Pure design/theory. Nothing has been built. Every performance figure on this page is a physics-grounded estimate extrapolated from real component technologies — electron microscopy columns, e-beam lithography, MEMS wafer packaging, cathodoluminescent quantum-dot research — never a measured spec. Theory-vs-measured status is flagged wherever numbers appear. The core thesis: CRT was abandoned for market-fit reasons (bulk, weight, power at living-room scale), not physics reasons — and VR is the one application space that punishes none of CRT's weaknesses while rewarding its exact strengths.
01 — System Overview
HEADLINE DESIGN TARGETS
A single steered electron-gun assembly per eye — six converged beams, one per primary, Trinitron converged-assembly lineage — writing onto a litho-patterned quantum-dot phosphor faceplate inside a wafer-level-sealed vacuum envelope. No fixed pixel grid. No per-pixel drive electronics. The beam goes where the eye is looking.
Screen Size (Per Eye)
2–5inch diagonal
The scale that erases every historical CRT weakness: deflection angle, vacuum load, envelope depth all collapse.
Addressing Model
VECTOR
Random-access beam steering (oscilloscope / Vectrex lineage). No fixed raster, no fixed grid — resolution is a scan plan, not a panel property.
Refresh Ceiling
MULTI-kHz
Electrostatic deflection is capacitive — multi-MHz drivable. Ceiling moves to phosphor decay + gun modulation. Theory estimate.
Pixel Density Target
1000–2500+PPI-class
Set by phosphor-stripe lithography + color bleed — NOT by gun or yoke, which carry 100–1000× precision margin.
Brightness (At Phosphor)
500–1000+nits
Phosphor-saturation-limited, not gun-limited: Schottky current headroom is 2–4 orders of magnitude beyond need. Theory estimate.
Response
SUB-ms
Impulse-driven phosphor flash + decay. Zero sample-and-hold persistence blur — CRT's most legitimate historical advantage, fully retained.
Color System
RGBMCY 6-Primary
Dual-triangle hex close-packed cluster (RYG / MBC rows). Gamut hexagon structurally exceeds any RGB triangle. Zero PPI cost.
Blacks
True Zero
Unstruck phosphor emits nothing. MicroOLED effectively ties here — no overclaim.
The Structural Advantage
Hardware-Native Foveated Rendering
Fixed-grid panels must refresh every physical pixel every frame — foveation there saves GPU only. A vector-scanned beam physically skips low-priority space: peripheral resolution savings convert directly into central refresh-rate gains. No flat panel can replicate this at the hardware level.
02 — Physical Architecture
TUBE CROSS-SECTION — GUN TO EYE
Every subsystem below is a locked-in architecture decision. Component lineages are real and shipping in other industries; the combination is not.
Fig. 01 — Schematic, not to scale. One assembly per eye. Beam paths animated for illustration; real scan order is the vector scheduler's problem (§04).
S1
Electron Source — Schottky Field Emission (ZrO/W)
Zirconium-oxide-coated tungsten tip at ~1800 K; extraction field lowers work function via Schottky effect. The standard source in modern SEM/TEM columns. Single steered gun assembly with 6 converged beams — one per primary — explicitly NOT per-subpixel emitter arrays (the 2000s FED manufacturing nightmare, and it defeats the steered-beam architecture). Thermally self-cleaning → stable and long-life; nm-scale coherent virtual source; <1 W heat plausible.
Source
S2
Deflection — Electrostatic MEMS Plates
Capacitive flat electrodes, lithographically batch-fabricated — no wound copper yokes. Electrostatic deflection's one historical weakness was voltage cost at wide angles; at 2–5″ screens the required angle collapses, removing it. Beam position = plate voltage = a DAC bit-depth problem. E-beam lithography demonstrates nm-scale electrostatic addressing — 100–1000× margin over subpixel pitch.
Steering
S3
Vacuum Envelope — Wafer-Level Sealed + NEG
Anodic / glass-frit bonding of gun-side substrate to phosphor faceplate under vacuum — standard MEMS process (gyros, bolometers). Thin-film non-evaporable getter coatings on internal cavity walls scavenge continuously; no flash-getter mirror. Short throw at this scale → short scatter path → forgiving vacuum spec. Scattering closed as a non-issue at this scale.
