MSNBC on Infinera's Optical Chip

pnoti writes: "This article at MSNBC is a loose overview of Infinera's new chip with circuits that control the flow of light instead of the flow of electrons. 'If this chip performed as they hoped, it would shatter many of the theoretical limits regarding the behavior of light in optical communications networks.'" Update: 04/10 04:26 GMT by T : That's MSNBC, not The New York Times -- oops.

Comments

Makes you wonder

[cscx]

Is such an implementation of ICs reliable, when compared to the trusted-and-true silicon ICs of today? I mean, I understand the huge quest for faster and faster chips, but I hardly see this making an impact into the IC industry. Maybe 10-20 years down the road, but not now. Plus, the manufacturing has to be a lot more expensive.

Re:Makes you wonder

[Soko]

From the article:

Hardly. Infinera's thumbnail-size chip is the first integrated photonic circuit. Though Infinera won't reveal the chip's cost, when built with manufacturing techniques used by chip makers like Intel and Advanced Micro Devices, it likely could be made very cheaply. The savings in manufacturing in turn would lower the cost of network equipment by half, perhaps even more. Beleaguered network carriers like Level 3 Communications and bankrupt Global Crossing could build networks for much less, and run them more efficiently and at a lower cost--maybe even profitably. For consumers, Infinera's chip could be instrumental in allowing communications companies to offer high-speed Internet access at affordable prices. And one day this technological breakthrough could lead to a device capable of projecting a holographic display, as on the TV series Star Trek.

Pretty much refutes your points.

The best thing about photonics is the absence of (photon) migration, which is a big problem with small trace size electronics (electron migration). (Aside: If a silicon engineer knows better, please correct me.) No migration happens because photons have 0 rest mass, and therefore don't have intertia. This means they are a lot less likely to over shoot the switching mechanism, and maintian signal. This is in addition to thier electrical interference resistance.

Commercial products may take a while to come to fruition since there will have to be some major re-tooling at the fabs, but with so many huge benefits, it'll come sooner than you think.

Now, where to put that Holodeck....

Soko

Re:Makes you wonder

[J. J. Ramsey]

I'm not so sure that photons not having mass means that they have no inertia. Photons have momentum, can exert a (very) slight pressure, and can be pulled on by gravity. Given that, their inertia is likely to me small, but nonzero.

Re:Makes you wonder

[harvardian]

Being pulled-on by gravity doesn't mean anything. Gravity (according to Einstein) is the warping of space-time, so things that are massless still experience it. According to Newton's equations, that wasn't the case.

Re:Makes you wonder

[gTsiros]

Photons don't have inertia ?!?

Where did you come from?!

Photons DO have inertia!
Inertia has nothing to do with rest mass, but it has with mass.

photons have mass

they have inertia.

they are particles after all (this is the basis of QM. particles can be seen as waves and vice versa)

Re:Makes you wonder

[cornjones]

gravitons (the particles we hypothesis are responsible for gravity but have not been able to observe) are generated in proportion to the mass of an object. the object they are working on doesn't (i believe) need to have mass

But gravity is such a weak force that the mutual attraction (ie both objects be attracted to each other) is often necessary for any significant effect.

That being said, light can be/is observably effected by large gravity producing bodies (stars, etc...) and would stand to reason that there is some effect by smaller gravitational pulls, even if that effect is, as yet, unobservable to us.

Re:Makes you wonder

[sluke]

You state:

The best thing about photonics is the absence of (photon) migration, which is a big problem with small trace size electronics (electron migration). (Aside: If a silicon engineer knows better, please correct me.) No migration happens because photons have 0 rest mass, and therefore don't have intertia. This means they are a lot less likely to over shoot the switching mechanism, and maintian signal. This is in addition to thier electrical interference resistance.

Actually photons can still "overshoot" switching devices etc. Photonic Band Gap materials have all of their properties because of they way they exploit the quantum nature of the light. So, just like semiconductors, there is a finite probability that the light will tunnel to another place. The reason that I don't think this will be a problem any time soon is that for the wavelengths of light people are interested in making "photonic circuits" with, the features of the material are so small and at this point difficult to manufacture, that the tunneling barriers are effectively infinite. Once methods are developed for working effectively with arbitrary 3-D fabrication at the precision we are accustomed to in silicon, photon migration will likely become a problem.

Re:Makes you wonder

[QuantumFTL]

No migration happens because photons have 0 rest mass, and therefore don't have intertia.

