4. Optoelectronic Acquisition System Design The present chapter deals with the design of the optoelectronic (OE) system required to translate the obtained optical modulated signal with the photonic acquisition system (described in Chapter 3). The OE module includes: a transimpedance amplifier (TIA) and a filtering and amplification stage. Processing of the bioelectric signal is one of the most important stages of a bioelectric sensor since it allows to isolate the signal of interest, while preserving its content and integrity. Although filtering and amplification are crucial in most of bioelectric signal measurements, the focus of this chapter is the design and optimization of the front-end component between the photonic stage (Chapter 3) and the signal processing electronics. 77
4.1 OE Conversion Module Photonic Platform for Bioelectric Signal Acquisition on Wearable Devices The OE conversion module refers to the signal translation and conditioning required to transform the photonic stage output optical signal into a bioelectric signal output. Part of the information contained in the overall bioelectric signals may not be readily available from the raw signal, requiring further analog or digital processing for feature extraction. Therefore, the design of OE system for bioelectric signals must comprise appropriate instrumentation to guarantee the integrity of the information, whereas minimizing interference or artifacts. In this way it s possible to guarantee that the acquired signal is useful for health condition monitoring or diagnostics, since a corrupted signal could provide misleading or meaningless information. In Chapter 1, the typical optical sensor acquisition diagram was presented (Figure 1.4), which included a photodetector, amplification, filtering, A/D converter, processing unit and a possible wireless data transmission component. The OE system may include all, or some of these components, being the mandatory for bioelectric signal acquisition: photodetection (photodiode & transimpedance amplifier), amplification and filtering. The remaining blocks in Figure 1.4, are optional, although in the specific application on wearable devices, the use of A/D converters and wireless transmission data are fundamental 4.1.1 Current-to-Voltage Conversion OE conversion is divided in two blocks: photodetection and current-to-voltage (I-V) conversion. The first was explained in Chapter 3, and is the bridge between the light modulation and the electronics used to process and interpret this phenomenon. The electrical current obtained at the output of the photodiode needs to be converted to a readable voltage in order to make the signal compatible with the remaining electronics [1, 2]. TIAs are used to convert the current signal from the photodiode, (i ph ), into a voltage signal, (v out ), providing a high-transimpedance gain, (G TIA ), over the desired bandwidth. Figure 4.1 shows an example of a simple TIA circuit with the photodiode in the photovoltaic mode. As shown in Figure 4.1, the basic TIA comprises a combination of a photodiode in series with an opamp and a feedback resistor (R f ), being the latter responsible for providing amplification during I-V conversion. In addition R f prevents the development of any significant voltage at the circuit input, restoring the virtually zero voltage at the opamp input. The TIA circuit in Figure 4.1, however, presents disadvantages for applications where gain, bandwidth and stability are crucial. 78
Photonic Platform for Bioelectric Signal Acquisition on Wearable Devices Chapter 4 Figure 4.1 Standard circuit of a transimpedance amplifier with photodiode in the photovoltaic mode. The main challenges when designing TIA for high-gain applications are related with: DC offset errors, bandwidth and stability. In fact, the value of R f strongly influences the overall TIA performance. Therefore, compensation is often required to provide a stable TIA operation. This holds true particularly for TIA with high sensitivities, which is the case for the present application. High-sensitivity Limitations A common problem in TIA performance is related with DC offset errors mainly originated by two error currents that flow through the feedback component: input bias current of the amplifier and the leakage current of the photodiode. In fact, these currents will experience the same gain as the desired current signal, superimposing the output. Ideally, an opamp would cause!!! to flow through the R f. However, what really happens is that some of this input current is driving the opamp input terminal, causing the input bias and photodiode leakage current to overlay the desired signal. Therefore, the selection of the opamp must