One Radar on Glass, Another in a Phone: Why Each Was Missing Something Critical

Introduction

In the rapidly evolving landscape of embedded sensing, two recent projects have captured the attention of hardware engineers and AI developers alike. One places a radar module directly on a pane of glass; the other integrates a radar chip inside a smartphone. Both aim to bring contactless detection — of gestures, vital signs, or occupancy — into everyday devices. Yet, as disclosed in a detailed technical write-up on Habr, each design exhibited a distinct set of limitations that prevented it from achieving its full potential. This article examines the two approaches, the trade-offs each team faced, and the broader lessons for engineers building sensor-fusion systems.

The Glass-Mounted Radar: Transparency Meets Signal Attenuation

The first project involved embedding a frequency-modulated continuous-wave (FMCW) radar module — operating in the 60 GHz ISM band — onto a sheet of architectural glass. The intended use case was smart-room occupancy detection and touchless gesture control for interactive displays. The developers chose a 60 GHz radar because of its high range resolution (down to a few centimeters) and its ability to detect tiny movements, such as breathing or finger flicks.

Technical Implementation

The radar system consisted of a single-chip transceiver (similar to Texas Instruments IWR6843 or Infineon BGT60TR13C) with integrated antennas. The module was affixed to the glass using a low-loss adhesive, with the antenna side facing outward. The signal path therefore traveled through the glass twice: once on transmission and once on reception. The developers calibrated the system to compensate for the dielectric constant of glass (typically ~4–6 for soda-lime glass), which shortens the effective wavelength and alters the beam pattern.

Observed Limitations

Despite successful proof-of-concept tests, the glass-mounted radar exhibited three critical shortcomings:

Limitation Technical Detail Impact on Performance
Signal attenuation Glass absorbs ~3–5 dB of 60 GHz energy per 4 mm thickness Reduced maximum detection range from 5 m to ~2.5 m
Multipath reflections Internal reflections within the glass pane created ghost targets False positives in motion detection up to 30% of the time
Thermal expansion mismatch Glass and radar module PCB have different coefficients of thermal expansion Delamination risk after 500+ thermal cycles (−20°C to +60°C)

The authors noted that while the glass-integrated design was aesthetically appealing (no visible sensor), the performance degradation made it unsuitable for commercial smart-building deployments. The signal-to-noise ratio (SNR) dropped by nearly 6 dB compared to a free-air installation.

The Smartphone Radar: Compact but Power-Constrained

The second project described in the same source article involved embedding a millimeter-wave radar chip inside a smartphone chassis. The goal was to enable in-air gestures (swipe, pinch, tap) without touching the screen, as well as presence detection for power-saving features. The chosen radar operated in the 60–64 GHz band and was mounted behind the front glass, near the earpiece speaker.

Technical Implementation

The radar module measured approximately 6 × 8 mm — small enough to fit inside the phone’s top bezel. It used a phased-array antenna with 4 TX and 4 RX elements, capable of steering a narrow beam. The developers integrated the radar with the phone’s application processor via SPI interface and wrote custom firmware to process range-Doppler profiles in real time.

Observed Limitations

The smartphone radar faced a different set of constraints:

Limitation Technical Detail Impact on Performance
Power consumption Radar chip drew 150–200 mW during active sensing Reduced battery life by ~8% when always-on
Interference from display OLED panel emitted electromagnetic noise at 60–80 MHz harmonics False detection rate increased by 15% near bright screen areas
Limited field of view Beam could only scan ±30° due to phone chassis shielding Gesture detection only worked directly in front of the phone
Thermal management Continuous operation raised chip temperature by 12°C above ambient Risk of throttling after 10 minutes of continuous scanning

The developers attempted to mitigate power consumption by duty-cycling the radar (waking it only when the phone’s accelerometer detected motion), but this introduced latency of 200–400 ms, which was unacceptable for real-time gesture control. The final prototype achieved 85% gesture recognition accuracy, compared to 97% for the same radar module mounted externally.

Comparative Analysis: What Each Approach Missed

When placed side by side, the two projects highlight a fundamental tension in embedded radar design: every integration choice involves a trade-off between form factor, performance, power, and reliability.

Key Trade-Offs

Parameter Glass-Mounted Radar Smartphone Radar
Effective range 2.5 m (limited by glass loss) 1.0 m (limited by power budget)
Gesture accuracy 92% (with multipath correction) 85% (with interference filtering)
Power consumption 250 mW (wired power) 150–200 mW (battery)
Thermal stability Poor (delamination risk) Moderate (throttling risk)
Aesthetic integration Excellent (invisible) Good (behind glass)
Commercial maturity Prototype stage Prototype stage

Root Cause Analysis

The article’s authors identified that each project team had prioritized different aspects of the design without fully considering the system-level implications:

  1. Glass team focused on optical transparency and mechanical integration but underestimated signal degradation caused by the dielectric medium. They did not account for internal reflections that created coherent ghost targets.

  2. Smartphone team prioritized miniaturization and low power but failed to anticipate electromagnetic interference from the display and the thermal constraints of a sealed chassis. They also neglected the beam-shaping effects of the phone’s metal frame.

Both teams would have benefited from a more holistic approach — using simulation tools (e.g., CST Microwave Studio or HFSS) to model the entire electromagnetic environment before fabrication, and conducting early thermal and reliability testing.

Lessons for Engineers Building Sensor-Fusion Systems

These two case studies offer actionable insights for anyone integrating radar — or any high-frequency sensor — into consumer or industrial products:

  1. Model the entire signal path early. Dielectric materials, metal housings, and even paint can significantly alter antenna patterns. Use full-wave electromagnetic simulation before committing to a PCB layout.

  2. Plan for interference mitigation. Displays, processors, and Wi-Fi/BT antennas all emit noise in the 60 GHz band’s harmonics. Include band-pass filters or time-division scheduling in the system design.

  3. Budget for thermal and power trade-offs. Active radar sensing generates heat. In battery-powered devices, duty-cycling is necessary but introduces latency. Consider using a wake-on-approach algorithm with a low-power infrared sensor to trigger the radar.

  4. Validate with real-world scenarios. Lab tests with perfect line-of-sight often miss multipath and interference issues. Test with the final enclosure, under varying lighting conditions, and with multiple users.

  5. Consider hybrid sensor fusion. Neither radar alone was sufficient. Combining radar with a low-resolution camera or ultrasonic sensor could compensate for each technology’s weaknesses. For example, radar provides depth and motion, while a camera adds classification (e.g., hand vs. object).

Conclusion

The two projects — one radar on glass, another in a phone — represent innovative attempts to embed contactless sensing into everyday surfaces. Yet each design fell short because of overlooked system-level constraints: signal loss through glass, power limitations in a phone, thermal stress, and electromagnetic interference. The key takeaway is that successful sensor integration requires more than selecting the right chip; it demands a thorough understanding of the entire operating environment — electrical, mechanical, thermal, and signal-processing. By learning from these failures, future developers can build systems that are not only invisible but also reliable and performant.

For a deeper dive into the technical details and raw data from both projects, refer to the original article published on Habr: Source.

Note: This article is based on publicly available information from the cited source. No proprietary data or confidential designs were used.

← All posts

Comments