Polymer Cables Could Replace Thunderbolt and USB, Deliver More Than Twice the Speed (appleinsider.com) 80
A user shares a report from Apple Insider: Researchers are working on a cabling system that could provide data transfer speeds multiple times faster than existing USB connections using an extremely thin polymer cable, in a system that echoes the design path of Thunderbolt. Presented at the February IEEE International Solid-State Circuits Conference, the research aims to develop a connection type that offers far better connectivity than current methods. In part, it aims to accomplish this by replacing copper wiring with something else.
Copper is typically used for wires like USB and HDMI to handle data transfers, but it requires a lot of power to work for high levels of data transmission. "There's a fundamental tradeoff between the amount of energy burned and the rate of information exchanged," said MIT alumni and lead author Jack Holloway. While the "increasingly bulky and costly" copper could be replaced by fiber optic cables, that introduces its own issues. As silicon chips have difficulty dealing with photons, this makes the interconnection between the cable and the computers more challenging to optimize. Combining the benefits of copper and fiber optic conduits, a plastic polymer is used by the researchers. This makes it cheaper to manufacture than copper wires, which could be an attractive proposition for cable producers.
The polymer can also use sub-terahertz electromagnetic signals, which is more energy-efficient than copper at high data loads. It is believed this efficiency brings it close to that of fiber optic systems, but crucially with better compatibility with silicon chips. Low-cost chips are paired with the polymer conduit that can generate the high-frequency signals powerful enough to transmit into the conduit directly. As such, the system is expected to be manufactured with standard methods, which also makes it cost-effective to produce. The cables themselves can also be extremely thin, with the cross-sectional area of the interconnect measuring 0.4 millimeters by a quarter millimeter, smaller than typical copper variants. That small hair-like cable can be used to transport data over three different parallel channels, enabling it to achieve a total bandwidth of 105 gigabits per second. Bundling conduits together could bring the cables into the terabit-per-second range, while still remaining at a reasonable cost.
Copper is typically used for wires like USB and HDMI to handle data transfers, but it requires a lot of power to work for high levels of data transmission. "There's a fundamental tradeoff between the amount of energy burned and the rate of information exchanged," said MIT alumni and lead author Jack Holloway. While the "increasingly bulky and costly" copper could be replaced by fiber optic cables, that introduces its own issues. As silicon chips have difficulty dealing with photons, this makes the interconnection between the cable and the computers more challenging to optimize. Combining the benefits of copper and fiber optic conduits, a plastic polymer is used by the researchers. This makes it cheaper to manufacture than copper wires, which could be an attractive proposition for cable producers.
The polymer can also use sub-terahertz electromagnetic signals, which is more energy-efficient than copper at high data loads. It is believed this efficiency brings it close to that of fiber optic systems, but crucially with better compatibility with silicon chips. Low-cost chips are paired with the polymer conduit that can generate the high-frequency signals powerful enough to transmit into the conduit directly. As such, the system is expected to be manufactured with standard methods, which also makes it cost-effective to produce. The cables themselves can also be extremely thin, with the cross-sectional area of the interconnect measuring 0.4 millimeters by a quarter millimeter, smaller than typical copper variants. That small hair-like cable can be used to transport data over three different parallel channels, enabling it to achieve a total bandwidth of 105 gigabits per second. Bundling conduits together could bring the cables into the terabit-per-second range, while still remaining at a reasonable cost.
Re: (Score:2)
Microwave Wave Guide (Score:5, Informative)
No this appears to be an compact wave guide system for sub THz signals - so basically an extremely small and flexible microwave wave guide.
It's certainly interesting work but I can't find much information on it than what is listed here. For example, what the losses are in this material and any mechanical properties (min bend radius, temperature range etc). Also the purported big advantage over fibre optics - that the baseband interface is simpler/cheaper - isn't a given. This is going to require a ~100GHz RF front end, and from my limited knowledge of high frequency RF circuitry, that will require pretty exotic silicon. It might be cheaper than a laser diode/detector, but I can't see it as being something you can stuff onto an existing mixed signal chip design with ease.
