Sea · Mk 58 Mod 0 / Mk 72 Mod 0 · Autonomous Undersea Effector

Deadfall

A hundred and eighty days on the seabed at 105 milliwatts — and it cannot arm without a signed authorization that expires on a date certain.

Design · no CDR held 533 MM · 180 DAYS · 2-OF-3 SIGNED All sea systems

Mission

Revenant Ledger

Hold a strait, a chokepoint or an approach with an effector that costs a fraction of a ship, needs no crew, no power and no attention for six months — and, this is the part that matters, can be laid without ending a career.

The United States has a naval mine inventory it is politically unable to use. So does most of NATO. The reason is not that the weapons do not work. It is that no combatant commander wants the headline, no ally wants an autonomous weapon in their exclusive economic zone that somebody else controls and nobody can audit, and no lawyer has ever been given an answer to the question of who decided, at the moment a six-month-dormant weapon woke up and fired.

DEADFALL is two variants of one architecture — a mobile heavyweight effector and an encapsulated seabed effector — built around the answer to that question.

105mW Measured dormant draw against a <2 W requirement — 19× margin.
180d Seabed dormancy on 454 Wh, tiered wake, no transmissions.
2of 3 Threshold signature required to arm. No one human can arm a field.
30m Self-surveyed emplacement accuracy — the minefield chart is a by-product.

Parameters

DEADFALL-T · Mk 58 Mod 0 — mobile heavyweight effector
Diameter / length533 mm / 6.12 m
Mass1,540 kg
Warhead190 kg PBXN-103, contact and influence
LaunchLEVIATHAN magazine by swim-out, submarine tube, STYX deck cell, or air drop
Transit / attack speed18 kt / 42 kt
Range32 nm at 18 kt; 6.4 nm at 42 kt
GuidancePassive and active seeker, wake-homing, optional fibre-optic guidance wire to 24 km
Dormancy40 days bottomed — a larger propulsion pack costs more state-of-charge maintenance
Dormant power190 mW measured
Unit cost target$1.4M
DEADFALL-M · Mk 72 Mod 0 — encapsulated seabed effector
Capsule533 mm × 3.90 m, titanium Grade 5
Mass780 kg
Effector324 mm swim-out vehicle, 44 kg shaped charge
Emplacement depthTo 480 m bottomed
Lay methodsLEVIATHAN conveyor (40 per payload fit), STYX chute, submarine tube, or by hand from a fishing boat
Terminal6.2 nm at 40 kt from the emplacement
Dormancy180 days at 105 mW — 454 Wh on a 1.10 kWh lithium thionyl chloride bus
Sensor headBuoyant, on a 2.4 m tether above the sediment plane; burial depth inferred from tether tension
ArmingRotary out-of-line safe-and-arm. No software path arms this weapon.
Post-expiry900 mW acoustic responder for 90 days off a reserved cell, so clearance is a survey and not a search
Unit cost target$310k

The 180-day power budget, done honestly

The requirement says under two watts. Two watts for 180 days is 8.64 kWh — roughly twelve kilograms of lithium thionyl chloride. It is a generous requirement, and meeting it is not the achievement. The interesting question is where the milliwatts actually go, and the answer is not where you would guess.

A classifying receiver cannot be always on, so the wake is tiered. Tier 0 is a mechanically resonant piezo stack tuned to the 40–400 Hz band where ship machinery lines live, rectified into a supercapacitor. It draws nothing — it harvests — and it trips a latching relay above roughly 96 dB re 1 µPa. It is also a dumb detector with a mechanical threshold: a whale, a seismic airgun and a container ship all trip it, and on a busy shelf we expect eight to twenty false wakes a day. The budget below assumes sixty, because building margin into the trigger model is cheaper than being right about it.

