Receive new posts as email.
This site operates as an independent editorial operation. Advertising, sponsorships, and other non-editorial materials represent the opinions and messages of their respective origins, and not of the site operator.
Entire site and all contents except otherwise noted © Copyright 2001-2009 by Glenn Fleishman. Some images ©2006 Jupiterimages Corporation. All rights reserved. Please contact us for reprint rights. Linking is, of course, free and encouraged.
Atheros thinks single-stream 802.11n has potential to replace 802.11g: Atheros has introduced the Align, a family of chips that use a single antenna to bring some 802.11n advantages without the spatial multiplexing, improved receive sensitivity, further transmit range, or antenna diversity, among other characteristics. The company told EE Times that they wanted to get beyond 802.11g for future devices to bring the advantages of newer designs. This should allow G prices with some improved N features.
This won't break 802.11n compatibility, as 802.11n can hear a single spatial stream just as well as it can multiple ones. In fact, 802.11n provides the flexibility to have multiple streams sending the same data redundantly, which is what Quantenna has opted to do with its consumer gear--sacrificing raw speed for resilient performance.
Atheros is claiming 50 Mbps in TCP throughput with 20 MHz channels and 107 Mbps with 40 MHz. This isn't out of line with the base raw symbol rates in 802.11n (65 Mbps instead of 54 Mbps). TCP throughput still has overhead, of course, so it's likely that single-channel N will be about twice as fast as the 20 Mbps or so 802.11g could achieve.
The folks at Quantenna made a splash with their "1 Gbps" Wi-Fi announcement today: Venture-backed chipmaker Quantenna says that they have a tiny chip that should make it easier and cheaper to push high throughput Wi-Fi around a home using wall-outlet adapters. The company claims 450 Mbps of throughput from the highest-end Draft N standard (600 Mbps raw), and that it has a 1 Gbps wireless offering that uses multiple bands and channels to achieve throughput. There's not enough detail to know how proprietary that is, or if it's a form of channel bonding.
Quantenna announced three chipsets and a reference design: simultaneous dual band at raw rates up to 1 Gbps, 5 GHz at up to 600 Mbps, and 2.4 GHz at up to 450 Mbps. The reference design is for a compact wall outlet Wi-Fi extender.
The company said it's using a proprietary version of the 802.11s mesh protocol to allow devices to interact with each other. Quantenna's focus appears to be on spreading signals across a house, such as with streaming high-definition, where lots of bandwidth will be needed as telcos, satellite operators, and cable firms deliver HDTV into homes today, but plan much more in the future. Storing HD and then being able to have multiple live streams sent among devices is apparently the wet dream of those involved in home entertainment.
You can be clever about pushing HD around a home (like Ruckus) or brute force it by flooding an area with high throughput like Quantenna, which isn't a bad strategy, but it's an interesting one. The fact is that there are already market solutions that don't require 450 Mbps of net throughput. The segment they're looking at seems too well developed and small for them to capture a sizeable chunk when products based on their design are released in mid-2009. And as a startup, their ability to sign deals with firms that sometimes take 1 to 2 years to negotiate and sign makes me wonder; their investors might be brokering those deals to make them conclude faster.
Small, integrated chips make a big splash because they reduce the battery drain on mobile devices, allow the use of these chips in handhelds, and can dramatically drop the cost of manufacture both through a reduced bill of materials and reduced assembly costs. Quantenna told several sources that they expect to charge $20 for a single-band chipset and $40 for a dual-band chipset in quantity. For chipmakers these days, that can mean from 100,000 to 1m before the price drop happens. (It used to mean much more, but efficiencies have improved in smaller lots of chipmaking, apparently.)
I've followed chip announcements in the Wi-Fi space for years, and small startups that have unique offerings tend to either get swallowed up in short order (Airgo into Qualcomm) or disappear (the very promising Engim\). Atheros, Broadcom, Qualcomm, Texas Instruments, Marvell, CSR, and a few others own the market, and that's just how that is. Chipmakers in this industry segment needs millions and then tens of millions of sales to make it possible to recover their R&D costs while sinking money into future R&D for the inevitable next generation.
