Contribution to time domain readout circuits design for multi-standard sensing system for low voltage supply and high-resolution applications
- Autores: Francisco Javier Pérez Sanjurjo
- Directores de la Tesis: Enrique Prefasi Sen (dir. tes.)
- Lectura: En la Universidad Carlos III de Madrid ( España ) en 2018
- Idioma: inglés
- Tribunal Calificador de la Tesis: Pieter Rombouts (presid.), J. Alberto Rodríguez Pérez (secret.), Dietmar Straeussnigg (voc.)
- Programa de doctorado: Programa de Doctorado en Ingeniería Eléctrica, Electrónica y Automática por la Universidad Carlos III de Madrid
- Materias:
- Enlaces
- Tesis en acceso abierto en: e-Archivo
- Resumen
This research activity has the purpose of open new possibilities in the design of capacitance-to-digital converters (CDCs) by developing a solution based on time domain conversion. This can be applied to applications related with the Internet-of-Things (IoT). These applications are present in any electronic devices where sensing is needed. To be able to reduce the area of the whole system with the required performance, micro-electromechanical systems (MEMS) sensors are used in these applications. We propose a new family of sensor readout electronics to be integrated with MEMS sensors.
Sensing a physical property has been one of the main field of investigation since a long time, sensing temperature or pressure have been possible since XVIII century. However, the world has change a lot since then, different kinds of sensors and connections between them have been developed. In fact, precision, size and connectivity have been the main goals of improvement. Nowadays, a new family of sensors have been used and tested all over the world. They are called Smart Sensors. It is because they are able to calibrate, compensate and transmit data with microprocessor circuits, opening the possibilities of use the sensing data. The evolution of this kind devices has helped in the growing of a new market called, Internet-of-Things (IoT). In the market of IoT, all the devices are connected through the network, sharing all the information. In the past decade, the applications related with the Internet-of-Things (IoT) have grown exponentially. In this field, smart sensors for environmental measurements (humidity, pressure, temperature and gas) are one of the most demanded products. All the previous examples have in common the bandwidth that they share. The main consequence is that most of the sensing activity will have similar readout outputs. This implies a big effort in developing new kinds of interfaces that are able to work with different type of sensors. To try to make as similar as possible the circuitry for different kind of sensors with a low-cost and energy efficient solution, Micro-Electro-Mechanical Systems (MEMS) sensors have arrived as the next generation of sensors for this purpose. MEMS have the property of being very small in size keeping high performance. Also, they have a low cost per unit thanks to their process of manufacturing. For these reasons, MEMS are mainly used in IoT applications. Similar to a microelectronics component, MEMS are able to be produced in big numbers. To try to make the output behavior similar between different sensing activities, capacitive MEMS are mostly used in IoT. This is one of the reasons why Capacitive-to-Digital Converters (CDC) are selected as one of the main products to used with sensors. They are able to digitize the signal of the MEMS with low power and high resolution which is the main interest in the market. A CDC is composed by an acconditioning signal circuit plus an analog-to-digital converter (ADC).
Capacitance-to-Digital Converter (CDC) is the most used electrical circuitry with capacitance sensors. Any kind of sensor that has an output in the capacitance domain is a candidate to be used with this topology. However, the trend to create circuitry that is able to be connected with different interfaces is the main reason of the upcoming interest in CDC topologies, specially inside the market of IoT. This is possible thanks to some constrains in the applications that were mentioned before. All the variables that are going to be measured of different applications have something in common. The bandwidth of all these signals is close to DC (BW < 50 Hz). It helps to target the bandwidth of interest to this value inside the CDC parameters of conversion and therefore to be able to measure all the different sources.
There are a lot of different topologies of converters used in this type of applications. The amount of converter types is just a consequence of all the types of applications that demands different performances or with different level of relevance. The main constrains for IoT applications are: resolution, power consumption, area, flexibility, voltage power supply, complexity and robustness. Finding a solution that satisfy all of them with a high acceptance level is still out of the limits of the solutions developed. Nowadays, depending on the priority of each constrains, the designers choose a different topology to optimize their design. The main types of converters used are: Sigma-Delta (ΔΣ) ADCs Its high resolution and linearity together with its intrinsic tolerance to analog non-ideal behavior make this option very popular for these applications. However, to reach the target resolution, large area and power demanding blocks are needed. In addition, when low voltage supplies are used, the performance of SC ΔΣ converters is reduced.
Successive approximation register (SAR) converters are good for scaling the CDC with low power consumption.
Incremental Data Converters (IDCs) are simple and flexible to different inputs. Also, can reduce the power consumption.
Voltage Controlled Oscillator (VCO). It’s a converter that mainly uses digital blocks. The power consumption is one of the lowest of all the converters. In addition, the flexibility of digital programmability gives the chance to be connected to multiple sensors.
Integration converters, like Dual-Slope (DS) are used for low BW application reducing the area and power consumption even further.
Hybrid converters (like SAR + Incremental) takes the properties of the previous topologies.