DEADFALL-M dormant power — busy-strait traffic model
Tier 0 · resonant detectorAlways on, energy-harvesting, mechanical threshold — 0.00 mW
Tier 1 · line tracker2.5 s wake, 2 channels at 4 kHz, 2,048-point FFT, DEMON and LOFAR line extraction. 310 mW × 60/day — 0.54 mW
Tier 2 · classifier12 s, narrowband line set and blade cadence against the onboard library. 900 mW × 22/day — 2.75 mW
Tier 3 · track and decide120 s, array processing, track formation, modem in receive, ledger write. 6.4 W × 3/day — 26.7 mW
HousekeepingMCU deep sleep at 65 µW, pressure/tilt/leak at one sample per minute, tamper accelerometer — 0.32 mW
Command windowOne 40 s receive-only listen per day for a REVENANT MESH order — 0.51 mW
Commanded transmit6 s at 178 dB, 42 W electrical, limited to four per deployment — 0.07 mW
Sensing subtotal30.9 mW
Propulsion-pack SOC maintenance74.0 mW — and this is the line that dominates
Total average105 mW · 454 Wh over 180 days · 19× margin against the requirement

The dormancy bus runs on lithium thionyl chloride, which self-discharges at about 0.7% per year at 20 °C and less on a 4 °C seabed. Over 180 days that is negligible. But the effector's propulsion pack is lithium-ion, and lithium-ion self-discharges at two to three per cent per month. Left alone for six months a terminal-run battery arrives at the fight at 82% state of charge, which is eighteen per cent off the terminal range, which is the difference between a hit and a miss. So the primary bus trickle-maintains it: 1.78 kWh × 18% over 4,320 hours is 74 mW — 2.4 times the entire sensing budget.

That is the finding. What makes a 180-day undersea dormancy hard is not listening. It is holding a battery at charge in the cold and the dark for half a year. It does not belong in a brochure, because it is not an achievement, it is an admission — and it is the number that governs the design.

And the power budget is not what actually kills a 180-day emplacement. In descending order of what has ended real seabed deployments: seal and connector life, which is why the capsule is titanium with no dissimilar-metal couples, anodes sized for 400 days, and a 40 µW conductivity cell watching the O-ring interspace so the weapon knows it is leaking before the leak reaches anything; sediment burial, which costs 6 to 14 dB of aperture and is why the sensor head sits on a tether above the mud; biofouling on the transducer face, about 3 dB by day 120 in warm shallow water; and the three-year magazine life before the weapon ever gets wet, which is a surveillance-lot problem and not a mission one.

Wake on signature — and both numbers, not one

Classification runs on three independent features, fused: narrowband tonals — shaft rate, blade rate, gear mesh, diesel firing rate, turbo-alternator lines, where shaft rate multiplied by blade count is close to a hull-specific pair; DEMON cadence, which recovers shaft rate and blade count directly and survives at lower signal-to-noise than the tonals do; and broadband spectral shape, which separates a diesel merchant from a gas-turbine combatant from an electric-drive vessel.

Classification performance — published in full
Warship vs. merchant, SNR ≥ 6 dB94 %
Warship vs. merchant, SNR ≈ 0 dB71 % — the 94 % is the number that sells; this is the number that gets people killed, and it is why the authorization architecture below is not dressing
Classification range, merchant4–18 km at a 165 dB source level
Classification range, quieted combatant1.5–7 km
Why the spreadSea state, sound-speed profile, bottom type and burial depth. This is a probability distribution, not a specification, and any vendor quoting a single number is quoting the best day they ever had.

Hague Convention VIII of 1907 is the only treaty written specifically about naval mines. It is 119 years old, it addresses automatic contact mines, and it requires that a mine become harmless when it breaks its moorings or when control over it is lost, and that belligerents record where they laid their fields. It has nothing coherent to say about a weapon that classifies its own target. The San Remo Manual of 1994 adds that mines must be recorded, must not cut neutral shipping off from the high seas, and must have a mechanism to render them harmless when they no longer serve a military purpose. DoD Directive 3000.09 requires appropriate levels of human judgment over the use of force — which, for a weapon dormant on a seabed for six months, is close to unanswerable by conventional design, because the human is not there when the decision happens.

