.3 meters from a parked van. The acoustic evidence supports a coordinated multi-source event involving at least three spatially separated locations: the rooftop, the tent, and the van.

The physical evidence, video analysis, gas imaging, acoustic triangulation, source distance computation, and five-signature acoustic fingerprint converge on one conclusion: this was a coordinated, multi-source event

The gas-first temporal sequence. The radial motion pattern. The epicenter at the transmitter location. The component trajectories. The glass fragmentation in the transport vehicle. The necklace physics. The wound characteristics. The stereo ITD polarity pattern proving a near-field source. The audio-visual propagation delay showing 0–2 frames (not 11). The 4940Hz Strouhal tone localized to a van 3.3 meters away. The cavitation collapse pulses at 5.4ms intervals. The chest cavity resonance at 50–133Hz. The muzzle blast arriving at +233ms from 127 meters. Five acoustic signatures from five different sources. Each piece of evidence independently supports the same conclusion. Together, they demand examination.

Tyler Robinson's life hangs in the balance. This evidence deserves to be examined. The defense deserves to present it. And the truth — whatever it ultimately is — deserves to be found through rigorous forensic analysis, not assumptions.

Tracking the circuit board

The RØDE Wireless PRO TX transmitter housing suffered an energetic event that compromised the case. Enhanced video analysis — using CLAHE contrast enhancement, multi-scale gradient mapping, and false-color rendering — tracks the liberated circuit board from the moment of case breach across the subject’s chest, through capture in the shirt collar, and down inside the shirt to a final resting position at the waistline.

Six independent lines of evidence converge: multi-angle video, CLAHE enhancement, multi-method 3D surface analysis, colorimetric spectral matching, dimensional calibration via the MagClip GO clasp, and before/after housing comparison.

Key Finding

The TX housing bulge visible in pre-event footage is absent in all post-event frames. The circuit board, liberated from its housing by the energetic event, is tracked across the subject’s chest, captured momentarily in the shirt collar, and observed descending inside the shirt to a resting fold at the waistline.

TX Housing Intact

Multiple camera angles confirm the RØDE Wireless PRO TX is mounted on the subject’s chest via MagClip GO, creating a visible rectangular bulge (44×45.3mm) under the white “FREEDOM” t-shirt.

CLAHE Enhanced4 Camera AnglesTX Bulge Present
Pre-event CLAHE enhanced views
Exhibit 13-A
Pre-event TX housing bulge visible across all four camera angles.

Case Breach

The energetic event compromises the TX housing, liberating the internal circuit board. Enhanced slow-motion video captures head reaction and body response.

2.mov — Camera 2CLAHE Ultra-SlowHead Reaction
Exhibit 13-B
2.mov — CLAHE enhanced ultra-slow head/body reaction.
Exhibit 13-C
IMG_6368 — CLAHE ultra-slow event capture.

Across the chest

The freed circuit board travels laterally across the chest under the shirt fabric, creating textural deformation patterns inconsistent with normal fabric drape.

Close-up CLAHEShirt DeformationPCB Dimensional Match
Exhibit 13-D
Close-up CLAHE super-slow — textural deformation across the chest.

Caught at the neckline

The circuit board is momentarily captured at the shirt collar. Colorimetric analysis confirms the object’s blue/red ratio (B/R = 0.567) is shifted toward blue compared to surrounding skin (B/R = 0.546), consistent with a dark blue PCB solder mask viewed through white cotton.

Neck DeformationColor Forensics B/R MatchDimensional Analysis
Exhibit 13-E
Neck deformation — rectangular object at collar opening matches RØDE TX PCB.

Gravity takes over

Released from the collar, the circuit board descends inside the shirt. Two independent camera angles capture the downward travel, eliminating the possibility of a visual artifact.

Front ViewSide ViewIndependent Corroboration
Exhibit 13-F
PCB descent — front view.
Exhibit 13-G
PCB descent — independent side view corroboration.

