Emerging IoT Applications Require Careful Consideration

No IoT device will be effective if the battery power is insufficient.


The Internet of Things (IoT) is creating opportunities in the existing space held by traditional semiconductor applications, typically falling into categories that include industrial, fitness, health and lifestyle to apparel, safety and productivity.

But there also are new, inventive devices. In the last several months, we’ve seen the launch of a canine Fitbit, a hybrid dog collar, and dynamic fabrics that will soon be found on the Paris Runways during Fashion Week. Leaf and soil moisture sensors will be welcome in drought regions like California. Sensors used for RF energy harvesting have piqued the interest of consumers with “free” energy.

While it seems like anyone with a novel idea can get into the IoT business, it’s worth exploring what’s needed for a robust architecture to support the invention and ensure success. Decisions about hardware and software hardware are important and not insubstantial. For example, the electronics could be custom or off the shelf, a choice that may be based on budget. Selecting the brains or the microcontroller is equally important and decisions about power and the peripherals must be considered here. Of course, with IoT applications, the type of sensor and quantity are factors. The power source could be battery, harvesting or management. Next comes the casing’s material, presumably prototyped by an industrial designer.

As is the case with SoC designs, software is the key differentiator and where consumer features are defined. Careful consideration must be made for code structure and size, the communication protocol, and computation offloading. Security is of the utmost importance, as is the memory.

No IoT device will be effective if the battery power is insufficient and doesn’t last at least one year or more. It should be 250mAh with 30uA as the max draw. The trend is toward using an ARM processor core that will take approximately 10uA, leaving 20uA for everything else.

The system’s memory is a major contributor to the power budget. Generally, static RAM (SRAM) or embedded non-volatile memory (eNVM) is used in IoT applications. SRAM is following Moore’s Law, while eNVM has not been keeping pace. Depending on the eNVM technology, some scales with process, others not as well.

The standard code size for IoT device is less than 128KB. For embedded devices with determined functions, eNVM offers advantages, including lower power than Flash or a Flash/SRAM combination, security, lower manufacturing costs and reduced bill of materials. However, as the microcontroller, foundry library and customer logic migrate down the process nodes, eNVM power consumption will become a big issue. Even now, a large portion of the power budget is being allocated to eNVM.

Obviously, the need to reduce power in flash-less eNVM is acute, and a vendor of one-time programmable eNVM has an offering worth exploring. Its OTP eNVM is application specific and has been qualified at multiple foundries, making it 100% compatible with manufacturing rules. It scales to 20nm and beyond and beyond, and the technology and IP design follows CMOS logic technology scaling. Moreover, its features include security, low-active and standby power. It cannot be hacked using passive, semi-invasive or invasive methods because of a strong layer of protection at the most vulnerable physical layer.

The decision to develop an IoT device may be intuitively easy, but developing the system spec is more complicated and multiple decisions need to be made. Selecting the right on-board memory may be among the most critical. The choice should be to implement a low-power OTP eNVM.


Steve Wald says:

One time programmable NVM is insufficient for many applications. Where battery power is unavailable, energy must be harvested from the environment and will be intermittent — e.g. sunlight and RF. In this case, working state and parameters must be saved in NVM. The problem is that writing a typical NVM requires extra time and a large dose of energy, both of which are hardly compatible with an unsteady power supply. Supercaps can be used to extend power-down time, but add complexity, size, cost, and stiffness. For the IoT idea to bloom into pervasive use, we need cheap embedded NVM which is fully writable at high rates and low voltage.

Leave a Reply

(Note: This name will be displayed publicly)