Q&A with Tony Chan Carusone

We met Dr. Chan Carusone during his last visit at McGill, where he gave a guest lecture about the ongoing research in his lab at UofT. He and his students are designing nanoscale electronic chips for the communication of information. Some of our readers will recognize him as the co-author of the textbook Analog Integrated Circuit Design.

In your talk, you mentioned that optical communication is becoming more practical for shorter distances. What has permitted this reduction?

In the past, optical fibers were ultrafine and very delicate strands glass that required careful installation by highly trained personnel and tight mechanical tolerances. The cost of such installations could only be justified for transoceanic telecommunication or similar long-haul communication.

Advances in optics have recently allowed thicker and more bendable optical fibers to carry data at rates of 10+Gb/s. That’s fast enough to transmit an entire blue-ray disc in under 30 seconds. Using a thicker fiber relaxes tolerances everywhere in the system, making fiber optic installation easier, cheaper, and more robust.

Why is it difficult to do optical communication at very small distances?

The challenge is to make optical links economical and practical for use over very small distances. Currently, inexpensive optics are capable of communicating at data rates up to around 14 Gb/s, with research progressing towards commercial systems at 28 Gb/s. However, the optoelectronics at either end of the links (i.e. components responsible for converting the data to & from electrical signals) are very similar to those used 20 years ago when optical communication was reserved for long-haul links. The cost of these optoelectronic components are limiting the application of optical communication in areas that are cost-sensitive.

One area of research in your lab is CMOS photodetectors. Why is using CMOS an improvement?

CMOS is clearly the technology of our age. It has given us an ability to mass produce high-performance transistors at such low cost that it has transformed the world.CMOS has not only advanced computer chips. Digital image sensors were, for a long time, manufactured using CCD technology. But what really made image sensors ubiquitous was the discovery that by embedding a small circuit alongside each pixel of the sensor, CMOS image sensors can have a quality comparable to CCD sensors. Today, CMOS image sensors and digital cameras are everywhere, and creative uses for them continue to emerge.

Similarly, CMOS photodetectors with GHz bandwidth will enable a whole new set of applications for optical communication with far-reaching impact on our modern information age. Not only will they make optical communication less expensive, but more importantly they will enable optical links to be mass produced and integrated seamlessly into computing, memory, and wireless technologies using nanoscale CMOS manufacturing technologies.

In your talk, you mentioned that it is difficult to put photodetectors on CMOS. Why is that?

CMOS technology has been refined over decades to facilitate the fabrication of very high performance transistors. Unfortunately, the requirements of high performance transistors conflict with the requirements of high performance photodetectors: Tiny transistors require very thin interfaces between n-type and p-type silicon, whereas photodectors perform better when these interfaces, called depletion regions, are thick enough to absorb all photons incident on the detector.

When photodetectors are made using narrow depletion regions, many photons penetrate right through the depletion region resulting in a slow persistent current that can obscure the received data.

What has been done to circumvent these problems?

Several labs are trying to develop new manufacturing technologies that will permit the manufacture of both high performance transistors AND photodetectors. Unfortunately, those approaches imply increased cost. Our approach is improve the performance of CMOS photodetectors by making clever use of the high performance transistors already available in today’s CMOS at no additional cost. This is analogous to the advances that permitted CMOS technology to revolutionize image sensors.

What’s the difference with light travelling on a chip and when it does in optical fiber?

Light is significantly attenuated as it propagates through electronic chips, even after only a few millimeters. Light can travel along optical fiber for kilometers with little or no appreciable attenuation.

What other projects do you work on?

Another project my lab (the Integrated Systems Laboratory) is currently working on is to improve the energy-efficiency of distributed supercomputing environments by targeting the interconnections within them. The total energy per year consumed by compute servers is 220 TWh, roughly 10% of which is attributable to I/O.

Hence, even research that improves I/O energy efficiency by only 1% in these installations yields a savings equivalent to the average electricity consumption of 20,000 homes. Our research promises improvements far exceeding 1%