Your browser is not supported. Please update it.

4 June 2008

Intel latest to distance itself from WiMedia, as wireless HD market heats up

An Intel-backed start-up, Ozmo, is the latest to throw the cat among the pigeons in the personal area networking (PAN) market, pushing Wi-Fi into applications traditionally dominated by Bluetooth. And while Intel has been the greatest advocate of basing next generation PANs on UltraWideBand, and particularly the WiMedia standard it supports, it now seems likely that it will put this platform on the back burner, at least for now, as its partner Texas Instruments has already done.

While Wi-Fi is seeking to eclipse Bluetooth in short range applications like PC-peripheral connect, its major focus is on high speed home media networking, a market also targeted by WiMedia. In this sector of the PAN market, where the holy grail is a wireless equivalent of HDMI, there are several contenders emerging, including WiMedia, other UWB systems like Pulse~Link’s C-Wave, the 60GHz WirelessHD, and Wireless High Definition Interface (WHDI) running in 5GHz spectrum like Wi-Fi.

The problem for WiMedia – which is based on the Multiband OFDM technology originally backed by Intel and TI, and has received significant input from Sony and Philips ‘ is that so far it has failed to live up to its performance promises. While it has been adopted as the basis for the next generation Wireless USB system, in the home media networking sector it is not supporting the speeds that HDTV and other applications demand. Hopes that WiMedia would be a common physical network underpinning a wide range of short range wireless standard protocols have been dashed by the lacklustre results of early trials. The Bluetooth SIG changed plans to base the next iteration of its standard on Wi-Fi 802.11n rather than WiMedia, and recently TI itself said the UWB technology, in which it owns significant IPR, would not make an impact on its key mobile device market for at least three years, and put it on the back burner in favor of 802.11n (see Wireless Watch May 7 2008).

Now it seems that Intel may be the latest UWB cheerleader to bring 11n to the forefront and put WiMedia product plans on hold, at least for the time being. It has revealed details of a development program focused on putting Wi-Fi into mobile devices and PANs where Bluetooth lives now, and where WiMedia had expected to fit. It has been working in this roadmap with Ozmo, which will this week unveil its chip and software. These promise to support wireless peripherals and short range connections with more bandwidth and lower power for less cost than Bluetooth.

Ozmo takes advantage of the incorporation of standard Wi-Fi into many mobile devices from notebooks to media players and handsets, and promises peripheral makers a stripped-down, dual-band 802.11 chip with a 10-meter range that costs about the same as a Bluetooth chip, and will communicate with any of these devices. The 130nm Ozmo chip measures 6mm x 6mm and features a simplified receiver that does not require support for roaming or scanning to link to an access point.The chip uses Wi-Fi Protected Set-up to establish a 9Mbps link over 2.4GHz or 5GHz bands to any client running the Ozmo Wi-Fi driver software. This fits into existing Wi-Fi and USB software stacks as an additional driver layer. Battery life is two to three times that of Bluetooth ‘ up to nine months for a mouse, or 20 hours’ talk time (compared to four months, or six hours, with Bluetooth).

Ozmo does not yet have any announced customers though Belkin is testing its products and, of course, it is working with Intel ‘ as part of the chip giant’s Cliffside research program, described at the Intel Developer Forum in April. This aims to use standard Wi-Fi protocols to handle PAN tasks such as syncing notebooks with MP3 players, digital cameras and projectors. Intel claims its technology can connect up to eight Wi-Fi devices to a notebook on a PAN while the computer is on a Wi-Fi Lan. This could improve the quality of media streamed between a computer and a TV, because it eliminates the latency of going through an access point. Cliffside should result in a commercial chip – a modified Intel Wi-Fi chip with additional buffers to switch between PAN and Lan modes ‘ within a year, and will be available on Centrino.

As well as pushing into ultra-low power markets, Wi-Fi’s credentials are increasingly being established in high speed media networks, as the 100Mbps-plus 802.11n nears finalization, and as vendors, even at the draft standard stage, start to slash prices of consumer and embedded gear. But the home high definition network will be a toughly fought battle, and TI summed up the dilemma of many consumer electronics chip and equipment makers, saying: ‘In home networking we have to decide if we are using Wireless USB, 1394 Firewire, or DLNA-on-IP over UWB, Wi-Fi or Bluetooth, before we can make any headway.’ Demand for wireless connections between HDTVs and audio-visual or PC equipment is rising, and will be driven further by the recent emergence of TV designs that separate the monitor from the tuner, enabling very flat screens (Hitachi and Sharp already have screens 1.37-inches thick with external wireless adapters).

Along with Wi-Fi and WiMedia, the more cutting edge contenders to carry any of the high speed wireless PAN protocols are starting to show their colors, all claiming to deliver the magic hat trick ‘ high transmission speed (at least 80Mbps for compressed video, up to 3Gbps for uncompressed 60 frames/sec video at 1080p resolution); low latency (supporting channel change at less than 100ms); and high image quality, even with the interference and obstacles common in a home environment.

