Converting Light into Sound

prostoalex writes "Researchers at Duke are trying to solve the problem of speeding up fiber-optic connections by converting light into sound, then converting it back into light. From the Nature News article: 'To get the information from the acoustic wave out again, a third light pulse, the 'read' pulse, is sent in. When it reaches the part of the fibre being affected by the acoustic wave, the light scatters in such a way as to regain the information that was left behind by the initial pulse. The newly-formed data pulse leaves the fibre, resuming the journey in the same direction as the original pulse, taking the same information with it.'"

I'm confused

[flar2 (938689)]

Doesn't light travel faster than sound?

Re:

[by Anonymous Coward]

Yeah, except on Monday.

Re:I'm confused

[the_humeister (922869)]

It does, but they're using the interference from 2 light beams to "store" a sound wave that is still within the cable. Then they use a third beam to retrieve that info stored in the cable.

As the light pulse leaves the fibre, the acoustic wave with all its inherited information lags behind: the speed of light in the fibre, at some 200 million metres per second, far exceeds the more sluggish 5,000 metres per second of sound. "The acoustic wave is essentially stationary over the duration of our experiment," says Gauthier, whose work is published in Science

In the tests done so far, a 2-nanosecond pulse could be held in the fibre for up to 12 nanoseconds

Re:

[flar2 (938689)]

Thank you, that makes sense

Re:I'm confused

[gumpish (682245)]

It does, but they're using the interference from 2 light beams to "store" a sound wave that is still within the cable. Then they use a third beam to retrieve that info stored in the cable.

lol wut

Re:I'm confused

[tomz16 (992375)]

After reading the summary of the article linked here, it seems as if this would be used as a buffer to temporarily store information. Currently, when you need to hold onto some information for a little while, you use electronics.

light -> electronics -> light

This would allow for

light -> sound -> light

which offers a few advantages. Namely, power and cost come to mind.

Re:

[by Anonymous Coward]

And buffering information in a line is faster than letting it travel the whole line without buffering how? Seems to me that this signal was delayed 12ns, or 6-fold. I wouldn't call increasing transmit time six-fold an "speed up."

Re:I'm confused

[TheRaven64 (641858)]

Multiplexing. Most signals do not travel over a single fibre, they travel over a network of connected fibres. If two messages arrive at once to go down one saturated upstream pipe (tube) then you have to delay one of them. You want to delay it just long enough for the other to be transmitted, no longer. Turning it into an electrical signal and then back again is time-consuming. This conversion time adds to the total latency. Being able to delay a fibre message for a shorter period enables (in theory) lower total latency.

Re:

[cynicsreport (1125235)]

Delaying an optical signal is as simple as having a specific length of a particular type of fiber. A signal can be delayed with respect to a different signal fairly easily by adding the extra fiber. In cases in which a high degree of precision is required, the signal can be sent through free space, and the distance can be adjusted with a micrometer. Alternatively, the fiber can be wrapped around a piezoelectric drum, which responds to a voltage my changing its diameter. The latter method is an effective way

Re:

[AndrewHowe (60826)]

Forgive me, I don't know this area well, but aren't all those methods static rather than dynamic delays? The piezo drum thing is presumably intended to work at very low frequency so we can consider it static.
Whereas the subject of the article is dynamic, so that the receiver can delay or not, choosing at a high frequency. It's as if an extra 2.5m of fiber were inserted, but how would you practically achieve that?

Re:

[smallfries (601545)]

Splicing and adding duct tape... really quickly

Re:I'm confused

[AndrewHowe (60826)]

I had considered the homunculus approach. But the little guy would need a lot of tape, and a constant supply of good amphetamines.
Unfortunately Homunculi are in short supply these days as processor manufacturers need them to sort incoming instructions into superscalar execution units. That used to be my job but I guess it got outsourced...

Re:I'm confused

[quanticle (843097)]

Doesn't light travel faster than sound?

The problem here is that light travels too fast. Fiber optics can carry so much information so quickly that its impossible for the computer on the receiving end to transcribe it into bits quickly enough. This technique turns light into sound first, an intermediate step that speeds up the digitization process. This is more about removing the computer as a bottleneck than speeding up fiber optic communication itself.

Re:I'm confused

[Sen.NullProcPntr (855073)]

This technique turns light into sound first, an intermediate step that speeds up the digitization process.

While this is implied in TFA it's a little misleading. As others have pointed out the light -> sound conversion is being used more as a delay line [wikipedia.org] possibly as a method to synchronize a packet into an active data stream. This is normally done in a router as a conversion from optical (light) to electrical. Then the data is held in memory until needed and sent back out as optical aka "store and forward" [wikipedia.org].

What is interesting is how the interaction between the data pules and the "write" pulse setup a vibration in the fiber (electrostriction [wikipedia.org]) causing an interference pattern (Stimulated Brillouin [wikipedia.org]). Then up to 12 ns latter the "read" pulse can be passed through the interference pattern and it comes out with the original data stream.

