News

Broadcom "misread 802.11g specs" says Texas Instruments

There is no need, say technologists at Texas Instruments, to have fast Wireless LAN nodes slow down to legacy speeds if a legacy user moves into the area. And the fact that current 802.11g "preliminary" chips from Broadcom have this defect, is Broadcom's fault for jumping the gun.

Today's Broadcom-based 54g wireless LAN products definitely do slow down from 54 megabits to 11 megabits each time they see a slower 802.11b packet. But this, say angry WiFi Alliance members, is not because the 802.11g specification is meant to work like this.

"You incorrectly assert that in a mixed mode network consisting of 11g and 11b nodes, that 11g devices will have to revert to a maximum of 11mbps," commented Bill Carney, business development director for Texas Instruments, in response to our recent analysis - "But this is completely incorrect."

Carney told NewsWireless Net that it is understandable how this can be concluded - because this is the way Broadcom-based systems from Buffalo and Linksys and D-Link operate.

"Your conclusion is based on your experience with products that use the faulty, pre-11g implementation of Broadcom's 54g chipset," said Carney, who subsequently made it clear he was not speaking on behalf of the WiFi Alliance, even though he is a board member there.*

The inability of the WiFi Alliance to police this standard has caused angry letters from customers. Typical of the response was this, from reader, Chris Koveleski: "The WiFi alliance are a bunch of morons. The 2.4 Ghz bandwith is overcrowded with cordless phones, micro communicators, microwave ovens, and now BlueTooth. Look at the roster of members to find out the reason why their incompetence is only exceeded by their ignorance," he said angrily.

Clearly stung by the accusation that the WiFi Alliance has failed buyers, Carney accused Broadcom of developing their own chips using an early version of the draft standard, and making "a key interpretation mistake" as to how a mixed mode network is to behave.

"In fact, the IEEE 802.11g committee spent numerous hours discussing the ways to preserve the performance advantages of the higher rate devices, in presence of legacy 11b devices," Carney explained. "Several mechanisms were included in the draft specification to allow this. It's unfortunate that Broadcom's incorrect implementation is now setting a market expectation that one must upgrade the entire network to 11g, else suffer the performance fallback to existing speeds."

Other chip vendors - who have interpreted the draft specification correctly, says Carney, offer performance advantages for 11g nodes in a mixed network. "Because of this and other interoperation issues, the 54g technology is quickly being dropped in favor of other solutions. In this case, being first to market with a pre-standard offering ended up doing more harm than good."

TI is still smarting from accusations of being "late" with a product offering when it announced prototypes back last October - it felt it was being ambitiously early, and that attempts to market the product should be delayed even further.

Its own products in this arena include one single-chip solution, which plays in all three current technologies; that is, 11 megabit at 2.4 GHz, 54 megabit at 2.4 GHz, and 54 megabit at 5 GHz bands - roughly, 11b, 11g and 11a standards under IEEE 802.11.

The TNETW1130 solution is capable of handling 11g users at full speed even if serving packet requests from 11b users, says the company. Samples are now being shipped to builders of wireless systems.

An excellent interview with Carney on the subject of 802.11g and pre-release standards was published by Hometoys.com recently.

* In earlier editions of this story, we stated that Carney was speaking on behalf of the alliance; he has asked us to make it clear that this isn't the case. Photo supplied by Texas Instruments publicity department.