Requiem for Magnificent Failure
i/oZONE's Product of the Year Awarded Posthumously to Transpectrum's
XPT110 All-CMOS all-in-one 10-Gbi/st Transceiver
by Lee Goldberg
Boole wanted to prove the existence of God with his "calculus of reason." Beethoven wanted to write music that transcended fate. Einstein wanted to fully understand the nature of the universe. I just wanted to send a spacecraft to Mars. In the end, the only thing we have in common is that all of us failed in those efforts. But even when we fall short of the mark, the effort itself is a magnificent act that still has value and significance. So, how should a person or organization be remembered if it set off on a worthy quest, did its best, and met with failure? This is the dilemma I faced after discovering that Transpectrum, the creator of a 10 Gbit/s transceiver I had nominated as a Product of the Year for 2002, had gone out of business late this fall.
The part was reviewed June 2002 in i/oZONE, and despite my misgivings about the company's ability to deliver on time, or on all its promises, it received a reasonably acceptable Vapor Index Rating for such a cutting edge device (3 saltshakers.) Normally, when I see a small startup tackling a task formidable enough to scare away the likes of Agere, Agilent, and Infineon, I am immediately suspicious of their motives. But these folks seemed to be different than the many shady silicon snake-oil companies that filled to bursting the industrial parks of the greater San Jose region for most of the '90s. Rather than focus on producing high stock prices at their IPO, or getting bought up by a larger company for a ridiculous price, Transpectrum seemed to be in business to deliver commercially viable products.
There were several clues that, whether or not Transpectrum could deliver, they were certainly clever and earnest in their attempt. First, Fred Mohamdi, their CEO, was rather candid with me about the design challenges they faced in producing such a complex, high-speed part in vanilla CMOS. He seemed to understand and acknowledge the problems involving substrate coupling, threshold drift, and jitter that he'd be up against, and explained (under NDA) in some detail how his team would deal with them.
The other thing that allowed me to "review their part with a straight face"* was that Mohamdi and crew were doing due diligence by fabricating nearly every element of the final product on a series of test chips. While I cannot recall precisely, I believe he told me they were taking advantage of the "shuttle" opportunities provided by a major fab to run single-wafer prototype batches of the test chips at regular intervals to verify the latest elements of the design as they were completed. I only wish more companies developing ambitious designs used similar practices.
At the end of the year I looked through all the products I wrote up for i/oZONE and asked myself which ones showed the highest levels of technical innovation plus offered significant cost and performance advantages over the current state-of-the-art solutions. There were several products that grabbed my attention and competed for the two available slots. Among other products, TriCn's solidly-engineered SPI 4.2 interface IP core and Accelerant's extremely innovative 5-channel, 6.25-Gbit/s MLT-based backplane transceiver both made my choice rather difficult.
In the end, the Transpectrum part won out, thanks in equal parts to its technical audacity, and its potential to lower the cost of 10- and 40-Gbit/s I/O channel links. It was only while trying to locate the proper person to notify about the award that I came upon Craig Matsumoato's obituary of the company, which closed its doors in October 2002 . From what I can tell Transpectrum could have probably delivered the parts it set out to build, but a combination of a tight capital supply and a market that was not really ready for high-volume OC-192 and OC-768 I/O elements did the company in prematurely. It seems that they expired so quietly that they escaped my notice. I'm sorry I didn't have the chance to send my condolences to Fred.
When faced with the tough decision about whether to re-assign the award, I finally chose to leave it with Transpectrum. While we'll never know for sure if they could have pulled it off, I sincerely believe that with a few more months of funding, they would have had a commercial part that would have raised the bar for the industry.
As a former engineer I have experienced my share of failures. Whether it was the world's first all-CMOS math co-processor that never made it to market, or the interplanetary probe I worked on that blew up 2 days before arriving at Mars, I've tasted my share of "the agony of defeat." And if you've been in this field for more than a little while, I'm sure you have poured your heart into a project, overcome tremendous odds, and still had to endure the pain of seeing your work end unceremoniously tossed into the cosmic dustbin.
I sincerely believe that even in failure, a wholehearted attempt at a noble goal is something to be proud of. I also feel that quite often these failures are the teachers, the tuition, and the footsteps of pioneers that someone else will follow the way to eventually achieving the goal we set out for ourselves. In this way, the good is rarely lost, even if we do not benefit directly from it.
For this reason, I decided to leave the award where it was and award it posthumously to Transpectrum's stillborn XPT110. Hopefully, it in some small way stands as a tribute to some of the other magnificent failures (technical and otherwise) that help pave the way for the human adventure.
Comments, questions, tales of your own magnificent failure? Write me at: lgoldberg@green-electronics.com.
* My precise words in the original review