The housing bulge is gone

Post-event imagery confirms the TX housing bulge at the original chest mounting location is absent — the MagClip external magnet remains, but the housing is gone. The PCB has come to rest in a shirt fold at the waistline.

Bulge Absent Post-EventAll Angles Confirmed4 Camera Views
● Pre-Event — Housing Present
Pre-event TX housing bulge
Rectangular bulge consistent with RØDE TX (44×45.3mm) visible under shirt across four camera angles.
● Post-Event — Housing Absent
Post-event TX housing absent
Flat shirt profile at original TX location. Housing and PCB no longer present.

Multi-method enhancement

Six independent enhancement techniques each confirm the 3D form change between pre-event and post-event frames.

Enhancement comparison grid
Exhibit 13-H
Six enhancement methods: Original, CLAHE, Multi-Scale Gradient, Normal Map, Height Map (SFS), Emboss. Top: pre-event rectangular form. Bottom: post-event diffuse profile.
Color forensics
Color Forensics — Collar vs. PCB
Object under collar (B/R = 0.567) shows blue shift vs. skin (0.546). Consistent with dark blue PCB solder mask through white cotton.
Dimensional analysis
Dimensional — MagClip Scale Calibration
MagClip GO clasp (26×17mm) as in-situ scale: Object ~138×135px (ratio 1.02), matching RØDE TX body 44×45.3mm (ratio 0.97).
CLAHE close-up
CLAHE Close-Up — Pre-Event Housing
High-resolution CLAHE of pre-event TX housing showing rectangular 3D form with RØDE dimensional specifications.
Housing relief
CLAHE Grayscale — Housing Relief
Contrast-enhanced grayscale isolating rigid 3D housing relief under fabric.

Physical Evidence Chain — Trajectory Reconstruction

  1. Pre-event: RØDE TX mounted via MagClip GO, 44×45.3mm bulge confirmed across four angles.
  2. Energetic event: TX housing compromised, circuit board liberated.
  3. Transit: PCB travels across chest producing visible textural deformation.
  4. Collar capture: PCB caught at collar. Colorimetric (B/R = 0.567) and dimensional (~44mm square) analysis confirm identity.
  5. Descent: PCB descends inside shirt, tracked by two independent camera angles.
  6. Resting position: PCB at rest in waist fold. TX housing bulge absent from original location.
Reproducibility

Each stage is independently verifiable from raw source video. All enhancements use standard forensic techniques (CLAHE, gradient mapping, Shape from Shading) applied uniformly. The full interactive timeline with frame-stepping controls is available for detailed examination.

The spatter that was not there

The prosecution's case requires a high-velocity .30 caliber rifle round impacting a vascular neck target. Peer-reviewed fluid mechanics — Comiskey, Yarin, Kim and Attinger 2016 (Phys. Rev. Fluids 1, 043201) and Comiskey, Yarin and Attinger 2017 (Phys. Rev. Fluids 2, 073906) — predicts three independent observables for that scenario: a dense back-spatter halo of fine droplets around the entry wound, a forward-spatter cone reaching surfaces downstream, and visible vascular bleeding onset within tens of milliseconds.

None of the three are present in the recorded evidence. The bleeding-onset interval measured directly from video2_1.mp4 is 429–461 ms — 30× to 110× longer than the physics predicts for any major vessel.

video2_1.mp4 frame-by-frame bleeding onset timing

Exhibit BS-04 — video2_1.mp4 timing reference: F68 (T+0) → F70 (+66 ms) → F81 (+429 ms, first mark) → F82 (+461 ms, blood unambiguous)

~22M
Forward-spatter drops predicted
From 1.5 mL of ejected blood through the exit wound. The model predicts a visible 16 cm-radius cone pattern on any surface within 2 ft downstream. The available recordings show no such pattern on the banner or backdrop.
4–14 ms
Predicted bleeding onset
For an exposed carotid or jugular wound at the stated impact velocity. Set by Bernoulli–Torricelli jet velocity through a wound channel of bullet diameter and a photographically resolvable 1 mL blood volume.
429–461 ms
Observed delay
First shirt movement at Frame 68 (t=2.239s); first visible blood at Frame 81–82 (t=2.668–2.700s). 30× to 110× longer than vascular flow physics predicts.
3 / 3
Predictions failed
Back-spatter halo: absent. Forward-spatter cone: absent. Bleeding onset within tens of ms: absent. Three independent hydrodynamic predictions fail simultaneously.
Forensic significance