(Airgo, I might note, was sucked into Qualcomm and sunk without a trace, although it's likely their patents were part of what was of interest; their approach to building MIMO systems was probably integrated into other product lines and multi-standard chips.)
Atheros has released an open-source version of the driver software that talks directly to its chips: The company has long maintained that it required a closed HAL (hardware abstraction layer) to prevent rogue developers from changing settings in its Wi-Fi chips that would cause the chips to perform activities that were against its interest. For instance, it's a/b/g chips can use the 4.9 GHz band, which is illegal in the U.S. and many other countries, but allowed in Japan.
Those objections must have been overcome, as the firm is providing a full, ISC-licensed free software code base for their HAL for 802.11a/b/g chipsets. This should allow the ath5k project to create a fully Linux kernel integrated driver for Atheros chips with no reverse engineering or licensing issues.
This opening up of the HAL allows laptops and handhelds running versions of Linux to have more effective use of the Wi-Fi adapters built in or that can be added on. Note that Atheros hasn't opened up its 802.11n chips yet.
This HAL isn't the same as the one used by the Madwifi project, headed for several years by Sam Leffler. Leffler was able to start Madwifi up by signing an agreement with Atheros that let him write a binary HAL that could be released alongside open-source or free drivers. Leffler reiterated a few days ago on a mailing list that his HAL still wasn't available for release. And, at this point, the Madwifi project appears to be deferring to the ath5k folks. (Confusingly, information about ath5k is all noted at and accessed via links on the Madwifi site, but it's a separate project.) [news via Thomas Gee, Canard WiFi]
Upping the ante for mobile devices, Atheros offers a series of chips that consume almost no standby power: In recent years, every new chip design for mobile devices focuses on three factors: integration, or the number of features backed into one chip to reduce the cost, form factor, and power use of multiple chips; size; and standby/idle power. That last can be the killer. You can have tiny chips, but if they pull several percentage points of the in-use power to maintain status on a network or scan for networks, it's hard to get out of the gate.
With less power consumed, the longer lived a mobile device is, and the more likely a manufacturer is to design high-bandwidth uses. Atheros's AR6002 series (single-band g, dual-band a/g) consumers what the company calls "near-zero standby power," and 70 percent less than competing offerings in active mode. Their two examples are that the chip could be used on a standard phone to provide 100 hours of VoIP or download 200 GB of data.
Chips will ship in quantity in the first quarter of 2008.
The U.S. International Trade Commission bars imports of newer handsets containing Qualcomm 3G cell data chips: This ban stems from a patent dispute with Broadcom, in which the commission found that Qualcomm infringed on Broadcom patents. Handset models previously imported may continue to be brought into the country from overseas manufacture. However, no chips or modules containing these chips, nor any device released after June 7 that contains Qualcomm chips may be imported. Qualcomm also must halt some domestic activities, too.
This should not affect Apple's iPhone, which uses so-called 2.5G EDGE technology that doesn't appear to be affected by this decision. Apple may have, in retrospect, had a stroke of luck by not including UMTS or HSDPA, GSM flavors of third-generation (3G) cellular data networks that might have wound up using Qualcomm chips. (W-CDMA, while a GSM standard, contains technology patented by Qualcomm; Qualcomm also makes UMTS and HSDPA chips.)
While Qualcomm has little impact currently on the Wi-Fi market, they have patents and technology that cover all major third-generation (3G) cell phones data networks and handsets. Disputes have arisen in the US and Europe over Qualcomm's extent of claims of what technology they control through patents, and their licensing fees. Broadcom and a number of handset makers have a variety of lawsuits against Qualcomm and Qualcomm against them.
Qualcomm purchased Wi-Fi chipmaker Airgo, the earliest mass developer of multiple-in, multiple-out (MIMO) antenna technology to supplement 802.11 specifications; and has staked out contrary positions around mobile WiMax, initially completely opposed to it and waging a propaganda war against it, and later purchasing a firm that had WiMax equipment in its portfolio.
President Bush can overturn this order.
