VCO-based ADCs for low power instrumentation applications
Abstract
Switched capacitor circuits have dominated instrumentation readout circuits for over three decades. On the other hand, the advancement of MEMS technology required integrated circuits capable of amplifying and converting all kinds of low level signals. MEMS combined with switched capacitor ADCs have enabled the inclusion of sensing capabilities in portable devices, in special in cellular phones and wearables. A vast knowledge base has been built around this combination in 0.35u or even 0.18um CMOS technologies at supplies above 1.2V.
However, as technology nodes and supply voltages scale down, switched capacitor circuits become more and more problematic. Low voltage levels and poor analog performance of MOS devices impose a limit on the resolution of switched capacitor oversampled data converters. Moreover, the new trend in the sensor market is to integrate the sensing part with a simple preprocessing stage (see figure 1).
The architecture of figure 1 enables the “always on” operation of the sensors discharging the computing elements of a portable or IoT device form staying awake. The pre-processing elements require both ultra-low power and quite dense logic circuitry to implement machine learning algorithms. These features can only be accomplished by integrating the analog interface together with digitally intensive logic in a new nanometer CMOS process.
VCO-based ADCs were known since the beginning of sigma delta modulation. However they have not received much attention by the industry due to low performance and excessive power consumption for wide bandwidth communications applications. The situation can be very different when the applications are biomedical or instrumentation signals where the speed of digital circuitry is well beyond the signal bandwidth. VCO-ADCs implemented with ring oscillators exhibit an outstanding sensitivity, allowing LSB resolutions of a few microvolts which enable direct MEMS to ADC interfaces without instrumentation amplifiers. On the other hand, most of its power consumption comes from the digital processing part, while the ring oscillator itself is a smaller part of the power consumption. Therefore, digital logic can be driven with a smaller supply than the analog blocks, saving power. The inherent problem to VCO-ADC is the lack of linearity. However, in sensor interfaces, this can be solved by either using only a small portion of the dynamic range of the ADC or by use of feedback, as in conventional continuous time sigma delta modulators. Recent papers show designs in nanometric CMOS technologies which are able to compete with the FoM of conventional switched capacitor ADCs but operating around 1V in deep submicron processes, which paves the way for the systems described in figure 1.
We consider that this flavor of VCO-ADCs is still in its early development stages and that it may become a promising industrial solution. This is the reason why we propose the special session with leading experts from both the industry and academia.
Switched capacitor circuits have dominated instrumentation readout circuits for over three decades. On the other hand, the advancement of MEMS technology required integrated circuits capable of amplifying and converting all kinds of low level signals. MEMS combined with switched capacitor ADCs have enabled the inclusion of sensing capabilities in portable devices, in special in cellular phones and wearables. A vast knowledge base has been built around this combination in 0.35u or even 0.18um CMOS technologies at supplies above 1.2V.
However, as technology nodes and supply voltages scale down, switched capacitor circuits become more and more problematic. Low voltage levels and poor analog performance of MOS devices impose a limit on the resolution of switched capacitor oversampled data converters. Moreover, the new trend in the sensor market is to integrate the sensing part with a simple preprocessing stage (see figure 1).
The architecture of figure 1 enables the “always on” operation of the sensors discharging the computing elements of a portable or IoT device form staying awake. The pre-processing elements require both ultra-low power and quite dense logic circuitry to implement machine learning algorithms. These features can only be accomplished by integrating the analog interface together with digitally intensive logic in a new nanometer CMOS process.
VCO-based ADCs were known since the beginning of sigma delta modulation. However they have not received much attention by the industry due to low performance and excessive power consumption for wide bandwidth communications applications. The situation can be very different when the applications are biomedical or instrumentation signals where the speed of digital circuitry is well beyond the signal bandwidth. VCO-ADCs implemented with ring oscillators exhibit an outstanding sensitivity, allowing LSB resolutions of a few microvolts which enable direct MEMS to ADC interfaces without instrumentation amplifiers. On the other hand, most of its power consumption comes from the digital processing part, while the ring oscillator itself is a smaller part of the power consumption. Therefore, digital logic can be driven with a smaller supply than the analog blocks, saving power. The inherent problem to VCO-ADC is the lack of linearity. However, in sensor interfaces, this can be solved by either using only a small portion of the dynamic range of the ADC or by use of feedback, as in conventional continuous time sigma delta modulators. Recent papers show designs in nanometric CMOS technologies which are able to compete with the FoM of conventional switched capacitor ADCs but operating around 1V in deep submicron processes, which paves the way for the systems described in figure 1.
We consider that this flavor of VCO-ADCs is still in its early development stages and that it may become a promising industrial solution. This is the reason why we propose the special session with leading experts from both the industry and academia.
Figure 1