Real-World High-Speed FPGA Use Cases: Melt Pool Monitoring, Laser Tracking, and 3D Metrology
Several types of laser-based applications in machine vision illustrate why FPGA-level processing speed matters in practice, and why fast enough in these applications usually means responding in the millisecond or sub-millisecond range rather than waiting tens of milliseconds.
Melt Pool Monitoring
In additive manufacturing, a laser progressively melts and fuses layers of metal powder to build a part. Monitoring the resulting melt pool as it forms, and adjusting laser power in response, functions as a closed control loop rather than a one-time measurement: an undersized pool signals incomplete melting, while an oversized one can spread beyond the intended boundary. Because the system must process the data as it arrives and tune laser power on the fly, the imaging and processing have to keep pace with the build in real time.
Laser Tracking and 3D Metrology
A related requirement appears in laser tracking systems used for defense and scientific positioning applications, where an object's location has to be measured and corrected on a continuous basis, and in 3D metrology, where a projected laser line's displacement across a surface is used to calculate height at each point along a scan. Since every captured frame contributes only a single line to the finished 3D profile, the imaging system has to operate at high speed to assemble a complete and accurate scan. In each of these cases, the compute must be done before the next frame arrives; anything slower and the result trails the process it is supposed to steer or measure.
How These Projects Get Built
Projects like these usually start with either a working model (Python, MATLAB, whatever the customer has already prototyped) or just a description of the problem. Concurrent EDA models the algorithm in software first, then moves it into the FPGA inside the camera. Cameras with capable FPGAs can run that compute in real time, and the customer doesn't need FPGA expertise of their own. What they get back is a custom camera, whether it is something like a GigaSens camera from Concurrent EDA, or their own.
Working on an application with similar real-time demands? Contact Concurrent EDA to talk through your idea. Watch Ray Hoare discuss this and much more on FPGAs and real-time processing here.
Email