At least that is what I got from the all too brief article.

Re:

[ockegheim (808089)]

So the beauty of this is that the read pulse can be sent whenever it's needed until the standing wave dies? In other words, it could be a delay of up to 12 nanoseconds?

Re:

[juiceboxfan (990017)]

Yes, this is a variable delay line.
From the abstract [sciencemag.org];

These stored pulses can be retrieved later, after a time interval limited by the lifetime of the acoustic excitation. In the experiment reported here, smooth 2-nanosecond-long pulses are stored for up to 12 nanoseconds with good readout efficiency: 29% at 4-nanosecond storage time and 2% at 12 nanoseconds.

What a "2% readout efficiency" is is anyones guess. Maybe only 2% of the modulation amplitude is recovered?

Bad summary

[Valdrax (32670)]

From reading the article, it looks like this is more of a way to handle the energy inefficiency of storing backlogged data as electrical signals. Theoretically, storing them as sound would create less waste heat.

It has little to do with speeding up transmission. It's actually about slowing it down so that systems can process the data received on the other end. The problem is that the trick requires light pulses far higher in power than most equipment is rated to handle.

Re:

[jamesh (87723)]

Theoretically, storing them as sound would create less waste heat.

But more noise. And you could really say 'wow... that switch is really humming!'

A trained ear might even be able to tell the nature of the traffic passing through the switch :)

Re:Bad summary

[beav007 (746004)]

You get used to it. I don't even hear the hum. All I hear is Blonde, Brunette, Redhead...

Re:

[dbolger (161340)]

Well the article was a bit lacking in specifics, but I think I have the general idea. They argue that light is a poor medium compared to sound for the transmission of Bad News. Bad News, as you will no doubt be aware, travels faster than the speed of light as it obeys its own special laws. Research done on Arkintoofle Minor suggests that this method results in unwanted side effects on receiving systems (packets rejected as unwelcome, etc), hense the need to convert it back into light before it reaches its f

Re:I'm confused

[SamP2 (1097897)]

Doesn't light travel faster than sound?

When you sit at the intersection and the red light 5 meters in front of you turns green, it'll take about 15 nanoseconds for it to reach your eyes. But rest assured, the honk from the car 5 meters behind you will reach your ears faster.

Re:I'm confused

[ajs318 (655362)]

Well, when you see the red and amber on together, you're supposed to change into first gear (second if facing downhill) and bring the clutch up to the biting point in readiness.

Extra bonus points* if you can manage to turn your right foot sideways to press down both the brake and gas pedals together, remove the handbrake and hold your car steady like this.

* However, this manoeuvre will not find you any favour with one of Her Majesty's driving examiners, so don't do it on your test.

Re:

[bjorniac (836863)]

Alas he's posting from the US where the red-amber doesn't exist before green. Lights go green, amber, red-green here. I've never understood why they don't have the "get ready" signal. It would save a lot of sitting behind minivans for two minutes after the light goes green...

Re:

[vux984 (928602)]

I've never understood why they don't have the "get ready" signal.

We watch the lights for the other drivers, when they get their amber, we know our green is imminent. I thought everyone knew this. ;)

Re:I'm confused

[springbox (853816)]

Yes, and that's probably why people in front of me keep jumping into oncoming traffic.

Re:

[zippthorne (748122)]

Because people think that the "Stop! (unless you really can't, then don't.)" buffer signal means, "Time to stop, soon. Speed up if you think you can make it." So we need to have an additional buffer after the official stopping cutoff. If crosswise drivers were poised to accelerate the instant the light turned, there wouldn't be as much of a tertiary buffer. (there's also a slight delay between the yellow->red and red->green transitions)

It's almost as if people don't bother taking *any* drivers ed

Re:

[bjorniac (836863)]

Well, it usually takes them a few seconds to notice in all honesty. So if it took a few seconds to notice red+amber (which usually, conveniently for my argument lasts a few seconds) they'd be ready when the green came along.

Re:

[blahplusplus (757119)]

"Doesn't light travel faster than sound?"

What I want to know is how fast does sound travel *in a superfluid*, now *that* is an interesting question. Does anyone know?

Re:

[woodrad (1091201)]

Depends on how strong the aether wind blows.

Re:

[Malevolyn (776946)]

Scientists increased the speed of sound this past summer. You probably didn't get the memo.

Re:

[evilviper (135110)]

Doesn't light travel faster than sound?

Not necessarily. The speed of sound in a vacuum is about the same as the speed of light in steel...

*cough*

Fiber optic synesthesia.

[sleeplesseye (113792)]

"Researchers at Duke are trying to solve the problem of speeding up fiber-optic connections by converting light into sound..."

Fiber optics on acid?!

Ummm

[Finallyjoined!!! (1158431)]

Isn't this what happens anyway? Pick up the phone, dial, talk. Sound converted to light, transmitted down a fibre optic cable, then converted back to sound. OK there's a bit of digital wizardry in-between :-)

Re:Ummm

[gardyloo (512791)]

Yes, but then you don't typically transmit 10 Gbit/second by speaking. Except when one is drunk, of course.

