Will optics ever replace copper interconnects? We asked this silicon photonics startup Science fiction is littered with fantastic visions of computing. One of the more pervasive is the idea that one day computers will run on light. After all, what’s faster than the speed of light? But it turns out Star Trek’s glowing circuit boards might be closer to reality than you think, Ayar Labs CTO Mark Wade tells The …
> After all, what’s faster than the speed of light?
It depends a lot on what material the light is traveling through.
However that is a distraction. The better question is how fast can the carrier (be it: light, RF, x-rays) be modulated. Because it is the speed of changes in whatever property of the carrier is used to transfer information , that is important.
What none of these "Star Trek" technologies consider is the physical interconnects at either end.
The hard lesson from USB, ethernet and all other pluggable, high-speed, connectors is that they are the weakest link - literally: the link. Anything that can eliminate the general crapness, cheapness, vulnerability to crud on the contact(s) and mechanical failings of these connections will get my vote.
Given how ludicrously fragile fibre optic connectors are - and that you still need power - it seems that copper connectors are going to be around for a very long time.
Optical is great for connectors you make once then don't touch for years. Less so for something you make and break several times a day, then stick in your dusty pocketses
Short interconnects can be plastic or very fine glass. I don't know what FTTH uses, but it seems no more fragile than coax or in-wall Cat 6.
Low loss Cat 5 and 6 are actually really fragile. It's a hard springy copper so the twists stay in perfect form. Sharply bend the cable a few times and it snaps. It's a lot different than patch cables made of stranded soft copper.
It all makes you wonder whether Transputers may have had a proper uptake if there had been some decent optical interface at chip level at the time. After all they were designed on the basis that serial comms can be faster and more reliable than waiting for all the parallel bits to sync correctly.
Nope, they'll still get dirty enough to drop the signal even when not touched for years. "Dirty optical jumper" is a rather well used closure code in the world of telecom. I don't know how the dirt gets into the connector to begin with, but would not be surprised if it wasn't degraded glass on the edge of the jumper, degraded by the light passing through.
I was thinking about that maybe they meant Aluminum Nitride (AlN) which does not contain any Oxygen, which has a refractive index of about 2 (light would travel at about 150,000,000 m/s through it).
Still scratching my head looking for any advantage, other than the sale of solid gold power cables to audiophile angle.
> Transparent Aluminum fibre optics is the future
(I'm going to ignore the silly "Oxygen-free" faux pas, which would not be Transparent)
Transparent Aluminum would at a guess be white Sapphire with a refractive index of about 1.77, compared to glass which is typically used in fibre optic cables with a refractive index of about 1.5. (For reference air has a refractive index 1.000293 and vacuum has a refractive index 1).
I'm curious, what advantage do you see from light travelling ~77% slower through white Sapphire (~169,000,000 m/s) compared to ~50% slower through glass (~200,000,000 m/s) ?
Also, I think I'm right in saying that because light doesn't go through fibre optic in a straight line, the end-end transmission time is actually slower than electricity, in the order of 0.67c vs 0.8c.
Fibre more than makes up for it in bandwidth, but if you want to get a small amount of data to its destination as fast as possible, wireless is actually the quickest way to do it.
Depending on whether you are talking about single-mode or multi-mode fibre. Multi-mode fibre does have the light bounce from side to side at a range of angles hence taking different length paths, which is why short pulses spread out over distance, limiting the bandwidth. Single-mode only allows a very restricted range of transmission paths - possibly not just 'straight down the middle', but not far off.
The velocity of propagation for electromagnetic fields in copper are typically 0.5c for most PCBs and about 0.67c in coax cables.
It is down the relative permittivity of the materials involved.
Strictly speaking, the permeability (magnetic part of this) should be considered but the relative permeability of copper compared to a vacuum is within parts per million.
Considering the speed we can transmit data through copper has increased year on year for decades, I think this point is key here. Copper itself hasn't massively changed, but the tech to utilise it properly has. Eg. compare USB 1 to USB 3.1. Massive increase in capability, but it is still just a copper wire.
Just thinking idly for a moment... For short range do you need any fragile connections?
I can seen what's on my screen here with no fibre optics getting in the way and, as Pete implies, using air rather than glass the light is travelling faster. A broadcast or, at best loosely aimed set of transmitters and receivers would do just as well.
At first sight this is OK if there are only a pair of end points trying to communicate in this way, otherwise there would be conflicts. But we already have the technology to deal with such conflicts - CSMA/CD. Optical Ethernet.
Indeed it probably is. But a glass light guide has other advantages. A single mode waveguide preserves polarisation* and avoids the 'signal smearing' that can occur if the light takes multiple paths due to interference by anything arbitrary in the optical path. Such smearing would lower practical bandwidth to quite a bit below theoretical maximum.
*however I heard recently that polarisation shifts have been detected in undersea fibre data links, caused by earthquakes. Apparently there's a plan to use this phenomenon as an earthquake detector alongside the primary data comms purpose of the links.
I found the last paragraph about "someday" having board level optical waveguides puzzling. I saw that done in the '70's on R&D MCM's when I was a summer intern at a large US aerospace corporation. I don't recall what the MCM's did as I wasn't working on the project but I was working in the same lab. I presume that's not common today as other ways to skin the cat won out.
given the fact that BT (or their contractors) laid Fibre to the poles in my street over a year ago, they don't have any plans to switch it on this side of 2025 (or so on BT guy told me)
That all seems like a tick box job. Lay the fibre for FTTH and get paid by the Government. Nothing about making it work...
'The Epoch
So... the answer for us users at home is 'The Epoch' or Armageddon, whichever is sooner.
I've heard talk of optical interconnects for, literally, more than a quarter-century, and they are *always* "just around the corner"; I expect they'll ship around the same time as laptop fuel cells, which have been promised for just about as long, with shipping product perpetually imminent, but never arriving.
Yep. Startups have been making noise about using optical interconnects to eliminate crosstalk in high throughput applications for decades at least.
Maybe this one will actually ship? the chiplet approach might be a big improvement since it means whatever weird process they need for making the lasers doesn't matter to the rest of the silicon that goes in the package, you put drop it in and bond a lot of very short wires to it.
* With states now held with a few electrons, could light be used to discharge the state directly? ( Like a CCD)
* Could these few electrons switch a nano liquid crystal?
* Could one have a 2D optical data plane that allows a single data source to be "seen" by many data sinks?
* Could one have a 2D optical data plane that allows a single data sink see many data sources (More difficult with channel conflict, but useful for sparse channels)?
thus directly processing without any current switching, just a power line to charge the gates.
And a whole new architecture for a computing platform
Given that (copper clad) PCB and PCA production has been optimized to within an inch of it's life I highly doubt optical on the actual board will ever be a thing. We don't have even the slightest clue about how one would go about producing such a thing outside of single unique cases with lots of time consuming custom set-up and trial-and-error. Sure it can work for board to board interconnects, but how often do you NEED something like PCIe at that level of throughput over a distance of more than a few hundred millimeters at best?
I'll have to find my notes, but the IPC (institute of printed circuits) had chip to chip optical interconnects on their emerging technologies roadmap at a symposium I attended about 15 years ago. ISTR the tech was at least to the "we believe it will be done like this" stage. I don't recall if there were any actual implementations shared at the meetings.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
