Performance Matters -
Lessons On 802.11 Radios And Customer Satisfaction From The Saga Of Madman
Muntz
by Lee Goldberg
With all the promise the wireless LAN market holds, it is in grave danger of killing its own prospects by delivering poorly-performing products that do not even come close to meeting the range and rate specifications of the 802.11 standards. And the sad part is that this failure is often preventable if chip makers and OEMs valued performance enough to add as little as $0.23 to their BOM.
We can learn a lot about taking cost-cutting to extremes from the wild saga of "Madman Muntz" and his famous television sets. Earl Muntz, a true-life legend in the 1950s and 1960s was a colorful character who, amongst other things, co-developed the 8-track tape system with Bill Lear (of Lear Jet fame), an exceptional production hotrod called the Muntz Jet, and was an early pioneer in the concept of cheap electronics. Besides the 8-track tape players that adorned the dashboard of almost every car in the early 1970s his most famous contribution to the consumer electronics world were his incredibly low-priced TVs, which cost under $200 back in the early 1950, when everything else was at least twice the price. As we shall see, however, the low price tag was made possible by sacrificing other, less obvious things.
Without going into painful detail (which you can get from an excellent column by National Semiconductor's Bob Pease - What's all this Muntzing stuff anyhow?), Muntz's key to success was a ruthless war on the TV's bill of materials. He would quite literally walk around the lab and clip out resistors, capacitors, and vacuum tubes he believed to be unnecessary from prototype circuits, and then challenge the engineers to make them work again. Once the circuit was back in operation, he'd clip a few more components again, and repeat the exercise until the engineers couldn't revive the circuit.
While quite effective at cutting the cost to the consumer dramatically, the Muntz philosophy lead to some clever, if not downright bizarre design trade offs that compromised the set's performance and serviceability. For one thing, his TVs had very poor signal sensitivity and would only work in areas close (10-20 miles) from the transmitter. Of course, the millions of people in metropolitan areas did not much care about this, but a few other surprises awaited owners of Muntz TVs as their sets aged and began to need adjustments. It turns out that the tuner section substituted a factory-set resistor for the normal fine tuning circuit, requiring an expensive call from a repairman as the values drifted over time. Similar problems plagued other sections where hand-selected resistors were inserted at the factory to save the cost of a potentiometer.
In the same way that Muntz's customers got a TV set that had a low up-front price, but had to contend with its hidden costs, I worry that many people who purchase 802.11 wireless LAN equipment will pay dearly for a few pinched pennies on the manufacturer's end. We're already seeing this phenomenon with the return rates of Wi-Fi equipment at Fry's and other electronics stores running as high as 40%-50%. While I'll guess that some percentage of these returns is due to the poor excuses for installation software provided by so many manufacturers, I'm still pretty sure that much more than half of the product comes back because the damned radios don't reach from one end of the customer's house to the other.
And the appalling part is that we're even seeing this problem with name-brand products, and not just the cheap 3rd-party discount units of undetermined Asian origin. My friend Bill, for example, bought a Linksys wireless router outfit for his house because he is very familiar with the excellent value and performance their regular wireline products offer and was willing to spend the extra $20-$40 over bargain brands. You can imagine his surprise when the unit could not talk to its matching access card when it was in the living room, a mere 35 feet way, separated by only two layers of drywall and a wood floor. While I would not expect anything close to the full 100 m open-air range that is usually possible for a decently-constructed 100 mW system, this sort of pathetic performance is unacceptable and, sadly, all-too-often the norm as products race to the bottom of the price curve.
With all the supposed advances in 802.11 semiconductors and design, I still have yet to see a system that performs even close to my three-year-old Lucent Orinoco system, or my friend's even older Aironet network.
There is enough blame to go around between both chip designers and the folks that use them in their products. Some chip makers have gotten very aggressive with their designs, employing zero-IF architectures, all-CMOS front ends, and radical architectures that, at least theoretically, eliminate many of the filters and other passives that usually surround a radio design. Now some of these designs work nearly as-claimed, some require a few extra tweaks or components (an external PA or LNA is often a good start), and some others fail miserably. More often than not, their relative degree of success can almost always be predicted by looking at how they balance their cost-shaving measures against performance issues. But since performance is not simply a function of a single variable, it takes some very careful evaluation and research to figure out who's telling the truth.
And even when a chip set can deliver decent performance, designers and manufacturers are often tempted to fiddle with a working reference design to squeeze the last few pennies out of it. And in doing so, they often severely compromise range, rate, multipath resistance, or other performance factors. One very reputable manufacturer of RF front end chips said that his products appear in many reference designs and are specified into many products, but that his sales force must spend half its time convincing overseas manufacturers to keep using his products instead of substituting poorly-matched, but less expensive, components.
Another 802.11 chip set manufacturer said that his product delivers excellent range and multipath tolerance when coupled with what he feels is the highest-quality 5-GHz antenna available. The problem is that most manufacturers he's talked to balk at the prospect of spending $0.23 for it, and are quite content to suffer 3 dB or more worth of loss with the more traditional stripline antenna etched into the PCMCIA card's PCB material. Now, given a 5X markup on the antenna, I wouldn't even blink at paying an extra $1.25 or so for a Wi-Fi card that actually reaches my whole house, or better yet, out in the gazebo where I like to write when the weather's good. And even if it meant another $5 or $10, I don't think there are many consumers who would disagree with me on this.
Going the cheap route in 802.11 products will hurt both customers and the market. For every 802.11 unit returned to Fry's, Best Buys, or The Wiz, I expect that there are 5-10 of his or her friends who have heard of the problem and have suddenly lost interest in buying one for themselves. At this rate, it won't take long to kill, or at least put a severe damper on one of the markets that many of us are counting on to help pull the electronics industry out of its doldrums. We have to do a better job of defining the performance of wireless LANs in meaningful ways, and aggressively incorporate things to deliver at least baseline levels of reach and rate in a cost-effective manner.
But performance is a tricky thing to define because there are so many factors besides output power and receiver sensitivity - such as antenna design, noise and multipath immunity, and MAC and baseband processing efficiency, to name a few. This is one of the reasons I've set up a panel session to specifically to address these issues at the upcoming 802.11 Planet Conference held December, 3 - 5 in Santa Clara. I don't think we'll come up with many definitive answers, but I expect that the panel will make some good progress on defining the questions we should be asking.
It would be great to see you there, but if you can't make it, I'd love to hear about your experiences with 802.11 equipment - as either a designer or end-user. I'd also welcome any articles you can supply that would educate me and the rest of the analogZONE community on this important topic.
Comments? Questions? Suggestions for where to buy decent 802.11 equipment,
or used 8-track cartridges? Write me at: lgoldberg@green-electronics.com.