No abstract

A system-on-chip passive UHF RFID tag with embedded temperature sensor is developed in a standard 0.18µm CMOS process for the EPC Gen-2 protocol from 860-960MHz [1]. Flip-chip technology is used to bond the developed tag IC to an antenna to realize a complete tag inlay, which is successfully demonstrated and evaluated in real-time wireless communications with commercial RFID readers. Figure 17.1.1 shows the block diagram of the tag IC. Multiple supply voltages, generated by a power management unit (PMU) using a separate-storage-capacitor technique to save area, are employed to optimize the performance of individual building blocks, while minimizing the total power consumption. A dualpath clock generator is used to generate on-chip clocks for signal processing to support both applications with either very accurate link frequency or very low power consumption. A low-power temperature sensor is also embedded with a gain-compensation technique that makes use of current correlation from the same bandgap reference (BGR) for the clock generator to reduce the temperature sensing error due to process variations.Figure 17.1.2 shows the PMU. The rectifier (RECT1) supplies the 670nA BGR while the core RF-DC conversion is performed by a triple-output rectifier (RECT2). Further supply regulation is done using low-dropout regulators (LDRs). For the duration of PW (=Tari/2) [1], as the RF input is significantly attenuated due to the data modulation, the rectifier becomes inactive, and a storage capacitor C S is employed to supply both the load current I L and the reverse leakage current I leak that flows back to the rectifier, which results in a ripple voltage of VR CS =(I L +I leak )×PW/C S . In existing work [2], all the blocks share one single capacitor C S , which needs to be large enough to meet the ripple voltage requirement of the most noise-sensitive block. To provide a highly stable current for the dual-path clock generator, the BGR is required to have a small supply ripple of 0.1V, and, as a result, the single C S would need to be at least 1.75nF to supply the nominal total I L of 14µA when Tari=25µs. In our work, three separate capacitors, C Sx (x=1,2,3), are employed for different blocks with different optimal ripple voltages (0.4V, 0.25V, 0.1V, respectively), which helps to significantly reduce the total capacitance to 805pF, even with >10% margins. The switches S x (x=1,2,3) are controlled by the demodulator's output to cut off I leak during PW, which helps further reduce both the sizes of C Sx and the required input power for replenishing. M ST1 and M ST2 are used for start-up. High voltages of 3.5V and 7.8V, used for programming the OTP memory, are generated by three charge pumps (QPs). During the WRITE operation, a VCO is activated to regulate the QP_VPP's output at 7.8V with an output current up to 20µA. With sufficient input power, the power detector (PD) sends a Power-Good (PG) signal to turn on S 4 to power up LDR 3 for the injection-locked frequency divider (ILFD) in the dual-path clock generator. Figure 17...

Abstract-This brief presents a CMOS temperature sensor suitable for ultralow-power applications. With a MOS transistor operating in the linear region, a linear relationship between delay and temperature can be obtained. A differential sensing architecture is utilized to reduce the signal offset and increase the effective signal-to-noise ratio. A design methodology concerning power optimization and improved sensor linearity is also presented. The sensor, which occupies 0.0324 mm 2 , is fabricated using the TSMC 0.18-µm one-polysilicon six-metal (1P6M) process. Measurement results show that the sensor consumes 405 nW with a 1-V supply at 1 ksample/s at room temperature. An inaccuracy value of −0.8• C to +1• C from 0 • C to 100 • C after calibration is achieved.Index Terms-CMOS temperature sensor, ultralow-power applications.