How embedded hardware development has changed over the past 20 years
Every now and again I see a "blast from the past" that makes me feel a tad nostalgic. For example, one of the things we've been doing for the past 20 years is our Embedded Market Studies/Surveys in which we ask embedded designers from around the world all sorts of questions relating to embedded hardware, software, design and verification tools, and... the list goes on.
One such survey on its own is valuable, but having surveys going back for multiple decades is invaluable because this allows you to compare and contrast things and spot developing trends. Having said this, in certain cases, discovering the fact that some aspect of design or component use is not changing is as interesting as spotting something that is changing.
For those who are interested, we actually maintain the Embedded Market Surveys Archive here on Embedded.com. In addition to the results of the surveys by year, this also includes links to articles based on the survey results.
Quite apart from anything else, it's illuminating to see how the process of actually taking the survey has evolved over the years. Take the case of the 1997 Survey, for example, which was orchestrated by Embedded Systems Programming magazine (a print publication -- "So 20th Century, my dear") and the Embedded Systems Conference (ESC).
In those days of yore, the survey was conducted by mail by disk (i.e., a disk-by-mail survey). Furthermore, the incentive offered to complete the survey was a crisp two-dollar bill, which was mailed along with the disk and the postage-paid return disk mailer.
The results from the survey were presented in tabular form -- not even a spiffy little pie chart to brighten things up. How things have changed. Today, users take the survey over the Internet from the comfort of their office chairs, we have all sorts of tools to cross-probe and analyze the correlations between different questions, and the results are presented in graphical form.
I just took a quick skim through the results from the 1997 survey. Even though I'm not a software guru, the fact that ~80% of respondents were using C and ~70% of respondents were using assembly language in their designs struck me as interesting (C++ came in at ~36%).
In the case of MCUs and MPUs, in answer to the architectural question "What have you used in past 12 months or plan to use in next 12 months?" a whopping ~60% of respondents said they were using 8-bit processors.
Of particular interest to me was the fact that the 1997 survey referenced components and component manufacturers -- also tools and tool vendors -- that I'd either never heard of or that I'd completely forgotten about. In the case of tools, for example, ICE (in-circuit emulation) was of interest to a lot of embedded systems designers. I wouldn’t be surprised to hear that a lot of younger engineers have never even heard this term.
Of course, in 1997, there were no questions relating to topics like artificial intelligence (AI), deep learning, embedded speech, embedded vision, wireless mesh networks, or the Internet of Things (IoT), all of which are to be found twenty years later in the 2017 Survey. Similarly, security wasn't a big topic back in 1997, but it's of significant concern here in 2017.
Take the case of the IoT, for example. As a concept, this wasn't officially named until 1999, which was two years after the 1997 survey (I pride myself on my mathematical prowess). By comparison, in the 2017 survey, ~54% of respondents say that they are currently designing devices for use on the edge of the Internet; ~64% of companies have projects devoted to the IoT; and ~18% of all projects will be primarily devoted to the IoT.
With regards to advanced technologies, which is one of the things I'm focusing on at the moment, ~47% are planning on including cognitive (thinking, reasoning) capabilities in their next project. I don’t know about you, but I was surprised at how high this number is.
When I was a lad, 8-bit MCUs and MPUs were the "bee's knees," as it were. On countless occasions over the years I've heard pundits predicting "The death of the 8-bit processor." However, although they have passed their heyday, 8-bit processors keep on hanging in there, with 12% of current embedded projects using an 8-bit engine as their main processing engine.
Now I want to go back and look at other things, like the way in which system clock frequencies have changed over the years (13% of today's systems sport a 1GHz clock and 4% are running at 2GHz+); also, if there's any correlation between the decline in the use of dedicated DSP chips and the rise of advanced FPGAs. Chat amongst yourself -- I may be gone some time.