For novel ideas about building embedded systems (both hardware and firmware), join the 27,000+ engineers who subscribe to The Embedded Muse, a free biweekly newsletter. The Muse has no hype and no vendor PR. It takes just a few seconds (just enter your email, which is shared with absolutely no one) to subscribe.
This was written for the final issue of Embedded Systems Design magazine in May, 2012.
If you've made it to this column, at the very end of the magazine, by now you know that this is the last installment of Embedded Systems Design.
ESD, like so many other publications, is a victim of the Internet. In its glory days it ran well over a hundred pages, but in recent years has been hollowed out to the slender magazine you hold in your hands. In the publishing business it's all about ad revenue, but that has deserted print in favor of those on-line eyeballs.
Print is pretty much dead, at least in terms of free trade magazines. Those I get are all just ghosts, hints really, of their former glory. EDN, EETimes and so many that have been so important to working engineers now play second fiddle to their on-line versions. Does anyone remember Electronics or Computer, both done in even before the web?
The failure of print is not good for the industry. Hard-copy publications are very expensive to produce and mail, which is both their strength and their Achilles' heel. The high costs means a decent editorial staff is needed to cull weak or vendor-serving articles. So quality is high. But the costs undermine profitability. The relationship between a reader and a hard-copy magazine is very different from that of the same reader and Internet sources. With the former one relies on the magic of serendipity; new or different ideas explode off the page as one flips through the magazine. Readers of on-line content largely get narrowly-targeted responses to specific queries.
Web alternatives are sometimes great, but too many are designed as Google magnets. Content is needed to draw searches to the site; lots of content, so quality actually has little benefit to the bottom line. Blogs and blather replace detailed analysis and the exposition of complex concepts.
The supreme irony of on-line publishing is that we've become very efficient at ignoring the ads that pay the freight. Who looks at those flashing graphics? A click and they're gone. Print ads, on the other hand, are large enough to contain real content. We engineers peruse them and absorb information that's useful to us and worthwhile to the vendor.
The only magazines that thrive are those with paid subscriptions, which is a pretty small set when it comes to embedded systems. Circuit Cellar Ink and IEEE Embedded Systems Letters are the only ones I can think of that are published in the USA.
ESD started as Embedded Systems Programming. In 1988 Ted Bahr, one of the founders, visited my company in Maryland soliciting an ad for the first issue. He left us with a stack of demographics about the potential circulation. It seemed a gutsy idea, almost goofy. After all, the word "embedded" was pretty new, at least in the context of computer systems. Though the microprocessor was almost two decades old most of us referred to this business as "the micro industry" or used other more cumbersome terms. But Ted's pitch was persuasive, and we took a small ad in the first issue. A .PDF of this inaugural ESP is here. (In another irony, I can only point you to the first print edition via an on-line source.)
The ad worked pretty well, and we eventually took full page displays at something like $7k/issue, which was quite a bit for a small outfit. Pre-web all of the magazines had "bingo cards," postcards with an array of numbers, one per ad. Interested readers circled the numbers associated with products they wanted more information about, and mailed the card to the magazine. (Via snail mail - remember that quaint concept?) This was in turn passed on to the advertisers. You'd be surprised how few leads were generated: our full-page ads pulled in 15-25 per month. But ESP appealed to a very specific audience and the lead quality was very high.
We would ply the salespeople with dreams of more ads to find out how many leads the competitors' ads generated. They'd hem and haw, protesting that they weren't allowed to share this information, but eventually would let it leak. I'm sure all of the advertisers did the same.
Just about everyone in the embedded tools business advertised in ESP. One editor told me that each page of advertising supported one of editorial content, so lots of ads meant plenty of articles, most of which were really quite good. ESP thrived.
Though the web dates to the early 90s, by the turn of the millennia even business types took notice of it. Startups abounded and venture capitalists would fund any crazy notion that had an Internet component. Price/earnings ratios soared (in the cases where there were any earnings, that is) and the stock market became irrationally exuberant. At the peak in October, 1999, the Dow wasn't much lower than it is today. High valuations meant publicly-traded companies' stock was very valuable, and waves of acquisitions followed. In the embedded world, Wind River, then the 800 pound gorilla of the industry, bought lots of tool vendors. One friend worked for five companies in a six week period without changing jobs, all of them absorbed into the Wind empire.
As a result ESP's advertising dried up. Many of the companies vacuumed up in the acquisition frenzy had taken good-sized ads in the magazine. Those were replaced by a single, generic, Wind River ad. Page counts took a hit. Then the dot-com insanity was replaced by the inevitable bust and tech in general lost a couple of years of good times. That didn't help the magazine, either.