Ahh, yes... a very common mistake by the non-physicist. It seems to make sense that something that has zero rest mass cannot possibly have inertia. This is, however, completely wrong. The problem is that massless particals travel at the speed of light, which is where some interesting things happen in the equations. We start out with the general equation:
E^2 = m^2*c^4+p^2*c^2

Substituting 0 for m, we can solve for p = E/c. It's well known that photons carry energy, and thus they must carry momentum. (There are other methods of deriving this, however I will not get into them... pretty much all waves carry momentum, one way or another).

How else would projects like the Astronomical Society's Solar Sail function?

As for this being the primary reason that optronics are better than electronics, I'm not entirely sure... definately massless particals are in general better for things like this (where you want maximum information carried for a minimum amount of energy, in a minimum amount of time). Photons typically propagate faster than electronic signals, and optical circuits usually have a much higher bandwitdth due to frequency-level multiplexing.

Also, it is possible to use physical properties of photons to compute fast fourier transforms, which are especially important for digital signal processing. Not to mention the amazingly fast access times of ultra-huge holographic databases.

Dislaimer: I'm not a physicist, but I'm studying to become one.

-Justin

Re:Makes you wonder

[jo2nathan]

Electron migration is only a problem because it knocks (migrates) metal atoms out of a section of wire when you have high current densities. Basically it leads to an break in the connection. Light can travel through 'empty' space. If it one photon knocks something out of the way the next one can still come through. The thing is you can't use the same manufacturing processes to create a photonic chip. Plain/doped silicon (that intel and AMD use) isn't a practical way to deal with light. Eventually the processing for something like this could get to the point where it is as inexpensive but only after Billions of dollars of investment in processing techniques. And then the fabrication facilities will still cost billions. This kind of thing has great potential but nothing that is going to happen overnight, or even in a year or two. Probably not in the next 3 years, not unless Intel/IBM/ (your choice of multi-Billion dollar corporation) invests majorly in it.

Re:Makes you wonder

[n9hmg]

That's not what this one is for. This is about the LSI of an optical communications set, whether a multiplexer, a booster (can't remember what they call them... the repeater thing), whatever.... all these things that involve lots of big discretes under control of asics, now can be built using lithographic processes, instead of robots, soldering irons, and people. Righ now, if you lose one of those suckers, you fix it. with this breakthrough, you replace it.
It's like in the old days, when you'd have a hard drive fail, a tech would come out with an oscilloscope and debug it... time lost, labor bought. Now, if you have a hard drive fail, you tell the software it's offline, flip the lever and pull out the old drive, push in the new drive and flip the lever, tell the software to set up the drive and put it back in the pool.

Re:Makes you wonder

[spike hay]

Of course photons have inertia. Haven't you heard of light sails? Those are pushed by sunlight.

What about beer?

[VirexEye]

Yes but when will they have microchips than can control the flow of beer?

Re:What about beer?

[Nilatir]

He who controls the beer controls the universe...

Re:What about beer?

[lambsonic]

Yes but when will they have microchips than can control the flow of beer?

You don't need any particular chip, just the right protocol.

Red Herring

[The Gardener]

Yeah, Red Herring carried the story, and with a little lower "fluff factor". At least, it seemed to me . . .

The Gardener

Re:Red Herring

[JFMulder]

Yeah, this story is the same as on MSNBC, except for the layout is less good.

Re:Red Herring

[zeno_2]

Yeah, this story is the same as on MSNBC, except for the layout is less good.

less good?

Re:Red Herring

[tswinzig]

Yeah, Red Herring [redherring.com] carried the story, and with a little lower "fluff factor". At least, it seemed to me . . .

That's strange, since the byline on the MSNBC article says they got it from Red Herring. I guess the MS in MSNBC has their editors used to trimming articles for the lowest common denominator?

Re:Red Herring

[bryan1945]

Yeah, this story is the same as on MSNBC, except for the layout is less good.

less good?

Would you prefer "more bad"?

Re:Red Herring

[zeno_2]

"Yeah, this story is the same as on MSNBC, except for the layout is less good."

How about..

The layout is not as good =)

Re:Red Herring

[BigBir3d]

most writer's freelance. write a good story for Red Herring, MSNBC says "hey, we want that too" and the writer thinks "ka-ching!"

Re:Red Herring

[rot26]

You mean "worser".

Re:Red Herring

[JFMulder]

How about, I'm not english, so I might suck sometime at grammer? :-)
No offense taken, anyway, thanks for the tip.

