networkZONE Products for the week of February 24, 2003
IceFyre Says…
Chip-Tease - IceFyre Samples Pre-Pre-Production Versions
Of the SureFyre 802.11a PHY
IceFyre Semiconductor, Inc., is now sampling of the first member
of its SureFyre WLAN PHY Integrated Circuit (IC) family, the ICE5350 802.11a
PHY IC. Designed to drive the industry's lowest power, highest performance
802.11a and dual-band 802.11a/b/g full-system solutions, the ICE5350 will
be shown in private demonstrations at the Wireless Systems Design Conference
and Expo in San Jose, California, the week of February 24th. Demonstrations
will include wireless end-to-end connectivity at all 802.11a data rates.
"With the SureFyre 802.11a family, IceFyre is setting the standard for performance and low power consumption in the WLAN market," said Allen Nogee, senior analyst, In-Stat/MDR. "With a number of innovative features backed by a strong intellectual property portfolio, IceFyre is in a strong position to address the requirements of PC-based WLAN products and drive emerging audio visual, last mile wireless and WLAN-enabled consumer device markets."
Using patented 5-GHz Orthogonal Frequency Division Multiplexing (OFDM) modem technology, the SureFyre 802.11a Physical layer (PHY) family delivers up to a 75 percent reduction in power consumption, a significant total system cost savings over conventional designs and exceeds all 802.11a performance specifications including Receive Sensitivity and Output power. SureFyre 802.11a PHY solutions anchor 802.11a and 802.11a+b full system reference designs that deliver power consumption, performance and high data rate, multi-cell deployment economics far superior to competitive 802.11a and dual-band 802.11a+b system solutions. In addition, the innovative SureFyre family provides the foundation of a solid product roadmap that includes the industry's lowest power, best performing dual-band 802.11a/b/g and higher data rate offerings.
"Our lead customers and partners are extremely excited about having
the ICE5350 in-hand. This is the first in our family of innovative wireless
LAN IC products," said Dan Mathers, president and chief executive officer
of IceFyre Semiconductor, Inc. "Our core intellectual property and
innovative architecture has now been proven in silicon. We are very confident
that our products will be the industry's lowest power, best performing 802.11a
and 802.11a/b/g full system solutions. Having achieved wireless end-to end
connectivity at all 802.11a data rates with our first silicon spin places
IceFyre in a unique position among current market competitors and speaks
to the skill and determination of our team."
analogZONE Says . . .
Editor's Note: It's both exciting and challenging to have both Envara and IceFyre, two of my favorite dual-band 802.11a/g solutions, making their debuts in the same week. I've been eagerly awaiting both announcements for some time, and despite being overloaded with other projects, I'll do my best to give them both the attention they deserve. You can start here and read all about the IceFyre chip set, but be sure you click over to see what Envara's been up to as well.
I've been waiting for the chance to review IceFyre for some time, and now that they actually have some pre-pre-production (also known as proof of concept) chips available, I have my excuse. The folks at IceFyre have been teasing me with some fascinating details of their unique RF technology that they believe will offer significant price and performance advantages in the 802.11a WLAN market. I've even been to their labs up in Kanata, Canada and had a firsthand look at some of their prototype circuits. But since I normally do reviews on products, and not technologies, I've had to stifle myself and wait until they brought something to market. Although the ICE5350 not precisely a production chip, it validates IceFyre's technology enough that I feel comfortable talking about the products they are planning to introduce late this year and early in 2004.
IceFyre's strategy has been unique from the get-go, choosing to build only the PHY portion of an 802.11 radio (the RF and baseband sections), and leave the MAC to a third-party for the moment. We'll discuss why this may be a very smart move later but, for the moment, we'll concentrate on what makes the chip tick. Their original plan was to make a one-chip, 5-GHz radio in CMOS, which included an integrated power amp. Instead of trying to correct the massive non-linearities that CMOS produces at these frequencies, they envisioned a series of parallel RF amplifiers that generated different pre-distorted outputs which were summed into a clean waveform in a clever on-chip RF combiner. The resulting design also included a proprietary circuit that eliminated the traditional antenna switch and raised output power at the same time.
I was skeptical of the scheme, even when IceFyre let me in on some of the more important technical details, but my partner Paul McGoldrick seemed very impressed. And on the rare occasions when Paul is impressed, I pay attention.
A Bold Decision Yields Superior Results
By the time I had a chance to visit with IceFyre and see their labs this past December, their plan had changed a bit, and they had moved their PA off-chip and fabricated it in SiGe. It turns out that this was a very brave, and a very smart move for a variety of reasons. Rather than holding fast to the common wisdom that an all CMOS, single-chip solution was the ultimate goal, IceFyre decided that performance took precedent over fashion and. And in the process, they have come up with a much better product that will carry a similar, or perhaps lower price tag.