Additionally, waveguides will have similar problems as fibre optics at any connection interfaces (dust, contamination, alignment). This was one of the big negatives against fibre connections. I remember when TOSlink was common on sound cards. It wasn't prohibitively expensive, but in the end the connectors just kinda sucked and Firewire got fast enough.
Re: (Score:2)
Despite its slim size, it can carry a hefty load of data, since it sends signals over three different parallel channels, separated by frequency.
This would also suggest that it isn't using any type of waveguide transmission since the cross section of a waveguide is designed specifically for a single frequency and is hollow inside. The paper also suggest increasing the data transfer rate by bundling multiple polymer conduits together, p
Re: Microwave Wave Guide (Score:1)
So can we settle on vaporware? Since apaprently they do not even manage to tell others what it is. (Or is there a paper?)
Re: (Score:2)
There is a press release [mit.edu] about a conference presentation. Our THz dielectric waveguide link work, originally presented on ISSCC 2021, is reported on MIT homepage. [mit.edu]
So it appears to be a waveguide, based on scant information and no paper. In the absense of falsifiable specifics, vaporware is a fair call.
Re: (Score:2)
Re: Microwave Wave Guide (Score:2)
Waveguides for high power signals for antennas have different constraints than short haul communications. Losing a couple 10s of percent of power because of bending is unacceptable for the former, irrelevant for the latter.
Sub wavelength plastic waveguides for double/triple digit GHz signals seem a pretty active research area. Usually hollow or foam core.
Re: (Score:2)
waveguide, and the rectangular cross section suggests polarization specific.
Re: (Score:2)
I remember when TOSlink was common on sound cards. It wasn't prohibitively expensive, but in the end the connectors just kinda sucked and Firewire got fast enough.
Firewire did not kill TOSlink. Putting the same signal on a piece of coax instead of using fiber did. There was never any need for fiber to carry digital audio signals, it doesn't require enough bandwidth for that.
Re: (Score:2)
TOSlink was about noise suppression more than anything. Bypassing the often incredibly noisy analogue parts of the path at the source, eliminating the noise cheap analogue cables pick up - especially on long runs. Note: I didn't say sound quality, just noise, though it was certainly sold that way.
AFAIK Firewire didn't kill anything but consumer SCSI, all my PC's, disk players, set top boxes, TV's (inc one bought 2 months ago) and sound cards still have TOSlink, none this century still had Firewire.
HDMI will
Re: (Score:2)
TOSlink was about noise suppression more than anything. Bypassing the often incredibly noisy analogue parts of the path at the source, eliminating the noise cheap analogue cables pick up
No. And also no. Going to a digital signal was about noise suppression more than anything. It doesn't matter if it's on coax or twisted pair or optical. That's why TOSlink was always dumb.
HDMI will kill completely kill TOSlink eventually
What about non-video devices? It still makes more sense to just have a coax digital connector, which is just a nice cheap RCA jack, than to use optical.
Re: (Score:2)
And all this time I thought TOSLink avoided another D/A-A/D conversion in the chain... Yeah, whatever, I still use my Minidisc recorder, and it's better than damnable MP3.
Re: (Score:2)
And all this time I thought TOSLink avoided another D/A-A/D conversion in the chain
It doesn't, unless you have a stereo with a digital amplifier. Such a thing is certainly possible, although I'm not aware of anyone building one. A friend of mine built one out on a breadboard once. It took digital optical in and then did 100% digital amplification with a MOSFET. It sounded fine to me, but no analysis was performed to determine whether the end result was actually faithful to the recording. Besides avoiding noise in an analog signal, all it does is change the location where the D to A conver
Re: Microwave Wave Guide (Score:2)
Digital amplification? Oh, you mean a Class D amp? Maybe all gussied up with a PWM input stage?
Re: (Score:2)
Digital amplification? Oh, you mean a Class D amp? Maybe all gussied up with a PWM input stage?
Dunno, probably. I only ever dabbled in electronics, so I don't remember the specifics of how he did it. I just remember it took SPDIF input.
