FPV drone work — HaShomer

Threat

Why your counter‑UAS
stack misses it

Nothing to jam

Guidance and video run down a hair-thin optical fibre unspooling behind the airframe. There is no control link to detect, jam, spoof or take over. Every RF-defeat system in the inventory is blind to it by design, not by failure.

Nothing to spoof

The route is flown by a human at the far end of the wire, watching the nose camera in real time. No waypoints, no satnav dependency, no geofence to enforce. Spoofing a receiver that isn't in the loop changes nothing.

Nothing to track

A 30 cm airframe flying at 40 m above ground level, terrain-masked behind relief, returns almost nothing. Ground-based radar sees the ridge. By the time the target clears it, the engagement window is measured in seconds.

Principle

Small things are hard.
Long things are easy.

A quadcopter at range is a handful of pixels — statistically indistinguishable from a bird, a bag, or sensor noise. Detectors that hunt for it drown in false positives.

The fibre is the opposite kind of object. It is continuous, near-perfectly straight, and long enough to cross the entire sensor frame. You are no longer classifying a blob, you are fitting a line through thousands of correlated pixels. Noise does not form lines. Birds do not form lines.

And a line has two ends. The near end is the aircraft you have to shoot down in the next few seconds. The far end is the launch team you can hand to fires at your leisure.

Sensor frame · simulated

Toggle the search mode

EO channel · 820 × 500

Swipe the frame to see all of it →

— Toggle to compare

Blob search scores every small bright region on its own. Most of what it returns is noise, and the true target scores no better than the clutter around it.

Line search accumulates evidence along a hypothesis. Thousands of weak pixels that happen to be collinear sum into one strong detection, and the two endpoints drop out of the fit.

Illustrative render — not a measured detector output.

System

Three parts.
One mast.

A helium aerostat carries the eye above the terrain that hides the threat, and the shooters ride with it. Height does two jobs at once: it clears the relief the drone is using to mask itself, and it puts the sensor at a depression angle, so the fibre is seen against ground clutter rather than lost against a bright sky.

A directed-energy mount sits behind the anchor for everything that comes close enough to burn.

Engagement

Run an interception

Set the threat range, pick an effector and commit. The engagement plays out from first contact to launch-point cue, sensor view included. Drag the stage to take the camera.

Ranges, closing speeds, effector selection and standoff are all inputs to a parameterised model — change them and the outcome changes with them. It is a model, not a measured result.

Engagement simulation · 72 s

Parameterised model

Launch

T 00.0 / 72 s · ×1 real time

Phase

Standby

Threat range

—

Aerostat at 300 m. Sensor sweeping the approach corridor north of the line.

Threat + fibre Line fit Laser Protected area

Drag to orbit

Gimbal · EOSEARCH

Effector

Detection range 7.0 km

Sequence

Result

Outcome—

Kill range—

Standoff from town—

Effector used—

Launch point—

Terrain is generic. Aircraft are drawn oversize to stay visible at scale. The approach phase runs time-compressed and engagement timings are indicative.

Design targets

Modelled.
Unbuilt.

300m

Sensor altitude on tether

8km

Fibre line detection — side-on, nominal air

<8s

Detect to defeat

4+∞

Kinetic rounds plus laser magazine

TBD

Time on station — pending envelope leak test

2.5km

Standoff of kill point from built-up area

Detection range is a range, and two of its inputs have never been measured.

Sensor altitude, detect-to-defeat, magazine and standoff are illustrative concept figures — chosen to size the system, not derived from test data or from a fielded article. Line detection range is the exception: it is the output of the detection model.

The fibre trails downwind behind the drone, so range depends on the angle you view it from and on how much the air is shimmering. Side-on, in nominal air, the model gives 8 km; viewed end-on it foreshortens to about 6 km, and a hot afternoon over sand costs a further kilometre or two. The working envelope is roughly 4 to 9 km.

Two inputs to that model have no experimental backing: how strongly the fibre stands out against the sky, and how precisely the sensor can be held steady on a tethered platform. Both need measuring before any range here is claimed as demonstrated rather than modelled.

Deployment

Masts watch along the line, not across it.

The fibre streams downwind behind the drone, so a sensor aimed straight at the target is looking up the length of the thread and sees almost none of it. Sites are therefore placed in a chain and aimed along the border at one another, which puts every crossing point side-on to somebody.

Roughly five sites at 16 km spacing hold a 60 km frontage, and the overlap means a target on a seam is held by two line fits.

One trailer, one crew

Anchor, winch, shelter, generator and the laser mount move on a single towed platform. Setup is a crew task, not an engineering task.

Fits behind your existing layer

HaShomer handles the class of threat the RF layer cannot see. It publishes tracks and launch-point cues into the same picture rather than replacing it.

The exchange ratio inverts

A laser engagement costs power. A kinetic round costs a fraction of an interceptor built for aircraft. The attacker is spending more per drone than the defender is spending per kill.

Deployment schematic

Not to scale

Swipe the schematic to see all of it →

Key

▲: Site — aerostat on tether, laser mount at anchor
◌: Coverage arc, 16 km site spacing
▬: The line — hostile launch area to the north
■: Protected built-up area, south of the arcs

Why your counter‑UASstack misses it

Small things are hard.Long things are easy.

Three parts.One mast.