Pulse~Link’s implementation of UWB, which uses a traditional pulse-based approach rather than OFDM carriers like WiMedia, has performed far better than its better supported rival in early trials, and also runs over coax cable and powerline as well as wireless.

Meanwhile, WirelessHD is the main group of vendors looking to use the license-exempt 60GHz spectrum, which supports very short range but very high speeds (up to 4Gbps, not far off the 4.95Gbps typical of cable HDMI 1.2 connections). The lead technology developer is SiBeam, and the would-be standard is backed by Intel, LG, Panasonic, NEC, Samsung, Sony and Toshiba.

The other major contender ‘ though commercially proven products, as for WirelessHD, lie 2-3 years in the future at best ‘ is WHDI, whose lead developer is Israeli company Amimon. This technology has reportedly achieved data rates of 3Gbps.

WirelessHD addresses concerns over signal robustness in high frequencies by using two channels ‘ a high rate PHY to transmit an HD image at high data speed, and a low rate PHY to handle control signals between transmitters and receivers, and manage the antenna’s directionality, optimizing signal-to-noise ratio for the high rate channel and adapting to any changes phased-array beamforming technology.

At the Consumer Electronics Show earlier this year, Panasonic and Toshiba demonstrated WirelessHD systems, and the WirelessHD Consortium will conduct protocol certification and interconnectivity tests from this summer. So far, Toshiba, Hitachi, Fujitsu and Sharp have initiated R&D on chips for the platform, though SiBeam and NEC remain in the forefront, and like IBM, have succeeded in addressing a major cost obstacle of 60GHz integrated circuits by implementing them in CMOS rather than gallium arsenide or BiCMOS. NEC will commercialize its CMOS RF transceiver chip for WirelessHD in 2010 and is also working on a baseband/MAC embedded digital IC.

WHDI is even further down the track, as so far only Amimon can make the technology, and is just now forming an industry consortium, preparing to share its technology with partners. The system works in 40MHz channels in 5GHz and can transmit and uncompressed 1080p HD image, as it reaches a peak data rate of 3Gbps. It uses 4×5 MIMO antenna schemes combined with a modulation approach devised by the inventor. This combines Amimon’s own single-level modulation with an existing signal processing technology called joint source channel oding, which applies various levels of error correction, which vary in intensity depending on how recognizable the errors are to the human eye (down to no correction if the errors are not deemed to be detectable). WHDI does not guarantee complete transmission of data, as WirelessHD does, but claims there is no difference between HD image quality using HDMI cables or WHDI.

Sharp is the first vendor publicly to adopt the Amimon chipset, and audio-visual products from Funai and others will debut this summer, but several vendors are said to be planning to join the consortium, spurred by the promise of Amimon licensing its systems for larger companies to design their own chips. Interested parties reportedly include several Japanese, two South Korean and at least one US equipment manufacturer.

Pulse~Link:

For almost eight years now Pulse~Link has had the technology wherewithal to send massive amounts of video from one device to another around the home, either wirelessly or over a wire. If the chips were made in serious volume, they could cost as little as a few dollars (currently between $10 and $20 depending on order volumes) and create a genuine digital home. But its pulse-based UltraWideBand has gained less political clout than rivals such as WiMedia, MoCa, HPNA and simple Ethernet.

Companies like Samsung and Sony and Panasonic, plus all the major pay TV operators like Comcast and DirecTV, as well as companies like Apple, have the need for such a technology, but there have been no major design wins forthcoming, just some smaller ones with Westinghouse Digital and Geffen.

There are two major use cases for such a technology. One is that High Definition TV signals need to travel around a home in the cheapest way possible, from set-top or Blu-Ray player to a TV screen or a PC. The other is simplified rapid handset sideloading of video files from a home network. One requires high bandwidth due to the requirements of the application; the other needs it to allow sideloading to be automatic or last minute (and copying a one-hour TV program to a handset will need 1Gb in SD, and 250Mb in QVGA, which might need to transfer in the time it takes to put on your coat and shoes on in the morning).

Other technologies don’t look like achieving this duo in the near future. Wi-Fi can send video files around a home, and even through a wall, but not every time and not HD files. Ruckus Wireless has improved this to the point where HD files are supposed to be able to be streamed around a home, but that takes the price of each device into the multiple hundreds of dollars, when the price point for putting this in every device needs to be around $2 each.

It may be possible for Wi-Fi to have a more sophisticated quality of service software layer – in fact we know some proprietary Wi-Fi that uses the WiMAX QoS schema – and it may be possible eventually to take the MIMO antenna and the packet resend software and drop them into a cheap SoC. But that isn’t today and given all the interference from which 2.4GHz suffers, there hardly seems any point.

The core complaint from Pulse~Link founder and CTO John Santhoff is that many of his rivals use a contention-based protocol, in that they support CSMA/CD, which is part of the Ethernet. This was designed for a data bus that allows anything attached to it to begin sending at random, stop if it collides with other sends, and then wait a random period before sending again. Contention-based protocols are fine when they are operating at the low end of their capacity, but up at the high end, performance can become unpredictable, which is why today the Ethernet world has shifted to Gigabit, so that it has plenty of capacity.