envisage a small bias current, since as higher this parameter is, less photodiode current goes through the R f. Also, in this way, higher current errors are generated. Another way to solve this, is to use a compensation resistor in parallel with a capacitor, which together develop an offset component that neutralizes the existent error. Another common problem when dealing with high-gain TIA is a phenomenon called gain peaking that induces an unstable behavior. Gain peaking is characterized by a ringing effect on the output voltage of the circuit. Any noise signal can trigger an unstable circuit into oscillation. In order to maintain stability, and therefore avoid gain peaking, a feedback capacitor (C f ) is placed across R f, creating a pole (f p ) in the Noise Gain (NG) function. The NG corresponds to the non-inverting closed-loop gain, and represents the amplification of the 79
Photonic Platform for Bioelectric Signal Acquisition on Wearable Devices inherent voltage noise of the opamp. A graphical analysis provides a better understanding of the effect of C f in gain peaking prevention (Figure 4.2). Figure 4.2 Bode plot of NG and opamp Open Loop Gain. The inset shows the gain peaking effect on the I-V response curve. As shown in Figure 4.2, the NG exhibits a flat response until a pole is created due to the input capacitance of the TIA and R f. After this pole, the NG begins to rise, but still within the open-loop gain of the opamp, whereas this curve simultaneously drops. At a certain point, the opamp lacks sufficient gain to support the feedback demands, i.e. exceeds the open-loop gain of the opamp. This is the critical point of the circuit that defines the gain-peaking threshold. At frequencies beyond this, i.e. frequencies higher than f GBW, the opamp doesn t have sufficient gain to support the feedback (R f //C f ) demand, and the circuit s I-V response rolls off, leading to circuit instabilities and gain peaking (Figure 4.2). The selection of C f and R f should, therefore, be made in order to position the high-frequency pole (f p ) of the TIA inside the open-loop gain opamp curve. This leads to an unconditional stability. The only associated problem regards to bandwidth limitations, which is not the case for bioelectric signal applications, where frequencies are in the range of few Hz to khz. Figure 4.3 shows the TIA circuit with phase compensation to prevent gain peaking, as well as the equivalent model of the photodiode. As shown in Figure 4.3, the photodiode is represented by an equivalent model that includes a current source (i D ), a shunt resistance (R sh ), a junction capacitance (C j ), a diode (D) and a series of resistance (R s ). The latter passive component is related with the semiconductor material used in the photodetector and is generally negligible due to its low values. Regarding TIA overall performance, the most important photodiode elements are C j and R sh, since they define the stability of the circuit, as 80
Photonic Platform for Bioelectric Signal Acquisition on Wearable Devices Chapter 4 well as the sensitivity and bandwidth. These elements will be further discussed in the design considerations section. Figure 4.3 TIA circuit with phase compensation and photodiode electrical equivalent. Figure 4.3 also shows a compensation component formed by the parallel between R C and C C, which main role is to develop an offset component that will neutralize the existent DC error produced. However, the R f value for high-sensitivity circuits is higher, and R C may introduce greater diode leakage. This leads to an increase of the net offset error. Therefore, this counteracting offset effect may not be the most appropriate solution, and it depends on the application and the settings used in the TIA design. The overall transfer function of the circuit in Figure 4.3 is described as [1]:!!"# =!!!!!!!!!!"!!!! + (!!"#$!!! ) (!!"#$!!! +!!"#$#%"!! (4.1) where!!"#$! and!!"#$! are the input bias current into the non-inverting and inverting inputs of the opamp, respectively. The leakage current (!!"#$#%" ) of the photodiode is also present at the TIA output and amplified by the feedback component. As explained before, this current is induced by the compensation resistor R C, through the relationship:!!"#$#%" =!!! 1!!!!"#$!!!!!! (4.2) where!!! is the thermal voltage. 81
