Home » IoT Embedded Systems » Featured Stories » Analog, Sensors and Low Power Impacts the IoT

Analog, Sensors and Low Power Impacts the IoT

Here’s the “rest of the story” from Internet-of-Things experts  at Freescale, Rambus and ARM on end-node challenges and favorite applications.

By John Blyler, Editorial Director

The Internet-of-Things (IoT) is synonymous with sensors. That’s why sensor technology growth projections are so high, including both MEMS and non-MEMS implementations (see Figure 1). Individual sensors are being replaced by sensor hubs to achieve denser integrations and lower cost in more vertically oriented systems, (see, “The Rest of the Story: Sensor Integration and Processor Selection”)

Figure 1: MEMS market trends as presented by Jeremie Bouchaud of the HIS at the MEMS Congress, 2013.

Figure 1: MEMS market trends as presented by Jeremie Bouchaud of the HIS at the MEMS Congress, 2013.

What challenges await designers and implementers on the monolithic mixed signal sensor side of the IoT equation? Several experts from the IoT ecosystem have differing viewpoints on these questions including Patrick Gill, Principal Research Scientist at Rambus; Ian Chen, Marketing, Systems, Applications, Software & Algorithms manager at Freescale; Pratul Sharma, Technical Marketing Manager for the IoT at ARM; and Diya Soubra, CPU Product Manager at ARM. What follows is a portion of the responses. — JB

Blyler: Many of the end nodes of IoT will be previously unconnected objects, e.g., sensor systems. Low power, connectivity and even security will be key design factors. What analog IP is needed to enable these kinds of sensors?

Gill: The big three are power regulator ICs, wireless communications, and gating sensor events.  Good switching regulators are important for devices where power is at a premium, for instance where power is scavenged from the environment or the battery won’t be recharged often (or ever).  Power-efficient wireless communication, especially at low bit rates, is going to be very important too.

There’s some interesting work in academia on radios with a very low duty cycle as in Figure 2 (see, “Ultra Low Power Impulse Radio Based Transceiver for Sensor Networks”).  The trick to having power scale down with data rate is to have the sender and receiver wake up at precisely the same time.  It’s still the early days , but I expect IP around ultra low-power radio protocols and hardware to mature within the next 5 years – that’s when we’ll see interoperable useful radios with a household reach that burn well under 100 uW.

Figure: Sensor network nodes implementing the pulse coupled oscillator (PCO) state function are guaranteed to converge towards phase lock, with no need for a PLL or any high speed components.

Figure 2: Sensor network nodes implementing the pulse coupled oscillator (PCO) state function are guaranteed to converge towards phase lock, with no need for a PLL or any high speed components.

There’s a whole class of applications where the IoT device is constantly monitoring the environment looking for changes that it might need to react to. These are known as gating sensor events. If a very low power sensor can be put into sentinel mode and wake up a higher-power function only when there might be an event of interest, the total system power draw could be slashed dramatically.

Chen: Whereas networking thinks of sensor systems as end nodes from a topology perspective, sensor systems could be seen as source nodes from a data collection perspective. In short, they are responsible for converting the physical world into data people can use. As such, we will need precision analog to digital converters with offsets stable over temperature ranges, wireless and wired connectivity, and intelligent power management for optimal system power consumption. Many of these IPs are integrated into advanced sensor products but continuous improvements are always necessary.

Soubra: In addition to all existing types of analog IP, many new types will be needed to satisfy specific endpoint requirements for every vertical market. After successful field trials with a few thousand nodes – before the millions of nodes are installed – cost will be the next big factor. There will be a cost reduction exercise where the (sensor) module and the SoC are stripped of all items that are not required for that specific vertical market. This will be needed since we cannot claim that one 16 bit A/D or specific pressure sensor type will be enough for application in each different vertical market. We need to start with a whole variety of devices (e.g., A/D converters) and sensors (e.g., pressure sensors). Mass deployment dictates cost reduction which dictates specialization. That’s why a general purpose block to catch multiple markets will burden each with the added cost.

Sharma: Some of the key analog IP needed for end-point sensors includes on-chip voltage regulators; temperature sensor; leakage sensor; brown-out detector; data converters; and phase-locked loops (PLLs).

Blyler:  What is your favorite or most challenging example of an IoT end-node application?

Chen: One of my favorites is the tire pressure monitoring sensors (see Figure 3). Fleet managers are requiring data about the conditions of their trucks to be uploaded to the cloud to help improve business efficiency. A tire pressure monitor includes pressure sensors, up to 2 accelerometers, a short range RF transmitter and MCU for signal processing all in a 7 x 7 x 2.2 mm package operating on a coin cell battery for a 10 year life.

