ISSCC 2019 / SESSION 22 / PHYSIOLOGICAL MONITORING / 22.

22.1 A 769µW Battery-Powered Single-Chip SoC with BLE for To achieve a high effective DR without relying on >16b ADCs, a digitally assisted
Multi-Modal Vital Sign Health Patches DC current-cancellation loop is used [4]. The DC component is cancelled with a
6b current DAC directly at the input of the readout, resulting in a 75μA maximum
input current. The PPG readout channel consumes 121μW while achieving a 28pA
Mario Konijnenburg1, Roland van Wegberg1, Shuang Song2, rms input noise in 20Hz band and 111dB effective DR with a 13b SARADC (pulse
Hyunsoo Ha1, Wim Sijbers2, Jiawei Xu1, Stefano Stanzione1, frequency at 2kHz, 16× oversampling). Figure 22.1.3 shows a PPG waveform
Chris van Liempd1, Dwaipayan Biswas2, Arjan Breeschoten1, recorded using a finger sensor with infrared light and an SpO2 waveform derived
Peter Vis1, Chris Van Hoof1,2,3, Nick Van Helleputte2 from red and IR PPG during breath holding. In an HR detection only scenario the
total LED power is reduced to 205μW because a 0.67% duty cycle is used which
1
imec - Netherlands, Eindhoven, The Netherlands is a 52% reduction compared to the state of the art [5] (Fig. 22.1.6).
2
imec, Leuven, Belgium
3
KU Leuven, Heverlee, Belgium A 13b SARADC (Fig. 22.1.4) is reused in ECG, BIOZ and PPG channels due to its
flexibility in exchanging power for resolution via oversampling. The BIOZ and PPG
Continuous vital-sign monitoring is of paramount importance in remote heath channel with a low BW of interest require a larger DR than ECG, and hence a
monitoring or rehabilitation environments for chronic diseases. Medical-grade higher oversampling ratio is employed. However, oversampling only reduces the
wireless and wearable bio-sensor systems that can be used at home offer a much noise, providing no improvement in linearity. Therefore, the voltage DAC
more attractive solution than hospital-based monitoring systems. We report an employs randomization dynamic element matching in the 6 MSBs by a Pseudo-
all-in-one battery-powered SoC designed for low-cost single-use health patches Random Number Generator (PRNG) controlled thermometer-coded elements,
(Fig. 22.1.1), allowing continuous monitoring in a home setting to improve patient which turns the harmonic distortion into a noisy spectrum in the Nyquist band.
comfort and reduce cost of care by, e.g., reducing hospital stays. In addition to Apart from quantization, sampling noise and nonlinearity of the DAC, the
medical-grade signal quality, low power consumption is key in such a health patch comparator thermal noise also contributes to the total noise spectrum. The low-
system, to enable a comfortable form factor with miniature battery size and power dynamic pre-amplifier and latch-based comparator in [3] suffers from
prolong the operational lifetime to at least several weeks. time-varying and limited gain (8 to 12×). The CM voltage of the decision point
(Node 1,2 in Fig. 22.1.4) is difficult to set, because for the latch it should be low
The SoC (Fig. 22.1.1) has built-in support for concurrent and synchronized ECG to guarantee high gain, but for the pre-amplifier it should be high enough to keep
(2 channels), Bio-impedance (BIOZ) and PhotoPlethysmoGraph (PPG). Data the input devices in saturation to reduce IRN. Figure 22.1.4 shows the high-gain
processing is performed by dedicated digital filters and accelerators for DMA, comparator.  By inserting  cascode  transistors MN4/MN5 and gain boosting
FFT, matrix calculations and sample-rate conversion. In addition, an ARM Cortex- transistors MN6/MN7, the gain of the dynamic preamp is increased to 52× and is
M4f microprocessor performs system control and further data processing. less time-dependent. A high gain reduces the noise contribution of capacitors at
Secured boot and data transmission are important features in medical systems node 1,2 and the latch. As a result, the IRN of the comparator is reduced (2 to
for privacy and safety reasons. The SoC includes cryptography IP that supports 3×) compared to [3] for the same power consumption. The SARADC consumes
a wide range of protocols combined with an SRAM-based PUF solution that 0.9 μW for an ENOB of 12.8b in a BW of 0.5 to 150Hz at a sample frequency of
guarantees a chip-specific unique cryptographic key. The SoC has several 32 kS/s (OSR=64).
interfaces including a BLE radio and an USB2 interface. Two on-chip LDOs
generate from a single battery source (1.3 to 2.0V; e.g., a ZnAir battery): a Figure 22.1.5 shows the measurement setup executing a realistic typical (health
configurable voltage for the digital core (0.6 to 1.2V) and a fixed voltage (1.2V) patch) application. It covers concurrent and synchronous data collection from