Quantum-dot-doped phosphors, emission wavelength tuned by dot size (quantum confinement) → narrow peaks, purer primaries than legacy P22-class grains. ALD-deposited atomically-thin oxide shells per grain resist e-beam degradation and raise quantum efficiency. Photolithographic stripe/dot deposition — Sony aperture-grille geometric lineage with zero mechanical wires. The system's real bottlenecks live here: stripe-pitch lithography and beam-scatter color bleed set max PPI.
Faceplate curvature is a physics byproduct (constant beam throw distance), and it lands in the favorable direction: convex toward the viewer for wide-FOV eyepiece field-curvature correction. Relay: small array of Sony camera-grade aspherical multi-coated elements at fixed deterministic distance. Structurally simpler than pancake stacks correcting flat panels — avoids the sweet-spot, ghosting, and brightness-tax problems documented in shipping 2026 headsets.
Optics
03 — Electron Source Trade Study
THE GUN SHOOTOUT — FOUR CANDIDATES, ONE SURVIVOR
Select a candidate. Decision is final unless reopened; ratings are qualitative engineering-judgment scores from the trade study, not measurements.
CHOSEN — THE SEM/TEM COLUMN SOURCE
ZrO/W tip at ~1800 K with strong extraction field. 2–4 orders of magnitude more current density than thermionic; nm-scale coherent virtual source; thermally self-cleaning — adsorbed gas bakes off continuously, so emission is stable for years, not hours. Forgiving vacuum requirement, non-toxic materials, <1 W heat plausible.
Consequence: the gun is never the brightness bottleneck (phosphor saturation is), and spot size carries 3+ orders of magnitude headroom over the ~5–10 µm subpixel pitch needed for 1000–2500+ PPI.
Current Density
9
Stability / Life
9.5
Vacuum Tolerance
8
Power / Heat
7.5
System Simplicity
8.5
TIER 2 — BEST RAW EFFICIENCY, WORST HABITS
CNT / Spindt / nanodiamond cold cathodes: best raw efficiency and instant-on. But emission current drifts as gas molecules adsorb onto the tip — the classic cold-FE reliability problem — demanding ~10⁻¹⁰ torr and periodic tip flashing.
Manageable with modern getter technology, but it trades away the one property a consumer display must have: set-and-forget stability. Kept on the bench as the fallback if Schottky heater power ever becomes the binding constraint.
Current Density
8.5
Stability / Life
4
Vacuum Tolerance
2.5
Power / Heat
9.5
System Simplicity
7
TIER 3 — SOLVING A BOTTLENECK WE DON'T HAVE
Laser-triggered photocathodes are the fastest, most precise electron sources in existence — free-electron laser and ultrafast electron diffraction heritage.
But they add a laser driver and an optical excitation path inside the envelope, for temporal precision the display never needs: gun modulation already carries massive headroom over phosphor decay. Complexity purchased, nothing gained.
Current Density
7
Stability / Life
6
Vacuum Tolerance
4.5
Power / Heat
5.5
System Simplicity
2
REJECTED — THE LEGACY CATHODE
The barium-oxide cathode that powered every living-room CRT: ~800 °C heater, 5–10 W continuous draw, 30–60 s warm-up, and toxic barium/strontium oxide chemistry.
Every one of those properties is disqualifying in a head-mounted device. Rejected without regret — this is precisely the part of CRT heritage the design leaves behind.
Current Density
2
Stability / Life
6.5
Vacuum Tolerance
7.5
Power / Heat
1
System Simplicity
6
Precision Headroom — Spot Size
100–1000×
Schottky spot size (sub-nm to low-nm in electron microscopy) vs. the ~5–10 µm subpixel pitch this display needs. The gun is never the resolution bottleneck.
Current Headroom vs. Need
10²–10⁴×
Brightness ceiling is set by phosphor saturation, not beam current. Theory extrapolation from SEM/TEM column data.
