> "What I cannot create, I do not understand." > — Richard Feynman
Forget terms like "phased array," "pulse compression," and "Doppler FFT" for a moment. Let me start with a more fundamental question: what does a radar actually do?
The Essence of Radar: A Bat Using Echolocation
When a bat flies in the dark, it emits ultrasonic pulses and listens to the echoes to figure out whether obstacles are ahead, how far they are, and which direction they are moving. A radar does exactly the same thing — except it uses electromagnetic waves instead of sound.
You emit a pulse of electromagnetic energy. It hits an aircraft, a car, or a drone, and bounces back. By measuring the time between transmit and receive, you learn the target's range. By measuring the frequency shift of the return (the Doppler effect), you determine whether the target is approaching or receding. By measuring the tiny time differences at which the echo arrives at different antennas, you find the direction.
That is the whole story. Every complex term is a wrapper around three basic questions:
- Range: time × speed of light / 2
- Velocity: frequency shift
- Direction: phase differences between antennas
- Nexus: 3 km range, 8x16 patch antenna array, ~1W × 16 power
- Extended: 20 km range, 32x16 slotted waveguide antenna array, 10W × 16 GaN power amplifiers
- Hardware: full schematics, PCB layout, Gerbers, BOM
- Firmware: FPGA signal processing (VHDL/Verilog), STM32 control code
- Software: Python GUI, map integration
- Documentation: system architecture, debug guides, test reports
- Hardware under CERN-OHL-P (explicitly covers hardware patent risk)
- Software under MIT (maximum flexibility)
Phased Array: The Magic of No Moving Parts
A traditional radar is like a searchlight — the antenna physically rotates and sweeps across the sky. This is slow, mechanically clumsy, and prone to wear.
The insight behind phased array radar is simple: if you have a row of antennas, and you precisely control the time delay (phase) of each antenna's signal, you can make the beam add constructively in one direction and cancel elsewhere. Nothing moves — it is all done electronically.
Imagine a line of people clapping in unison. If each person delays their clap slightly, the sound combines into a loud boom in one direction, while canceling in every other direction. Control the delay, and you control the beam direction. This is beamforming.
AERIS-10 uses 16 antenna elements, with each element's phase controlled by an ADAR1000 phase shifter, achieving ±45° of electronic scanning. A stepper motor provides 360° mechanical rotation to cover the full airspace.
PLFM: Why "Chirp" Signals?
AERIS-10 uses PLFM (Pulse Linear Frequency Modulated) modulation. In plain language, instead of a fixed-frequency "beep," it emits a signal whose frequency slides from low to high — a chirp, like a bird call.
Why do this?
The resolution trade-off. To see far, you need a long pulse (more energy). But a long pulse has poor time resolution — you cannot distinguish two closely spaced targets because their echoes overlap.
The cleverness of the LFM chirp: every instant has a different frequency. Even if two target echoes overlap in time, their frequencies differ, and a matched filter (pulse compression) can tease them apart.
It is like two instruments playing at once — because their pitches differ, you can still tell the violin from the piano.
AERIS-10's FPGA executes this pulse compression in real time — ADC captures the echo, performs I/Q downconversion, decimation, FFT, and matched filtering to produce high-resolution target information.
Two Versions: Why Nexus and Extended
AERIS-10 ships in two versions:
This split is practical. The cost of a phased array radar is dominated by two things: the antenna array and the power amplifiers.
Nexus uses patch antennas — cheap, easy to fabricate, suitable for short-range applications like drone avoidance and local surveillance. Extended uses slotted waveguide antennas — more complex, more efficient, and combined with GaN PAs they reach 160W total output, ideal for long-range search.
Splitting the two versions lets users pick the right tool for their scenario without paying for capabilities they do not need.
The Significance of a Full-Stack Open Release
What impresses me most about this project is not any single technical spec — it is its completeness:
The licenses are also carefully chosen:
16.4k stars on GitHub signal one thing: many people wanted to play with radar, but there was no door before.
Signal Processing Pipeline: What Happens Inside the FPGA
Once an echo enters the system, the FPGA (XC7A50T) executes this pipeline:
1. ADC capture: digitize the raw RF signal 2. I/Q downconversion: shift the RF signal to baseband, separating real and imaginary components 3. Decimation: lower the data rate while preserving useful bandwidth 4. CIC/FIR filtering: remove out-of-band noise 5. Pulse compression: matched filter to disentangle overlapping echoes 6. Doppler FFT: extract velocity information 7. MTI (Moving Target Indication): filter out stationary background (ground, buildings) 8. CFAR (Constant False Alarm Rate): adaptive thresholding — only report real targets
This pipeline runs in real time on the FPGA — not in post-processing, but within microseconds of pulse transmission.
The STM32 handles housekeeping: power sequencing, GPS/IMU interface, PA temperature monitoring, stepper motor control. The FPGA focuses on signal processing; the STM32 focuses on system management — clean separation of concerns.
Limitations and Blind Spots
1. Alpha stage
The README explicitly marks Status: Alpha. Some features are still under development. This is not a criticism — that is how open-source projects work. But if you expect "buy it and it just works," you will be disappointed.
2. Assembly barrier
The project requires "PCB assembly experience." This is not an Arduino — there is 10.5 GHz RF circuitry, PA thermal management, and FPGA bring-up. Even with all the files open, not many people can successfully bring it up.
3. Regulatory issues
10.5 GHz at 160W total output requires a radio transmission license in most countries. The project documentation does not appear to discuss spectrum compliance (FCC Part 15, ETSI, etc.). Open-source hardware being "usable" is not the same as being "legal to use."
4. Cost is still not low
Even though it is marketed as "low-cost," 16 ADTR1107 front-end chips, 4 ADAR1000 phase shifters, an ADF4382 frequency synthesizer, and an XC7A50T FPGA are not cheap parts. A full BOM could run into the low thousands of dollars. "Low-cost" is relative to military-grade radars, which cost millions.
5. Competing ecosystems
The project overlaps with Analog Devices reference designs (such as ADAR1000 evaluation boards) and SDR community projects (such as USRP + GNU Radio). AERIS-10's advantage is integration — a complete, full-stack system — but flexibility may be lower than a pure SDR setup.
A Feynman-Style Summary: What Is Actually Understood
AERIS-10's core insight is small but fundamental:
The high barrier to phased array radar is not that the principles are complex — it is that no complete, reproducible, end-to-end engineering implementation was previously available.
The principles are in textbooks. The chips are purchasable. But stringing together clock distribution, phase calibration, pulse compression, thermal management, PA bias closed-loop control, and GPS attitude correction — all in one open, reproducible package — that had not been done before.
As Feynman said: "What I cannot create, I do not understand." This project makes it possible to *create* a radar system, and understanding follows.
As for terms like PLFM, CFAR, and MTI — they are just names. Knowing a chip is called ADAR1000 is not the same as understanding how phase control steers a beam. AERIS-10's value lies in giving you the chance to genuinely understand the latter.
---
Project: AERIS-10 / PLFM_RADAR (https://github.com/NawfalMotii79/PLFM_RADAR)