Fantom's answer is to separate the decision from the authority, and to make the authority a cryptographic object with an expiry date.

01 · The interlock is metal

There is no code path that arms this weapon

DEADFALL ships, transits and lays with its initiation train mechanically interrupted — a rotary out-of-line safe-and-arm device physically misaligns the detonator from the booster. Moving it in-line requires a torque the arming actuator cannot produce unless the arm authorization validator releases a hard interlock.

Software cannot arm it. A corrupted mission plan cannot arm it. An adversary who owns the autonomy stack cannot arm it. The cryptography's only job is to permit a solenoid. That is deliberate, and it is why this argument is not vapour: the guarantee is a piece of metal.

02 · The authorization

A signed capability token, not a setting

A REVENANT LEDGER arm authorization carries the identity and delegated command chain of every signing human; a 2-of-3 threshold signature, so no single person can arm a minefield; the engagement polygon to 30 m; hard not_before and not_after bounds; the target predicate the classifier must satisfy — for example surface combatant, displacement over 2,000 t, not squawking an authorized IFF, classifier confidence at least 0.92; a consumable maximum engagement count; and the behaviour on expiry.

The token is loaded before lay, or delivered in-mission over acoustic mesh by an authenticated node. With no valid token the weapon is a sensor. It still listens, classifies, tracks, records and reports when interrogated — and it physically cannot fire.

03 · Expiry

A minefield that expires on a date certain is one an ally will let you lay

When not_after passes, the validator revokes the interlock, the safe-and-arm rotates back out of line, and the weapon is permanently and mechanically a sensor. It then reports, scuttles, or waits for recovery according to the token.

That is Hague VIII's render-harmless obligation and San Remo's no-longer-serves-a-military-purpose requirement satisfied by construction rather than by promise. It is the whole product. Everything above it is plumbing.

04 · Fail-disarmed

On every fault, always, in that direction

Expiry depends on the weapon knowing what time it is, six months in, on a temperature-compensated oscillator. A clock running fast expires the field early, which is annoying. A clock running slow leaves live weapons in the water past their authorization, which is unacceptable. So the design fails toward expiry: the validator treats the token as expired at not_after minus three sigma of accumulated clock uncertainty — for a 0.5 ppm oscillator over 180 days, a worst case of 7.8 seconds — and it re-disciplines the clock at every authenticated contact.

And loss of the housekeeping clock's integrity counter for more than six hours forces disarm, full stop, irrespective of the token. A weapon that does not know the time is a weapon that disarms.

05 · The chart

The minefield record is a by-product of laying the field

Hague VIII Article 3 and San Remo both require belligerents to record where they laid mines. Historically those records are appalling, which is why clearance in the Baltic and the Gulf has taken decades.

Each DEADFALL-M knows its own emplaced position to about 30 m — from the laying platform's inertial solution at release plus an acoustic self-survey against its neighbours after settling — and signs that position into the ledger at emplacement. The field produces a tamper-evident, machine-readable minefield record as a side effect of being laid, and at the end of hostilities it is handed to a clearance authority as a file. To a mine-warfare officer that is worth more than any performance parameter on this page.

06 · The record

The minefield can be audited after the war

Every observation, every classification and its confidence, the token presented, the humans who signed it, the moment the safe-and-arm rotated in-line, the moment it rotated out, and — if it fires — the complete track and the classifier's confidence at the instant of release, are hash-chained inside the weapon and exfiltrated on recovery or on authenticated interrogation.

By a board of inquiry, by a treaty body, by a Congressional committee, by a plaintiff. A weapon that wakes up six months after the last human touched it should be able to say exactly what it saw, who authorized it and when that authority expired. This one can, and the record is signed where it cannot be quietly edited.