Re: Microwave Wave Guide (Score:2)
Yeah, probably an amp with digital inputs, not a 'digital amp'. SPDIF, AES/EBU, these are digital encodings, and aren't useful to try direct d/a conversion. Advantages over analog, especially with digital sources like CD or DAT. Lossy formats like MiniDisc or MP3 don't benefit as much. I think HDMI includes a digital audio transport.
Re: (Score:2)
Well, he explicitly told me that he was amplifying a digital signal. I thought he said it was PWM or PCM... Unfortunately he's no longer with us so I can't ask him, not that he would likely remember. He was one of the smartest guys I knew. He gained momentary geek popularity when he created OmniRemote for Palm Pilot...
Re: Microwave Wave Guide (Score:2)
Sure. SPDIF could be decoded into PWM. But no digital audio format I'm aware of that permits recording and playback is anything but a true digital encoding. It will need to be converted into something a typical amplifier can use, PWM, analog, for instance.
If your amp had an SPDIF interface, you didn't have a 'digital amp'. And like many formats, SPDIF could be lossy.
Re: (Score:2)
It was also about not having any electrical path between two pieces of equipment, so no ground loops.
Toshiba TOSLINK (Score:5, Interesting)
> a plastic polymer is used by the researchers. This makes it cheaper to manufacture than copper wires, which could be an attractive proposition for cable producers.
Cool, they're re-introducing Toshiba's 1983 TOSLINK technology for general purpose communication.
https://en.wikipedia.org/wiki/... [wikipedia.org]
Re: (Score:2)
Who thought it used lasers? Please be specific.
Audiophiles use laser to avoid attenuating the bass frequencies through jitter.
Re:Toshiba TOSLINK (Score:5, Informative)
Low-cost chips are paired with the polymer conduit directly
Leads me to believe that the connectors will still be copper interconnects, and there will be a chip inside each end of the fiber cable doing the transmitting/receiving. I guess they think it can have more bandwidth than copper per line?
Also modern connectors can provide a LOT of power, how are they going to do that without copper lines as well? Complexity for the sake of complexity.
Re: (Score:2)
but copper power lines can be separate thing that don't have to deal with high frequency.
Re:Toshiba TOSLINK (Score:4, Informative)
Almost, but this quote:
Low-cost chips are paired with the polymer conduit directly
Leads me to believe that the connectors will still be copper interconnects, and there will be a chip inside each end of the fiber cable doing the transmitting/receiving. I guess they think it can have more bandwidth than copper per line?
That sounds like the optical HDMI cables that I use for field production work. They're fragile, but they can carry a high-bandwidth signal for longer distances.
Also modern connectors can provide a LOT of power, how are they going to do that without copper lines as well? Complexity for the sake of complexity.
Who says they are? This is what a fiber optic HDMI cable looks like:
https://cdn.shopify.com/s/files/1/0011/3449/3775/products/fiber_optic_hdmi_1080x.png [shopify.com]
Notice that the four fibers (one per HDMI video channel pair) are wrapped in a small bundle of copper wires for power, ground, audio return/Ethernet/hot-plug detection, etc. The details probably vary from manufacturer to manufacturer, but they all contain at least a couple of copper wires; they just don't use them to carry the actual video data.
Re: (Score:2)
The idea of combining the data transmission and power transmission has been problematic. Though I love the idea, I also realize the amount of risk presented.
Bubye expensive computer:
https://arstechnica.com/gadget... [arstechnica.com]
Re: (Score:1)
So they can charge more.
Re: (Score:2)
I didn't read the fine article, but from the summary it sounds to me like more of an RF waveguide sort of deal than anything else.
Re: (Score:2)
Strength in numbers. (Score:3)
If it's more durable then this may solve my headphone longevity woes.
Re: (Score:2)
Your headphone woes are caused by ultra-thin wires already. The makers of your headphones know that the wires are too thin. They want you to buy another pair soon, so are looking to make them even thinner, instead of doing the right thing and going the other way.
Re: Strength in numbers. (Score:2)
Which is also ridiculous, given that the manufacturing cost differences are completely negligible at this point. "What, we'll have to pay 0.002 CENTS for the wires! unacceptable!"
Re: (Score:2)
People still use wired headphones?
