The movement of information using light is nothing new. Fiber optic communications networks are commonplace in data and telecommunications systems. Advances in IC technology and fabrication are now moving optic—or more accurately, photonic—communications to the device level. These devices, capable of moving, modulating, and detecting light, integrate tens to thousands of components into a single chip, and offer very high-bandwidth performance at low power. That’s cool. But what does that really mean? In traditional integrated circuits, we use some sort of metallic substrate, usually copper, to transfer electrons through the circuit. And with Moore’s Law, which dictates basically that we’re shrinking these circuits smaller and smaller and shoving more electrons through this substrate, the most common side effect is heat. Just as an overloaded outlet gets hot, so will a chip that has too much going on in it. Heat, and therefore power, is the problem. Enter the concept of moving photons instead of electrons. (We’re looking at light as a particle, in this instance.) Using traditional CMOS fabrication processes and materials, photonic ICs (PICs) carry photons through a silica fiber instead of electrons through copper. When transferring photons instead of electrons, you can transfer at least 10X more information in the same amount of time than you can using a metallic substrate, creating 50X less heat. Paul McLellan wrote a Breakfast Bytes post about Silicon Photonics a couple of years ago, and I take some his text to explain further: … with photonics the dream is to get as much as possible of the optical interface on-chip. … This is what silicon photonics is. We build chips that convert electrical signals into light that is emitted directly into a fiber, and at the other end we have light coming down a fiber that directly is converted into an electrical signal. … these solutions allow us to leverage the huge investment in fabrication facilities and technologies for semiconductors, rather than requiring completely new factories. Using existing techniques in fabrication facilities, it is relatively simple to add a photonics component to an existing CMOS chip. Why is this post so short, you ask? Here’s the thing. When I talk to people about this topic, or when I do research online, it’s either super basic, as I have described above, or it goes rapidly over my head, to the point that I can’t explain it. (One of the best explanations came from Paul, linked above; also the Cadence Photonics page has a nice explanation.) This is complicated stuff and it requires some advanced understanding of physics and IC design. We’re standing on the shoulders of giants, here, and the technology required to Make It Go is not easy to capture in a blog post. See? Very cool.![]()
↧