Reliable systems for micro aerial vehicles -- Adaptive clocking and future work
Editor's Note: Embedded designers must contend with a host of challenges in creating systems for harsh environments. Harsh environments present unique characteristics not only in terms of temperature extremes but also in areas including availability, security, very limited power budget, and more. In Rugged Embedded Systems, the authors present a series of papers by experts in each of the areas that can present unusually demanding requirements. A separate excerpt of the book addresses fundamental concerns in reliability and system resiliency.
Elsevier is offering this and other engineering books at a 30% discount. To use this discount, click here and use code ENGIN317 during checkout.
Adapted from Rugged Embedded Systems, Computing in Harsh Environments, by Augusto Vega. Pradip Bose, Alper Buyuktosunoglu.
CHAPTER 7. Reliable electrical systems for micro aerial vehicles and insect-scale robots: Challenges and progress (Cont.)
By X. Zhang, Washington University, St. Louis, MO, United States
5.4.3 Adaptive-frequency clocking with unregulated voltage
We now turn our attention to how adaptive-frequency clocking performs with an unregulated voltage generated by the SC-IVR operating in open loop. We used the same test setup with and noise injection via the on-chip generator. The failure rates and the average frequencies are captured in Fig. 15. Compared to the measured results in Fig. 14B, average frequencies are much higher, because DVDD settles to higher values (≈0.8 V) when the SC-IVR operates in open loop. Despite the high susceptibility to fluctuations on DVDD to load current steps as seen in Figs. 10 and 15A shows zero errors occurred even for D1⁄410. The higher DVDD voltage provides more cushion to avoid intermittent retention failure.
In order to illustrate the extended operating range offered by running the SC-IVR in open loop, Fig. 15B plots the average DCO frequency and average DVDD voltage for error-free operation versus battery voltage. These measurements were again made with 0–15 mA current load steps. As expected, the open-loop SC-IVR’s average output voltage scales proportional to the battery voltage. Moreover, the system can operate error-free even for battery voltages below 3 V, which approaches the 2.5–2.7 V lower discharge limit of Li-ion batteries. In comparison, assuming a target SC-IVR regulated voltage of 0.7 V, the system would only operate down to a battery voltage of 3.2 V and at a lower frequency across the battery discharge profile even with the adaptive-frequency clocking scheme. A fixed-frequency clocking scheme would lead to even lower performance.
click for larger image
FIG. 15 Performance under unregulated voltage from open-loop SC-IVR operation. (A) Failure rate and average frequency versus DCO settings and (B) measured average clock frequency and output voltage.
Finally, we look at the transient waveform of the supply voltage captured by the internal voltage monitor when the SC-IVR operates in open loop with the same periodic load step condition used in Fig. 15A. Fig. 16 shows both CLKOUT, which is a divide-by-2 signal of the internal system clock, and the supply voltage (DVDD). The DCO’s control code is set to 10 as suggested by Fig. 15A. In the zoom-in window of the waveforms, it clearly illustrates that the DCO frequency can respond promptly to a 82.1 mV supply droop within 6.82 ns, as the load current steps from 5 to 30 mA, so that no memory access error is recorded by the BIST module, suggesting superior supply-noise resilience of the SoC system due to the adaptive clocking scheme. Moreover, the waveform of the supply voltage demonstrates that supply noise in a typical microrobotic SoC with IVR is characterized more by the voltage droop and ripple in response to the load current steps and the IVR switching, instead of resonant noise cause by the LC tank in the power delivery path.
click for larger image
FIG. 16 Waveforms of the supply voltage (DVDD) and a divide-by-2 clock signal (CLKOUT) during IVR open-loop operation with load current (ILOAD) periodically switching between 5 and 30 mA.
In addition to validating the resilience and the performance advantages of adaptive-frequency clocking, our experimental results also reveal the synergistic properties between the clocking scheme and the IVR design in a battery-powered microrobotic SoC. The supply-noise resilience provided by an adaptive clock alleviates design constraints imposed by voltage ripple and voltage droop. Therefore, the IVR can trade-off its transient response for better efficiency or smaller area when codesigned with adaptive-frequency clocking.