Beam Configuration
6 Beams · 1 Assembly
Classic Trinitron "single gun" = one assembly, three converged beams, geometric color separation. CRT-VR: same converged-assembly principle at six — "6 guns, 6 dots" — with a litho-patterned grille.
04 — Addressing Model
RASTER vs VECTOR — WHY THE BEAM DOESN'T SWEEP
A fixed raster wastes beam time sweeping low-priority space at constant density. Vector / random-access steering (oscilloscope, radar, Vectrex lineage) jumps the beam point-to-point and skips sparse regions entirely — the enabler of hardware-native foveation (§05). Toggle the modes; the frame-time budget on the right is the whole argument.
Interactive · Scan Model Simulator
Simulated phosphor persistence · illustration, not a physical model
Mode
VECTOR
Beam Dwell Points / Frame
—
Raster must visit every grid position; vector visits only where content lives.
Frame-Time Budget Used
—
Unused budget converts directly into refresh rate or foveal density — nowhere for a fixed grid to bank it.
Lineage
OSCILLOSCOPE · VECTREX
05 — The Killer Feature
HARDWARE-NATIVE FOVEATED RENDERING
Shipping headsets already do eye-tracked Dynamic Foveated Rendering — but on fixed grids it saves GPU only: every physical pixel still refreshes every frame, so display-refresh time saved is zero. Here the beam physically skips low-priority space, so peripheral savings convert into central refresh gains. Move your cursor across the display below — it is the gaze point. Structural, not incremental: panels cannot close this gap by iterating.
Interactive · Gaze-Following Scan Density
Cursor = gaze · green = fovea · blue = midground (interlaced) · dim = periphery (full temporal rate)
Scan Points This Frame
—
vs. — for a fixed grid at equal foveal density — the grid must refresh all of them, every frame.
Beam Time Reclaimed
—
Converts directly into central refresh-rate gain.
Illustrative Central Refresh
—Hz-class
Illustration of the conversion mechanism from a 120 Hz full-field baseline — not a measured or committed spec (TRL 2–3).
FOVEA
No Compromise
Full/native resolution (AI-upscaled from ~1080p optional) at full progressive refresh. Acuity AND flicker sensitivity both peak here — nothing is cut.
MIDGROUND
Temporal Cuts Live Here
Low resolution, AI-upscaled, interlaced / temporally subsampled. Acuity is low and motion sensitivity has already fallen from its peripheral peak — the one safe place to cut temporal corners. Vector addressing means "interlace" can be any per-region subsampling pattern, not just alternating lines.
FAR PERIPHERY
Never Interlace — Perceptual Trap
Lowest resolution, raw — but FULL temporal refresh. Peripheral vision is rod-dominated: LOW acuity but HIGH flicker/motion sensitivity (evolutionary motion detection). Interlacing here reads as visible shimmer. Counter-intuitive; do not violate. Cut spatial detail only.
Precedent & Open Thread
Tiering precedent: Varjo's dual-panel foveated headsets, optically stitched — CRT-VR is one continuous beam-addressed surface, a cleaner solve. Honest open thread: the eye-tracking → beam-steering control-loop latency budget (saccade velocity vs. per-frame scan-plan recompute) has been raised and never worked out. See §08.
06 — Color Architecture
6-PRIMARY RGBMCY — THE GAMUT HEXAGON
Dual interleaved triangles per pixel cluster, hex close-packing (RYG / MBC rows). Same pixel pitch as 3-primary — each dot shrinks to ~half area (~30% diameter reduction) — so the sixth primary costs zero PPI. Mixing is spatial/eye-blended only, like classic subpixel rendering; colocated photon summation was rejected as unnecessary complexity. Precedent: Sharp Quattron (RGBY) and the multi-primary display literature.
Red
QD stripe · R
Yellow
QD stripe · Y
Green
QD stripe · G
Magenta
QD stripe · M
Blue
QD stripe · B
Cyan
QD stripe · C
Chromaticity Coverage — Hexagon vs Triangle (schematic)
Geometric schematic, not CIE-accurate — final primary wavelengths are open question №6 (§08). The structural point holds regardless: a hexagon through six primaries contains every triangle through three of them.