Figure 3 : In this example, Imec demonstrates it’s tire pressure sensor monitoring technology during the Imec Technical Forum.

Figure 3 : In this example, Imec demonstrates it’s tire pressure sensor monitoring technology during the Imec Technical Forum.

Soubra: My favorite example is the WiFi connected sprinkler system (see Figure 4). It checks the current weather conditions before turning on the water. This is lower cost approach and easier to do than putting a moisture sensor in every corner of the garden with a mesh network.

I am sure newer models will also measure the amount of water used so we can track consumption.

Figure 4: Here’s an example of a favorite IoT end-node application – the Wi-Fi/BlueTooth-based wireless water sprinkler. This one is controlled with an ARM®-based GainSpan chipset.

Figure: Here’s an example of a favorite IoT end-node application – the Wi-Fi/BlueTooth-based wireless water sprinkler. This one is controlled with an ARM®-based GainSpan chipset.

I am in a believer in the saying “if you can measure it, you can improve it.” If we can measure water consumption for a specific purpose then people will watch out for over consumption. The same argument is used for measuring electric consumption for appliances in the home.

Gill:  I like the idea of smart windows, ventilation, heating and air conditioning.  An automated home that knows the weather report (and air quality forecast) as well as when its occupants will be home will be able to maintain a suitable environment for its occupants using less energy.  Nest is a good first start, but there’s more to comfortable air than controlling the HVAC.

Solar power is sexy, but solar hot water generation can give an even better ROI.  (Admittedly less R, but a lot less I.)  A black hose and storage tank on the roof can operate autonomously, but if it knows when it will be sunny and when its humans will want a shower, it can operate even more efficiently in its decisions of when to clear the hose vs. when to let the hose bake some more.

Timed sprinklers only get you so far – sensing and context awareness could deliver better results for less human effort.  Humidity sensors in the soil and infrared spectrometers which measure a plant’s water stress could reduce water consumption, especially if these tied to a rain forecast and adjusted their watering appropriately.  Sprinklers aware of what they’re watering could help even more.  For example, tomatoes taste better when they’re watered sparingly as their fruit ripens.

Blyler: What analog-to-digital interfaces issues will be faced by designers? Will additional features be needed on the micro controller (ARM Cortex®-M) side to enable this analog end-node sensor data?

Chen: With MEMS sensors A-to-D converters must discern sub-picoFarad capacitive changes. Because of the small signal and low power requirements, these converters are normally integrated with the sensor and not on the Cortex-M processor.

Sharma: Low power will be a critical issue. Analog circuits typically have DC currents. The designer will need to cut the DC currents from the μA range to the nA range by making the analog circuits more energy efficient and design mostly to operate in sub-threshold. But decreasing the power supply will affect the voltage headroom and increase the design difficulty of the analog circuits. An additional challenge will be that threshold voltages increase at cold temperature which degrades analog circuits, thus making the voltage head-room even tighter. One solution is power gating of the analog circuit but that will increase the complexity of the validation process.

Soubra: Analog designers will be faced with having to become digital design experts. There are no (new) technological challenges; we just need to get that analog block on the Advanced Microcontroller Bus Architecture (AMBA®). [Editors Note: AMBA is an ARM supported, open-standard, on-chip interconnect specification for connecting functional blocks in system-on-a-chip (SoC) designs.] This approach may seem easy to do once we understand the sequence. In reality it is a bit harder on analog designers since they need to step out of the analog design context and into a mixed digital analog setting.

This means the use of new tools, new design flow, and more validation. (see, “Best Practices for Mixed Signal, RF and Microcontroller IoT” ) Luckily, the tools are 10X better than a few years ago. The Cortex-M processor already has what is required to connect to any analog core.

Gill: Picking up the earlier thread of a low-power sentinel, it could be useful for some chips to have configurable analog functions that detect changes in the input without needing to wake up an ADC.  These would make sense from a commercial perspective if they allowed the microcontroller to be able to monitor sensor data using only a few microWatts of power.  Also, if security is an issue (and it will be for all sorts of things), low-power crypto cores could be useful to help relay data to a cloud base station.

Blyler: Thank you.

Read the abridged version at: “Chip Design

Great information delivered straight to your inbox

2 comments

Leave a Reply

Your email address will not be published. Required fields are marked *

*