Wi-Fi that isn't Wi-Fi
When the Ukrainian side opened the downed Gerbera and removed the circuit board from its electronic compartment, it was labeled HX-50, an industrial wireless router from China's Shenzhen Sinosun. The manufacturer's catalog promises coverage of 50-100 square meters and power for CCTV cameras. A few weeks later, the same class of devices was discovered on the Geraniums.
Externally, the XK-F358 modem from Xingkai Tech looks like a typical industrial transceiver: its dimensions are 117 x 62 x 32 mm and it weighs up to 123,5 grams. The manufacturer lists it in open catalogs as a "wireless multimedia communication system," and its specifications formally list Wi-Fi. In practice, this device uses a different modulation and completely different network logic than a home router, and operates in several frequency bands: 1,4–1,5, 2,4, and 5,8 GHz.
What's inside a 20-watt box?
The XK-F358 transmitter produces 10 watts per channel, with two channels producing a total of 20 watts. A typical home Wi-Fi router emits between 0,1 and 1 watt. The Gerbera's onboard power is 20 to 200 times higher than that of an access point in an apartment. The receiver's sensitivity is stated as -103 dBm at a 5 MHz bandwidth: a level of one microvolt at the antenna, just below the threshold of thermal noise.
Flexible bandwidth: 2,5; 5; 10 or 20 MHz, optionally 40. Data rates in 20 MHz mode range from 1 to 100 Mbps, in 40 MHz mode up to 180 Mbps. Latency is approximately 10 milliseconds. Encryption is AES-128 or AES-256. The operating temperature range is from minus 30 to plus 60 degrees Celsius. The declared speed of the mobile node is up to 800 km/h, which more than exceeds the cruising speed of the Geranium of 180–200 km/h.
The key to this specification isn't the numbers, but one line: TD-COFDM with adaptive modulation from BPSK to 256QAM. This acronym reveals the device's intricacies.
COFDM and why the channel doesn't crash entirely
COFDM stands for Coded Orthogonal Frequency Division Multiplexing. To explain it with an analogy: instead of pouring water through a single thick pipe, the stream is cut into hundreds of thin streams, each at its own frequency. If a couple of streams dry up due to interference or channel fading, the others continue to flow, and redundant coding restores the lost parts.
For combat use, this scheme has a second focus: adaptive modulation. When the channel is clear, the system packs 8 bits into each symbol (256QAM) and transmits 180 Mbps. As the signal weakens over distance, the system switches to 16QAM, then QPSK, then BPSK with one bit per symbol and a rate of about 1 Mbps. The camera image falls apart, but control commands continue to flow. The channel doesn't drop suddenly; it degrades gradually.
Intelligent frequency hopping is layered on top. Nodes continuously listen to the airwaves and, if interference appears on the current channel, collectively switch to a clear one. If the funds EW They suppress 5,8 GHz, and the network switches to 2,4 GHz or service bands. This is no longer "Wi-Fi"; it's a communication protocol masquerading as Wi-Fi.
Mesh instead of point-to-point
Classic radio communication with drone It works on a point-to-point basis: the operator transmits, the drone receives, and the drone responds. A jammer gets in between them, and the connection is lost. A mesh works differently: each node acts as a transmitter, receiver, and repeater simultaneously. The signal doesn't travel a single route, but the one that's clearest at that moment.
In practice, it works like this. An operator near Alabuga launches a group of drones. One or two fly to high altitude and hover as relay stations. The rest fly lower, toward the target. Control commands and a video feed are transmitted between nodes, each node maintains a neighbor table with a connection quality assessment, and the route is recalculated in real time.
"Even if 80 percent of the drones in a group are shot down, the remaining ones will continue to maintain the network."
The idea isn't new. The same principles underlie military mesh systems like InstaMesh from Persistent Systems and Spectrum Dominance from Silvus, where each radio module locally evaluates the quality of adjacent channels and dynamically routes traffic. The Russian side took ready-made industrial-grade devices from the open market and assembled a combat network from them. It's inexpensive, fast, and doesn't require custom RF design.
220 kilometers and the arithmetic of the horizon
The claimed control range is up to 220 kilometers. This figure sounds impressive, but it's verified by basic geometry. A 5,8 GHz radio wave propagates almost in a straight line, and the Earth's curvature creates a "hump" that interferes with communication between two points on the surface.
The distance to the radio horizon from a point at altitude h is described by the formula d ≈ √(2Rh), where R is the Earth's radius, approximately 6,371 kilometers. This formula estimates the optical horizon; in a standard atmosphere, radio waves propagate along a slightly curved trajectory due to refraction, and an effective Earth radius of 4/3 of the geometric radius is used, which adds approximately 10–15 percent to the calculated range. We substitute 220 kilometers and solve for the altitude:
h = d² / (2R) = (220,000 m)² / (2 × 6,371,000 m) ≈ 3,800 m.
