Over the past couple of months, I’ve been devoting a lot of attention (and blog real estate) to Gene Frantz’s series, Everything You Ever Wanted to Know About the Internet of Things, which ran last summer on Embedded Computing. I’ve done so for a few reasons. One, the IoT is so vast and ubiquitous, it’s essential that, as engineers, we all have a good understanding of it. And then there’s Gene’s writing style – clear and interesting. He’s a really good explainer. (I hope that readers are drilling down on my summaries and reading his full posts.) Finally, I’m a long-time admirer of Gene Frantz, the Father of Digital Signal Processing.
In my earlier Gene Frantz-related posts, I wrote about his overall view of the IoT (which he breaks down to aggregators, the cloud, and smart sensors), about aggregators, and about the cloud. Now we’re getting into the area that’s nearest and dearest to my heart: smart sensors.
In his first sensor post, Gene begins with an example of the “ultimate smart sensor,” in this case one designed for artificial vision.
Block diagram of a smart sensor
He then breaks down what he calls the P’s & Q’s. The P’s are the three components that make up a sensor: Performance (blue), Power (Green), and what he calls Personality (Red) – the purpose of the sensor. For each of the P’s, Gene lists the Q’s – the questions that need to be addressed when designing that block. For Performance, the questions are around the process, communications, security, etc. For Power, they’re about method, buffer, management, distribution… For Personality, the questions are specific to the function. In the case of his example image sensor, they would include pixels, bits per pixel, b&w vs. color.
Gene’s second post on smart sensors is largely a rumination on smart dust, those tiny MEMS that get their name because they can be as tiny as grains of dust. Gene comes up with one application for them: smart pain that would let him “change the color of the walls of [his] house using a remote control.” Interesting…(And now I’ll have to think of an app for it…)
In his next post, Gene goes back to the block diagram (shown above).
The ultimate goal of a smart sensor is that it be completely autonomous. That means it sources its own energy, performs all of its functions and communicates with the outside world wirelessly… My mental view of how this plays out is to have three independent subsystems in the smart sensor. One of the subsystems handles all of the power management assets, one handles all of the performance aspects and one subsystem handles all of the personality aspects of the smart sensor. Each of these three independent subsystems can be connected together to create the smart sensor. With this flexibility, various methods of energy management can be developed then mixed and matched with various processor systems. Finally, different personality boards with different arrays of sensors can be attached to the other two subsystems to create different smart sensors.
He then goes into some detail to provide answers to the Q’s associated with the 3 P’s (Performance, Power, Personality.)
Next up, Gene discusses the never-ending issue of performance vs. energy efficiency.
Now, in the IoT system we have a need for ultra-low powered smart sensors along with energy efficient high performance cloud computing devices….The answer is smart sensors will need just enough performance to obtain the input from the sensors, process the signal and send the results to the communications system to be transmitted to the aggregator.
Gene then asks a few provocative questions on performance-power tradeoffs, including whether we need to go back to assembly code, whether security is a luxury, should we go clock-less, and – my favorite – “is it time to go back to analog computing?”
What do you think?