CONOPS

Phase 01 · Authorize

Before the weapon goes anywhere

Two of three named humans in the delegated command chain sign an arm authorization defining the polygon, the window, the target predicate and the engagement count. The token is loaded. Nothing about the weapon's hardware changes — it is still mechanically safe.

Phase 02 · Lay

And the chart exists from this moment

By LEVIATHAN conveyor, STYX chute, submarine tube or fishing boat. Each unit settles, deploys its tethered sensor head above the sediment, self-surveys against its neighbours by acoustic ranging, and signs its own position into the ledger.

Phase 03 · Dormant

Up to 180 days at 105 milliwatts

Tier 0 harvesting, Tiers 1 to 3 gated, one 40-second command window a day, and no transmissions at all. A mine that phones home is a mine you can find.

Phase 04 · Wake and classify

Four tiers, and every step written down

Tier 0 trips, Tier 1 confirms a propeller, Tier 2 classifies, Tier 3 forms a track. Every step is hash-chained whether or not anything follows it — including the ones where the weapon decided to do nothing, which are the entries an inquiry will care about most.

Phase 05 · Decide

Two tests, and both must pass

The classifier's output is tested against the token's target predicate, and the contact's position against the engagement polygon. If either fails, the weapon records and stays safe. If both pass and engagements remain, the validator releases the interlock and the safe-and-arm rotates in-line — for this engagement only.

Phase 06 · Engage

Swim out, ascend, run the terminal

The 324 mm effector runs 6.2 nautical miles at 40 knots. The full track and the classifier's confidence at the instant of release are signed into the chain before the effector leaves the capsule.

Phase 07 · Expire and clear

Then it becomes a survey problem

At not_after minus the clock margin the interlock is revoked, the safe-and-arm rotates out of line permanently, and the unit becomes a sensor. Every expired unit runs a 900 mW acoustic responder for 90 days off a reserved cell, which turns clearance from a search into a survey.

Why it is built this way

Every design decision on this weapon was made twice: once by an engineer and once by a lawyer, and the lawyer went first. That is not how ordnance is normally developed and it is the reason DEADFALL exists while a great deal of technically excellent mine warfare sits in a magazine in Yorktown.

The mechanical interlock is the clearest example. It would have been far simpler, lighter and cheaper to inhibit the firing circuit in software and audit the software. We did not, because a software inhibit is a claim about code and a rotary out-of-line safe-and-arm is a claim about geometry, and only one of those can be demonstrated to a sceptical reviewer in a room with the weapon on the table.

What is hard, stated plainly

Open engineering risks — DEADFALL
The catastrophic failureA false positive against a neutral merchant. No architecture makes it zero. It is stated in the legal section above and it is stated again here because it deserves to be said twice.
Recovery of unexpended unitsFour hundred expired DEADFALLs on a seabed are a clearance problem even when every one is mechanically safe. The 90-day responder is the mitigation; after 90 days it is a side-scan survey and a lot of ship time.
Shelf life before immersionA three-year magazine life on a primary bus, in a weapon that must then work for 180 days, is a surveillance-lot problem we have not fully retired.
Sediment and foulingBoth degrade the acoustic aperture across the deployment. The detection ranges quoted above are day-one numbers and we say so.
In-mission token deliveryA 200 bps channel with a five-second latency and a probability of delivery well under one. If a token has to arrive after the field is laid, plan for it not to. Pre-loading is the primary path; mesh delivery is the exception.
Wartime key infrastructureThe token architecture requires a functioning signing authority. If it is destroyed, every laid field runs to expiry and disarms. Correct failure direction, real operational constraint.

Related systems

All sea systems

Government & Primes

Request a DEADFALL briefing

The arm-authorization architecture, the safe-and-arm design disclosure, and the legal-review package prepared for weapons-review counsel are available to US government customers and allied ministries of defence under an executed agreement.

Request a briefing programs@fantom.aero

Every figure on this page is a design target derived from first-principles analysis. No test article exists and no design review has been held. No operational employment is implied. No export-controlled information is published on this site.