Spark Microsystems: LP On-Chip Radios

Spark Microsystems is taking aim at on-chip radios that continue to be the primary source of battery drain, even in power-conserving designs like Bluetooth Low Energy.

“If you wear AirPods, something like 80% of the power is going to power the radio, not the sound. That’s not the most efficient approach.” according to Frederic Nabki, co-founder of Spark Microsystems, and a former professor of electrical engineering and the University of Montreal’s École de technologie supérieure (ÉTS).

The company is developing an ultra-low-power transceiver and microcontroller called the SR1000, a proprietary, ultra-low power, ultra-short latency radio system whose overall with net power use is 20 times lower than the most efficient BLE units on the market, 35 times better than the BLE average, and 600 times more efficient than Zigbee, according to Nabki, who developed the technology as part of his university research and launched Spark with co-researcher and fellow EE professor Dominc Deslandes in 2016.

The transceiver can send a packet in 1/15th the time it takes a BLE device, and saves power by flipping on and off to broadcast in just 50 microseconds, compared to approximately 3 milliseconds for a BLE device, Nabki said. It has a range of 50 meters, broadcasts in the 3GHz to 6GHz unlicensed ultra-wideband spectrum and demonstrates latency 60 times lower than BLE. Its maximum data rate for TX/RX for a group of devices is 10Mbit/sec compared to the 227kBit/sec or so delivered on average by BLE, whose nominal throughput is 1Mbit/sec.

“We did things like the quick on/off, and decided to use very low-frequency quartz crystals like the ones you see in watches, rather than the 20MHz crystal most Bluetooth devices use, to get better power efficiency,” Nabki said. “But the secret sauce isn’t compression software. It is in a fundamental architectural innovation at the radio transceiver level that makes the transceiver significantly different than anything else out there.”

Nabki won’t discuss details of the key innovations, but the system’s development pathway is documented in his research, as well as Deslandes’ research, and in a series of patents for systems using modulated coded impulses, frequency and bandwidth hopping and wavelength-tunable optical components.

The SR1000 is a beta version of the package, built on a 65nm process at TSMC that Spark will be testing and developing with potential partners though much of this year before finalizing a mask for the production version during the second quarter, which should make production versions available during 2019.

Fig. 1: Spark SR1000 low-power receiver. Source: Spark Microsystems.

The power requirement of the transceiver is so low that it is possible to run it without a battery or any external power supply other than a low-cost solar-power cell or a module to harvest power from heat, vibration or other environmental variables, which would allow it to be added to another device or chipset at no cost in power profile.

The company initially is aiming the technology at applications within the IoT, automotive, smart-home and medical market for streaming data, audio and video, wireless devices and wearables. Its ultra-short latency and comparatively high bandwidth would make it a good alternative to Bluetooth wireless game controllers and headsets and for mainstream audio headsets, though gaming and IoT applications are more likely to have customers willing to use a dongle to make the connection with a new radio technology, Nabki admitted.

The signals have very low levels of EMI, so they’re difficult for other systems to detect and resistant to interference in radio-dense zones. Its signal doesn’t fade over distance, so Spark connections are robust as well as power-efficient, Nabki said.

The frequency is unusual for its purpose and the protocol is proprietary, so the only way to use Spark radios would be to have a Spark transceiver at either end of the connection, though Nabki expects the method and frequency to become de facto standards in time. Spark signals are not interoperable with Bluetooth or other protocols, but they are designed to link effectively with other network types through gateways, allowing Spark to fit into a hierarchy of devices or networks at the very lowest-power level.

“We see ourselves as the last-hop to the sensor in the IoT, for example,” Nabki said.

Currently, even most Bluetooth or WiFi gateways would need a dongle to connect to Spark radios, but that is likely to change as the product gets closer to production.

The business plan is to sell transceivers both alone and packaged with a microcontroller, a partner could build in a Spark radio but not include the MCU if they already provide that function.

The alpha version of the chip was funded by the founders’ university research. The second round was a combination of private investment and a $2.25 million (CDN) investment from Sustainable Development Technology Canada (SDTC) a government agency that encourages clean-technology development in Canada.

Spark has raised a total of $3.5 million (CDN) to cover the cost of development and to get the final mask set for the production version. Nabki expects the mask to be complete during the second quarter and ship off the final mask to TSMC during the fourth quarter.

“We are aiming at a few markets primarily, but this is a platform with pretty horizontal applicability,” Nabki said. “So growth will come as people in different markets recognize the value of a batteryless system with very low latency and high data rates.”