AERIS-10 Radar Claims 20 km, but Its Own Hardware Predicts About 10.7

AERIS-10, an open-source 10.5 GHz phased-array radar posted by Moroccan engineer Nawfal Motii, has 26.3k GitHub stars, 6k forks and a claimed range of 20 kilometres. The number that matters is not the star count but the ratio between its two variants: the long-range build claims 6.7 times the reach of the short one, and the radar equation says the hardware difference should buy about half that.

Hackaday covered the project on March 12, so this is not new; the repository was back on GitHub's trending list on September 30 with 466 stars in a day. Stars are attention, not adoption, and nothing I read reports a working build by anyone else.

What the repository contains

The README describes an electronically steerable radar using pulse linear frequency modulation (PLFM), with hardware, firmware and a Python interface. Hardware is under the CERN Open Hardware Licence version 2 in its permissive variant; software and firmware are MIT. The project is marked Alpha, with several features listed as work in progress.

Variant	Range claimed	Antenna	Transmit power
AERIS-10N (Nexus)	3 km	8x16 patch array	about 1 W x 16 channels
AERIS-10X	20 km	32x16 slotted waveguide	10 W x 16 channels (GaN)

The README credits Nawfal Motii of ABAC INDUSTRY, who began the project in Morocco. It lists 10 open issues and 9 pull requests against 26.3k stars and 6k forks, a thin maintenance queue for that much attention. I did not open the forks to see whether any of them changed the design or reported a working build.

Tom's Hardware's headline calls the extended model "95% cheaper than $250,000 commercial offerings" and gives its name as AERIS-10E; the README uses 10X. I could read only the headline of that article, so the price comparison is Tom's Hardware's claim, and the ratio implies a build near $12,500.

The signal chain is an Artix-7 XC7A50T FPGA and an STM32F746 microcontroller, 16 transmit and receive channels driven by four ADAR1000 beamformers, ADF4382 synthesisers, an AD9523-1 clock generator, a GPS receiver and an inertial unit. Beam steering covers plus or minus 45 degrees electronically, with 360 degrees of mechanical rotation. The processing pipeline lists pulse compression, Doppler FFT, moving-target indication and constant-false-alarm-rate detection.

1. AERIS-10N (8x16 array, ~1 W per channel)3 km
2. AERIS-10X (32x16 array, 10 W per channel)20 km

Source: AERIS-10 README on GitHub, accessed 2026-09-30

Does 3 km become 20 km?

Start with the radar range equation, in which range to the fourth power scales with transmit power times transmit gain times receive gain. Going from 1 W to 10 W per channel is a factor of 10 in power. Going from 8x16 (128 elements) to 32x16 (512 elements) is a factor of 4 in element count, and array gain scales roughly with element count on both transmit and receive, so 4 squared is 16. Multiply: 10 x 16 = 160, and the fourth root of 160 is about 3.6. My calculation puts the extended variant at roughly 3.6 times the short variant's range, or about 10.7 km, if pulse length, integration, noise figure and target size are unchanged.

The claimed 20 km is 1.9 times higher than that. There are honest ways to close the gap: the 3 km figure may be conservative, the long variant may use longer pulses or more integration, and the target for each claim may differ. The README does not say what target, at what radar cross-section, the ranges refer to. Without that, 20 km is a design goal rather than a measurement.

Regulation and the missing demonstration

Hackaday commenters raised two problems. One ham operator said that pulsed emissions are not authorised for amateurs in the 10.0 to 10.5 GHz band, and that operation needs radiolocation licensing (Part 90 in the U.S.). Motii's reply, as Hackaday reports it, called the project an experimental and research platform and said he was working with regulatory consultants toward certification. Sixteen channels at 10 W each add up to 160 W of transmit power before antenna gain.

Others asked for footage of it tracking aircraft or drones, and at the time of the Hackaday piece none was posted. One electrical engineer wrote that without a clear demonstration he had to assume it did not work. Motii also said he used DeepSeek to draft English descriptions while the hardware and firmware are his own, which Hackaday's commenters criticised.

That is the gap between a popular repository and a validated instrument. Compare VoiceStudio's 47,000 stars: software can be tried in a minute, while the README's own notes say assembly needs PCB manufacturing expertise and RF component sourcing. Stars measure interest; the chips behind physical A.I. are funded on shipped units instead.

What would change my read

A build by someone other than the author, logging a measured detection range against a named target such as a small aircraft, would settle the 20 km question either way. So would a bill of materials: the README does not list one, and the $12,500 inferred above is arithmetic on a headline. Until then AERIS-10 is a well-documented design with an unverified range claim and an open licensing question.