The standard high-velocity rifle-shot account predicts three independent observables, each derived from named equations in peer-reviewed work. None are present in the recordings. The simplest alternative that accounts for all three null observations simultaneously is a localized internal energetic event — documented elsewhere on this site through dense optical flow, gas-release imaging, acoustic source localization, and shooter-position analysis — which produces no high-velocity bullet wake through tissue and no major vascular jet, and therefore none of the predicted hydrodynamic signatures.

Full analysis & downloadable pipeline

Equations, code, exhibits, and the complete reproducible Python pipeline are on the dedicated page. All parameters — impact velocity, bullet geometry, blood properties, ejected volume — are exposed for independent investigators to vary and re-run. The predictions follow from the published equations and the prosecution's stated impact parameters; no parameter has been adjusted to fit a desired outcome.

What is proven, what is inferred

A category-by-category review of the State’s evidence — drawn from the sworn probable-cause affidavit, the charging Information, the unsealed ATF report, and statements made on the record in open court. Each item is graded by what the primary record actually establishes, against what rests on inference, characterization, or contested data the defense has not yet been able to examine.

What this is — and is not. This is an analysis of evidentiary weight and gaps, not a determination of guilt or innocence. “Inferred,” “contested,” and “unverified” do not mean “false.” Every grade below is anchored to a public primary document or an attributed on-the-record statement. Where the record is silent, that silence is noted as a gap — suggestive, not dispositive, given that discovery in this case reportedly exceeds 600,000 files.
Filter All Record establishes Inferred / contested Record does not show
Record Establishes

The Event

ShowsA single shot struck Charlie Kirk in the neck at ~12:23 p.m. on Sept. 10, 2025 — witnessed by thousands and captured on multiple recordings.
ShowsSurveillance filmed a figure in dark clothing crouch, then lie prone at the Losee Center roof edge with a line of sight to the courtyard. Not in dispute.
Record Establishes

The Recovered Weapon

ShowsA scoped Mauser 98 .30-06, wrapped in a towel, was recovered in the wooded strip along the northbound route. The rifle tested operable.
ShowsATF matched the fired casing found with it to that rifle. (Whether it fired the fatal bullet is a separate, unresolved question — see Ballistics.)
Inferred / Contested

Confession Texts & Discord

ShowsMessages describe non-public detail — a rifle drop point, leaving it wrapped in a towel, engraving bullets, changing outfits.
ButPhotographed off the roommate’s screen, not forensically extracted; sender identified only as the contact name “Tyler”. Discord told USA TODAY the rifle messages were not sent on its platform.
Inferred / Contested

DNA “Consistent With Robinson”

ShowsState asserts DNA consistent with Robinson on the trigger, fired casing, two unfired cartridges, towel, and screwdriver.
ButLead defense counsel stated in court the scene DNA is a mixture of at least five individuals. Defense has summary reports only — not the raw data. No contributor count or statistical model is public.
Record Does Not Show

Ballistics — the Fatal Bullet

ShowsATF identified the fired casing as fired in the recovered rifle. Rifle tested operable.
GapThe autopsy bullet was inconclusive — “could not be identified or excluded.” No casing was recovered at the rooftop, and no test dates any firing. FBI re-analysis incomplete.
Record Does Not Show

Identity of the Shooter

GapThe affidavit concedes the figure’s face was never clearly captured (cap, sunglasses, head down). No facial-recognition match. No eyewitness facial ID.
Rests onFamily likeness recognition of released images, the confession, and DNA on recovered items — not a visual identification.
Record Does Not Show

