Introduction

Compressed air is one of the most widely used energy carriers in industrial production. It powers machinery, tools, actuators, and processes across manufacturing environments — but it is also expensive to generate. When compressed-air systems leak, energy is lost continuously, compressors run longer than necessary, and maintenance effort increases.

To help industrial users address this challenge, CS Instruments developed the LeakCam600, a handheld acoustic camera for detecting compressed-air and gas leaks with real-time acoustic imaging. The device enables maintenance teams and energy managers to locate leaks faster, even in large, noisy or difficult-to-access production environments.

Behind this simple user experience lies a demanding embedded computing task. The LeakCam600 is based on a SoC-based ultrasound system that needs to capture ultrasonic signals from a 64-channel microphone array, process them in real time, run advanced beamforming algorithms, combine the acoustic result with a live optical camera image and display the leak location smoothly on screen. This required a compact, high-performance processing architecture capable of bringing complex acoustic imaging into a handheld industrial device.

The Challenge

Rising energy costs and increasing pressure to optimize resource consumption have led many industrial companies to focus on reducing energy losses and improving the efficiency of their operations.

Compressed air is particularly relevant because it is used in almost all industrial sectors and is costly to generate due to the thermodynamic compression process. A large share of this compressed air can be lost through leaks. By finding and repairing these leaks, companies can reduce energy consumption, lower operating costs, and reduce the load on compressors. This can also help reduce maintenance effort and extend maintenance intervals. The practical challenge is especially visible in large and highly automated production sites. These facilities often have extensive pipework, restricted areas, safety fences, dangerous logistics zones or hard-to-reach installations. In such environments, it is not always possible or efficient for technicians to stand close to every pipe, connector or valve. In some cases, users need to scan from 5, 10 or even 20 meters away and document leaks for later repair when the area can be accessed safely.

This is where real-time acoustic imaging becomes valuable. Instead of manually following kilometres of pipework at close range, users can scan larger areas from a safe distance and quickly identify where leaks are located.

The LeakCam600 does not rely only on standard beamforming. It uses a more complex beamformer that requires parallel computation. The system needed to accelerate an existing algorithm with the goal of achieving around 15 frames per second, so the user could move or turn the device and see what is happening in real time.

At the same time, the LeakCam 600 has to process data from 64 digital MEMS microphones, perform preprocessing and FFT, run the beamforming calculations, exchange large amounts of data between FPGA and CPU, receive the optical camera image, overlay the acoustic result and update the display smoothly. A pure microprocessor-based approach would likely have required compromises in frame rate, responsiveness, acoustic-map quality or overall product feasibility.

The Solution

To meet these requirements, CS Instruments needed a compact FPGA-based SoC module that could combine several types of processing in one architecture: FPGA acceleration for parallel computation, CPU performance for software-based calculations and system tasks, fast memory for large data volumes, and multimedia interfaces for camera and display integration.

They selected Enclustra’s Mercury+ XU5 as the central processing platform for the LeakCam600. The decisive factors included the size of the programmable logic, the high-bandwidth connection between FPGA and CPU, the four-core CPU, large and fast RAM, and the interfaces needed to connect the 64 digital microphone channels and handle the processing pipeline.

The programmable logic was particularly important. CS Instruments used around 80 percent of the available FPGA logic, showing how demanding the beamforming implementation was. The RAM was also essential, because the microphones and the different processing steps generate and move large amounts of data continuously.

Simulation results of six ultrasound sound sources (from the strongest to the weakest source a decreasing step of 10 dB was used) at 4 m distant from the array: (a) CBF for 64 microphones; (b) CBF for 128 microphones; (c) Power Beamforming set to FBF v = 32 for 64 microphones; and (d) Power Beamforming set to FBF v = 32 for 128 microphones.

The architecture uses the strengths of different processing resources. The FPGA handles the parallel computation and real-time signal processing required by the complex beamformer. The CPU handles calculations and system tasks that are easier or more flexible to perform in software, supported by NEON functionality. The GPU is used for alpha blending, meaning the acoustic image is overlaid on the optical camera image so the user can see the leak location directly in the visual scene.

MIPI interfaces were also important because the LeakCam600 requires a high-resolution optical camera and display. This camera-display integration is essential for the user experience: the ultrasound result alone is not enough; users need to see the acoustic information placed directly on top of the live visual image. The result is low-latency visualization that allows the operator to move the device and immediately understand where the leak is located.

The System-on-Module approach and the usage of Mercury+ XU5 helped streamline development and reduce project risk during the critical prototyping phase. Rather than investing significant time and resources into a fully custom hardware design from day one, CS Instruments were able to build and validate their concept using the Enclustra Mercury+ XU5 development kit in combination with their own custom hardware. This approach enabled rapid proof-of-concept development, early architectural validation and a smoother transition to the final product design. Enclustra’s PetaLinux environment and supporting hardware resources further accelerated the process, allowing the team to focus on application development.

Conceptual illustration of the SoC, representing the data flow regarding the core part of the hand-held system.

"The Mercury+ XU5 was essential for the LeakCam600. Without its FPGA-based SoC architecture, we would not have been able to implement this level of real-time acoustic imaging in a handheld device."
Thomas Blessing, Product Manager at CS Instruments

Applications

The LeakCam600 is relevant wherever large compressed-air systems, vacuum systems or technical gases are used, especially where extensive pipework and connectors might create potential leak points.

Typical application areas include:

• Automotive manufacturing and automotive OEMs
• Food and beverage production and packaging facilities
• Pharmaceutical manufacturing and processing plants
• Electronics and semiconductor manufacturing
• Chemical plants and process industries
• Large manufacturing plants and highly automated production environments
• Light industry, including LED and lighting production
• Facilities operating technical-gas systems such as helium, argon, nitrogen, oxygen, CO₂, methane or hydrogen, depending on pressure and flow conditions.

Typical users include energy managers and maintenance technicians responsible for compressors and the supply of technical gases or process media, as well as service providers performing compressed-air audits who benefit from being able to scan large areas faster and document leaks more efficiently.

The reporting functionality is especially valuable for companies working with ISO 50001 certification. These companies are required to monitor energy consumption and take measures to improve energy efficiency. By documenting detected leaks and repair actions, the LeakCam600 helps users prove that concrete measures have been taken.

LeakCam 600, a hand-held ultrasonic acoustic imaging system for leak detection, showing different views

Conclusion

The LeakCam600 by CS Instruments is a strong example of what the Mercury+ XU5 can enable in demanding industrial edge devices: real-time signal processing of 64 microphone channels, low-latency visualization, advanced beamforming algorithms, camera-display integration, and compact product design.

For developers of acoustic cameras, SoC-based ultrasound systems, industrial inspection systems, sensor-array devices or other industrial edge devices facing similar challenges, the project shows how an FPGA-based SoC module can help turn complex processing requirements into a reliable industrial product.