A Low-Power Noise Scalable Instrumentation Amplifier for Fetal Monitoring Applications S. Song, M. J. Rooijakkers, C. Rabotti, M. Mischi, A.H.M. van Roermund, E. Cantatore Department of Electrical Engineering Eindhoven University of Technology, The Netherlands Abstract-This paper proposes a low-power noise scalable instrumentation amplifier (IA) for fetal monitoring applications. The noise specification of the IA is made adaptive to the peak-peak value of the fetal electrocardiography (fECG) signal, which varies for different gestational age and measurement settings. Contrary to the currently available point solution IAs, the proposed IA is scalable for a noise range from 30nV/ to 250nV/ while consuming 15μW to 1μW respectively. A new IA architecture is proposed to achieve a better noise efficiency factor (NEF), while allowing noise scalability. The IA is designed in TSMC 0.18μm CMOS process. Simulation results show that the IA achieves a NEF of 3.4 to 5.5 over the noise scalable range, a CMRR of 100dB, and an input impedance (Z in ) of 1G. I. INTRODUCTION High-risk pregnancies are becoming more and more prevalent because of the progressively higher age of pregnant women. Regular monitoring of fetal electrocardiography (fECG) can reduce complications and mortality, but it would generate extremely high costs with hospital based systems. Currently available home-based fetal monitoring systems including a radio, an ADC and instrumentation amplifiers (IA) to be connected with cables to the electrodes consume high power, so a bulky battery is required making the system not suitable for ambulatory monitoring. Future fetal monitoring systems should be therefore low cost, low power, integrated and applicable to continuously portable monitoring. As the first step towards the implementation of an ultra-low power fetal monitoring system, the IA is being investigated. Intensive research has been done in biomedical circuits to minimize the power consumption of IA, resulting in several designs with different noise specs and noise efficiency factor NEF (this describes how many times the noise of a system with the same current and bandwidth is higher compared to that of an ideal bipolar transistor) [1]. The amplitude of the fECG strongly depends on gestational age, inter-electrode distance, and measurement orientation, while the required signal to noise ratio (SNR) of the fECG is fixed. Therefore a noise scalable IA covering a wide range of input-referred-noise levels would offer better energy efficiency than a point solution specified for the worst case. In this paper, a low power noise scalable IA is proposed and optimized in TSMC 0.18μm technology. A new architecture is used to obtain high CMRR and input impedance (Z in ) while minimizing the power. The IA noise can be scaled from 30nV/ to 250nV/ while consuming 15μW to 1μW with a NEF from 3.4 to 5.5. The paper is organized as follows: in section II, the signal characteristics of the fECG are analyzed and the specifications for the IA are determined. In section III, IA topologies are compared and a new architecture for low power is proposed. The power consumption of the IA is analyzed and optimized, enabling noise scaling with good NEF. In section IV, the IA noise scaling is discussed. Simulation results are given in section V, and conclusions are drawn in section VI. II. FETAL ECG SIGNAL AND IA SPECIFICATIONS Various physiological signals play an important role in fetal health monitoring, the most important of which is the fECG, which can be used to determine the fetal heart rate (fHR) with beat-to-beat accuracy, and also allows for analysis of the fECG morphology [2]. Most of the spectral energy of the fetal QRS-complex, which has an amplitude between 3μV and 20μV when measured on the abdomen, is contained in the range between 20 and 60Hz [3]. Reliable detection of the fECG poses a problem due to the plurality of interference sources which can be present in ambulatory abdominal measurements. The most prominent ones are the maternal ECG (mECG) and the electrohysterogram (EHG) signal that originates from muscle activity during uterine activity. To enable reliable fECG detection, the influence of these interference sources needs to be reduced. This can be achieved to a large extent with suitable algorithms implemented in a digital signal processor (DSP) [2]. Fig. 1: SNR of the fECG in six bipolar derivations as show by the diagram on the right for a 20 minute segment The amplitude of the fECG strongly depends on gestational age, inter-electrode distance, and measurement orientation [2]. Fig.1 shows an example of the influence of measurement direction on the SNR of an abdominal fECG, which is defined as .    where V PP.QRS is the peak-peak value of the fetal QRS complex and  is the total integrated noise in 0.5 to 100Hz band. The SNR of various electrode pairs can differ significantly and change over short time due to the movement of fetus. Therefore, a dynamic selection of the electrode pair with the best SNR during the measurement is essential to obtain reliable fECG detection. To determine the SNR needed to obtain reliable fetal R-peak detections, increasing amounts of white noise were added to fECG signals with low interference obtained from abdominal recordings, finding out the detection error rate (DER) at various SNR given by the algorithm in Ref. [4]. This research is supported by the Dutch Technology Foundation STW 978-1-4673-5762-3/13/$31.00 ©2013 IEEE 1926