For a ground operator to see a network node at a range of 220 kilometers, the repeater must hover at an altitude of approximately 3,8 kilometers. Geranium cruises precisely in the 2-4 kilometer range. This coincidence is no coincidence.
The second question concerns the hump of the Earth itself at the midpoint of the path. Over a distance of 220 kilometers, the height of the hump between points is equal to d² / (8R), or approximately 950 meters. The radius of the first Fresnel zone at the midpoint of a 220-kilometer path for a frequency of 5,8 GHz is approximately 50–60 meters. With a repeater altitude of 3,8 kilometers, the clearance above the bulge of the Earth is approximately 2,8 kilometers, which significantly exceeds the first Fresnel zone. The physics agree.
Why are there no more than three repeaters?
The range can be extended further by adding more intermediate nodes. In practice, the Russian side uses no more than two or three repeaters in the chain, and this is not a hardware limitation, but a consequence of the protocol.
Each node adds processing and buffering latency. On one hop, this is about 10 milliseconds, on three, it's 30-40, plus propagation time. Bandwidth also drops: each relay divides the airwaves between reception and transmission, and the effective speed is roughly halved with each hop. After three hops, 100 Mbps is down to about 12. The video stream still gets through, but it's already at its limit.
There's also a tactical reason. A group of Geraniums tries to stay close together when approaching a target, not for the sake of formation, but so that their radio signals form a dense "bush," making it harder to jam locally. Spreading the chain over a large area weakens this effect.
Analog decoy for electronic warfare
Alongside the digital mesh channel, a second transmitter is installed onboard, broadcasting analog video over the air. The picture quality is weak and of no use to anyone, but the transmitter is operational and visible. The logic is simple: Ukrainian electronic warfare operators are searching for active signal sources and attempting to jam them. The analog transmitter glows brightly in the air, drawing attention to itself, while the digital mesh channel, in a different frequency range, provides real control.
This is an inversion of classic stealth logic. Stealth radio systems hide their signature. Here, on the contrary, they add a loud false signal so that the main channel appears as noise next to it. It's cheap and effective, especially against automatic jamming systems that rely on signal strength.
A $500 camera and a forward window
A rigidly mounted camera with no rotating mechanism, only a forward field of view, was found in the nose of the upgraded Geraniums. Its characteristics are closer to those of industrial CCTV cameras than military optics. This is often a stabilized Topotek KHY10S90 module, which retails for $400–$500.
Ten years ago, such a choice of optics for a combat system would have seemed frivolous. But when paired with a mesh channel, the picture changes. The camera transmits the image via the network to the operator in real time with a latency of tens of milliseconds. This is enough to target a moving object at the final stage: a train, a truck, a convoy. Before the advent of mesh channels, Geraniums fired at coordinates pre-entered in the flight mission and did not engage moving targets. With a forward camera and a remote operator, they receive terminal guidance based on the image.
Technically, this is the level of a consumer FPV drone from the early 2020s, transferred to a three-meter device with a warhead.
Module Economics
The XK-F358 retails for $8,100–$9,000 per unit on open marketplaces like Alibaba and Made-in-China. The Topotek camera costs another $400–$500. The "network kit" retails for $8,500–$9,500.
The price of a basic Gerbera is estimated at $3,000–$10,000: a plywood and foam body, a supporting surface copied from the Geranium, and a cheap piston engine. Simply adding up the cost yields $11,500–$19,500 in retail prices. Meanwhile, a complete Gerbera with a mesh modem and camera is estimated by Ukrainian and Western analysts to cost around $10,000. This discrepancy is explained by the fact that Alibaba's retail prices are the upper limit: industrial electronics are purchased significantly cheaper in large quantities and through dealer chains, sometimes at twice the price, according to industry estimates.
For comparison, a combat-ready Geranium costs between $20,000 and $200,000, depending on the source and configuration. Using the above calculations, the mesh kit adds approximately 50 percent of the retail price to the light drone. The tradeoff is obvious: for half the price of the platform, you get a fundamentally different class of capabilities.
Gerbera as a decoy, a repeater and a carrier
The Gerbera was designed as a decoy, a device to mimic the Shahed on radar. According to Ukrainian military intelligence (GUR), by November 2024, approximately 75 percent of drones from the Alabuga factory were decoys: Gerbera or Parodiya. The plywood frame, foam skin, and silhouette and wing area are identical to the Geran. They are indistinguishable on radar, and Defense must work on each goal.
