01 March 2010

The Lessons of Wi-Fi #7: You Don't Need Unobtainium To Build Great Wi-Fi Products

Introduced in July 1979, the Zilog Z80 was an 8-bit microprocessor that operated on 1, 4, 8, or 16-bit data, had a 16-bit address bus, generated its own RAM refresh signals, and would run programs originally designed for Intel’s 8080 CPU. The flexibility of the design made it suitable for a very wide range of consumer, industrial, and military applications spanning from the Tandy TRS-80 computer to programmable logic controllers to naval weapon systems. Prices fell as volumes rose, and the Z80 was one of the most popular 8-bit CPUs for many years following its original introduction.

One of the wonders of semiconductor technology is that a standard part like the Z80 can find its way into so many different applications. The very same CPUs, memories, amplifiers, voltage regulators, and/or transceivers found in consumer products in your home might be found in automobiles, office equipment, factory production lines, airplanes, or ships. What differs is how the part is applied, packaged, and tested. In other words, you don't always need custom parts made of unobtainium to perform specialized tasks in demanding environments.

What happened to the Z80 in the 1970s is happening today with 802.11n chip sets. Chip set vendors are designing a common set of 802.11n parts for use in enterprise, SMB, gateway, and home access point and router products. Doing so drives up the volume of sales, resulting in production economies that boost profit margins for chip vendors even as prices fall for end users.

One of the largest Wi-Fi chip vendors – Atheros – sells its AR9002AP-4XHG chip set for all of the above referenced applications. The chip set features extensive component integration, a small form factor, and low overall cost. The fact that the AR9002AP-4XHG finds its way into such a diverse range of applications speaks volumes about the potential flexibility and robustness of the design. I say potential because whether the objective is realized or not depends on the implementation of the final Wi-Fi device.

Those who forget the lessons of Wi-Fi are doomed to repeat them. Lesson #7: you don't need unobtainium to build great Wi-Fi products.

Just as naval weapon system vendors leveraged a common Z80 design to create very unique and rugged products, so, too, has Aruba leveraged an 802.11n chip set targeted at a broad market in the design of its unique AP-105 802.11n Access Point. The AP-105 was tailored to demanding enterprise applications, and special care was taken in the design of the packaging, antennas, power supply, and security features to make the product both robust and exceptionally fast. A great AP, with a great standard 802.11n chip set, selling for a great price.

The result is an enterprise-class 802.11n access point that has higher throughput and more features than Cisco access points,
yet sells for roughly 40% less money. So much less that Cisco felt compelled to pull apart the AP-105 to find out what makes it tick (they did the same when Aruba's high-end kick-ass AP-125 802.11n Access Point was released).

Their conclusion? The AP-105 is unobtainium-free and therefore no better than a consumer product. You know, like that
cell phone you rely on for emergency calls 24x7, or that iPod that has delivered faithful service every day at the gym. Comparing the reliability of the AP-105 to that of a consumer product is not an insult. At the end of the day, Cisco still has to explain why the AP-105 is faster, more feature rich, less expensive, and easier to install than its own run-of-the-mill, over-priced, unobtainium-based access points.

So with the wind of good design at our backs, and unobtainium nowhere to be seen, the AP-105 is flying off the shelves, charting a path the Z80 followed.