Sensors and MEMs at ISSCC2012
February 21, 2012, ISSCC, San Francisco–No matter how much we move into digital signal processing, the world and physics are still analog. Four papers looked at capacitive sensor interfaces and the other four papers described advances in temperature sensors.
Converting mechanical motion into an electrical signal is a key component in the increasing functionality in consumer devices. Unfortunately, the desired signals are degraded by component and parasitic parameter drift. The circuit from Stanford University was designed to compensate and cancel bond wire drift in accelerometers. They added an additional reference, a calibration cycle, into the conversion process and then shift that signal out of band. The resulting signal is then applied as an offset to the normal readout and force feedback cycles.
Higher precision capacitance to digital converters make mechanical servo systems possible. Delft University of technology built a 17.2 bit converter with a resolution of 65 aF from a 10 pF offset capacitance in the presence of another 10 pF parasitic capacitance to ground. Although this is not the highest resolution CDC, its 1.95 p joules per step energy requirement and 20 µs conversion time indicate much better performance in these areas than any previous published circuits.
NXP Semiconductors described in active biasing circuit for a mems microphone. The need for precision requires a load on the capacitance in the G? range. Any digital feedback loop for offset cancellation coupled with a pair of bipolar diodes achieves this impedance without adversely affecting microphone performance. The digital implementation provides improvements in the area of power and settling time compared to analog feedback methods.
Robert Bosch demonstrated a pressure sensor for automotive engine compartment functions. To isolate the measurement, they created a resonant transformer between the sensor and the measuring electronics. The resonant frequency of the sensor tank, comprised of the capacitance sensor and a five turn, on-chip coil, decreases with higher pressure. The measurement system digitally tunes the oscillator to match the resonant frequency, providing both a readout and temperature compensation.
Temperature contributes to the variability in the MEMs sensors and other integrated circuits. Temperature sensors must have high resolution and high accuracy to provide signals of value to the rest of the system. Delft University of technology achieved plus or -0.4° C. over us -72° C to +200° C. range using a time domain temperature sensor based on differing diffusion profiles between silicon and silicon oxide. Their sensor is fully self-contained and does not require an accurate external time reference.
MEMs frequency references, have testing challenges due to their inherent materials temperature coefficients. Masdar Institute of science and technology implemented a thermistor-based temperature sensor to stabilize a MEMs frequency reference. Unlike temperature compensated crystal oscillators, this error correction system modulates the drive signal to adjust frequency for a stability of less than 0.5 ppm over the industrial temperature range.
Although the world is fully into CMOS processes, Delft University of technology and Intel both reported on bipolar transistors as temperature sensors embedded in standard deep submicron CMOS. The first design emphasizes post calibration accuracy of +/- 0.15° C, over the full military temperature range and dissipates only 27 microjoules per conversion. The Intel design emphasizes speed and area rather than accuracy and resolution to allow the embedding multiple temperature sensors into a microprocessor to enable clock throttling below 50° C, and more drastic temperature management operations up to the thermal limits.
