The small form factor of IoT sensor nodes places severe energy constraints on computing platforms. To save energy, duty-cycling helps reduce the power of the sensor node, but this needs an ultra-low-power timer for reliable timed shutdown and wake up. Crystal, MEMS, and LC tank-based oscillators need off-chip components or entail extra fabrication costs. This leaves the RC Oscillator (RCO) as the preferred choice that can be easily integrated with such systems.
Obviously, the quality of the RCO’s frequency directly influences the accuracy of a timer. To maintain good frequency stability against environmental variations and noise, a generic RCO, as depicted in Figure 1. This expends significant powerup fundamental circuit blocks including the RC network, voltage reference, and comparison. The latter two factors present a trade-off between power consumption and overall frequency accuracy.
Figure 1: A generic RCO and its fundamental circuit blocks
By recapping the phase-shift RCOs, which were documented in the Fair child Semiconductor Application Note 118 in 1974. And, the one improved and published by Texas Instruments (TI) in 2014, our recent work relaxes the requirement of oscillation frequency stability, which leads to an 82.6% power saving, as described in Figure 2.
Figure 2. Power breakdown comparison of the proposed topology (right-hand side) with baseline design (left-hand side)
The fundamental idea of the power reduction technique (phase 1) depicted in Figure 3 is to manage the RCO subthreshold operation. This uses a much lower supply voltage (VDDLOCAL) along with resistor scaling. This allows the RC voltage swing to be aggressively scaled and exploits a quadratic reduction in power. In this instance, the normalized active power can be decreased to 0.23×, compared to the baseline design adapted from the prior art. The proposed bias current (IBAIS) generator is crucial to implement the subthreshold region operation.
Figure 3. Proposed power reduction techniques versus frequency stability performance
A further power reduction technique (Figure 3, phase 2 shown on the bottom right diagram) can decrease power to 0.18×. The switched-capacitor network (SCN) reduces the drop-out power loss from VDD (original supply voltage) to VDDLOCAL (a much lower supply voltage to bias the phase-shift RCO). A level converter is required to overcome the voltage gap. The combination of power reduction techniques that were described earlier can scale power down by approximately 5× in total. This is despite the power overheads of both the SCN and level converter.
The downside of deploying these power reduction techniques (Figure 3, phases 1 and 2) is the resulting deterioration of temperature sensitivity (353 ppm/ºC in simulation). In other words, the oscillation frequency of phase-shift RCO presents reduced stability against temperature change. This is due to the high turn-on resistance in subthreshold operation, where the temperature dependency significantly affects the oscillation time period. The forward-body-biasing (FBB) technique (Figure 3, phase 3) improves temperature sensitivity to 79 ppm/ºC in simulation. This demonstrates its effectiveness by reducing the threshold voltage and turn-on resistance of the RCO devices, so the temperature variation becomes manageable.
In this blog, I have presented power reduction techniques for an on-chip oscillator that would otherwise dominate the shutdown power of a microcontroller. The 5x improvement in energy efficiency has been achieved using these techniques could significantly increase IoT device longevity. And, increasing the utility of such devices, prolonging the life of battery powered IoT sensor is a significant consideration in the development of environmentally friendly microelectronics. When we are heading towards a world containing one trillion IoT devices, even small power savings in individual IoT nodes can have a huge impact on overall energy consumption. Let alone the 82.6% saving demonstrated with these techniques.
This work was recently presented at the IEEE European Solid-State Circuits Conference (ESSCIRC) 2019. Read the full paper using the following link, and please feel free to contact me if you have any questions.
Read the Paper Contact Philex Fan