The Clothing Change

GapNo camera captured a clothing change. Nothing independently ties the dark-shirt rooftop figure to Robinson’s filmed maroon-outfit arrival that morning.
Rests onAn inference, plus a reference in the contested texts (“where I changed outfits”).
Record Does Not Show

Rifle on the Rooftop

GapNo camera identified a rifle — before, during, or after. The affidavit’s fleeing object is “an item, whose identity is not clear from the surveillance.”
NoteLater coverage upgraded this to “what appeared to be a rifle” — a characterization drift from the sworn wording. (No roof casing is expected: the Mauser is bolt-action.)
Record Does Not Show

Rifle Deposit in the Woods

GapSurveillance stops at the road crossing, then switches to “investigators discovered” the rifle. No footage shows the figure reaching, entering, or depositing it in the treeline.
Rests onDirection of travel plus the contested texts — not footage of the act.
Inferred / Contested

The Screwdriver (Rooftop Anchor)

ShowsState’s strongest rooftop tie: DNA said to match Robinson on a screwdriver recovered from the roof.
ButAbsent from the charging Information. Touch DNA on a portable object proves contact at an unknown time — not presence at 12:23 p.m. Function and exact location never explained. Falls inside the five-contributor mixture problem.
Record Does Not Show

Fingerprints & Palm Print

ShowsPrints and a smeared palm print found at the northeast rooftop edge; used as a DNA source.
GapNo friction-ridge identification of Robinson disclosed; no disclosed DNA match from the prints. The shoe impression is a Converse class match — not individualizing.
Record Does Not Show

Cellphone Location Data

ShowsPhone location data and a Google Maps route place Robinson’s device near campus on Sept. 10.
GapDoes not place the phone on the rooftop or fix it to 12:23 p.m. — ties the device, not necessarily the person, to the area at unspecified times.
Inferred / Contested

Custody & the Discord Timeline

IssueA defense filing reportedly shows Miranda read 6:25 p.m. with counsel invoked — predating the 8:57 p.m. Discord “surrender” post and the 10 p.m. booking. The affidavit’s own arrest field reads 04:00 the next day.
OpenAt least three inconsistent custody times. Innocent explanations (clerical, jurisdictional, time-zone) remain unresolved. Primary transcript not yet public.
Record Does Not Show

Gunshot Residue

GapNo GSR test appears anywhere in the public record. GSR cannot date a firing or identify a shooter in any event.

15.1 — The Chain of Inference

The State’s narrative is a sequence of links. Several of the load-bearing ones are established by inference or contested data rather than by direct, individualizing proof. Isolated, in order:

15.2 — On the Record

“DNA evidence that was seized from the scene consisted of a mixture of at least five different individuals.”
Kathryn NesterLead defense counsel, open court, Feb. 2026 · reported by The American Prospect
The autopsy bullet “could not be identified or excluded” as fired from the recovered rifle.
ATF ReportDated Sept. 17, 2025 · unsealed via defense filing, Jan. 2026
The fleeing suspect “appeared to be carrying an item, whose identity is not clear from the surveillance.”
Probable-Cause AffidavitSworn by Officer Brian Davis, Utah AG · Sept. 12, 2025
Read Before Citing
  • This is a weight-and-gap analysis, not a verdict. The strong evidence (confessions; DNA on the recovered rifle and towel along the escape path) ties Robinson to the weapon and to the event — the contested question is whether it independently places his body at the shooting position at 12:23 p.m.
  • Some anchors are as-reported, not yet as-read. The 6:25 p.m. Miranda transcript and the per-item DNA contributor data have not been verified against released primary documents.
  • A probable-cause affidavit is a summary. Its silences are suggestive, not dispositive; full discovery may fill gaps in either direction.
  • Every claim here is sourced. This page deliberately excludes unverifiable, conspiracy-adjacent assertions in favor of primary documents and on-the-record statements.

SOURCES: Charging Information (Utah County Atty.) · Redacted PC Affidavit · ATF Firearms Report · Defense Motion to Continue (Mar. 2026) · In-court statements, Jan–Apr 2026 hearings.  |  Graded from primary documents only.