Photonic Platform for Bioelectric Signal Acquisition on Wearable Devices The relationship between the feedback element and the opamp of the photodiode amplifier determine the well functioning of the TIA. There is a relationship by which an optimum C f is calculated to a specific R f or TIA gain, expressed by:!! =!!! 1 + 1 + 4!!!! (4.3) where C i combines the three input capacitances of the circuit: photodiode (C j ) opamp common mode (C cm ) and differential (C diff ). The capacitor C c can be considered a virtual capacitor determined by the relationship of R f and the gain-bandwidth product of the opamp (!!"# ):!! =!!!!!!!"# (4.4) Design Considerations In order to guarantee proper photodiode amplifier circuit operation, some basic requirements for the photodiode and opamp must be met. These requirements seek to increase the signal of interest, in regards to the noises imposed by the photodiode, opamp, and other elements, such as powering and electromagnetic interference (Table 4.1) [1, 3]. Table 4.1 Design consideration for TIA design. Element Parameter Considerations Photodiode Opamp Biasing technique Capacitance Dark current Input capacitance Gainbandwidth product Bias current In order to increase sensitivity, the photodiode must be used in the photovoltaic mode, i.e. with zero-bias operation. In this way, the dark current offset errors are not generated by the leakage current. This should be as low as possible since together with the R f, forms a noise-gain zero (feedback pole), affecting frequency and noise. Should be as low as possible, since this current is responsible for DC errors at the output. This parameter contributes to the total capacitance that defines one of the poles in Figure 4.2. As high as possible to prevent gain peaking. A trade-off must be made with the gain and feedback elements used. This value should be set as low as possible, since bias current causes voltage offset errors, specially in the presence of large- R f. 82
Photonic Platform for Bioelectric Signal Acquisition on Wearable Devices Chapter 4 Feedback component Resistor Capacitor Should be as high as possible in order to increase sensitivity. In general a high R f is better than implementing a T-network. Although the gain is in principle higher, the T-network sacrifices performance due to the generation of higher noise current. In addition, input offset voltage, drift and voltage noise are multiplied by the ratio of the T-network, which constitutes a drawback for high sensitive applications. Nevertheless, using an opamp with low bias current, can overcome the limitations imposed by T-networks, providing better gain adjustment. Usually a small capacitor is added to suppress oscillation, although for low-frequency signals, larger values can be used. However, the pole frequency introduced by this capacitor should be larger than the frequency of the desired signal in order to prevent attenuation. The design considerations shown in Table 4.1 strongly depend on the type of signals to measure, particularly regarding bandwidth and amplitude. It s important to follow these guidelines since the result of this stage, i.e. TIA output, provides with the voltage signal that will be use for further processing and feature extraction. 4.1.2 Signal Processing The actual output of a bioelectric signal acquisition device consists of several components: desired bioelectric signal, e.g. Electrocardiogram (ECG), Electroencephalogram (EEG) or Electromyogram (EMG); other bioelectric signals, for instance EMG may appear in ECG recordings; interference signals from power-line; motion artifacts and other interference from tissue/electrode interface; Inherent noise of the device. The acquisition device, in this case the photonic system described in Chapter 3, should inherently eliminate the undesired components, without requiring further processing. However, in real acquisition systems, these undesired components are present and can overlay the signal of interest, particularly in the case of EEG, where amplitudes are in the range of µv. In Chapter 2, an overview of the noises and interferences involved in wearable bioelectric signal acquisition was shown in Table 2.2. Those noises have amplitudes close, 83
Photonic Platform for Bioelectric Signal Acquisition on Wearable Devices and in some cases, higher than the bioelectric signals of interest. For example, skin motion artifact limit the gain factor of the amplification stages, since their amplitudes are typically between 5 to 15 mv. Thus, it s important to remove this component at an early stage [4]. Therefore, after OE conversion it s important to include a processing block featuring low noise amplification and filtering. This stage is also important in order to match the input range of analog-to-digital converters and to limit its bandwidth to the desired range. Filtering can be performed in two ways: analog or digitally. At a first stage, signals are filtered using analog technology that is implemented in the acquisition platform. These analog filters generally include band-pass and notch