Lighting Fast – High Speed Optical Connectivity

Once relegated only to datacenter and telecom environments with high price points, optical technology may soon find its way into mainstream client systems, consumer electronics, and even handhelds. A new technology was announced at Intel Developer Forum (IDF) which provides initial data rates of 10 Gigabits and potential scalability to 100 Gigabits and beyond; something copper IO won’t be able to achieve.

Light Peak also supports multiple simultaneous protocols which will allow bandwidth aggregation of the various interconnects used in systems today onto a single high speed, thin, flexible, and long cable and small connector. Imagine being able to connect to your camera, display, docking station, or external hard drive through a single, thin connector!

Light Peak makes this possible by moving the next IO speed increase to optical and getting away from the electro-magnetic interference (EMI) and thickening and shortening of cables that are plaguing copper IO technologies today. Unlike the current high cost optical technologies in the datacenter, Light Peak will bring the benefits of optical in a mainstream client-ready cost footprint.

Light Peak is in the developmental stages, but Intel Executive Vice President

General Manager, Intel Architecture Group Dadi Perlmutter showed a demonstration of real silicon transmitting storage, LAN data and display (1080p) data across a single thin, 30m fiber optic cable.

Intel will be working with the industry to determine the best way to make this new technology a standard and to accelerate its adoption on a plethora of devices including PCs, handheld devices, consumer electronic devices and more. The end goal is to make Light Peak a complement to existing I/O technologies by enabling them to run together on a single cable and at higher, and more scalable speeds.

With its potential for future I/O speed increases, and as rich multimedia proliferates, Light Peak can enable technologies and systems that share data both in the home and office to continue to deliver full speed external IO that can keep pace with internal compute device bandwidth.

The future of computing looks as bright as the past… and optical technology is helping to light the way.

—-

Ben Hacker

More resources:

http://techresearch.intel.com/articles/None/1813.htm

http://www.intel.com/go/lightpeak

35 Responses to Lighting Fast – High Speed Optical Connectivity

  1. Igor Nikolaev says:

    Imho all photos was photoshoped. Why not connected all contacts on worked device? Star-wars light ray from fiber ends lotred me ;-)

  2. Duff says:

    Why does this not provide for power? an extra copper cable to provide electricity? Yer, it’s great that we’re getting rid of all these other cables, that i’ll only have that one cable… as well as a power cable! think about plugging in a USB cable for your phone, great, now we have to plug in the Light Peak Cable, as well as a power cable? it’s brilliant that we get faster speeds, but what about power? sure, years down the track we won’t need power cables, but if it’s being deployed in 2010, i’m pretty sure my iPhone will still need to be charged!
    Please Please PLEASE! some one look at this deficit in the technology, i’d hate to see this technology fail because everyone has to choose between getting one cable to charge and sync, or two cables, individually charging and syncing! sure i’ll have to leave my phone plugged in for that little bit longer for it to charge, but that’s better than having to plug one cable in to sync, then wait for it to be done (regardless of how quickly that is) and come back to unplug it to charge it, then realize i forgot to sync something so i have to change the cables over again!
    NOT COOL!

  3. beez1717 says:

    after reading and learning about this technology I was very excited about this and thought about uses, and I can see using these cables to connect external hard drives for quick painless backup via mac Os 10.5 and 10.6’s time machine. I can also see this used to get faster displays and to transfer music and videos to an ipod and iphone at high speeds :)

  4. Jeremy says:

    How does the latency and interrupt tolerance compare to conventional technologies (particularly display and audio interconnects)? Can this technology support even cheaper and simpler connectors for dumb peripherals?

  5. alan geering says:

    I work with digital audio electronics. Firewire has for a long time been the connection of choice when moving audio and video in real time. USB is usually only used when the datarate required is much less than the total USB provides.
    Will lightpeak be master-slave or peer-peer in architecture? Will real time applications be able to reserve bandwidth?
    Basically, how apporpriate will lightpeak be for connecting multichannel digital mixers/A-D converters to computers for recording?
    Currently I use a mix of Firewire (32 channels 24 bit @ 48khz), USB (4 channels 24 bit at 96khz) and PCI adapters with inputs for many channels (via ADAT, TDIF, AES, etc.).
    An aside:
    Comparing some (admittedly older generation) optical fibres with copper cables I’ve some concerns. Only the latter survived being crushed by a ladder. Copper cables have a toughness I worry optical fibres might not. What has been done to mitigate this in the design of Light Peak?