for SRAM, BLE, AFE, and the on-chip PLL. The very high integration level of the ECG, BIOZ, PPG (2 streams, red and infrared LEDs, duty-cycled 10%) and
SoC allows a minimum of system component count (Fig. 22.1.1), which saves accelerometer. Additionally, on-chip calculation of HR from ECG, respiration rate
area, cost, power and weight. from BIOZ, and SpO2 from the PPG streams is performed. The accelerometer data
is processed to track activity for further data processing, e.g., motion artefact
Ultra-low-power, low-noise biomedical analog frontends are traditionally designed reduction. During data processing, the FFT, matrix processor, and sample rate
in older technology nodes due to their better noise, matching and leakage converter accelerators are used. The relevant processed data is sent via a BLE
performance than (deep) scaled technologies. To allow co-integration with link (duty-cycled 5%) to a viewer application on a computer. The SoC consumes
advanced low-power digital and RF blocks in a single-chip solution, this SoC is 769μW (from the regulated LDO outputs). The total average system power drawn
designed in a 55nm technology node. The ECG and BIOZ readouts are based on from the battery, including LDO, LEDs and accelerometer, is 2mW allowing an
a power-efficient current-feedback IA [1]. For the PPG channel, the main design operating time of more than 3 weeks on 2 ZnAir batteries in series. Finally, the
challenges are achieving a high dynamic range (DR) to keep signal integrity in impact of the system on channel noise is demonstrated (Fig. 22.1.5). In general,
presence of large DC component and low power consumption, which is dominated a slight increase of noise from 10.3 to 11.7mΩrms is observed. Figure 22.1.6
by the LEDs [1,2]. The LED power is usually as high as several mW, dominating benchmarks the SoC, the PPG channel and the SARADC. Summarizing, an all-in-
the system power consumption. one (highest level of integration) low-power SoC is proposed for low-cost
miniaturized wearable devices (e.g., Health Patch). The improved comparator in
A pulsed LED synchronized to the readout reduces the power consumption. The the SARADC for all readouts, and specifically the differential PPG readout, assures
signal quality is determined by the peak current and frequency of the LED pulses. medical standard signal quality for ECG, BIOZ and PPG with lowest power
Hence short pulses and fast readouts should be used to reduce LED power consumption among reported solutions (Fig. 22.1.6). Figure 22.1.7 shows the
consumption. The achievable pulse widths are usually limited by the die, sized 4320×4320μm2.
PhotoDetector (PD) capacitance and amplifier input impedance. Existing single-
ended readout solutions provide a DC bias for the PD that reduces its parasitic References:
capacitance, however the DR is limited. Fully differential solutions provide a high [1] M. Konijnenburg, et al., "A Multi(bio)sensor Acquisition System with Integrated
DR but lack DC biasing capabilities for the PD [4]. Figure 22.1.2 shows the fully Processor, Power Management, 8 × 8 LED Drivers, and Simultaneously
differential PPG readout frontend, which achieves high DR and a bias for the PD Synchronized ECG, BIO-Z, GSR, and Two PPG Readouts," IEEE JSSC, vol. 51, no.
at the same time. This PPG readout includes a current-sensing stage and a 11, pp. 2584-2595, Nov. 2016.
transimpedance stage. In the first stage, the sub-amplifiers A1 and A2 keep the [2] P. Schönle, et al., "A Multi-Sensor and Parallel Processing SoC for Miniaturized
voltage on both nodes of the PD to the desired value, set by VBIASP and VBIASN. Up Medical Instrumentation," IEEE JSSC, vol. 53, no. 7, pp. 2076-2087, July 2018.
to 0.6V DC bias voltage can be provided, resulting in about 30% reduction in the [3] M. van Elzakker, et al., "A 10-bit Charge-Redistribution ADC Consuming 1.9μW
parasitic capacitance. Moreover, the input impedance is reduced by both the local at 1 MS/s," IEEE JSSC, vol. 45, no. 5, pp. 1007-1015, May 2010.
and main feedback loop in the current-sensing stage (Fig. 22.1.2), resulting in a [4] J. Xu, et al., "A 665μW Silicon Photomultiplier-Based NIRS/EEG/EIT Monitoring
<100Ω input impedance. Therefore, the amplifier enables a power-efficient 10μs ASIC for Wearable Functional Brain Imaging," ISSCC Dig. Tech. Papers, pp. 294-
level LED pulse (Fig. 22.1.3) compared to 100μs pulses [5]. 295, Feb. 2018.
[5] A. Sharma, et al., "A Sub-60-μA Multimodal Smart Biosensing SoC With >80-
dB SNR, 35-μA Photoplethysmography Signal Chain," IEEE JSSC, vol. 52, no. 4,
pp. 1021-1033, Apr. 2017.