To save power and area keeping the same performance of high resolution CDCs, this thesis explains the development of a topology that is able to detect small capacitance changes while reducing ADC complexity, which is, typically, a critical block area and power wise in the CDC. To be area and energy efficient the proposed ADC is based on the single-bit noise-Shaping Integrating DS converter that maps the amplitude information into the time domain. Compared to traditional Dual-Slope ADCs, it introduces quantization error noise-shaping as in standard ΔΣ ADCs. But it does not require multi-bit circuits (i.e.: flash quantizers or n-bit digital-to-analog converters (DACs)) to keep high resolution and performance; instead, it uses single-bit circuitry to exchange amplitude by time resolution allowing to use lower supplies voltages. In addition, this topology has an intrinsically small sensitivity to temperature and process variations.
Within the ADCs, Dual Slope (DS) topology is very interesting to explore a new compromise between performances, area and power consumption. DS topology has been extensively used in instrumentation. The simplicity and robustness of the blocks inside classical DS converters it is the main advantage. However, they are not efficient for applications where higher bandwidth is required. To extend the bandwidth, DS converters have been introduced into ΔΣ loops. This topology has been named as integrating converters. They increase the bandwidth compare to classical DS architecture but at the expense of higher complexity. In this work we propose the use of a new family of DS converters that keep the advantages of the classical architecture and introduce noise shaping. This way the bandwidth is increased without extra blocks. The Self-Compensated noise-shaped DS converter (the name given to the new topology) keeps the signal transfer function (STF) and the noise transfer function (NTF) of Integrating converters. However, we introduce a new arrangement in the core of the converter to do noise shaping without extra circuitry. This way the simplicity of the architecture is preserved.
We propose to use the Self-Compensated DS converter as a CDC for MEMS sensors. This work makes a study of the best possible integration of the two blocks to keep the signal integrity considering the electromechanical behavior of the sensor.
The purpose of this front-end is to be connected to any kind of capacitive MEMS sensor. However, to prove the concepts developed in this thesis the architecture has been connected to a pressure MEMS sensor.
To obtain experimental results, a prototype of the proposed CDC was fabricated in a standard digital 0.13 µm CMOS technology. The CDC was bonded together with a pressure sensor MEMS to minimizes the effect of parasitic capacitances between them. The bonding is done in a ceramic carrier of 64 pins. The cover of the package has a drill on top of it to give the possibility of controlling the air pressure inside this cavity. The CDC core has an area of 0.317 mm2 and a mother clock of 1.28 MHz (which gives a sampling frequency of 160 kHz). The chip is connected to a 1.5 V power supply and it consumes 146 µA. This current includes analog blocks, digital blocks and excitation signal generator blocks.
Ten different samples have been measured to study and verify the robustness of the design. To measure these samples the pressure controller was attached to the top of the package. A constant pressure with an error of ±0.1 Pa. The measured equivalent integrated noise over a bandwidth of 10 Hz is 4.5 µVrms. Using the main formula to calculate the SNR in CDCs the obtained integrated noise leads to an SNR of 103.9 dB or an equivalent ENOB ≈17 bits.
The purpose of this implementation is to reach high resolution for differential measurements. To achieve this, the chip needs to distinguish small changes of pressure from following measurements. For that, another experiment has been done. Keeping a constant pressure, different measurements have been done and processed. In an ideal design, the output of all of them (with the same input pressure) would be exactly the same. However due the noise and non-idealities of the circuits and implementation, some noise will affect the performance. Because of that, the different digital output values will be taken out of the CDC for the same input pressure. This data will follow a normal distribution with the average on the theoretical value of the transfer function and a deviation that will be related with the error of the CDC Taking in account that in this experiment the input data would be considered as a constant DC value through all the acquiring of data, an alternative way for measure the resolution of the chip can be used and contrast that the values obtained for static measurements (with the FFT) are valuable. For this reason, to measure the resolution of differential measurements, instead of using an FFT, the standard deviation of all the samples is used. Out of 100 of measurements, the results give a σrms= 4µVrms which leads to a ΔC= 7 aF and ΔP= 1 Pa.
Measurements in the whole range of input capacitance have shown a good level of accuracy for differential measurements. To give a constant resolution in the whole range of the application was one of the main concerns of this application. Differential measurements in continuous time mode have make possible to detect minimum variations of pressure in different parts of the transfer function In this work, an incipient new family of converters is presented. The proposed Self-Compensated noise-shaping DS device can achieve the same performance of the topologies that use ΔΣ properties but using one-bit circuitry and time domain for the conversion. This fact gives also the benefit of low power architectures like SAR or classic DS. In this scenario, the proposed CDC that works in the time domain using a hybrid architecture offers a power efficient solution using very simple and robust implementation. In addition, its digital control gives an easy scalability and because of that, the solution can be adapted to different sensors just by a digital configuration. The topology presented is an efficient solution for the applications mentioned inside the IoT market. Thanks to the time domain conversion this topology doesn´t suffer from reduction of the voltage supply which is one of the new main constrains for IoT applications. This feature plus its robustness, make the prototype presented in this work a good solution for high resolution demanded CDCs in the low frequency domain.