No trust required

Every claim on this site can be independently verified. Below are the original, unmodified source video files from six cell phone cameras, plus the complete Python analysis script. Download them. Run the code. Check our work. The science either holds up or it doesn’t — and we want you to find out for yourself.

The analysis requires Python 3.8+, numpy, scipy, matplotlib, and ffmpeg. Place the video files in a videos/ folder and run the script. It will extract audio, bandpass filter into five signature bands, perform onset detection, TDOA multilateration, stereo ILD analysis, cavitation pulse detection, and Strouhal verification — all automatically.

Analysis Script

acoustic_fingerprint_analysis.py

Complete five-signature analysis pipeline. TDOA multilateration with GPS-verified positions, Strouhal 4940Hz verification, cavitation collapse detection, stereo ILD directional analysis. 328 lines, fully commented.

Download Python Script

Source Video Files

Six cell phone recordings from the UCCU Center courtyard. These are the original, unmodified files. Each contains synchronized audio and video tracks. Together they provide the multi-angle acoustic data for TDOA source localization.

2.mov

iPhone 15 Pro • 59.94fps • 44.1kHz stereo • Closest to van position (6.4m). Strongest 4940Hz detection (+12.7dB L-channel). Primary cavitation pulse camera.

Download 2.mov

7.mp4

29.97fps • 44.1kHz stereo • Northwest position. Farthest from courtyard cluster. 123 N-wave zero crossings — nearest to supersonic trajectory.

Download 7.mp4

13.mp4

30fps • 44.1kHz stereo • Closest to tent (4.0m). Strongest RØDE event detection (+16.2dB energy spike). First Mach cone arrival.

Download 13.mp4

img_6368.mov

29.97fps • 44.1kHz stereo • East position. Strongest low-frequency (chest resonance) capture at -16.1dB. +6.8dB L-channel for 4940Hz.

Download img_6368.mov

video2_1.mp4

30fps • 48kHz stereo • East of tent. Significant pre-event ambient noise. Audio arrival delayed relative to other cameras.

Download video2_1.mp4

1.mp4

30fps • 48kHz stereo • North of tent. Contains close-proximity voice throughout recording. 9 stereo polarity flips — equidistant between two sources.

Download 1.mp4
How to Run the Analysis

1. Install dependencies: pip install numpy scipy matplotlib and ensure ffmpeg is in your PATH.
2. Create a folder called videos/ and place all 6 files inside it.
3. Run: python acoustic_fingerprint_analysis.py
4. Results are printed to console and saved as images in results/.

The script uses the same GPS-verified camera positions, event frames, frequency bands, and multilateration algorithm documented in the forensic report. If you get different results, we want to know. Contact: followtheepicenter.com

Download the evidence

★ Reasonable Doubt — The Defense Summary

The hardest evidence that casts reasonable doubt on the official narrative. Canon XA55 audio as centerpiece. The detonation thesis unified with the necklace, shirt deformation, shrapnel, acoustic signatures, 53Hz chest resonance, and decorticate posturing. Start here.

Open Defense Summary

NEW: Canon XA55 — 4-Channel Professional Audio Analysis

Video 15: 4-channel 48kHz uncompressed PCM from a Canon XA55 broadcast camcorder. The strongest acoustic evidence from the event — 733 sub-200µs N-wave zero-crossings (6× phone recordings), three resolved LF events placing the rifle at ~120m, and independent confirmation of every finding from the 10-camera analysis. Interactive visualizations generated from raw source audio.

View Full Analysis

⬢ NEW: Audio Evidence Visualized — Interactive Map

An interactive aerial map of the UCCU Center courtyard built from authoritative GPS coordinates. Eleven recorder positions, four acoustic sources, and animated wavefront propagation at 343 m/s. Click any pin to see precise arrival times for every signature. Press play to watch the Mach cone, muzzle blast, stage detonation, and 4940Hz Strouhal tone propagate across the actual scene geometry. The clearest visual explanation of the timing evidence available.