Low brightness → purest 1–2 primaries per hue. Hue discrimination peaks at low-to-mid luminance; saturation is protected where the eye can see it.
High brightness → recruit all relevant cluster dots for maximum luminance. Chromatic sensitivity falls at high luminance — Bezold-Brücke territory — so purity can be traded away invisibly.
Pure firmware/driver logic — the hardware doesn't care.
Crossover points and perceptual validation: open question №11 (§08).
Gamut-Mapping Pipeline (Host-Side)
Real-time RGB→RGBMCY conversion, metamerism-aware: multiple valid drive combinations exist per color — optimize for brightness / efficiency / power.
Natural home: the VR render pipeline, which already performs heavy per-frame distortion and color correction. Compute cost lives here, not in display hardware.
Software-prototypable today — flagged as a good first build task (§08, №5).
07 — Competitive Position
CRT-VR vs THE FIELD — MID-2026
Disclosure, stated every time this table is reused: the CRT-VR column is physics-grounded theory at TRL 2–3 — nothing built. Every other column is shipping or mass-produced technology as of mid-2026 (Pimax Dream Air: Sony 4K microOLED, 3840×3552/eye, <170 g; Pimax Crystal Super: 116° horizontal FOV, 57 PPD; Apple Vision Pro class: Sony 1.3″ 4K OLED-on-silicon; refresh record: XREAL X1 at 240 Hz, 1080p only). That asymmetry is real.
Metric
CRT-VR (theoretical)
MicroOLED
OLED (TV-class)
True RGB LED
MiniLED LCD
Peak Brightness
~500–1000+ nits at phosphor — phosphor-saturation-limited, not gun-limited. Relay optics lose less than pancakes → at-eye advantage even at equal panel output.
1000 nits panel-rated (Sony ECX344A); substantially less at eye through pancakes
Persistence blur; strobing helps at brightness cost
Persistence blur
Depends on drive
Worst of group
Burn-In
QD + ALD passivation designed against it — unproven
Real, documented on static HUDs in 2026 product
Real
Minimal
None
Weight / Packaging
Unknown — honest open question №1
107–170 g shipping
N/A
N/A
N/A
Power
Single emission point + deflection — no per-pixel tax; plausibly better per-lumen
Millions of driven diodes; scales with lit pixels
Same
Same
Backlight-dominated
Foveated Rendering
Hardware-native — beam skips sparse regions; savings become refresh
GPU-side only; every pixel still refreshes
GPU-side only
GPU-side only
GPU-side only
Maturity / Cost
TRL 2–3; unknown yield, cost, lifetime
Mass production, $899–$2,200 street
Mature
Mature
Mature
Hover rows to compare. Where microOLED flatly wins: it exists — proven longevity, mature supply chain, no vacuum envelope, no HV near the face, silicon-backplane sensor integration, solved per-eye packaging. Every measurable-today metric defaults to shipping product.
The "Window-Through" Sum
Impulse motion clarity — under fast head motion (VR's constant condition) CRT should read as glass, not screen.
No fixed pixel structure — phosphor point-spread function is a natural reconstruction filter: no screen door, no hard edges, graceful non-native scaling. Analogous to a DAC reconstruction filter. Friendly to weak GPUs.
MicroOLED delivers blacks and near-invisible pixels; it cannot deliver impulse motion or the gamut hexagon.
Verdict Framing
The pitch is not "better specs."
The pitch is: "the only display architecture where foveated rendering is physics instead of software."
Keynote guardrail — never "we shoot your eyes with an electron gun." Preferred framing: "immersive photon-native display." Product name plays it straight: Cathode. Ray. Tube.
08 — Engineering Honesty
OPEN QUESTIONS — PRIORITIZED
Unresolved threads, stated plainly. Perceptual constraints already established and not up for re-litigation: never interlace the far periphery (rod-dominated, flicker-sensitive); the midground is the correct place for temporal subsampling; the fovea gets no compromise; Bezold-Brücke justifies multi-primary recruitment at high luminance.