Santhoff has issues with his other rivals too. ‘HPNA and powerline use 30MHz of spectrum or less on the cable, and to carry more data they would have to use more power, which would push out more emissions, which would break the rules of the FCC spectral mask.’ The issue here is that the home is a hotbed of signals, and that any radio signal can interfere with any neighboring signal if it is too powerful.

Which is the beauty of using a pulse of radio energy across a very wide spectrum. The FCC defines UWB (ultra wide band) as anything which is spread across 500MHz or more. It can be arranged as multiple carrier waves spread in this way, using Fast Fourier transforms to disentangle them, just as OFDM does. This is the direction the WiMedia Alliance took when it defined its version of UWB, based on Multiband OFDM.

Pulse~Link instead puts a single pulse across an entire 1.35GHz of spectrum, sending on all the wavelength at once. The duration of the pulse is around 23 centimeters long, which at the speed of radio (light) is exceptionally short, coded in simple BPSK. The signal can be sent more than once, which improves robustness and distance, but in essence because the data can be picked up from anywhere across the spectrum, if there is any fade or interference, it can be ignored. Sending once gives a native data rate of around 1.35Gbps, sending twice halves that.

The UWB of the WiMedia Alliance suffers from the fact that, because it is arranged in a series of separate carriers, to increase capacity requires modulating them at higher frequencies, and that means increasing total power use, and that means breaking the FCC emissions rules. To Santhoff it’s just obvious and incontrovertible.

On WirelessHD, he says: ‘It uses 36 different antenna arrays, which means that there are 36 individual receivers and 36 analog to digital converters. That’s going to take a long time to integrate and get right, and it will cost a lot, but it’s the only way that this line of sight technology can get around obstacles.’

‘UPA Powerline from DS2 is about the most honest effort out there and we have tested its performance at around 68Mbps in real world conditions, which is 10% to 12% better than any of the other technologies. But it uses QAM 1024 as a modulation which means it has to use a very robust FEC scheme.’

Inventing its own way of transmitting radio hasn’t made Pulse~Link popular but over the course of eight years, the company has convinced a number of eminent people that it is on the right track. The company recruited the top engineers from Freescale’s effort in UWB, XtremeSpectrum, once that was closed down, and it has employed Henning Harmuth, one of the pioneers of modern UWB and the CDMA Walsh codes, as well as the founder of a number of the WiMedia UWB companies and even a WiMedia board member.

While technologists from theWiMedia camp may be open to accepting the pulse-based view, in the short term, politically, they need to continue to stand apart. This is partly because Wi-Fi has a grip in the home already, with over 100m Wi-Fi-enabled notebook shipping annually, and forecasters saying that around 240m Wi-Fi-enabled consumer electronics devices will ship in 2011. So Wi-Fi is protecting its base, even though it cannot send multiple HD streams around the home in its current iteration (802.11n).

But the replacement of communications chips inside 1.25bn TV sets, and in around 400m set-tops, and a similar number of DVD players, is a market that Wi-Fi can only dream of. Never mind the three billion handsets that would be brought to bear if instantaneous video sideloading could be made possible. With his eye on these sectors too, Santhoff insists that Wi-Fi just can’t go any faster.

All this has shaped Pulse~Link’s recent strategy. It no longer expects many people to use its radio technology, but the same CWave chips can now be used with cables, pretty much any kind of cables, but certainly coax and potentially power cables. ‘Our approach now is to be able to deliver HDMI, 1394 (Firewire) and Ethernet over coax on top of our signal, and for that signal to coexist on the same cable with MoCA or a satellite signal,’ said Santhoff. This works for the same reasons that the radio works ‘ it’s the same wide pulse approach as basic signaling, and then capturing pure HDMI, 1394 or Ethernet bit streams and mapping them across the pulsed network.

‘Back in 2002 we started looking at all home network technology and none of them have a silver bullet, and we decided that what was needed was a hybrid which combines two or three media types. Today we have HDMI in and HDMI out, we have 1394 in and 1394 out, the same with Ethernet. But in our labs we have 1394 in and Ethernet out or any other combination and that’s useful if you have a TV output in HDMI and you want to send it to a PC which is used to Ethernet ‘ so that’s where we are headed with this.’

In the end, how can this tiny company (that has soaked up over $80m in VC funding so far) convince the market politically that its approach is right and that, of itself, it’s not a threat to established giants. In our view it only takes one giant. This could be a Sony, to say that it wants a networking system to join up its PS3, its PSP, its Bravia TV sets (with set-tops embedded) and its high speed video cameras and its Vaio PCs – then it has to think about HDMI, Ethernet and 1394, at video copying speeds. Or it could be an Apple, if the iPhone maker wants to fight mobile TV, by offering iTunes sideloading onto handsets for files that are gigabyte class. Apple virtually invented Firewire and could virtually guarantee itself supremacy in copying TV or DVR held content to an iPhone to view on the go, by adopting a wireless or coax connector that can copy the film you recorded last night in under 10 seconds.