A Simple Front-End for Pulse Oximeters With a Direct Light-to-Time-to-Digital Interface Circuit

Pulse oximeters are electronic devices that provide the values of two physiological parameters: heart rate (HR) and functional O2 saturation, SpO2. Acquiring this information requires a photodiode and a set of red and infrared light-emitting diodes (LEDs) to generate two plethysmograms, which are digitally processed to find the HR and SpO2 values. The circuits currently used to generate these plethysmograms require multiple analog modules, such as voltage references, transimpedance amplifiers (TIAs), voltage amplifiers, and comparators. Analog-to-digital converters (ADCs) or switched integrators (SIs) are also needed to digitalize the signals that provide the HR and SpO2 values when introduced in a digital processor (DP). This article proposes a new circuit that eliminates the need for all these analog modules or converters, replacing them with a single capacitor and two resistors. The circuit is based on a light-to-time-to-digital conversion performed using the DP, which does not require any special characteristics to carry out this task. As a proof of concept, the new design has been implemented using a field-programmable gate array (FPGA) as the DP. The results show plethysmograms with good detail regarding amplitude and time, allowing the device to be used for clinical purposes. A comparison with two conventional commercial pulse oximeters shows that the new circuit provides similar HR and SpO2 values.

[devices capable of offering data on, at the very least, heart rate (HR) and functional oxygen saturation (SpO 2 )] have become a category of electronic tools that immediately attract attention.Mass use of these devices means efficient hardware and power consumption design is more important than ever.
Conventional pulse oximetry uses two light wavelengths [red and near-infrared (IR)] transmitted/reflected through the distal phalanx of the finger or the earlobe [4].A photodetector senses these transmitted/reflected lights.Pulse oximeter functionality is based on hemoglobin (Hb) color changes from dark to bright red when oxygenated.Oxyhemoglobin (HbO 2 ) absorbs higher amounts of IR light and lower amounts of red light than Hb [5].It is fortuitous that HbO 2 and Hb have significant differences in red and IR light absorption, as these two wavelengths effectively permeate tissues, unlike others that experience substantial absorption by nonvascular tissues and water [6].This difference in light absorption properties between HbO 2 and Hb is exploited in conventional pulse oximeters using two small light-emitting diodes (LEDs): red at 660-nm and IR at 940-nm wavelengths.A broadband photodiode then detects the light transmitted/reflected through Fig. 1.Classic circuit for light-to-voltage-to-digital conversion in a pulse oximeter.Fig. 2. Light-to-voltage-to-time-to-digital conversion scheme in a pulse oximeter.
the finger or earlobe.Variations in the amounts of red and IR light received at the photodiode determine the oxygen saturation ratio of Hb in arterial blood (SpO 2 value).The frequency in the change of the amount of IR light received is also often used as the HR value [1].
Conventional pulse oximeters first perform light-to-voltageto-digital conversion, followed by digital signal processing that provides the HR and SpO 2 values [7].There are systems of different degrees of complexity to perform this conversion.For example, in [8], a complex chopper network converts the photocurrent signal into a voltage.However, this is usually done using a transimpedance amplifier (TIA) [9], followed by a voltage amplifier that feeds the signal to an analog-to-digital converter (ADC) [10], [11], [12], [13], [14], [15], [16], [17].A general layout for this kind of circuit is shown in Fig. 1.The TIA transforms the information provided by the photodiode into the form of a current into a voltage [18].Fig. 1 also shows the output capacitance of the photodiode (C PD ), which may slightly distort the output value of the TIA.The amplifier's positive input is used to supply a reverse-bias voltage, V bias in Fig. 1, for the photodiode, such that the current in this element, I PD , is mainly due to the amount of light it receives.The design presented in [19] (also based on using a TIA followed by an ADC) is particularly worthy of note, as here, the LEDs are replaced by silicon nanowires, while several modifications are introduced in the circuit to reduce reading errors.The circuits are complex for all of these designs since, together with the ADC, several operational amplifiers (OAs) and discrete elements are needed for implementation.
A different approach is presented in [20], where a lightto-voltage-to-time-to-digital conversion is carried out (Fig. 2).In this circuit, a voltage mode switched integrator (SI) follows the TIA.The switch (SW) in Fig. 2 is closed when no LEDs are activated, such as V SI = V Ref1 .When an LED is activated, SW opens, and V SI increases linearly over time, with a constant of proportionality related to the current flowing through the photodiode.V SI is fed into a comparator that generates a binary signal that takes a "1" value for a time inversely related to the value of I PD .The same circuit layout is used in [21] with some minor changes [22], [23], [24] also use this architecture with different types of TIAs.Although circuits based on this type of architecture are simpler than those based on TIA-ADC, they are still complex circuits with OAs, switches, and additional reference voltage sources.An excellent review of the state of the art of photoplethysmographic sensors based on the structures shown in Figs. 1 and 2 can be found in [25].
All the aforementioned circuits require a digital processor (DP), which, based on the digital values of the output signal, provides HR and SpO 2 .In these implementations, the DP merely receives the digital information without being actively involved in acquiring data.
However, direct interface circuits (DICs) [26], using DPs as part of an analog front-end, allows circuits to be simplified.DICs have been widely used in resistive, capacitive, and inductive sensor readout circuits [27], [28], [29], [30].In DICs, the magnitude of the variable to be measured is established based on the discharge times of one or more capacitors that have previously been charged to a voltage set by one of the DP's output pins, performing a magnitude-to-time-to-digital conversion.
Despite its simplicity, to the best of our knowledge, the ideas underlying DICs have never been used to design a front-end of a pulse oximeter.This article proposes a new front-end based on such DICs.The new architecture significantly reduces the hardware without any tradeoff with other device parameters.As proof of concept, a pulse oximeter has been implemented using a field-programmable gate array (FPGA) as the DP.The results show that the plethysmograms acquired with the new design are high in quality, clearly showing the morphological characteristics of the pulse waves that provide relevant information.Furthermore, tests performed with the new pulse oximeter estimated HR and SpO 2 values very similar to those provided by other commercial pulse oximeters.