And they complain at all about cable thickness?
Power delivery? (Score:4)
What, do you just pull hard on the polymer cable to start the two-cycle generator?
dielectric waveguide (Score:4, Interesting)
it's quasi optical.
you get a cylindrical waveguide mode. since the frequency is so high, the bandwidth of the mode is also very large, so it' can carry a lot of data. the other advantage is that you can use true modulation, like QPSK for example, instead of simple on-off keying and get large BW efficiency. however that may not be worth the cost in terms of the complexity of the TX and RX circuitry. That's ok, stick with OOK and you'll still get ridiculous bandwidth.
It will definitely work, the question is what the loss characteristics will be, but if it's 3-5 feet sort of connection, then loss becomes much less of an issue.
Re: (Score:2)
--
All opinions are highly speculative and may not be indicative of future results.
Re: (Score:2)
the question where are the supposed advantages.
They seem to be trying to spin that copper is bulky/expensive, which is quite obviously pure BS for usb cables/etc.
They are claiming 'faster' while not comparing to equivalent encoding/modulation techniques over copper, which would be simpler to implement.
Basically their advantage seem to be 'hey, this is different, gimme research bux!!!'
Thunderbolt Optical Cables (Score:2)
Have they forgot that Thunderbolt already supports optical cabling, can buy a 60m cable on Amazon right now if you want. The real question is if this provides significant cost savings compared to those optical cables. If the whole thing is going to cost as much or more then the whole thing is just headline fodder.
Re: (Score:2)
What a deal at only $999! I’d love to run 10gb fiber on my lan because nics and cabling are cheap. 10gb switches are a different story. I even looked at running IP over FC but nobody does it. I guess I could run dual nics and have a few systems daisy chained together but that’s kinda hacky. Infiniband hardware is cheap too but not really meant for IP and can have high cpu usage.
Re: (Score:2)
What a deal at only $999! I’d love to run 10gb fiber on my lan because nics and cabling are cheap. 10gb switches are a different story. I even looked at running IP over FC but nobody does it. I guess I could run dual nics and have a few systems daisy chained together but that’s kinda hacky. Infiniband hardware is cheap too but not really meant for IP and can have high cpu usage.
If you can justify the performance gains, the additional cost shouldn't be a concern. If you're still adding up the pennies, then you really don't have a justification beyond "would be cool". Time is Money if you have employees sitting around wasting dozens of hours every month waiting for terabytes of data to be flung between systems.
Speaking of costs, used switch hardware will embarrass any new car when it comes to clocking depreciation times. Many a "cool" network upgrade has been built by eBay at dot
Re: (Score:2)
Its all for personal use. A 10gb nic is like $35 with the transceiver. The cheapest 10gb switch is $350 new, a factor of 10x the other hardware. I did see someone selling 8 port 40gb infiniband switches for $65! Yes a 40gb switch. But again nobody runs IP on it and the switches can be software locked for functionality. IB cards and cables are so cheap and its so tempting.
Re: (Score:2)
Its all for personal use. A 10gb nic is like $35 with the transceiver. The cheapest 10gb switch is $350 new, a factor of 10x the other hardware.
Uh, the 10x switch cost factor, is due to the 16x - 24x port capacity? I guess I'm not quite understanding the complaint here. And considering the age of Gig-E, you'll replace that 10gb switch in what, a decade? $35/year isn't too bad when you consider the lifetime of a $350 switch.
Re: (Score:2)
Try $350 for a new 5 port switch. I’m just saying the switch prices are out of line compared to the rest of the hardware.
Re: (Score:2)
I think you will find that most if not all Infiniband setups run IPoIB. No idea if Windows supports that by Linux most definitely does.
However you will need to put in a Infiniband to Ethernet bridge somewhere if you want to communicate with the wider world.
Re: (Score:2)
What a deal at only $999! Iâ(TM)d love to run 10gb fiber on my lan because nics and cabling are cheap. 10gb switches are a different story.
Since you can do 10GbE on copper I'd just wait for that to get cheaper if you don't have a business case for actually needing it.
