“Prediction is very difficult, especially about the future.” —Niels Bohr
I’ve spent most of my career designing the analog front ends for microcontroller-based systems. That was the branch of electronics I took when 32 years ago I thought writing the control software for a Mostek 3870 would be fun. It was a time when software was just becoming an engineering discipline and everyone was trying to figure out how best to define, design, and document control software.
By 1985, C was the de facto standard for firmware, and the problems I found interesting had been solved. I say this because although I have used programmable logic, my outlook on the future is shaded by my past experience. Last year I wrote a column on my thoughts about the future of programmable devices. This year I will make some observations about developments so far.
Big Companies Still Do What They Know Best
You’re starting to see different companies move out of their comfort zone and integrate more onto silicon. Unfortunately, they view their specific expertise as core and concentrate on it at the expense of the new features. For example, programmable logic companies are starting to add analog circuitry. Unfortunately, it isn’t the best it can be.
Likewise, traditionally analog companies are now starting to include microcontroller functionality but are limited in RAM and ROM space. Microcontroller companies are starting to include some programmable logic, but its integration into the total system hasn’t always been well thought out.
It reminds me how much aviation advocates had problems getting the Army to take aerial warfare seriously. It wasn’t until they spun off and became the Air Force that airpower reached its full potential.
Sensors, Sensors, Sensors
This is the decade of sensors. Many products will start including more of them. Sensors will need to measure:
- Capacitance for button replacement, touchscreens, and some microelectromechanical-systems (MEMS) devices
- Acceleration to measure G forces for shock protectors
- Magnetic fields for compasses and dc current measurements
- Temperature management by means of diodes, thermistors, and thermocouples
Many of these measurements will require signal-to-noise ratios (SNR) greater than 70 dB. This means more attention will have to be given to analog layout, both within a system and the chip itself. Analog gets a lot pickier above 12 bits and is going to require practicing design engineers to update their layout skills. Look for more seminars from your favorite analog suppliers.
This also means more use of differential inputs. Start looking for parts that offer a complete differential signal path from input amplifiers to the analog-to-digital converter (ADC). Mixed-signal will become the standard for combined digital/analog systems. This means delta-sigma modulation (DSM). With built-in filtering, flexible resolution, and sample rate combinations, DSM will become the affordable ADC of choice.
Successive approximation registers (SARs) will always offer the lowest power for a given sample rate. However, you gain speed at the expense of flexibility. If you’re uncomfortable with DSM, get comfortable or risk losing your effectiveness as a designer. I have a five-part column that will better explain it.
Continue on next page
It’s A 32-Bit World After All
People have been predicting this for years, but this time it’s true. Honest! As companies move to the 32-bit domain, they want to reduce risk and standardize on a common architecture. No longer will companies leave themselves to the mercy of software companies for their compilers and other software tools.
ARM will be the common architecture that will win. The relative simplicity of these processors makes them suitable for low-power applications. Different versions of these processors are presently used in more than 98% of cell phones. Of the top eight companies selling microcontrollers in 2009, four offered some sort of ARM architecture.
Since their smallest part (Cortex M0) can be implemented with only 12,000 gates, there’s no more advantage to sticking with an 8-bit or 16-bit architecture. Any professor developing a microcontroller program should consider the benefit of students gaining experience directly applicable to their first employer. This should be one of the first questions that students choosing an engineering program should ask during their campus visit.
Programmable Logic Stuff
For the last 20 years, every complex system I’ve assembled has had both a programmable logic device (PLD) and a microcontroller. It is only natural for them to be combined on a single chip. The problem is that they are seen as individual pieces that share a common substrate.
It makes sense that the registers used to configure a PLD should be in the memory space of the microcontroller. This combination allows for intelligent reconfiguration of the digital peripherals. Hardware will become an allocatable resource.
Frankly, using 100% of the resources 75% of the time is just as wasteful as using 75% of the resources 100% of the time. Combine this with direct memory access (DMA), and this CPU PLD fusion offers features yet imagined. Such mixed architectures are also going to require new design tools that understand the concept of peripherals having a time dimension.
These are my opinions and predictions. Let’s look back in five years and see just how accurate they are.