filters, which contribute for the removal of low and high frequency noise, as well as power-lines interference centered at 50/60 Hz. However, there must be a careful definition of the analog filter settings, since they may cause some distortion to the significant frequency components of the monitored bioelectric signal. Therefore, the design of analog filters must envision a linear phase frequency response over the spectral components included in the band of interest. In addition to this filtering stage, bioelectric signals can also be processed in its digital format, after A/D conversion. These filters have several advantages over analog filtering since they are easily handled in terms of settings. In fact, in the case of analog filters, it may be necessary to change passive and active components that may be already soldered to the circuit board. In addition, they are susceptible to environmental factors, such as temperature and corrosion. Nevertheless, it s preferable to first filter the signals in the analog format and if necessary, include digital filtering techniques. 4.2 OE Conversion System The output signal of the photonic stage, described in Chapter 3, needs to be converted back to an electrical value in order to perform signal processing and possible feature extraction. In addition to OE conversion, the TIA also performs amplification of the signal into a readable value. Nevertheless, this amplification is limited by the settings of the previous stage, i.e. photonic stage, such as input optical power and DC current levels. 4.2.1 TIA Design 84 Current-to-voltage conversion is performed by a high gain TIA, since it needs to detect even the small variations of current produced by EEGs. In fact, for input optical powers of
Photonic Platform for Bioelectric Signal Acquisition on Wearable Devices Chapter 4 10 mw, the resultant variations of electrical current at the output of the photodiode reach minimums of a few na. Nevertheless, the DC component of the modulated signal is higher in several orders of magnitude. For instance, considering the photonic setup settings specified in Table 3.4, the DC input current to the TIA would be 1.3 ma. This current imposes a limit on the G TIA, since in order to prevent opamp saturation, the output voltage to each terminal, should not exceed its supply voltage. In addition, high-level DC signals can also induce unintended biasing voltage on the following signal processing stages In order to prevent TIA saturation, an additional feedback block is used to compensate the high DC levels introduced (Figure 4.4). This DC suppression block works as a high-pass filter, and the low-cut frequency is adjusted according to minimal frequency component of the monitored bioelectric signals (Table 2.1). The TIA design used was based on Figure 4.3, i.e. including the C f to avoid gain peaking, although with the additional feedback block. Figure 4.4 TIA circuit schematic, with DC suppression block and compensation block. The overall transfer function (TF) of the circuit in Figure 4.4 takes into account both blocks and follows the typical feedback control system. Thus, TF(s) can be expressed as:!"! =!!!!!!!!!!!!!!!!!!!!!!!! (4.5) where natural frequency is expressed as!! = 1!!!!!!!!. 85
4.2.1 Circuit Dimensioning Photonic Platform for Bioelectric Signal Acquisition on Wearable Devices The TIA circuit design is performed according to the considerations shown in Table 4.1, whose main purpose is to prevent unsuitable operation of the OE conversion stage. First, it s essential to define the required TIA performance characteristics, particularly gain and bandwidth. As it was explained before, an upper limit to the G TIA has to be determined in order to prevent the opamp to reach a stage of saturation. On the other hand, there s a minimum value for this setting, since modulated currents that carry information about the monitored signal can reach a few na. Thus, dimensioning of the circuit in Figure 4.4 needs to take into account the requirements shown in Table 4.2. Table 4.2 TIA circuit requirements for gain and bandwidth. Lower (recommended) threshold Transimpedance Gain These calculations were based on: a minimum AC current of 10 na, respective to a measurement of an EEG; as well as a minimum output voltage of 1 mv. Upper threshold The upper threshold is only valid for DC currents, facing the high DC levels produced by an input optical power of 10 mw!!"# = 1x 10 5 V/A!!"# (@DC)= 1.25 x 10 4 V/A Bandwidth This value is estimated according to the lower and upper frequency components of interest of the main bioelectric signals: ECG, EEG and EMG. BW = 0.2 Hz 1 khz Besides requirements shown in Table 4.2, it s also important to take into account the remaining design considerations shown in Table 4.1. As it was explained before, highsensitivity TIA circuits are prone to oscillate facing inappropriate phase compensation, i.e., when the frequency pole of the transfer function (Figure 4.2) falls outside the stability region. In order to achieve this, the selection of the minimum recommended! needs to be calculated according to (4.3), and considering the following settings: 86