  6. C says:

    I’m a photographer and not an engineer.
    One of the issues with digital photography today is high ISO noise (i.e., digital artifacts when shooting in low light conditions). Since (light) signal is weak in low light conditions, noise is high due to signal amplification in the sensor (particularly in the Blue channel). Noise is also generated from sensor heat (byproduct of an electrical current).
    So, I got to thinking about whether this technology can be incorporated into the design of the camera’s sensor (i.e., switch from an electric signal to an optical signal) and whether doing so would both reduce noise and increase the sensitivity and light efficiency of a sensor for better photographs in low light situations?
    Also, is it possible to incorporate this technology into the camera’s internal circuitry and into the memory cards as a way to free up the memory buffers and speed up the write times so one can shoot continuously without stopping?
    That would be awesome if it’s doable.

  7. Neil says:

    This system is designed to work upto 100m
    Adding power cables to such a length would not be a good idea due to the loss of power due to resistance over such a lenght of cable.
    However for short distances, they should have the possibility to provide power. And maybe to add power locally for long cable distances.
    Mind you with ‘wireless’ power you won’t need a power cable soon, plus solar capabilities starting to come out.

  8. Britt says:

    PLEASE PLEASE PLEASE make a locking variant of the plug! I can’t tell you how many times I’ve been called over because some HDD or peripheral has gone “missing” and all it was was that they knocked the USB cable loose!
    Also, PLEASE make sure that it can do DMA and isochronous transfers like firewire does…
    Oh, and one other thing — can we get an open, no-cost device ID scheme? To get a USB ID number, you have to pay $4000 per year to the USB-IF… small developers can’t afford this. Instead, I would suggest using a domain-name based scheme, kinda’ like Apple does for application identifiers (so for example a device might identify itself as ‘kbd1234.intel.com’ if it was a keyboard controller model 1234 made by Intel — microcontrollers now have the space and performance to support this, we don’t need a small device-id address space like was needed in decades past).

  9. john says:

    Yeah of course power is the existential question. The really futuristic solution would be light energy running a tiny electrical generator on the other end.

  10. Jon L says:

    Obviously this will have a sender/receiver setup which will be powered on each end…
    As in, one box to plug your display, mouse, drive, etc. into, which has it’s own power brick to provide power where necessary. ??
    What type of fiber cable ends will this tech use? LC? SC? New proprietary?
    Hmm…

  11. Dear Ben
    I am very interested on this technology, both due to the package and the bandwidth. I would like to have more information and even have a test-setup of such a link…when you think that it will be possible to have engineering samples? (or even when it will come to market)…I have a immediate application for this link on particle physics applications and also on medical (PET Scanner). please keep me updated on all developments of this light peak tech. Best Regards
    JC

  12. Ben Hacker says:

    Thanks for all the comments guys!
    @ Michael: There are many possibilities with this technology. Their certainly may be some energy savings in the future.
    @ Duff: Power is obviously a requirement for a universal connector, and that is something we are looking at closely. Don’t panic yet!
    @ Jeremy: We are not sharing any lower level of technological detail at this time.
    @ Alan: I’ll share the same comment as I gave Jeremy above, but these issues are things we are aware of.
    Regarding locking cables, and length of optical only, vs. optical+power cables, we are aware of these issues and are looking at various options.
    On the other questions about additional specs, and specifica dates, we are not providing details at this time, but please stay tuned going forward.
    Thanks,
    Ben Hacker
    Intel Corporation

  13. Michael Jahn says:

    Hi Ben,
    I recall explaining bandwidth and data transfer to my non techie freinds in using a water metaphor.
    Think of data as something like water, only it comes in ‘packets’ of watter, each ‘packet’ being about 1 cup – 8 fluid ounces. with that in mind, think of your mouth as your USB Port, firewire, what have you, and your stomach as your harddrive.
    While we can reasonably assume that we might be able to sustain and input rate of 2 cups per minute, we simply can change the delivery mechansim from a drinking straw to a garden hose – our “data (fluid) processor may not withstand this increased speed of input. Now, what we seemed to be speaking of is changing out our stomach – actually, the entire digestive system – from what we have to this new technology. It is NOT about simply replacing the garden hose with a fire hose – we can certainly do that, but no ones moth or digestive system can handle fire hose-ish bandwidth, even though the telcos can certainly deliver 42 gig a second today.
    The requirement is to change how the stuff INDSIDE can process data. Thats what make this Light Peak so important. Once we have that inside, we can indeed hook up a fire hose up.