Atheros announces fast, two-radio gateway, USB adapter, revised single-radio gateway: Chipmaker Atheros announced today that it has dramatically expanded the variety of its Draft N reference designs to include the smallest form factor USB 2.0 after-market adapter and two new routers, including a dual-radio access point that can achieve 400 Mbps in aggregated TCP/IP throughput. Reference designs are licensed to manufacturers which modify and package them as unique products.
Atheros faces sharp competition from Airgo, Broadcom, and Marvell in the general market for providing Wi-Fi chips to manufacturers of consumer and enterprise equipment - the so-called OEM or original equipment manufacturer - and additionally from Intel in putting Wi-Fi into laptops. Intel would prefer its computer-making partners buy the whole Centrino Core 2 Duo shebang from them, Draft N chips included. These new designs are clearly aimed to ensure Atheros's manufacturing partners have the largest range of possibilities with the least amount of independent engineering.
In a briefing last week, Atheros's vice president of marketing Todd Antes said the firm sees the inflection point for Draft N products outpacing 802.11g products coming by 2008 as consumer products with Draft N become less expensive and more available, along with integration of Draft N adapters in notebooks and computers. "It's no longer just the early adopters," Antes said, who use Draft N.
Atheros designs Bluetooth chip aimed at PCs: Most Bluetooth chips used in computers are repurposed from mobile applications, Atheros claims. Their new product is more efficiently designed with a lower cost of goods and integrated flash memory.
Also features the Solid Gold Dancers: Broadcom said that they will offer a single chip with Bluetooth, Wi-Fi, and FM radios on board. The chip uses a 65-nanometer (nm) CMOS process, which means its circuits are tightly packed using the most common manufacturing techniques. Size has a relationship to power requirements. The Wi-Fi is a/b/g; the Bluetooth 2.0+EDR with 2.1 upgrades possible.
Update: CSR on Feb. 7 also announced a Wi-Fi, Bluetooth, and FM converged chip platform. The company released specific throughput figures, rare in the industry, noting that Wi-Fi by itself could achieve 23 Mbps in their chip designs, and Wi-Fi and Bluetooth together using "collision detection logic" would drop Wi-Fi down to 18 Mbps of net throughput.
On Feb. 7, Texas Instruments also announced a triple-threat, this time with 802.11n.
Broadcom announced a family of 802.11g Wi-Fi single-chip systems designed for mobile devices: Without being an electrical engineer, having a lab, and perhaps manufacturing products, it's tricky to evaluate power and performance claims on the chip level as made by wireless chipmakers. Broadcom states that their latest device has the best power performance, best coexistence with Bluetooth, and improved radio sensitivity compared to competitors' offerings, and their own previous options. Broadcom says that their software architecture controls power at every stage of data transactions, using just 270 milliwatts in active mode, which they state is the lowest in the industry.
With a 50-square-millimeter footprint, Broadcom expects the chips could be embedded into the smallest Wi-Fi devices or be part of low-power modules. The single chip includes the radio, baseband, computer interface (media access control), and power management circuitry.
Swing the noisemakers, folks: Yes, this is just a typical industry press release. But it's also a good milestone to mark. Broadcom was the first company to ship production 802.11g in advance of the ratification of that standard, and has now shipped 100m 802.11g chipsets. With that mark in mind, it's likely that we must be over 500m 802.11 chipsets of all kinds from all vendors.
Texas Instruments introduces a Bluetooth chip that rocks, dude! Its BlueLink 6.0 platform couples FM radio reception (mono and stereo) with Bluetooth in a single chip. This module also co-exists neatly with Wi-Fi. The notion is that a handset could be an FM tuner without additional chips or integration; this feature must be a top request as music players are added into phones. One analyst predicts 400m units with FM reception by the end of the decade.
Of course, if I put on my other hat, I know that HD Radio, a digitally encoded form of AM and FM radio, has begun making headway in the market. HD Radio uses unused guard bands around the primary analog frequencies to deliver crisp, even multi-channel audio. It makes a lot of sense in about two years to have HD Radio-only AM and FM tuners in handsets. About 700 stations broadcast HD Radio today and only a few car receivers, one high-end home receiver, and one tabletop radio can tune in these broadcasts. A few thousand stations will have added HD by 2007.