Re:Red Herring

[zeno_2]

Sorry I wasn't really trying to offend you in anyway, none taken so no biggie. And don't worry, people that have spoken english thier whole life still say that.

When I posted the reply, I was REAL bored at work, and it caught my eye =).

Cheers!

Re:Red Herring

[JFMulder]

I bet you we're real bored at work again to reply to this also. So am I right now :-)))

Read before you whore for karma

[Neil Blender]

The story is on msnbc. Not really sure where the NYT comes in.

Photonic bandgap technology

[Vireo]

Well, they are certainly not the firsts to make photonic chips. Optical mux/demux (cascaded couplers) are routinely built as planar waveguides on semiconductor materials. However, the size of their chip seems really small, which suggests that they use photonic bandgap technology, which uses very small arrays of refraction index changes in which light at certain wavelenght can't propagate to make it perform tricks, like turning at 90 degrees on very small distance. However, I didn't saw any mention of this in the article. Anyone can confirm it is the case?

Re:Photonic bandgap technology

[Anonymous Coward]

Your are missing the point.
Of course, optical components have been out for years. But no body was able to create an optic-based IC. That is, until Infinera.

Re:Photonic bandgap technology

[Anonymous Coward]

Photonic IC in InP can use low-loss curved optical waveguides with bending radius down to 30 microns. There are no commercial photonic IC's at the moment that are based on photonic bandgap technology.

For an example of what's possible in today's InP technology have a look at these circuits from Delft University and ThreeFive photonics in the Netherlands. They show a photograph of the bare chip containing a 4-wavelength optical crossconnect on 1.5 mm by 3 mm! This is without use of photonic bandgap structures (which could in principle reduce the size even further).

Re:Login for New York Times

[Rebel Patriot]

Ha! I just saw New York Times and karma whored it. Boy do I feel stupid. :^)

Cheaper, maybe, cheap? No.

[bertok]

Slashdot is not the only publication with bad editors! To quote the article: "Though Infinera won't reveal the chip's cost, when built with manufacturing techniques used by chip makers like Intel and Advanced Micro Devices, it likely could be made very cheaply."

Well, not quite. You see, the article later mentions that Infinera used InP (Indium Phosphide) chips instead of silicon, probably because they needed it's superior electrical and optical properties. With InP, it's possible to make 100 GHz circuits, but not cheaply. Certainly not for the same cost as a modern, silicon CPU.

Re:Cheaper, maybe, cheap? No.

[duct_tape_n_wd40]

Aw nuts, a Slashdot article that I can ramble on about ad nauseum shows up (I work in a related field, none of that sissy network administration stuff for me), and I have to catch a plane this afternoon...

Sure this sounds neat, and I'm not trying to knock anybody, but a few quick points -

i) as others have mentioned, this is a really nice press release disguised as a magazine article. They made some devices - congratulations!

ii) Fab - it sounds like they're doing this on bits of wafers in beakers on a wetbench, with presumably less than state-of-the-art litho. Great for proof of concept, but keep in mind that larger, better capitalized and more experienced outfits have trouble moving small-scale hero devices into assembly-line style production mode. It constantly amazes me how much compound semiconductor processing is still done by a combination of black magic and luck by a few process engineers with "golden hands".

iii) Related note: [these things will be cheap] when built with manufacturing techniques used by chip makers like Intel - try buying modern process equipment that will handle 2" and 3" InP wafers. I dare ya. All the modern interesting tools are built to handle acres of dinner-plate sized Si wafers, and can't cope with the teeny-tiny ones (hey, the Si market is about 100x bigger, can't blame the equipment manufacturers).

On the other hand, maybe I should take a look through Welch's publications in my ample free time...

Ouch...

[Duncan3]

"Some of Infinera's 700 or so competitors"...

Even if they do it, the cost competition is gonna make sure they never make any money....

Questions???

[Anonymous Coward]

Well suppose one day everything that is electronic today becomes optical.

What will our test instruments look like? What will be the units of measure?
How is work done in an optical device? Will we have 'fiber buss bars' a la Outland that carry 'DC light' everywhere?
Will we have to break open circuits to measure things a la current probe?

Will there be optical equivalents of everything electronic or will the optical stuff be a specialized peripheral of electronic devices?

Re:Questions???

[BakaMark]

What will our test instruments look like?

At least there is one current example of this today. There are devices that are used to tap optical fibre lines, that work by actually splicing into the line.