Open Interactive Map

⬢ NEW: Section 14 — Blood Spatter Hydrodynamic Analysis

Three peer-reviewed hydrodynamic models — Comiskey/Yarin/Kim/Attinger 2016 back spatter, Comiskey/Yarin/Attinger 2017 forward spatter, and Bernoulli–Torricelli vascular flow — predict three independent observables for a real .30 caliber rifle round to a vascular neck target: a dense back-spatter halo on the entry shirt, a forward-spatter cone on the downstream backdrop, and visible bleeding within 4–14 ms. Frame-by-frame analysis of video2_1.mp4 shows 429–461 ms onset, 30× to 110× longer than physics predicts. All three signatures are absent. Full Python pipeline downloadable for independent verification.

Open Blood Spatter Analysis

Van & Ballistic Gel Report

Supplemental report: Strouhal 4940Hz analysis, TDOA source localization, cavitation collapse detection, and why the battery hypothesis is less likely.

Download DOCX

Combined Evidence Report

Complete forensic evidence summary with all exhibits, methodology, analysis, and evidence tables.

Download PDF

Audio Analysis Report

Stereo source separation, acoustic triangulation, and supersonic signature identification.

Download PDF

Gas Analysis Package

Gas release imaging with enhanced frames, methodology, and temporal sequence documentation.

Download PDF

Collar ROI Analysis

Region-of-interest optical flow isolating collar displacement timing. 300ms temporal lead, spatial heatmaps, calibrated event markers.

Download PDF

Audio + Rolling Shutter Analysis

Integrated source localization via audio-visual propagation delay. Five cameras, sensitivity analysis, and calibration-independent distance estimates.

Download PDF

Acoustic Fingerprint & Helmholtz Analysis

4940Hz spectral identification, Helmholtz resonance calculation, source exclusion matrix, and frame-by-frame gas hiss correlation.

Download DOCX

Cross-Source Evidence Synthesis

Section 16 — Five datasets, four methods, two opposition reports. All findings cross-validated. Includes muzzle blast comparison, detonation timeline, 4940Hz coherence proof, and the 192m-underground stationary model result.

View Section 16

10-Phase Forensic Pipeline

Section 17 — FA-2026-002. Full standardized 10-phase forensic acoustic methodology applied to the UCCU dataset. Intake through reporting, with cross-validation against Maher (CNN), Ken Linke 21-recorder set, Canon XA55 PCM, and FR05m mic.

View Section 17

⬡ NEW: Section 18 — Shooter Position Analysis

Five independent acoustic lines of evidence — TDOA multilateration, crack-boom timing across 20+ recordings, Canon XA55 48kHz crack-to-boom interval, N-wave zero-crossings, and Dr. Rob Maher's CNN analysis — establish that the prosecution's claimed shot position is physically impossible. The prosecution's location is 67.1 meters and 24.5° from the acoustically-derived muzzle source, 54.5 meters off the bullet trajectory, and produces a crack-boom ceiling of 159 ms at the Canon XA55 — 43 ms below the observed 202 ms, a gap no geometry can close.

View Shooter Position Exhibit

⬡ NEW: Section 19 — Necklace Trajectory Analysis

Three independent camera angles document the trajectory of a 32-gram stainless-steel necklace during the September 10, 2025 incident. The chain travels upward from its rest position at the neckline, passes over the back of the head, and settles draped across the left shoulder — a trajectory that requires an impulse with both vertical and posterior horizontal components. A single broken link departs forward in the opposite direction, the conservation-of-momentum signature of an internal energetic source. Both observations are independently inconsistent with the prosecution's horizontal-bullet-impact mechanism.

View Necklace Trajectory Exhibit

Open Analysis Pipeline

Source videos, Python analysis code, and full methodology — verify the results yourself. Frame-by-frame. No trust required.

Open Repository

You can support my work by supporting my business

This independent forensic analysis is funded entirely through my small family farm. Every purchase helps keep this research going.

VISIT BLUE MOUNTAIN FARM →

Blue Mountain Farm — Haskell County, Oklahoma