Photonic Platform for Bioelectric Signal Acquisition on Wearable Devices Chapter 4 - The photodiode is used in the photovoltaic mode, which reduces the effects of dark current in terms of offset errors, becoming less affected by power supply-related noises. The compromise of using this mode of operation is an increase of junction capacitance (!! ). However, typical! values are often in the range of pf, depending on the bias voltage and active area of the photodiode. For calculations and estimations, a capacitance of 5 pf was chosen based on commercially available PIN photodiodes for optical-fiber applications. - The opamp selection was made through a compromise between!!"##,!!",!!"#, and input bias current and offset voltage. The three first settings contribute to the selection of the right capacitor to bypass the R f. A complementary metal-oxide-semiconductor (CMOS) opamp with low bias current (OPA2337, Burr-Brown) can be used, with!!"## of 2 pf,!!" of 4 pf and a!!"# of 3 MHz. According to these values and using (4.3), the minimum recommended C f value for a R f range of 0.5 to 10 MΩ was calculated, and results are shown in Table 4.3. Table 4.3 TIA phase compensation results for a selected range of!!.!! (MΩ)!! (pf)!! (pf)!! (pf)!!!"# (khz) 0.5 0.10 11 1.13 280 1 0.05 11 0.79 201 5 0.01 11 0.346 91.7 10 0.005 11 0.244 65.1 As shown in Table 4.3, it s possible to use capacitors of lower values, which not only reduces the capacitive coupling of external signals, but also allows for high flexibility in defining the high frequency pole. In order to reduce band-pass range of the TIA, a higher capacitor (!! = 22!") can be used towards a high-frequency pole of 1 khz. Since the photodiode amplifier is a very high impedance and sensitive circuit, it requires a good shielding and effective power supply bypassing. Therefore, this circuit will take advantage of the use of batteries instead of a power supply subjective to drift. 87
4.2.1 Performance Assessment Photonic Platform for Bioelectric Signal Acquisition on Wearable Devices The performance of the TIA circuit can be simulated in TINA (Texas Instruments) in order to determine the optimum values for each component in Figure 4.4, as well as to investigate the resultant frequency response, gain and SNR. From the simulations, the optimum values for each component are the following:! = 22 pf,!! = 5 MΩ,!! = 10 MΩ and!! = 5 kω. Regarding!!, this component determines the TIA low-cut off frequency, and a careful selection must be made since higher capacitors contribute for an increase of the capacitive coupling of unwanted signals as well as crosstalk between different points of the circuit. Table 4.4 shows the TIA performance results in terms of bandwidth, SNR and total output noise for different!! values. Table 4.4 Performance results simulated in TINA for different C 1 values. C 1 ( µf) Bandwidth (Hz) SNR (db) 0.2-1 khz 10 Hz Total output noise (µv) 1 16 597 131.76 72.8 100.5 0.258 229 4.7 3.5 597 130.74 72.8 93.08 0.292 229 10 1.68 597 128.29 72.8 91.75 0.387 229 100 0.17 597 113.3 72.8 90.7 2.18 229 As expected, the SNR decreases for capacitors with higher value, since the noise pickup is higher due to capacitive coupling and crosstalk effects. In some cases, it may be an advantage to compromise the low-frequency components over a stable and high-sensitivity TIA. In fact, even if the low-cut frequency is set to 3.5 Hz, the components from this point to 0.2 Hz are still converted over a gain from 95-130 db. 4.3 Electrical Processing Unit The processing unit includes the components needed to: remove unwanted components or signals, and amplify the resultant signal if needed. Figure 4.5 shows the block diagram of the acquisition electronics designed for the photonic platform. 88
