In the current frenzied wave of devices connecting everything from blenders to toothbrushes to the cloud, the IoT space is being controlled by low-cost integrated 32-bit microcontroller RF modules that offer small form factor solutions for a small number of sensor inputs. 

Communication protocol stacks for Wi-Fi®, NB IoT, and Bluetooth® are well suited for the 32-bit space, while also providing increased computing power to secure RF channels. However, as the number of sensor channels increases or the power consumption required for more remote locations decreases, it increases the complexity of the system design, when adding additional 8-bit MCUs as follows can add value, as shown in Figure 1. 

True 5V IO support and sensor aggregation 

The industrial environment is still dominated by the 5V power ecosystem, and while there are 32-bit MCUs that fully support 5V, most integrated 32-bit MCUs/RFs are devices that only support the 3.3V power domain. In the 5V power domain, allowing more efficient 8-bit MCUs to connect directly to 5V power sensors, switch contacts and actuators via GPIO without adding multiple level converters or adjusting analog voltage inputs to meet 3.3V voltage requirements. 

Now, only level conversion/adjustment operations are required for the communication channel between the 8-bit MCU and the 32-bit MCU/RF module. In some cases where the 32-bit MCU module has a 5V withstanding voltage input, level conversion may not be required at all, perhaps only some series resistor isolation is required. For cases where current isolation is also required, additional cost savings can be achieved by reducing the number of dedicated ICs needed to protect the RF portion of the system. 

Remote installations often require higher fault tolerance, which may lead to the use of multiple sensors or actuator controls to mitigate the effects of field failures. Redundant sensor interface connections mean more input/output pin assignments on pin-limited 32-bit MCUs/RF modules. 8-bit MCUs tend to offer huge interface pin density, allowing some intelligent fault tolerance to be added to the front-end sensor array. It does not require the use of machine learning algorithms to determine if one of the three temperature sensors has failed. These types of decisions can be made locally with faster event response. 

                                                                Industrial Sensor Integration © xiaoliangge - stock.adobe.com 

                                                                                          Figure 1 - 8-bit/32-bit system partitioning 

System partitioning 

Using an external 8-bit MCU interfaced to most sensors makes it easy to quickly interface a known working analog/digital front-end to a different RF module back-end. Integrated 32-bit MCU/RF modules are often accompanied by a large number of sample applications that demonstrate that connecting to the cloud is a snap, regardless of vendor. The application examples may not specify how to interface with sensors or actuators outside of the standard I2C or SPI bus. Validated known sensor/control front ends with consistent and well-defined interfaces also allow for more flexibility in selecting the right RF module by minimizing the migration process. Once the new physical layer on the new RF module supports the protocol layer between the two MCUs, the integration of the new system is almost complete. Development efforts can now be focused on the proper implementation of the new RF channel. 

Loosely coupled systems with fault-tolerant hot-swappable interfaces are a useful feature in industrial or remote environment setups. Sometimes an overall system swap cannot be avoided, but the ideal option is to minimize overall changes to a known reliable system. This loose coupling also allows trusted known RF platforms to support expanded system requirements without having to start from scratch. Keep the parts you trust and improve the parts that fall short. 

                                                         System partitioning and architecture © myboys.me - stock.adobe.com 

Smart Power Management 

Unfortunately, moving to smaller IC gate technology requires a trade-off between speed and quiescent current leakage. Gate oxide thickness in new process nodes is about to reach optimal thickness in terms of atomic numbers rather than nanometers. 8-bit MCU space is dominated by larger process processes that enable better static leakage rates. Since optimal low-power management techniques are by definition simultaneous power cut-offs, adding intelligent low-power management devices can improve low-power operation. Some 8-bit MCU devices operate at current with a standard 32.768 kHz crystal, which can leak current on 32-bit RF modules. This approach now adds a power management system based on precise timing, as well as the ability to charge the battery and monitor battery operation. 32-bit RF modules (especially Wi-Fi-based units) can have active currents of hundreds of milliamps. If the battery pack is about to run out of power, it may not be able to maintain the startup and transmission currents required to connect to the network. 

The 8-bit MCU-based power management system can now wake up the main RF module using a special wake-up command that reduces the required current demand, thus keeping the RF module online in optimal phase sequence. This special wake-up use case can now be used to eventually establish a connection to the network using reduced TX power. 8-bit MCU power management systems can periodically monitor peak start-up currents and voltage drops, and submit this data at each wake-up cycle. An appropriate cloud machine learning engine can use this data to better analyze the battery system and predict failures. 

                                                                   Low Power Remote Applications © aquatarkus - stock.adobe.com 

Programming Model/MCU Complexity 

Over the past few years, 32-bit MCU/RF modules have become significantly less difficult to program. Some of these modules offer Arduino-based support, which certainly helps speed up development, but programming difficulty increases when more customer sensors, power management, or other peripheral interfaces are involved. arduino support code is very large, but in many cases incomplete, and there are still some trust issues in the professional world. In addition, the IC vendors themselves provide support, but in the end, there is no way to avoid the additional complexity that comes with integrating 32-bit RF modules in the bare metal layer. All 32-bit based control registers seem to be too large for some control or status bits, and while this does happen when moving to 32-bit, at this point in time, not everyone can intuitively pick out the wrong bit in a peripheral control value like 0x23AA123C. 

The 8-bit MCU programming model presents common interfaces in the form of 8-bit blocks, sometimes extended to 16 bits for use in timer registers. In addition to enabling easier debugging of bit fields, peripheral sets on 8-bit MCUs tend to be easier to understand because they do not involve or provide more complex power reduction or bus interface synchronization functions. clock trees in 8-bit MCUs are also easier to understand and simpler to operate even when PLLs are provided in the clock tree. However, this is the whole point of using 8-bit MCU companion devices, to provide low-power, low-cost, intelligent but not smoothly supported IoT devices to handle all the background, power management and tedious tasks. 

Microchip offers several examples of 8-bit MCU devices, including the PIC18-Q41 family and the AVR DB family. Both families offer extensive analog functionality, including on-chip op amps and multi-level voltage GPIOs, reducing the need for additional external analog components and level converters. 

While the number of available multi-core 32-bit MCUs/RF modules is growing, adding 8-bit MCUs is still a viable option when designing robust low-power edge nodes in an IoT environment. They provide power and sensor management in a small package and therefore still play an important role in the 32-bit IoT space.