The platform works with all common cell phone standards (2G through 3G), as well as Linux, Microsoft, and Symbian operating systems. The chips in two modules are in sampling with devices expected in early 2007 based on the technology.
It's not as big a move as IBM/Freescale to Intel, but it's a shift, nonetheless: Broadcom scored an early trifecta with 802.11g back in late 2002 and early 2003 by signing Apple, Belkin, and Linksys for a round of 802.11g-based products. They also swept in Buffalo and several other firms (notably missing D-Link and NetGear) in that heady run-up to 802.11g ratification.
In the latest Apple products, the first to be based on Intel processors using the Core Duo chips, sources outside of Apple told me that Atheros chips have been incorporated: it's true, but Broadcom hasn't been abandoned. Both Atheros and Broadcom chips are specified in Apple documents and are shown in FCC filings.
It's not odd that with a new system architecture Apple would have reviewed chip suppliers, and they may have chosen to work with both Broadcom and Atheros to have competition for their business. There's a limited number of PCI Express-based Wi-Fi chips, which is what the internal, included AirPort Express hardware uses.
The MacBook Pro (the PowerBook replacement) and the Intel-based iMac support 802.11a for the first time, as well. Apple isn't emphasizing the 802.11a inclusion, and the technical specifications only say "802.11g standard."
Although Steve Jobs declared 802.11a "dead" back in Jan. 2003, it was clear he thought it was a non-starter in the consumer market, and the enterprise was far from a win. In Jan. 2006, 802.11a's place as a larger spectrum swath without legacy slower equipment as a way to run more dense, faster enterprise networks and handle campus-wide VoIP is pretty clear. Apple adding 802.11a lets them sell more easily into enterprises and academia that are adopting 802.11a.
One rumor cited by AppleInsider is that the demonstration of the MacBook Pro's built-in iSight video camera was carried over 802.11a to avoid conflicting with the many ad hoc 802.11b networks running at the keynote venue.
Broadcom has declined so far to provide any non-licensed access to its Wi-Fi chips: A project that has been working to reverse engineer access using legal means has released its first working drivers for Broadcom 4300 series chips. The project requires the use of the SoftMAC software as well to compile working drivers within Linux. The first successful use was documented in email Dec. 4 to the developer's mailing sent from a PowerBook running Linux with the project's drivers installed.
Atheros has allowed a third party to create a layer between the low-level functions of its chips and high-level drivers. The madwifi Hardware Abstraction Layer (HAL) prevents developers from having access to most of the radio functionality, which would might allow use of frequencies that aren't legal in particular countries, use of encodings that aren't allowed, and other regulatory problems.
The Economist magazine ran an article early this year critiquing the timidity of Atheros and Broadcom, noting that "if the firms are really worried, they could release most of the interface, keeping back those features that are legally sensitive." Neither Atheros nor Broadcom speak much publicly about this matter. [Link via Jim Thompson]
Atheros's latest Wi-Fi chip includes all 802.11g access point features in a single piece of silicon: They say it drops the component count 40 percent over its previous chipset. The cheaper ($12.50 each in quantity), smaller, and lower-powered these chips become, the more likely that APs shrink (they're still huge) or are found built-in to more equipment.
The company is working with Spansion, a flash memory maker, to stack Wi-Fi components: The new approach of stacking components vertically could allow Wi-Fi to be a layer in a package containing memory, reducing power and space needs. Spansion is a venture of AMD and Fujitsu.
The next-generation Intel chips will use less for more: This answers the conundrum posed by Apple's switch from Freescale/IBM chips to Intel, in which CEO Steve Jobs said Intel's roadmap offered dramatically improved wattage to computational power ratios. Next year's mobile processor will use one-third of its predecessor's power.
This increased performance with lower power requirements makes it easier to produce lighter, longer-lived mobile devices. Intel demonstrated some running a beta of Microsoft's next consumer release, Vista, with Wi-Fi and WiMax connectivity. (Since mobile WiMax is but a glimpse in the future, I'd be curious what actual chips were onboard.)