  14. Anders Sjögren says:

    Ben,
    when it comes to power is guess it would be easy to take the firewire-path of providing two connector types, one with power and one w/o, i.e. to have power being optional. I guess one problem is that the places where power is needed (cellphones, drives, etc), space AND duriability is also an issue so getting a small and robust connector is a must.
    In USB there are two levels of power, right, none of them with enough power for 3.5″ drives? It would be ideal if light peak would allow enough power to power even 3.5″ drives from it, perhaps using a fatter connector+cable.
    Looking forward to the new standard!
    Anders Sjögren

  15. Anonymous says:

    Ben,
    when it comes to power is guess it would be easy to take the firewire-path of providing two connector types, one with power and one w/o, i.e. to have power being optional. I guess one problem is that the places where power is needed (cellphones, drives, etc), space AND duriability is also an issue so getting a small and robust connector is a must.
    In USB there are two levels of power, right, none of them with enough power for 3.5″ drives? It would be ideal if light peak would allow enough power to power even 3.5″ drives from it, perhaps using a fatter connector+cable.
    Looking forward to the new standard!
    Anders Sj

  16. Bruce Trent says:

    Has cabling and/or connectors been tested for military applications and analysis done on the affect of noise/interference created by going through slip rings on gimbaled systems?

  17. Jon says:

    Ben – Light Peak does look like awesome tech, it’s amazing you’ve got the costs down to consumer levels. Thank you for answering some of the outstanding questions, but I’ve got a few more.
    Glad to hear you’ll be adding power over LP, any word on what that might be and over what length? Please provide real power, unlike USB… I’m hoping for at least 12V/1A/12W. In the eliminating cables catagory I would really like to not have to plug my 2TB 3.5″ external HD’s into the wall… as well as quickly recharging mobile devices (cameras, phones, etc)… Other I’m sure would like to power external DVD drives… Etc… Don’t short change us in power or this dream of one cable to rule them all will die quickly.
    Topology? Is this going to be hubbed or daisy chained? I’m certainly assuming the later but it’d be nice to know… Maybe it can handle both?
    The demo showed a connector that looked an awuful lot like USB A, will this be the final design? Seems like you could get much much smaller… Whatever you do please use a keyed shape, unlike USB A / like USB B and FireWire 400. Also some mechanism to reduce pull out would be great, not necessarily a positive lock (no screws, no catch like Ethernet) but something like magsafe or the friction springs on Apples iPod dock connecots (the newer version without buttons).

  18. Rob says:

    Having read as much as possible about Light Peak including the new Wikipedia entry I am curious about the multi-protocol support. Are the protocols transmitted side by side or is there a whole new method involved? I know that Light Peak is mostly meant for the PC/CE/Mobile sphere but I was thinking of using Light Peak for InfiniBand. What I am trying to ascertain is that since InfiniBand has 24 links at its maximum – 12 in each direction. How would Light Peak be able to accommodate the IB protocol? Would the cable need 24 fibers or would it need fewer fibers to accomplish the same thing?
    Thanks

  19. John says:

    I would appreciate some detail on how Light Peak differs from Intel Connects Cables (now Emcore Connects Cables). It sounds like Light Peak is a new connector and scaled 5Gb/s lanes to 10 Gb/s

  20. Struan Robertson says:

    What topology will Light Peak support and are any plans being made to increase transmission distances? The extra bandwith is great but from a broadcast production view no more than 100M transmission distance is limiting. 2km and you would be getting there. Single mode SFP,s give you 20km so …. Also will you be able to optically mutiplex Light Peak?
    Thx

  21. Rocko Docko says:

    There are already optical technologies that cover long distances. This is designed for the end-user and 100m will be more than enough in 99% of those cases.
    And that’s not a photoshop to the first guy posting. The used visible laser light to simulate what it would look like if you could see light peak signals, which you can’t as they are infrared.

  22. sho says:

    Dear Ben, I am very interested on this connection technology. I would like to have more information and please see our products on our website. There is some Optical Fiber Circuit Sheets. I believe growth to great tech if it join your light peak tech and our fiber sheet tech. Best Regards.

  23. John says:

    As an end user, how will Light Peak compete with USB3.0? It seems that users will just start to see USB 3.0 next year; which is the same year that Light Peak is being said to have products.

  24. Ed Palma says:

    Just for redundancy….If there’s one thing to be learned from USB, IEEE1394 and Ethernet, it’s that people will need power to run over the connector.
    Light peak looks so very promising we’d all hate for it to be limited by the absence of available power. My hope is that eventually this will replace USB, Firewire, Network, Display and S/PDIF, leaving just an AC plug and 6 to 10 Light Peak Jacks.
    I think that it’s ideal if devices get a bit of current from a host and can request more if they need it. Additionally there can be cheaper optical-only cables for longer hauls (projectors, soundsystems, etc) and situations where electrical isolation is desirable (robotics? lighning research?)
    But the jack itself needs to support power. Enough to power a flash drive, a mouse, a printer, a cellphone, a camera, audio capture devices (I love my MOTU Traveler for it’s power over IEEE1394 for recording in the Africa bush) etc etc. Having power over the line just increases the possibilities so much.
    On a more specific rant…I for one am looking forward to getting rid of XLRs and 1/4 in plugs in live audio. The only reason those things are still around is because they’re ubiquitous. The only reason digital alternatives haven’t take off is because they’re all proprietary implementations that nobody wants to play with. It is possible that Light Peak could solve this.
    Here’s hoping to a well planned connector!
    Ed Palma

  25. Hi Ben,
    A very interesting programme with potential applications in the aerospace community. Could you say what size optical fibre you are using, is it a multimode 50µm or a singlemode core?