P0 Feasibility-Deciding
01Per-eye packaging & weight budget▶
Two miniature vacuum tubes + relay optics covering 100°+ FOV vs. 107–170 g shipping microOLED headsets. No honest mass/volume estimate exists yet. This is where "smaller, thinner" is proven or dies. Inputs needed: tube envelope wall thickness at wafer-level-sealed scale, relay lens element count/mass, HV supply mass.
02Phosphor-stripe lithography ceiling & color bleed▶
The actual PPI limit of the whole system. Model beam scatter at short throw + QD stripe pitch vs. cross-talk. Gun and yoke carry 100–1000× margin; this subsystem sets max resolution.
03Phosphor saturation ceiling▶
The 500–1000+ nit estimates hinge on QD-phosphor saturation/degradation limits under e-beam flux. Needs a literature review of cathodoluminescent QD data at minimum; ideally lab data.
P1 Architecture-Completing
04Eye-tracking → beam-steering control loop▶
Latency budget: how fast can the foveal boundary shift before the beam schedule can't keep up? Saccade velocities vs. per-frame scan-plan recompute. Flagged in design sessions, never worked out.
05RGB→RGBMCY gamut-mapping pipeline▶
Real-time 6-channel conversion, metamerism-aware optimization (multiple valid drive combos per color — optimize for brightness/efficiency/power). Precedent: Sharp Quattron processing silicon. Lives in the VR render pipeline. Could be prototyped in software now.
06Six QD emission wavelengths▶
Lock the RGBMCY primary wavelengths for maximum hexagon area vs. the spectral locus, considering QD linewidths and eye sensitivity (CIE 1931 / CIE 170-2 observer).
07Scan scheduler design▶
Vector scan-plan generation per frame: foveal density map → point list → deflection DAC waveform. Tier boundaries, temporal subsampling patterns for the midground. Also software-prototypable.
P2 Productization
08High-voltage supply near the face▶
Acceleration voltage choice, insulation, failure modes, regulatory pathway. Solvable, but real engineering.
09Vacuum envelope lifetime at miniature scale▶
NEG getter capacity vs. outgassing over product life.
10Manufacturing yield / cost▶
Litho phosphor striping + MEMS electrostatic yoke + QD emitters + wafer vacuum sealing have never been combined in one device. Identify closest analog processes: FED post-mortems, MEMS bolometer lines, QD-OLED deposition.
11Drive-strategy tuning▶
XCD-LED-style purity-vs-brightness crossover points; perceptual validation of high-luminance primary recruitment.
† TRL 2–3 throughout. All figures are physics-grounded estimates extrapolated from real component technologies (SEM/TEM Schottky columns, e-beam lithography, MEMS wafer-level packaging, cathodoluminescent QD research) — never measured specs of this device, because this device has not been built.
† Refresh & brightness ceilings are headroom statements, not commitments. Electrostatic deflection is multi-MHz-drivable and Schottky current headroom is 10²–10⁴× — but the shipped ceiling would be set by phosphor decay/saturation, which needs lab data (open questions №3).
† Foveation refresh figures in the §05 simulator are illustrations of the conversion mechanism (peripheral scan savings → central refresh), computed from a notional 120 Hz full-field baseline. They are not performance claims.
† Chromaticity diagram in §06 is a geometric schematic, not CIE-accurate; final primary wavelengths are unlocked (open question №6). The structural claim — a hexagon through six primaries contains every triangle through three — is geometry, not projection.
† Cathodoluminescent quantum-dot phosphors are active research, not fantasy — but unproven at product scale. ALD passivation against e-beam degradation is design intent, not demonstrated lifetime.
† MicroOLED ties on blacks and leads outright on maturity, packaging, and proven density. Comparison rows state this without hedging; the asymmetry (theory vs. shipping product) is disclosed at the top of §07.
† "6 guns, 6 dots": classic Trinitron "single gun" marketing = one assembly, three converged beams, purely geometric color separation via aperture grille. CRT-VR extends the same converged-assembly principle to six beams with a litho-patterned grille — no mechanical wires, no damper-wire shadows.
† Adaptive drive strategy (purity at low luminance, recruitment at high luminance) originates in the XCD-LED display concept and is applied here as firmware/driver logic. See /Sony/XCD-LED for the originating architecture.