These devices have been around for a number of years now, and I have heard of one such device being able to tap an optical fibre bundle that has 50 individual optical fibres within it. Of course it will leave the optical pathways semi-intact, and detection is only by using complex test gear on either end that will tell you the consistency of the fibre as well as the points where the joins have been made. These things are usefull if you want to wiretap an optical fibre cable.

Of course removing such a device from the optical fibre bundle will effectively break the connection.

Re:Questions???

[benhaha]

If you squeeze optical fibres the leak light, so no problem there.

For chips, I would think you'd have to design in ports specifically for monitoring, as is done for regular chips anyway.

Is this really a photonic switch?

[Animats]

From the description, this is a switch or router on a chip with optical in/out, but the usual transistor processing. The innovation is that the lasers and receivers are on the same chip as the switch.

There are true optical switches (from Nortel, for example), although they're circuit switches for backbone links. An optical IP router is a ways off.

Light

[Anonymous Coward]

Couldn't this theoretically be thrown off by light nearby? Like internal LEDS for the power switch?

Re:Light

[BigBir3d]

wrong wavelength. any visible light would be of a wavelength that is way to huge to fit. besides, it will be a closed circuit, just like any electrical circuit should be.

change of address

[tux-sucks]

Indium Phosphide Valley, anyone?

Straight-thru

[lostchicken]

For this to be any good, the signal path must be pure optics, e.g. the same photon must go all the way thru the switch, and just be routed around. That means the switch would have to understand the rays of light out my 100baseFX network, or fibrechannel bus, and deal with it in photonic form.

This solves EMF issues, and other nasties. Electronics could be used for low speed control, and indicators, but fibre be used for ALL high speed stuff, including PCB traces and everything else.

Anybody developed optical solder yet... ;-)

Where's the beef?

[apk]

I read the article -- verbatim -- in Red Herring's printed rag. There's no meat to that article. What exactly is it that this thumb-sized chip does, and how/where will this device be used (to reduce cost, or increase functionality, or increase circuit density per rack, or...) in the optical systems being deployed by the optical carriers?

Does this chip offer SONET layer switching (or muxing/demuxing)? SONET layer Performance Monitoring? Does it bring anything to the DWDM playing field, in either the long-haul or metro arenas?

Optical carriers buy optical transmission and switching systems, not components, with accompanying network management platforms to operate, integrate, and manage it.

I ask again, where's the beef? As it is, this is just a glorified press release.

Andy

Re:Where's the beef?

[ckedge]

Absolutely.

This article is the equivalent of Bell Labs EXECUTIVES and CEO's claiming that they were in the process of single handedly pulling the transistor out of their a**es, before the transistor had even been created yet.

It ignores the 20-30 years of physics and engineering physics that came before it, it ignores the thousands of people and hundreds of groups who have been working at the dozens of different approaches to this EXACT problem, and it ignores the engineers who actually came up with the designs for the devices they are intending to use, and the related background between all of these.

I should know. I spent four years doing a degree on one possible approach to creating the exact components they claim they are working on. I worked with InGaAs/InGaAsP/InP Quantum Well structures and one possible method of creating a fundamental process to modify such a structure into the types of devices they are thinking about. We were thinking ahead to the exact thing that they are thinking of.

And we ourselves were basing our work on 10-15 years of other people's work. The first people who came up with the possibility of using non-silicon semiconductors was 3+ decades ago, and of creating fully integrated InP/etc based all optical ciruits is about 20-30 years old.

Re:Where's the beef?

[cornjones]

this may be nitpicky but when you say "we ourselves were basing our work on 10-15 years of other people's work" I think of two things.
1. DUH!!!
2. You are basing it on a lot more than 10-15 years of work. as the simplest example, did you come up w/ the mathematics that you were using. no? oh thanks newton, that calculus sure did help (to say nothing of the fancy things we have built on top of it). what about feynman and QED? not that I know about your work specifically but any scientific observation really owes itself to the last 4000 years (thanks euclid, aristotle, etc) of human scientific development.

Beer Chips? Old News

[Tokerat]

Yes but when will they have microchips than can control the flow of beer?

...This one keeps the flow going pretty steadily. :-D

skip it...

[doooras]

what we need are isolinear chips and optronic relays

maybe throw in a few bioneural gelpacks

Re:skip it...

[cybercomm]

LOL... i always wondered what a mixture od Star Trek TNG+Voyager would be like...though i presume gelpacks would be considerably faster than the IC chips =D

Oh well those things are far far away into the future, but there is one thing that isnt (or at least the article claims it is not) and that ia a hilodeck... /me starts planning holo-emmiter infrastructure....

