UC San Diego

Over the past several decades, early diagnoses and disease monitoring that rely upon biomolecular testing are the primary factors that have led to the substantial increase in average life expectancy. Molecular tests, which analyze patient samples for disease-specific biomarkers, are becoming the basis of the majority of diagnoses and therapy monitoring. Point-of-care (PoC) diagnostics uses a portable analytical device for accurate and fast tests to avoid frequent clinic visits and long turn-around time. Among biosensing techniques, magnetic sensors take advantage of the intrinsic lack of magnetic background in biological samples to achieve high sensitivity and are compatible with semiconductor-based fabrication processes to enable low-cost and small-size devices for PoC applications.

In this dissertation, magnetic sensor analog front-ends (AFEs) are designed to measure the signal from magnetoresistive (MR) sensors and overcome challenges such as small signal to baseline ratio, 1/f noise, and temperature drift. Two sensing techniques, magnetometry and magneto-relaxometry (MRX), are discussed and compared. Printed circuit boards (PCBs) and CMOS chips are designed to implement both techniques.

First, a CMOS chip based on magnetometry is presented, which reduces the baseline using a double modulation scheme and a reference sensor. The residual baseline from the sensor mismatch is further reduced using a high frequency interference rejection (HFIR) sampling technique embedded in the ADC. A fast settling duty-cycled resistor (DCR) is used to reduce the AFE settling time, thus enabling a readout time that is 22.7× faster than the state-of-the-art. This work results in sub-ppm sensitivity and a sensor mismatch tolerance of up to 10%.

While promising, the sensor mismatch still limits the baseline cancellation. MRX measures the relaxation signal after removing the excitation magnetic field, thus enabling baseline-free detection. PCBs, including an AFE and an electromagnet driver that can collapse the magnetic field within 10 μs, were designed to validate the time-domain MRX. The signal dependency on the sensor coverage, applied field strength, and magnetization time was investigated.

Lastly, a CMOS chip based on MRX was designed that uses magnetic or magnetoresistive correlated double sampling to reject the systematic 1/f noise. Moreover, a fast settling Miller compensation (FSMC) technique was presented to save the power, while maintaining the amplifier’s linearity and stability. As a result, this work achieves the best-reported magnetic sensor figure-of-merit (FoM).

These works enable ultrasensitive, broad dynamic range, and fast response magnetic sensing systems towards PoC diagnostics.