Original title: A single-precision floating-point systolic Givens-QRD Triangular Solver for MVDR Beamforming
Athi Ram R S, Alwin A, S. G. Sreejeesh, J. U. Kidav · September 2026
Ultrasound devices use a math technique called beamforming to sharpen signals and filter out noise. Calculating these settings quickly is very hard for computer chips to do in real time. To solve this, researchers built a specialized digital circuit on an FPGAField-Programmable Gate Array, a reconfigurable chip that can be customized for specific tasks, unlike a fixed processor. chip. An FPGA is a chip whose internal wiring can be reconfigured for specific math tasks. Their design uses an efficient grid of calculating units that pass data to each other, running math operations called Givens-rotation QR decompositionA mathematical method to break down a matrix into simpler parts to solve equations more easily.. When tested on a simulated 32-element ultrasound sensor, three of these parallel units ran at 100 MHz. Together, they produced 31,123 weight vectors/s. While a powerful 24-core Intel processor was about 15x faster in raw speed, the custom FPGA design was more energy-efficient. It delivered up to 2.3x the speed per watt of power, with highly accurate math matching standard computer simulations.
Imagine you're at a busy café, trying to have a conversation while many other conversations happen around you. It's hard to focus on one voice without straining or missing what's said. In this paper, an ultrasound array does something similar—it has many microphones picking up echoes from inside the body, and it needs to isolate signals from a particular point. The researchers built a chip that processes all these inputs simultaneously, using a mathematical shortcut to focus accurately while using minimal power.
This work could make medical ultrasound machines faster and more power-efficient, enabling real-time, high-quality imaging on portable devices. It affects doctors and patients by improving diagnostic tools that are easier to use in clinics with limited resources.
The abstract indicates that results are based on a simulated 32-element ultrasound array, not a real physical device. The comparison is only made against a specific Intel Xeon processor at one operating point, so broader performance claims may vary.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
A single-precision floating-point systolic Givens-QRD Triangular Solver for MVDR Beamforming — Athi Ram R S, Alwin A, S. G. Sreejeesh et al.
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.