Re:

[Finallyjoined!!! (1158431)]

Heh. You are, of course, quite correct :-)

Re:

[Glowing Fish (155236)]

And at such times, the signal-to-noise ratio can become a very big problem.

Signal Buffering

[Lifyre (960576)]

They aren't trying to speed up the signal but mitigate the heat and other restrictions for current signal buffering methods (light to electrical and back) used on commercial fibers.

They found that using pulses of light they can create a sound wave that will store the signal in the fiber instead of having to turn it into an electrical pulse. Very interesting and has potential especially as a bridge to any future optical processing.

The summary title is a little misleading but TFA is pretty good.

Re:

[gardyloo (512791)]

What's interesting to me is that this must (probably) be a holographic effect, and that requires very precise phase knowledge of the transmitting light signal. I don't know if those clever optics guys can do that yet, especially on rapidly modulated signals, without some high wizardry.

A back-of-the-envelope calculation says that the wavenumber of the light in the fiber is about 1.4*10^10 /meter; I don't know, but to capture all of the information via sound _without_ a holographic effect, wouldn't on

A bit of LSD

[nurb432 (527695)]

Will do the job just fine.

A step sideways

[bartoku (922448)]

If I understand the article correctly, they want to convert light to sound, because the light is transmitting data to quickly for processing.

Why bother with buffers then? Just limit the amount of data being transmitted over the fibre, to a rate that can be processed. No 100 watt sound buffers, that do nothing but slow the data down anyway.

Why not invest time in those materials, that we hear about every other week on Slashdot, that slow light down. They should find a way to store the data as light, and skip the whole light to electronics or light to sound waste.

Re:A step sideways

[The Master Control P (655590)]

Because that takes too long. With fiber optics, 10Gbps is routine. If the light moves down the cable at 2/3c, then each bit is only 2cm long. In the time it takes for the corner fiber box's first bit in "please slow down" to go 100 meters to your house, five thousand bits will have been processed. Even if your box acknowledged in the affirmative instantly, 5000 more bits before the slowdown reaches the fttc box.

There are already mechanisms in place to deal with aggregate bandwidth demand exceeding supply which do as you suggest, the overloaded machine asks it's peers to slow down. However, when operating near max bandwidth, there's a situation where you might need to store a packet for two microseconds while you process another since they chanced to arrive at the exact same moment. This allows you to do that without paying dollars-per-byte for ECL memory that can shift at 10Ghz or shunting the data through a seperate (electric) delay line.

Good for bursts, not for continuous data flows

[G4from128k (686170)]

This scheme does not change the speed of the data in terms of being able to process it. This system would convert 1 Gbps of optical data into 1 Gbps of sound data. Any processing of the data still needs to handle the rate of incoming bits (data speed), which is not changed by the conversion.

Yes, they can use this to buffer a short data burst (e.g., 1 kbit of 1 Gbps data would fit in 5mm of fiber, assuming they can increase the storage time from 12 ns to 1000 ns), but it would still need to be read at 1 Gbps and would not help handle a continuous data stream.

SWEET!!!

[EEBaum (520514)]

Now our server closets will actually sound like the giant computers from old sci fi flicks!!!!

The more things change...

[Angst Badger (8636)]

Wow, they've reinvented the delay line.

Back in the old days, tubes of mercury were used as memory devices. A transducer would set up a series of waves in a tube of mercury which would be read by a receiver on the other end of the tube and sent back to the transducer. The data could be read as it reached the receiver. This seems to use the same principle, albeit for different reasons and using air as a medium.

Did I ever tell you youngsters about how CRTs were originally memory devices? (hobbles off to find some Geritol)

Re:

[sjames (1099)]

An optical fiber is already analogous to a delay line, just like your mercury tubes.

Only in the sense that a wire is also a delay line, electricity doesn't travel at infinite speeds either.

In fact, this research HAS re-invented the delay line, this time for optical rather than electrical signals and operating on much smaller timescales. That in no way detracts from their work, it's likely to be quite useful and the fundamental mechanism of it's operation is quite novel.

Prior Art!

[garcia (6573)]

LSD accomplished this way back. Most notably in the 1960s.

RADAR and acoustic delay lines

[flyingfsck (986395)]

Hmm, I have some past experience with acoustic delay lines. These are awful things, since you have to keep the temperature of the glass constant.

I need an explanation..

[renoX (11677)]

Sound is based on the atoms movements, electricity is based on electrons movements.

Atoms are way heavier than electrons, so why converting light into sound is supposed to be more energy efficient than the convertion into electricity?

I would expect the opposite in fact..

Sound, you say?

[denmarkw00t (892627)]

So is the Internets going to graduate from "a series of tubes" to "information pipe organ?"

Is'nt that done with....

[Danathar (267989)]

Something called a "Compact Disc" or "DVD"?