Access Official Technical PDF

Share Summary:LinkedInEmail

VLSI signal processing links digital signal processing algorithms with the hardware architectures that execute them. For industrial automation engineers and system designers, the connection matters wherever embedded systems must process data under practical constraints such as timing, power use and available silicon area. Keshab K. Parhi’s career in electrical engineering offers a useful view of how algorithm design, architecture and implementation can be considered together rather than as separate stages.

Parhi is the Erwin A. Kelen Chair in Electrical Engineering and a Distinguished McKnight University Professor at the University of Minnesota. His research spans VLSI architectures and design methodologies for signal processing and communications, alongside work in areas including artificial intelligence, hardware security and computational neuroscience. Across these fields, a recurring engineering theme is the relationship between what a system needs to compute and how its hardware is organized to do so.

Why VLSI signal processing starts with architecture

A signal-processing algorithm describes a computation. A hardware architecture determines how that computation is scheduled, divided among processing units and supported by registers and data paths. A direct translation from algorithm to circuit is not always the best implementation: the system may need to process more data per unit of time, reduce power consumption or fit within a limited hardware budget.

Parhi’s research has addressed transformations such as pipelining, parallel processing, folding and retiming. These techniques change the organization or timing of computation while preserving the intended function. Pipelining, for example, divides a computation into stages so that different data items can be processed at different stages concurrently. Parallelism uses multiple operations or processing elements at once. The value of either approach depends on the algorithm, data dependencies and implementation constraints.

This makes architecture a design decision, not merely a final packaging step. Engineers evaluating a digital signal processing system need to consider throughput, latency, resource use and numerical behavior together. The architecture influences how the system performs, while the algorithm’s dependencies constrain which architectural choices are feasible.

Algorithm-architecture codesign for implementation trade-offs

Algorithm-architecture codesign treats the computation and its hardware structure as interdependent. Instead of fixing an algorithm first and asking hardware teams to implement it afterward, designers examine how changes to one side affect the other. This can expose useful options early, including alternative schedules, degrees of parallelism or ways to share hardware resources.

Parhi’s work includes high-level transformations, scheduling and allocation for DSP synthesis, as well as hardware-software codesign. These topics speak to a practical challenge for embedded systems: a design must satisfy its functional requirements while remaining implementable within available resources. Resource allocation determines which hardware units perform operations; scheduling determines when those operations occur. Together, these decisions shape the final architecture.

For a system integrator or controls engineer working with embedded processing, the broader lesson is to bring implementation constraints into the design conversation early. Requirements for response time, processing capacity, power and hardware availability should inform algorithm and platform choices. The source material does not prescribe a particular industrial control application, but the design principles are relevant to systems that rely on digital processing of signals or communications data.

Low-power design and communications hardware

Power-aware implementation is another established thread in Parhi’s research. His work includes low-power DSP approaches and architectures for communications systems, including error-control coding and decoding. Error-control methods such as turbo, low-density parity-check (LDPC) and polar codes are used to detect or correct errors in digital communication. Implementing these functions in hardware involves architectural choices that affect processing and resource requirements.

His experience also includes industry-facing work. He served as a senior principal scientist and technical director for DSP systems at Broadcom, was president of Leanics Corporation, and spent time at Medtronic. His professional record also includes technical presentations associated with IEEE 802.3 standards work on digital signal processing functions for Ethernet transceivers and optical channels.

For industrial automation, dependable communications and embedded computation are important parts of system design. However, the practical value of any particular hardware approach depends on the requirements of the application. Designers should assess the relevant signal-processing workload, communication standard, timing needs and implementation limits rather than assume one architecture fits every system.

Hardware security as part of system design

Parhi’s research portfolio also includes hardware security. The listed work covers approaches intended to prevent reverse engineering and intellectual property piracy, as well as device authentication using physical unclonable functions (PUFs). A PUF uses physical characteristics of a device to support identification or authentication. His patents include work on robust device authentication, functional obfuscation and feed-forward XOR PUFs.

Security considerations can intersect with hardware design choices: protecting a device or its intellectual property is not only a software concern. For technical decision-makers, this work highlights the range of questions that can arise when designing electronic systems, from how computations are implemented to how hardware identity and protection are addressed. Specific security measures still need to be selected according to the system’s threat model and requirements.

What engineers can take from this body of work

Parhi’s career combines academic research, teaching, industry experience, standards presentations, books and patents. His technical contributions cover areas including DSP architecture, digital communications, low-power design, error-control coding and hardware security. For automation professionals, the most transferable insight is methodological: connect system requirements to the algorithm and then to the architecture that will execute it.

  • Define implementation constraints alongside functional requirements.
  • Evaluate scheduling, pipelining and parallelism in light of data dependencies and timing needs.
  • Consider power and hardware resources as design factors, not afterthoughts.
  • Include communications and security requirements when shaping embedded hardware.

Conclusion

VLSI signal processing demonstrates why efficient embedded computation depends on more than choosing an algorithm. Architecture, resource allocation, timing, power and security all influence how a design can be realized. Parhi’s work across these areas provides a broad engineering perspective for teams developing processing and communications hardware. The practical implication is straightforward: involve hardware and system constraints early, and evaluate algorithm and architecture choices together against the needs of the application.

Access Official Technical PDF

About The Author