Re:Makes you wonder theWhy.com

[theWhy]

it will become reliable with time. IC industry folks need to keep all options open.

zero REST mass, in theory, and YMMV

[StandardDeviant]

But while the current theory holds that mass is invariant, the particle's _energy_ (which, when you think about it, is what you're worried about anyway) is most certainly not invariant. Since these little fellows are zipping along at a literally astronomic clip, even the "massless" photon has energy. IIRC, experiemental data held the mass of a photon as being something like 3.9x10^-(12?15?) that of the already quite svelte electron...
Electric fields generate magnetic fields, and both can in theory interfere with the propagation of electromagnetic waves, which are the other side of the photon coin (really, at that level, what is a wave? what is a particle? they're two ways of looking at the same thing. actually this is valid all size levels, but the wave/particle duality effects for anything larger than an angel's behind is vanishing, incredibly, stunningly small)
Besides, I was under the impression that quantum tunneling was the origin of some of the migration patterns in (or should I say through?) circuits. The lighter a particle is, the more prone it is to this "now I'm here, FOOLED YA! now I'm there" behavior... I'm too lazy to go dig up my pchem text, i'm sure somebody will follow up with a more precise explanation and some of the relevant equations. (I'm not a particle physicist, just a chemist, but we do rub elbows occasionally, much like every now and again a molecular biologist will talk to the chemists next door ;-))
This is not to say this isn't a cool advance. It's just that I'm even more curious now as to how they got the magic chip to work, given what I imagine the physical and technical hurdles were...

Re:zero REST mass, in theory, and YMMV

[rainwalker]

Ah, glad to see a fellow P Chem student here! I have my book handy, and so I figured I'd make a stab at posting that more precise explanation you asked for.

Tunneling has to do with the energy of the particle. In the one-dimensional particle in the box theory, the transmission coefficient T is given by an equation that I cannot reproduce due to the ascii art issues, but is a roughly second-order polynomial curve when plotted as T versus E/V0, where E is the energy of the particle and V0 is the "height" of the potential barrier, moving from T=0 at E/V0=0 to T=~1 at E/V0=~2. This would mean that the higher energy of the particle, the higher of a potential barrier needed to prevent tunneling. Now, free electrons have higher energy than photons, so it would be harder to contain them, as it were. Plus, this isn't the only issue. Photons have no charge, and so are undisturbed by electrical and magnetic fields, but electrons of course are negatively charged, which complicates things in 2 main ways, first they do not travel in a straight line (like photons), and having 2 narrowly separated regions with different charges (such as in a transistor) will lower the potential barrier. I believe this is why SOI (silicon-on-insulator) is useful, as it makes it more difficult for the electrons to migrate from region to region.

I hope this helps some, and if I am wrong, feel free to correct me!

There is some T&M stuff

[TalShiar00]

There are plenty of test and measurement equipment available for the optical market. Look around Agilent's T&M lightwave page.

Anybody developed optical solder yet...

[TalShiar00]

Unfortunately no. But that woudl be cool to breadboard a optical circut. Unfortunately we have to resort to special cutting and polishing tools just to connect a few components together.

How to piss a optical network admin off: go to the long haul switch and yank a few wires out, they will be there for hours redoing all the lines.

Buckets o' beer

[Carnivorous Carrot]

> Yes but when will they have microchips than can control the flow of beer?

I don't know about that, but AI thought experiments include a device that mimics a human brain made via buckets of water, poured. I suppose buckets of beer could work just as well keeping the brain moving along at a steady clip.

Re:Postus With the Mostus

[Carnivorous Carrot]

A fairly large, but cheap, Fresnel lens can reach a thousand degrees or more. Try that on Peta.

SSSCA workaround?

[autopr0n]

Hrm, If the SSSCA passes as is, it will dissallow "electonic digital" devices from being used without copy protection. But it dosn't say anything about optical digital devices :P

Re:If you keep clicking next enough times,

[stubear]

I think you'll find you're wrong. Open both side by side and scroll through the MSNBC article. You'll find they have the exact same sections which are exactly the same length and both end on the "Absolutely Fab" section. The ONLY difference I found in content was in the subtitle. The Red Herring article contained a semicolon followed by a qualitative comment about everything prior. The MSNBC article only contained the subtitle leading up to but not including the semicolon and qualitative comment.

Re:I think I know how it works

[katcoker]

Why are things so heavy in the future? Is there a problem with the earth's gravitational pull?