Time for the Intelligent Edge

4 mins read

IoT technologies are a key force enabling daily essentials from cities, buildings and transportation automation to retail experiences and healthcare, embedding smart capabilities that enhance convenience, safety, and efficiency.

Credt: Bipul Kumar - adobe.stock.com

As modern lifestyles move easily and continuously between the physical and digital worlds, the IoT is set to become more pervasive and powerful, massive and intelligent, to provide the support and immediacy people demand.

To support the rapid expansion of IoT, and the instinctive and seamless responses that people demand, data processing and AI must continue to move closer to the edge. The latency associated with transmitting information through the edge network, across the wireline core network, to the cloud datacentre, and back, is not practicable for real-time applications like virtual and augmented reality, Vehicle to Everything (V2X), the smart home and smart cities. In addition, the increase in wireless devices participating in the network will require an increase in bandwidth and, therefore, significant advancements in radio technologies like 5G-Advanced and forthcoming 6G.

These advancements in network infrastructure will drive two trends: edge datacentres, and data processing and AI integrated into wireless infrastructure and local edge networks in the near future. Just like cloud datacentres, edge datacentres are reliant on precision timing. In the datacentre, there are two applications that drive the need for precise time: end-to-end synchronisation and high-speed data transmission.

The accuracy of a network’s time synchronisation is heavily reliant on the environmental resilience of the local oscillator in each node, such as a server, switch, or router. The most common method of time synchronisation, IEEE 1588 PTP, relies on the local oscillator to maintain a stable frequency while the 1588 software loop filters out the best time packets. The more stable the local oscillator, the longer the loop’s time constant can be, and the more effective the filtering. MEMS TCXOs are 5x more stable than quartz TCXOs under fast temperature ramps, a common occurrence in datacentre servers when processors are heavily loaded, and fans turn on to cool the system.

Beyond time synchronisation, precision time is critical to transmit data at high bandwidths via Ethernet and optical links, requiring low phase jitter on the clock edge. As bandwidths grow from 200G to 400G, 800G, and beyond, the allowable phase jitter decreases by half at each step along the way. Compared to quartz oscillators, MEMS oscillators are 10x less sensitive to common sources of board-level noise that degrade phase jitter, such as power supply noise. MEMS oscillators also offer the greatest configurability, with flexible swing options to pair with non-standard low-voltage chips, reducing power by 30+ mA.

As 5G-Advanced and 6G are released and deployed, a significant portion of the data processing and intelligence will be moved away from the datacentre and embedded directly in the wireless infrastructure. This will enable ultra-low latency wireless AI for even the smallest and most power-constrained devices, such as smart wearable.

As timing in the wireless infrastructure becomes even more critical, and tighter synchronisation along with high-speed data transmission become paramount, MEMS oscillators will deliver the resilience and precision required for next generation telecommunication.

Pressures on IoT Endpoints

Inside the tiny IoT devices themselves, the timing challenges involve the right amount of performance at the right level of power consumption efficiency and size. Wearables, ingestibles, smart tags and electronic labels, sensors installed in appliances and almost any piece of equipment must usually be tiny and lightweight. In consumer applications, a visible sensor means the mystique is gone. In some industrial applications, the sheer number of sensors to be deployed ensures that maintenance is unfeasible, so the battery must provide enough capacity for a lifetime’s service. This leaves little space for bulky or inefficient circuitry.

Compact circuitry also enables more space for a larger capacity battery extending the battery life. There are always extreme limitations on positioning of the radio in small wireless devices, for example to ensure the clearance area required for the antenna to operate properly.

This can be a problem for timing because quartz crystals are inherently bulky. While in active mode, the clock frequencies that power communication circuits typically run in the tens of MHz. However, to reduce power consumption, many IoT end points will use a 32.768 kHz oscillator or resonator in sleep mode to maximize battery life since a much lower frequency clock consumes much less power. Unfortunately, physics dictates that lower-frequency quartz crystals demand larger resonators. MEMS (micro electromechanical system) resonators do not have this problem. In fact, the smallest MEMS resonators are about 10x smaller than the smallest quartz crystals of the same frequency.

MEMS XOs, TCXOs and resonators, with their power and footprint advantages, bring another valuable property to the IoT endpoint scene: the possibility for significant size reductions. MEMS resonators can be on the order of a few tenths of a millimetre in size, about 0.4 x 0.4 mm.  Moreover, having significantly lower mass than a quartz resonator, they are far more resilient to shock and vibration because an acceleration imposed on the MEMS structure results in a much lower force and frequency shift.

Lastly, today’s leading MEMS resonators exhibit more stable frequency over broad temperature ranges such as -40 to 125°C compared to their quartz counterparts, enabling a higher degree of timing precision which results in more efficient performance of an IoT system.

Ultimately, the miniaturisation now made possible by this technology could culminate with the integration of the silicon MEMS timing source in the same package as the microcontroller or system on chip (SoC) that supports the main functionality of the device. Co-packaging is a step too far for conventional quartz resonators, non-silicon technology, limited by the requirement to be housed in a separate ceramic or metal package in order to do its primary job of setting a stable frequency. Historically, environmental and mechanical issues such as the coefficient of thermal expansion (CTE) mismatch have proved impractical for co-packaging in a reliable and cost-effective manner. The pressure for progress exerted by today’s (and tomorrow’s) IoT applications is shining a light on this deficiency.

Packaging is another factor that impinges on designers’ wishes to continue shrinking the dimensions of IoT endpoints. While chip fabrication technologies are migrating to ever-smaller geometries, allowing more and more functionality in tiny SoC and ASSP ICs, a quartz crystal sits in splendid isolation in its own, hermetically sealed ceramic package, relegating it to remain on the PCB indefinitely. It’s fair to say that the quartz resonator or oscillator IC is becoming one of the limiting factors preventing further miniaturisation in an industry where tiny increasingly is the ticket.

MEMS-based products, including resonators and oscillators, will continue to blaze the trail and set new benchmarks for miniaturisation as product designers continue to bring new and more intelligent IoT devices to the market.

Author details: Eric Garlepp, Senior Director of Product Marketing and Parker Traweek, Senior Product Marketing Engineer, SiTime