Terahertz Wireless Chip Will Bring 30Gbps Networks 177
MrSeb writes "Rohm, a Japanese semiconductor company, has created a silicon chip and antenna that's currently capable of transmitting 1.5Gbps, with the potential to scale up to 30Gbps in the future. While this is a lot faster than anything currently on the market, the significant advance here is the reception and transmission of terahertz waves (300GHz to 3THz) using a chip and antenna that's just two centimeters long. Rohm says it will only cost $5 when it comes to market in a few years — a stark comparison to current terahertz gear that's both large and expensive. The problem with terahertz transmissions, though, is that it's highly directional — with a submillimeter wavelength, it's more like a laser than a signal. Terahertz waves might enable awesome device-to-device networks, but it isn't going to bring 30Gbps internet to a whole city block. More interestingly, submillimeter terahertz radiation is the next step up from the gigahertz radiation used in full-body millimeter wave scanners. Terahertz waves can not only see through clothing, but can also penetrate a few millimeters of skin."
Run it by a RF EE next time (Score:5, Interesting)
Run it by a RF EE next time, or at least an advanced ham radio guy.
using a chip and antenna that's just two centimeters long
a stark comparison to current terahertz gear that's both large and expensive.
with a submillimeter wavelength
First of all its hard from a RF perspective to make stuff thats more than a 1/4 wavelength long. Obviously possible, but much harder. For example, I'm working on a K band transverter and one nightmare is standard SMA connectors resonate at 18 GHz or so, making them quite exciting to use. Yes I already know about the expensive and complicated and almost but not quite SMA compatible connectors I can use. Aside from connector and feedline issues, Its actually EASIER to make small stuff than large stuff at high frequencies / small wavelengths. Cable attenuation makes you put the whole RF works at the dish feedpoint above 50 GHz or so, if you want decent performance. The smaller it is, the lighter it is, more or less, making the mechanical engineering job simpler. Its not like 50 GHz amplifier dies are currently the size of dinner plates and will someday be the size of rice grains... they're already tiny. Ditto this chip. Also the silicon is cheap, the tools are expensive. A new ultrasonic wirebond machine must be worth, i donno, tens to hundreds of thousands of cheap MMIC dies? When you buy MMIC dies, its not like they're blowing lots of money on packaging... And thats before you hire the rare skilled labor to set up and operate and maintain the already expensive wire bonder. Wirebonding zero ohm resistors wouldn't really change the overall cost vs wirebonding some fancy dies because of the huge fixed and variable costs of the technology, so changing the die cost from ten dollars to ten cents isn't gonna help if the overall project cost due to R+D and manufacturing and test gear averages out to ten grand per active device...
Secondly complete THZ systems are large and remain large and will probably always be "large". The internal chips are already small, and, frankly, relatively cheap. Antenna cannot be magically shrunk for same performance. Support gear like bias and main power regulators don't "know" they're powering microwave gear and should therefore be shrinking at a microwave pace. DSP processors don't "know" they're connected to a shrinking MMIC die and therefore they should be shrinking at a microwave pace. Support gear does shrink over time at the rate of normal support gear shrinkage, which isn't that fast. For example, not much has changed in the world of linear voltage regulators in the last 30 years... somewhat lower current references, MOS pass transistors instead of bipolar means lower voltage drop, um... thats about it?
Re:Getting tired of this.... (Score:5, Interesting)
To complain that network technology never really improves is the height of absurdity.
Despite what the blurb says, this technology may finally be a good competitor for wired ethernet to the home. It's directional, so it doesn't have to be shared among a huge number of houses, and at $5/pop you can build a "disco-ball" covered with them to blanket an area. It won't penetrate walls well, but will penetrate adverse weather better than laser light.
Short of replacing Comcast, at least we can finally have a wireless HDMI "cable" that is affordable, so I can hook any number of terminals to a computer without having to bunch them all together.
Re:The Future (Score:5, Interesting)
They said the same about broadband: "What could anyone possibly do with 20mbps? They barely use the 56k we give them!"
Give them the bandwidth - they'll find a good use for it. I can see it being very useful in a small/medium server room - 30Gbps makes it a competitive LAN system. Having a bunch of wireless cards would be much easier than running all that cable, even if some manual aiming and orientation of antennas is necessary.
I also imagine "the cloud" would benefit from this - even 1.5gbps is basically SATA speeds. Latency is higher, but the potential throughput gains are impressive. That may make it possible for "local storage" to be "operating system and cloud sync software", with everything being server-side somewhere. You and I may not join in (I don't like the privacy most of the cloud has), but many people don't give a shit about that.
Gaming might also benefit. Current online gaming depends a lot on synchronizing things, then letting the clients do a lot of the calculation. Updating the position of falling objects is almost always client-side, with the server checking every once in a while. It's a major headache, code-wise. With a suitably massive pipe, it becomes unnecessary - just send the coordinates every frame.
Or it makes video streaming work properly. Dealing with current streaming is rough on networks, as it needs to get there quickly. 30gbps to the home, and you can download an entire blu-ray, uncompressed, in two seconds. Latency can be looser - nobody's going to complain if it takes three seconds instead of two. There was an article on /. about that a couple months back.