  26. marco hernandez says:

    Your optical chip does not appear to have described the nature of interconnecting the fibers onto the detectors and lasers. the fibers would have to be angle polished to reflect onto active areas of laser & detectors. perpendicular fiber attachment would enlarge your footprint.the footprint for the mechanical infrastructure to hold the fibers aligned , the alignment & most important long term reliability & method to attach the fibers are not clear, freespace air gap I dont think a direct contact to the detector element is advisable. Also while the fibers are 125um they will need to be ruggedized for any practical use & the public will need to understand what happens when you bend the fibers . A retrofit package would have to at least support the current SFP & XFP cages that many thousands of network cards currently use, therefore the size will have to be the same if a retrofit solution is expected. I am still trying to understand the commercial use from everyday people and imaginge a bunch of 3mm fiber cable connected to various devices as you described. 3mm cable is a guesstimate to hold 4 fibers that need to be re-jacketed. your expected shipment in 2010 would be for what application? Wish I was in your lab to play with things. I pigtailed my firt fiber to a waveguide in 1984 building fiber gyroscopes so my interest is genuine. Regards

  27. Avi Levy says:

    Hi Ben,
    Light peak would be appropriate to replace high-speed board-to-board electrical connectors and at the same time introduce possible new “out-of-board” applications.
    For example a 113Gbps card may need to interconnect Tx&Rx 10×11.3Gbps between Client and Line cards (total 226Gbps).
    What do you think about this type of application? Please let me know how to elevate it.

  28. Dear Ben,
    1. First a little clarification, I know that Intel has been working on surface mount laser diodes for years now, if you could give a bit more history about the project (and officially let the Apple fanboys know that Apple was invited to the party, they did not start the party).
    2. Would it be possible to add an interface for fiber optic microphones and headphones to the specification, I know they are expensive now, but with it built into your chip it could drive the price down quickly. I wondered if this would be possible to finally get rid of one of the last vestages of analog electronics and everyone who has tried to work with audio recording and reproduction has suffered for their art for years. I don’t know enough about fiber microphones to know if it is possible (without just using Light Peak as the transmission medium instead of the pickup itself). I know that usb audio was desired to remove the need for another physical jack, so I was hoping that you considered this from the start. (note: it is not that I consider analog inferior to digital, it just seams that analog audio has yet to have a serious digital competitor)
    3. I assume that when you move the signal from copper to optical you are allowing for a physical layer bridge to be created through LP, so the source of data does not know or care that LP is in series, is this correct?
    4. Is there any truth that the optical connections is the pre-release USB 3.0 spec was going to be LP, except it would have delayed the official launch too long? If so, does this mean that the pre-3.0 connector might just be the LP connector (with direct USB 2.0 or 3.0 backwards compatability). I thought that it would seam that USB 3.0 + LP would be perfect, however I understand that the limitations of the original USB 1.0 housing may have too many drawbacks to continue forward. I personally like the DisplayPort connection, but I know you don’t need that big of a connector.
    5. Lastly, is ClearCurve actually going to be a necessary for lightpeak for is this jusk Wikipedia FUD?
    P.S. Some of comments are amazingly depressing, have these people never heard of PROTOTYPE or USB 3.0’s 0.9-1.8mA extended power handling capabilities.

  29. Michael says:

    How are the small diameter fibers terminated cost effectively? Do they not need to be cut, polished, and precisely aligned? Can the fibers be terminated in mass?

  30. analog_ says:

    I’m no expert, but best way I see power bundled with lightpeak if it’s a single hybrid connector (USB sized, with enough room for a fiber to pass in between). I know you can detected quite easily whether or not which cable is actually connected too, sell fiber only, hybrids and copper only cables. Using the same mgmt protocol which provides means for the remote device to request a certain level of power, the controller should check the cable if it matches within allowable specs of the cable. This way whatever you do, it should work as long you use the correct cable for the correct remote device. It’d be pretty easy to construct 95% of the cabling since it’s going to be 2M, 5V w/ optics (your average desktop cable).
    I haven’t read much into it yet, but i hope it’s some form of packet based connection type. It’d be great if the devices would request/offer proposals for device parameters and a single fiber strain could feed multiple devices at once (one fiber, aggregating multiple other single fibers). The speed should be load balanced of QoSed to make it work.
    apropos:
    o thick, 48V cable
    o thin, 5V cables (small peripherals, HID, storage)
    o 12V hybrids for small network devices
    o power injectors capability, with said mgmt protocol.
    o