Re: (Score:2)
Re: (Score:2)
The point is if this polymer is a drastic price decrease on the current optic cabling then that's great. If it's going to be running $1k a cable though then I'm back to "How is this different than the existing optical cables?"
It's not going to actually be faster on current runs, because that's going to be controlled by what the port itself can sustain, so it's going to cap out at the same point as Thunderbolt 3/4's spec (40 Gbit/s). Is it going to be more durable? Maybe, the article doesn't say. More co
Connectors, connectors, connectors (Score:3)
That's because they have to be aligned with pretty tight tolerances and even a mote of dust can cause enough degradation to ruin the SNR. And I don't see how this technology is going to solve that.
Re: (Score:2)
Re: Connectors, connectors, connectors (Score:2)
Millimetre wave has more relaxed coupling requirements than nanometre wave.
Re: (Score:2)
And as soon as I see the term 'millimeter wave' applied to this discussion, I think, 'hmm, big deal about mobile 5G and millimeter wave bandwidth, and they are making radios for my phone, so maybe, maybe these radios can fit in my computer...?'
Yeah, tiny stuff can be made.
Re: (Score:2)
The optical transceivers will be in the connectors, so the alignment issues are solved, as long at the silicon-polymer interface is rigid and secure inside the connector.
The bandwidth of the cable will be set by how many polymer wires are inside the cable, each at around 100Gbps. So these transceiver chips will likely have multiple polymer I/Os, say 4 out and 4 in initially (cables will provide 1 to 4 cables and price gauge on this, despite the cables being the cheapest part). They'll probably interface wit
Re: (Score:2)
The optical transceivers will be in the connectors, so the alignment issues are solved, as long at the silicon-polymer interface is rigid and secure inside the connector.
Which, at least in my experience with optical HDMI cables, it is until you tug the cable at the connector just a bit, and then it isn't. :-(
Re: (Score:2)
Only once had a problem with dirt on a fibre optical cable and that got dragged along under the raised floor and those stupid tiny little push protectors for the LC connector came off. Grrr I hate those but couldn't find any spare clip on ones. Anyway quick wipe on my cotton shirt and all was good.
Oh and LC connectors all have captive clip things unlike RJ45 which often get snagged and broken off unless you have purchased snagless cables. SC connectors are perfectly robust which is probably why they are use
Why would it replace USB/Thunderbolt? (Score:2)
Why would it replace USB/Thunderbolt? Those are just protocols that run on top of the physical layer. It doesn't matter if if the physical layer is copper, glass, or plastic. Just like TCP/IP doesn't care if it's packets are being sent across copper or glass or streamed across the spectrum.
Re: (Score:2)
Also there are no applications for USB or thunderbolt that need such a high capacity. Only the very fastest of high-end SSDs can saturate even USB3.0, and nothing comes close to the 20Gb/s of the stupidly-named 'USB 3.2 Gen 2x2.'
I can see this being used in some niche datacenter roles for connecting switches in a stack, maybe. Stick a QSFP56 connector on each end and you have a drop-in replacement for existing cables that can promise lower power consumption than copper and longer range than existing copper
Re: (Score:2)
Re: (Score:2)
Err. 20Gb/s is 2GB/s which is lower than decent NVMe SSDs already.
External GPUs could easily make use of more bandwidth than even 40Gb/s that current Thunderbolt external GPU connections have.
Re: (Score:2)
Re: (Score:2)
Audio acoustic research next? (Score:3)
I have witnessed cable differ in age, maybe quality, produce superior harmonics, over tones and that psycho-acoustic stuff of lore during music performance. This new cable addresses an intuition I have always had about two properties; speed and metalurgical alignment.
Stuff of dreams but disruptive as hell. A cable could nullify 6 figure hardware stacks.
Re: (Score:2)
Ummm.let me see. no?
Optical chips (Score:2)
Where are the promised optical processors? There would be no issue interfacing with those.
Conductive polymer? (Score:2)
Conductive polymer? [wikipedia.org]
Re: (Score:1)
It's the dual of transparent aluminum.
Sounds Fragile (Score:2)
They have obviously never see my wife running around the house withe a cable and charger dangling like a big tail from her cell phone... :-)