Photonic Platform for Bioelectric Signal Acquisition on Wearable Devices Chapter 4 Figure 4.5 Block diagram of the acquisition electronics, including an optional voltage amplifier. The power supply used may be a problem since it includes unwanted periodic oscillation called ripple. This artifact often has higher amplitudes than some of the bioelectric signals such as EEG. Therefore, to improve system stability as well as to remove this typical interference, batteries of 9 V, are used to power the acquisition electronics. 4.3.1 Band-pass Filtering In order to eliminate the interference from other bioelectric signals, as well as from environmental conditions (Table 2.2), analog band-pass filters with unity gain are used. The selection of filter range depends on the monitored bioelectric signal, as shown in Figure 4.5. Although both EEG and ECG have frequency ranges from 0.05 to 200 Hz, some clinical applications are only interested on data among a window of 0.2 to 40 Hz [5], [6]. In this case, the band-pass filtering allows to perform some extra attenuation of the 50 Hz power line interference. On the other hand, EMG signals have clinical importance considering spectrum components from 5 to 500 Hz [7]. The band-pass filter circuit implemented is shown in Figure 4.6 and is based on a Sallen-Key topology [8]. The transfer function of the filter is obtained by the product of the transfer functions of both cascaded components:!"(!) =!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! (4.6) where the low-pass cut-off frequency is defined as!! =!!!!!!! (4.7) 89
Photonic Platform for Bioelectric Signal Acquisition on Wearable Devices and the high pass cut-off frequency as:!! =!!!!!!!. (4.8) Figure 4.6 Circuit schematic of the Sallen-key band-pass filter. The selection of the values for resistances and capacitors is performed according to (4.7) and (4.8), and it s a good practice to first select the values for capacitors since the available values are more limited. Table 4.5 shows the calculated values for resistors and capacitors for each filter: 0.2 40 Hz and 5 500 Hz. The Integrated Circuit (IC) chosen for the band-pass filter were the low-noise high-speed JFET-Input operational amplifiers (TL071CP, Texas Instruments). Table 4.5 Optimum resistor and capacitor values for band-pass filter. R 1 R 2 R 3 C 1 C 2 C 3 ECG, EEG (0.2-40 Hz) EMG (5-500 Hz) 239 kω 119 kω 2.81 kω 9.58 kω 4.78 kω 225 Ω 4.7 µf 1 µf 2 µf In order to analyze the frequency response of the designed band-pass filter, the bandpass filter in Figure 4.6 was implemented in the software TINA, and the transfer function in (4.6) was simulated in Matlab. The results for both simulations are depicted in Figure 4.7. 90
Photonic Platform for Bioelectric Signal Acquisition on Wearable Devices Chapter 4 Figure 4.7 Frequency response of the band-pass filter for: ECG and EEG filter obtained in a)matlab from the transfer function and b) TINA from circuit simulation; EMG filter obtained in c) Matlab from the transfer function and d) TINA from circuit simulation. Arrows indicate the low and high cut-off frequencies. As shown, the results of both simulated and theoretical frequency response are similar for the two tested filters with unity gain (0 db) over the selected bandwidth. The cut-off frequencies are also very close to the expected and desired values, which suggest that the designed filters are adequate for each bioelectric signal acquisition system. 4.3.2 Notch Filtering A notch filter with an active twin-tee network topology is used to suppress the 50 Hz power-line interference. This filter is composed by two T-networks as shown in Figure 4.8. With this topology it is possible to achieve higher Q filters by using an operation amplifier configured as a voltage follower. 91
Photonic Platform for Bioelectric Signal Acquisition on Wearable Devices Figure 4.8 Circuit schematic of twin-t notch filter. The characteristic transfer function of this filter is:!"! =!!!!!!!!!!!!!!!!!!!!!!!!!!! (4.9) where the notch frequency is determined by:!!"#$! =!!!!!!! (4.10) The notch frequency should be centered at 50 Hz for European applications, and at 60 Hz if used in the United States. Figure 4.9 Frequency response of the notch filter obtained in a) Matlab from the transfer function and b) TINA from circuit simulation. Arrows indicate the notch frequency. 92 Similar to the band-pass filter, the circuit shown in Figure 4.8 was simulated in TINA and the transfer function (4.9) was plotted in Matlab. The values for resistors and capacitors