The prototype chip reflects this decision, and the seemed to do everything it was asked to do when I saw the first samples running in the lab a few days after they came from the fab in December of 2002. The GaAs PA chip has two class-B stages driving a class-G switch-mode final stage. The result is a final stage with a power-added efficiency of well over 50% that's unmatched by any other manufacturer. And just as important the switching PA maintains much of its efficiency as its output is raised or lowered away from its optimum bias point.
Of course, there is the obvious reduction benefit in overall power consumption, making the chip a prime candidate for cell phones, PDAs, and other devices whch don't have the luxury of half-pound batteries. But the punch-line bonus of moving the PA off chip is that this allowed the IceFyre designers to reduce the amount of chip area and pins devoted to isolating the 19 dBm PA from the rest of the chip. This led to a smaller chip with fewer leads, less expensive packaging, and fewer decoupling capacitors.
I suspect that most, if not all, of the added costs that were incurred by an external PA chip have been recouped by the smaller chip size and other savings. And when you factor in the dramatic reduction in passive components and board space requirements, I'm pretty sure that the overall "solution cost" will be exceptionally competitive. It's too bad that some other companies felt compelled to stay with an all-CMOS solution which resulted in disappointing compromises in performance with only marginal cost savings.
The ICE5350 does not simply rely on a good radio front end to guarantee its performance. It also uses the digital processing capabilities in the baseband to further enhance reception under marginal conditions. For example, the designers have paid a lot of attention to multipath issues, and made sure their product can deal with up to 150 ns delays spreads, or about 3X what many manufacturers work towards. It's great to see that companies like IceFyre, Marvell, Envara, and Resonext are serious about this issue. Even though it is difficult to quantify its value precisely, a receiver's ability to cope with delay spreads will greatly affect its range and performance in challenging real-world environments.
The wizards in Kanata have a few other tricks up their sleeves for improving performance, including something they call TrueSygnal spectrum analysis. This is a full channel equalization scheme that provides equalization for all carriers instead of equalizing only across the 4 pilot tones like many companies do. The full-channel scheme provides enhanced channel estimation based decoding that improves receiver sensitivity provides better performance in the presence of fading.
The production chip will also employ dynamic channel quality assessment, a feature that "sniffs" the airwaves to determine what's already on the channel. This allows the transceiver to actively avoid interference by lowering power or using another band. Access points that are part of larger networks can also use this to assess field conditions and take appropriate actions.
And while I don't want to provide an exhaustive list of its features, I'd be remiss if I didn't mention its ability to switch antennas in a per-packet basis. The chip's "IcePick" feature looks at RSSI for each antenna during the short symbol pre-amble and picks best antenna for that packet. While useful for PC cards and other terminal devices, it will be an even more important feature in access points and other applications where a radio might be talking to different terminals with each packet.
Segmentation Issues
IceFyre's strategy to offer a PHY-only solution is a double-edged sword. On one hand they are dependent on a third party for a MAC, something that could make them vulnerable unless they can team with a good partner for a reference design. On the other hand many applications will eventually absorb the MAC function into another ASIC (such as a broadband gateway or computer South Bridge chip) anyway, leaving many stand-alone transceiver makers with too much silicon in their chips.
While IceFyre is diligently working on offering a stand-alone MAC for its PHY, they have stressed system flexibility - i.e. the ability to work with other people's MACs in existing ASICs. To this end, they present a simple interface to the chip with an 8-bit parallel data path, 6 discrete control lines, and a SPI-bus based control channel. To make data transfers as efficient as possible, there are provisions to support both single byte and burst-mode transactions.
I think that as the market matures, most of their high-volume customers will supply their own MACs, but IceFyre is wisely working with several IC houses to develop commercial MAC solutions. Hitachi has publicly announced it's working on a family of MACs, intended for embedded and AV markets that will have both MIPS SH-4 and ARM-9-based products. Other products in the works include IP for ASIC-based solutions. At some point, I'd also expect to see IceFyre develop or license a MAC core, and perhaps even a microcontroller for stand-alone 802.11 applications in products which don't have a host processor of their own.
The only thing to remember with all of this is that what we are seeing today is proof-of-concept hardware, and not even pre-production chips. This means that we won't see pricing for a while, most likely until their pre-production 802.11a-only PA and PHY chips arrive in Q2 and Q3 of this year respectively. But by the time the production a/b/g solution begins to roll off the assembly line in Q1 of 2004, I imagine that their low passive count will help control their overall BOM and implementation cost enough to make them a very serious competitor.
It's nice to know that while production chips are nearly a year away, the proof-of-concept ICs that are here today verify that IceFyre's unique RF front end works, and works well. This, and their aggressive roll-out of reference designs and heavy-duty development tools for both 802.11a-only and a/b/g products should put you at ease. While I am always a bit cautious to endorse parts that are pre-announced, the presence of working prototypes, plus IceFyre's openness in sharing its technical data with me indicate that they are doing all the right things to bring their products to market on-schedule.
The good performance of their working prototype chips, and their extensive design verification program tempers IceFyre's extremely ambitious technical goals and long lead time to yield a surprisingly low Vapor Index Rating.
| |