Finally, of course, the web became the dominant source of information. ESP was quick to exploit it and early on got the coveted "embedded.com" domain, which is going strong today. (Please don't ask me why it redirects to "eetimes.com." I get lots of email about this and have no idea why).
In 2005 the magazine was renamed Embedded Systems Design, which was another gutsy move. It's always risky to rebrand a product, and the ESP brand by then was seventeen years old and well-established. But it was the right decision; in fact, one could argue that the original name was a mistake. The embedded industry is truly unique in that it's the synergy of both hardware and software. Probably the coolest part about this business is that we're always balancing the mix of the two. That's represented so well in the FPGA world where the tools let us seamlessly convert software into hardware or vice versa. The new name reflected this essential reality, and the content did change somewhat to reflect hardware tradeoffs. Unlike an EDN, though, I can't recall a single hardware-only article in the succeeding years. The editors did a great job of picking pieces that, when discussing hardware, also worked through the firmware implications.
ESD is owned by UBM, a British media conglomerate. As of 2010 the company produced 123 print magazines, but a summary of 2011 results shows that online revenue has doubled in the last five years while print's contribution is a third of what it was. The trend is clear. I doubt that any free trade magazine will survive the Web's inexorable force.
I contributed two articles "over the transom" (that is, unsolicited) in 1989, and the following year the very colorful Tyler Sperry asked me for a monthly column. He named it "Breakpoints," which has now run about 260 times. The first Breakpoints showed how to use extended memory on Hitachi's long-obsolete 64180 and 647180X processors.
Some of those columns stand out for being completely irrelevant today. The same year I penned one that showed how C code was substantially bigger than assembly, and cautioned that, while C made sense for a lot of applications, one had to be wary about using it. In 1990 the language had so many dialects that compatibility between compilers was a huge problem. Younger readers may be astonished that so many developers wrote their programs entirely in assembly language. Other than for applications using the very smallest microcontrollers, I wonder if anyone uses assembly exclusively any more.
In 1991 I wrote about the increasing use of prefetchers on some CPUs, like Intel's 186 (an embedded version of the 8086). Memory had gotten faster than processors, so vendors added hardware that would fetch a couple of extra bytes ahead if the CPU was busy executing an instruction, in the expectation that those queued bytes would be needed next. Well, since then the opposite has happened. Processors are oodles faster than memory today, so caches serve the opposite role. In a 2008 Breakpoints I showed how this memory/CPU mismatch now means that a lot of the multicore hype is marketing mumbo-jumbo that doesn't stand up to an engineering analysis.
Or how about the column on writing relocatable code? Before architectures that supported position-independent code came along one had to do horrible things to write apps that had to run in varying locations in memory.
Then there was the 1994 piece about the difficulty of using FPGAs. Gads, the tools were so awful then! Misrouting was extremely common, and one could count on them crashing many times a day. The FPGAs weren't much better: some had known flaws that would occasionally keep them from booting from their dedicated ROM properly. Some designers added a small microcontroller just to watch the boot sequence, and toggle the reset line if the part didn't properly come to life.
I wrote about the future of in-circuit emulators in 1997 and predicted that their market would narrow but they'd continue to be important. A month later an outfit made an offer for my emulator company and we took it. They bought a couple of competitors at about the same time, and just a few years later the emulator business imploded.
There were so many people associated with Embedded Systems Programming/Design. The names rush in as I write, people I came to know well. Some were authors; others editors, marketing folk and business development types. Some went on to greater successes, but there were sad stories, too. I never met anyone associated with the magazine I didn't like.
Regular columnists Dan Saks and Jack Crenshaw have had a fiercely-loyal following. Who knows how many others have contributed articles and commentary?
The readers sure had commentary. A lot. I've exchanged over 100,000 emails with ESD/ESP subscribers over the decades, and the on-line articles elicit plenty of thoughtful replies. Unlike pretty much every other Internet venue those comments are always polite and thoughtful. And often really, really smart.
As the magazine winds down I'd like to express my special thanks to Susan Rambo, my editor for seven years and now the magazine's Managing Editor. All of the previous editors stuck around for a year, maybe two, but Susan has been a constant presence, and she has markedly improved my clumsy sentences.
And I'd like to thank you, the readers. When I've gotten things wrong, you've been quick to correct me. And I have learned so much in my dialog with so many of you. I consider some frequent correspondents friends though we've never met.
The future is a bit murky, at least to me, but the staff is busily getting all of the old issues on-line, and are working on a revamped embedded.com. Meanwhile I plan to continue writing, at the very least in my e-newsletter, The Embedded Muse, and on-line at embedded.com as well. If you haven't explored that site, check it out. I can't wait to see what's in store for it, but have no doubt the redesign will result in an even better source of critical information for design engineers.
Embedded Systems Design is dead. Long live embedded.com.