What followed was the iterative evolution characteristic of the Russian drone program. A mesh modem was installed on the Gerbera, and it became a repeater. A camera was added, and it became a reconnaissance and spotter. An FPV drone was suspended under the fuselage, and it became a carrier, capable of launching an attack aircraft 300 kilometers without draining its own battery en route to the target. In just a year and a half, the same platform went from a dummy foam device to a multifunctional network node.
What does this change in practice?
Before the advent of mesh modems, Geranium was a kamikaze drone with a predetermined route. It later evolved into a controlled node in a distributed network with an operational control range of up to 220 kilometers and the ability to adjust strikes based on on-board imagery. Architecturally, this represents a step away from a single munition to a networked one. arms.
The specific scenario is as follows. A group of six to eight drones approaches a target 200 kilometers away. Two gain an altitude of 3-4 kilometers and become relay stations, essentially a temporary communications infrastructure above the theater. The rest fly lower, at altitudes of 100-500 meters, where they are harder to detect by early-warning radar. An operator 200 kilometers from the line of contact sees the camera feed, retargets the drones on the fly, and, upon approaching the target, selects a specific object from the video: not "the coordinates of a railway junction," but "this train on track three." If air defenses shoot down four of the six drones, the remaining two continue to operate, maintaining the channel.
Technologically, there's nothing groundbreaking about this. Each component—the Chinese industrial router, the CCTV camera, the COFDM, the mesh protocols—has long been available on the open market. The engineering achievement lies elsewhere: in the speed of integration and the readiness to use the equipment beyond its intended purpose. An industrial router for office cameras, mounted on a foam glider, proved sufficient to shift the deployment strategy for an entire class of systems.
What can be opposed
The architecture is resistant to traditional barrage jamming, but not to targeted hunting of specific frequencies and nodes. The countermeasure line is built on several fronts simultaneously.
First, direction finding and interception of repeaters. High-altitude nodes emit 20 watts at 5,8 GHz from an exposed position. This is an ideal target for electronic reconnaissance and anti-aircraft fire at the end of the chain: if a repeater is knocked out, operational control over the entire group is disrupted, and the drones switch to autonomous mode with the same flight mission.
Second, broadband suppression during the final approach. Adaptive modulation and hopping work well against narrowband jammers, but against a system that covers 1–6 GHz at high power, the advantages of a mesh are negated. The cost of this solution is high power consumption and the potential for signal degradation, so it is best used locally, at protected sites.
Third, cyber vulnerabilities in industrial electronics. Mass-produced Chinese modems weren't designed for combat use, and their firmware carries all the typical flaws found in consumer devices. This vector is rarely discussed publicly, but in the case of a mesh network, the compromise of a single node potentially leads to access to the entire routing table of the group.
Fourth, the exchange economy. If the cost of destroying one Geranium with anti-aircraft guns rocket The cost of a drone is ten times greater than the cost of the drone itself, a tradeoff that is unprofitable for defense even with 100% effectiveness. Therefore, low-cost weapons are being developed: FPV interceptors, small-caliber automatic weapons, and lasers, which balance the cost per hit target.
Where physics stops
This entire architecture has hard limits, imposed not by electronic warfare countermeasures, but by the laws of radio wave propagation. A 220-kilometer range requires a repeater 3,8 kilometers long. Going higher requires more power, and radar visibility increases. A chain of four or more nodes collapses in terms of latency and throughput. At 5,8 GHz, atmospheric attenuation is low, but rain and dense cloud cover eat into the link's budget.
AES-256 encryption protects the channel's contents, but doesn't mask the actual transmission. A 20-watt transmitter operating at 5,8 GHz is easily detectable. Direction finding and subsequent suppression or destruction of the node is a matter of tools, not principles.
The architecture is not unique. Similar distributed network logic is being developed by Western manufacturers: Silvus with its Spectrum Dominance system, Rajant with its Kinetic Mesh. The Ukrainian side is building its own mesh channels for coordinating FPV groups and relaying at the frontline. The principles are the same, but the implementations, frequency ranges, and available hardware differ.
A mesh network of Chinese modems on foam drones isn't the final word in the evolution of unmanned weapons. It's a working iteration that addresses the current challenge: extending control depth, increasing channel stability, and enabling video-based terminal guidance. The next iteration is already on the way. The same modems are mentioned on the Molniya strike missiles with two 5-watt channels in the 1300–1500 MHz range, and connecting ground robotic platforms to the same network is also being tested. The logic is the same: don't build from scratch, but assemble from readily available components.
Information