Intel also said that it would work with Cisco to make better connections with the networking giant's access points, including using special sauce that would allow an adapter to connect to the most available AP by load (rather than the most opportunistically available), and handle VoWLAN connections more fluidly. Cisco and Intel will have to update respective systems.
Reports are coming in from all over about Intel's breakthrough Wi-Fi chip design: But when you read a technical report, linked here, it's a not-yet-commercial design that simply demonstrates Intel's ability to incorporate 802.11a, b, g, and n within the same sort of flexible chip manufacturing process--CMOS--used for the largest wafer formats and highest yields. It's not that it's not interesting, but it's not yet a big deal given that 802.11n won't be finalized until what's looking like early 2007, and other chipsets already offer a/b/g in CMOS at low power.
Atheros will ship a/b/g and b/g chips for PCI Express in third quarter: This next-generation bus design supports a much improved architecture for maximum throughput across all cards, and Broadcom and Atheros are both interested in being on top of its deployment. Atheros says that they have a single chip solution that integrates into a single-sided PCI Express card; sampling is already underway to its best customers.
Broadcom announced in early April that they have a PCI Express chipset--ostensibly at least two chips--that's was in sampling then. I expect a war of the words over throughput, cost of goods, and other factors in the months ahead.
Broadcom claims first Wi-Fi chipset for PCI Express: PCI Express is the next next generation of motherboard-based expansion technology with a higher-speed switched bus architecture for better throughput. PCI Express (PCIe) can push 1 Gbyte/s of data; even the fastest proprietary flavors of 802.11 support about 3 to 4 Mbyte/s (30 to 40 Mbps), so it's not a stretch to have multiple cards in a single PCIe chassis, even. The Broadcom chipset is sampling now.
The Economist writes a call to open access to Broadcom and Atheros's radio technology for greater innovation (subscription required): The writer argues that by keeping their lower-level radio functions and any access to it close to their vest, they're discouraging wider uses of their chips and suppresses interesting projects from CUWiN and community wireless networking groups.
While the two companies produce Wi-Fi chips that don't use formally use SDR, they have aspects of SDR that make their concerns about opening up full control reasonable. And The Economist only suggests that more access than zero would be worthwhile. There is the Madwifi project which involves one programmer who was given access to the RF innards to write an intermediary, proprietary bridge between open-source drivers and the Atheros chips. But that's a pretty limited exposure.
Linux developers ask me all the time: when will Broadcom provide even that support? Perhaps The Economist's prod will cause both companies to think about how to sell more chips without incurring the FCC's wrath.
Mark Rakes notes that there's already an active thread discussing the article over the madwifi newsgroup.
Update: I want to clarify previous remarks a bit. From more technically minded types, I'm reminded to mention that the SDR that Broadcom and Atheros use doesn't allow access to all frequencies, as true SDR has the potential to do. Rather, it's SDR in the sense that there are several frequencies ranges, including both licensed and unlicensed, in certain chipsets.
Atheros and Broadcom should try to strike a balance in offering an abstraction layer which provides mediation so that open-source work could be built on top of it that still conforms to Part 15 rules but has a greater degree of flexibility than the current Madwifi project--and would allow any Linux use for Broadcom chips.
Another update: Sascha sent the link for the paper on which parts of the argument in the Economist argument are based, which he and two colleagues co-authored and delivered at a conference in Sept. 2004. I disagree with their argument that FCC sanctions a strawman; they can't be privy (nor can I) to the non-public aspects of working with the FCC and the issues surrounding partial SDR that might be part of the backstory to this issue.
Atheros will ship its sub-$13 chip in Q1 2005: The all-in-one chip includes everything a wireless access point manufacturer needs, making it easier to embed 802.11g into other products or produce even cheaper gateways that have a full range of features and performance. The chip includes Super G, a set that mixes proprietary and future 802.11e extensions to improve throughput; Atheros's distance-enhancing XR technology; and 802.11i with full AES encryption.