Photonic Platform for Bioelectric Signal Acquisition on Wearable Devices Chapter 4 were chosen according to a!!"#$! of 50 Hz, which results in a!! = 47 nf and in consequence a!! = 67.7 kω. Results for these settings are shown in Figure 4.9, where the simulated and theoretical notch frequency meets the expectations. Thus, the settings defined for each component in Figure 4.8 are the recommended for notch filter implementation. 4.3.3 Voltage Amplifiers Since bioelectric signals detected are commonly converted to digital signals for further processing and communication for instance to health monitors, amplification is usually required. Inverting amplifiers can be optionally included in this platform, with different gains depending on the monitored bioelectric signal. In particular, for EEGs high gains are required facing its typical amplitudes that can go down to 5 µv. The gain settings depend on the voltage gain produced in subsequent stages. The use of negative feedback over non-inverting amplifiers is related with their advantages such as stable gain and less signal distortion. 4.4 Performance Simulation of Overall OE System The overall acquisition electronics platform can be simulated in TINA in order to have a general picture of its performance. The implemented circuits are those shown in Figure 4.4, Figure 4.6 and Figure 4.8, with the described components. Regarding the TIA, the value for the capacitor C 1 was 100 µf, since the goal is to test the acquisition electronics setup for the best-case scenario where a low-frequency cut-off of 0.2 Hz is assured. First, the transfer function was plotted using both band-pass filter settings: 0.2 40 Hz and 5 500 Hz, being the results shown in Figure 4.10. Figure 4.10 Frequency response obtained in TINA for the overall acquisition electronics setup using band pass filter for a) ECG and EEG acquisition (0.2 40 Hz); and b) 5 500 Hz. 93
Photonic Platform for Bioelectric Signal Acquisition on Wearable Devices The results in Figure 4.10 show the evident 50 Hz power lines suppression performed by the twin-t notch filter, with an attenuation factor of 36 db. The band-pass range is according to the pre-set values for each bioelectric signal monitored. A noise analysis was also performed in order to assess the SNR of the overall circuit, as well as to determine the input noise, which only includes the noises of the components used, such as opamps, resistors, capacitors and photodiode. This analysis was performed for both band-pass filters, although only the results for ECG and EEG acquisition electronics setup are depicted in Figure 4.11. Figure 4.11 Simulation results obtained in TINA for the overall acquisition electronics setup in terms of a) Input noise; and b) SNR. As shown in Figure 4.11a, the input noise of the OE system includes the 1/f noise that is present in most active and passive devices. The input noise floor over the frequencies of interest is 1.88 pv/hz 1/2, which is in the limits described in Chapter 2 (EEG 1 µv; ECG 10 µv). Figure 4.11b shows the SNR for the overall circuit, which values are above the limit of 0dB, which means that, in principle, the designed acquisition electronics will overcome the input noises and provide sufficient resolution for bioelectric signal measurements. In fact, the minimum SNR (< 89.39 db) is obtained for frequencies above the highest component of interest for both ECG and EEG signals. Despite the results shown for the overall OE system, it s important to have in mind that in real applications, other sources of noises (e.g. photodiode noises) and interferences contaminate the signal of interest. These sources are not accounted in these results, although the efficiency of the filtering stage points for the effective isolation of the bioelectric signal monitored. 94
Photonic Platform for Bioelectric Signal Acquisition on Wearable Devices Chapter 4 4.5 PCB Design Once the OE system is well defined, is possible to design a printed circuit board (PCB) with the following. The PCB was designed using PADS (Mentor Graphics) and a single board was used for the entire OE system. First, the electrical circuit and the PCB layout were elaborate and are shown in Annex I. Figure 4.12 shows the resultant OE system equivalent PCB. The dimensions selected for the OE prototype were 4 cm x 8cm in order to fit the black box shown in Figure 1.5. Figure 4.12 PCB of the OE system designed for bioelectric signal acquisition. a) top view and b) bottom view. References [1] J. G. Graeme, Photodiode amplifiers. McGraw-Hill, 1995. [2] K. IIzuka, Elements of Photonics - Volume II. Wiley-Interscience, 2002. [3] Designing Photodiode amplifier circuits with OPA128, 1994. [4] E. Huigen, Noise in biopotential recording using surface electrodes, no. November, 2000. [5] V. K. Murthy, T. M. Grove, G. A. Harvey, and L. J. Haywood, Clinical Usefulness of ECG Frequency Spectrum Analysis, Proceedings of the Annual Symposium on Computer Application in Medical Care, no. 2, pp. 610-612, 1978. [6] E. Niedermeyer and F. H. L. D. Silva, Electroencephalography: Basic Principles, Clinical Applications, and Related Fields, vol. 1. Lippincott Williams & Wilkins, 2004, p. 1309. [7] K. Heinrichs, Introduction to Surface Electromyography, Journal of Athletic Training, vol. 10, no. 1, p. 69, 1999. [8] T. Hayes and P. Horowitz, The Art of Electronics - Student Manual. 95