360 • 2019 IEEE International Solid-State Circuits Conference 978-1-5386-8531-0/19/$31.00 ©2019 IEEE

Authorized licensed use limited to: CP Chen. Downloaded on August 03,2020 at 09:10:34 UTC from IEEE Xplore. Restrictions apply.
 ISSCC 2019 / February 20, 2019 / 8:30 AM

Figure 22.1.1: A fully battery-powered bio-medical multi-sensor acquisition

ASIC for wearable medical devices. A relatively low amount of external Figure 22.1.2: A two-stage fully differential PPG readout frontend together with
components makes miniaturization easy to be performed. a DC servo loop, providing a reverse DC bias voltage for the PD.

Figure 22.1.3: PPG measurement results, an IR PPG waveform, an extracted Figure 22.1.4: A 13b SARADC used in ECG, BIOZ and PPG channel. By adding
SpO2 waveform, the input referred noise spectrum when enable/disable the cascode transistors MN4 and MN5 and gain boosting transistor MN6 and MN7
radio and the differential output of the TIA, showing a 10µs level LED pulse the preamp gain is increased to 52× and becomes less time-dependent. An
width can be used. ENOB of 12.8b and an SNR of 82.8dB are achieved.

22

Figure 22.1.5: An health-patch setup executing a typical application (ECG, BIOZ,

and PPG) consumes on average 2mW, powered by two ZnAir coin-cell batteries.
The SOC consumes 769µW. The noise density of analog readouts is slightly Figure 22.1.6: Benchmark tables for SoC, PPG channel (readout and LED driver)
(~10%) affected when radio and digital processing are active. and SARADC.

DIGEST OF TECHNICAL PAPERS • 361

Authorized licensed use limited to: CP Chen. Downloaded on August 03,2020 at 09:10:34 UTC from IEEE Xplore. Restrictions apply.
 ISSCC 2019 PAPER CONTINUATIONS

Figure 22.1.7: Die photo. The size of the die is 4320×4320µm2.

• 2019 IEEE International Solid-State Circuits Conference 978-1-5386-8531-0/19/$31.00 ©2019 IEEE

Authorized licensed use limited to: CP Chen. Downloaded on August 03,2020 at 09:10:34 UTC from IEEE Xplore. Restrictions apply.