II. NEW PROPOSAL FOR A PULSE OXIMETER CIRCUIT
A. Design of the Pulse Oximeter Fig. 3 shows the new circuit to obtain plethysmograms from the red and IR wavelengths generated by the two LEDs.As explained in the following, the circuit obtains these plethysmograms by performing a light-to-time-to-digital conversion.In addition to the LEDs, the photodiode, and the three pins of the DP shown in Fig. 3, the circuit uses only two resistors, R IR and R R , and one capacitor, C. The DP can be a microprocessor, part of an ASIC that includes the circuitry shown in Fig. 3, or even an FPGA if the device is designed for research purposes, as in this article.The DP will also be in charge of processing the digital data it generates in order to obtain the HR and SpO 2 values.The new circuit eliminates the need for TIAs, voltage amplifiers, voltage references, ADCs, or SIs, bringing savings in terms of both hardware and area.Capacitor C in Fig. 3 replaces all these elements in the circuits of Figs. 1 and 2 (note that R R and R IR resistors, connected to a voltage source, are also required on the pulse oximeters in Figs. 1 and 2).
For the DP pins, O IR and O R should be outputs that allow the LED activation time intervals to be selected.The pin that will establish the time measurements, P C , must be able to be configured as input-output.If configured as an input, it should have a very high input impedance ("HZ-In" state).

TABLE I STEPS TO OBTAIN TIME MEASUREMENTS FOR RED AND IR PLETHYSMOGRAMS
Each data point in a plethysmogram generated with the new circuit is a time measurement from a C discharge process.These measurements are expressed in a number of DP internal clock cycles.The steps to obtain these measurements are shown in Table I, indicating the logic state of the DP pins in each step.The four-step sequence shown in Table I, Ch-DR-Ch-DIR, is repeated to obtain a datum for each plethysmogram.Thus, the plethysmograms are obtained in an interleaved manner, with each measurement of the red plethysmogram being followed by one of the IR plethysmograms.
The LEDs are off during Ch due to the "0" output of O R and O IR , meaning virtually no current flows through the photodiode, and C charges up to the maximum voltage, V DD , that the P C pin can provide when configured as a "1" output.The length of Ch may vary when previously C is discharged by the IR LED, T ch,R , or by the red LED, T ch,IR .Varying the charging times allows the maintenance of a stable sampling frequency to obtain the data.In any case, these times must allow a stable voltage to be reached in C and, therefore, depend on the current provided by pins O R and O IR , the value of the resistances, R IR and R R , and the capacitance, C.
Once C is charged, the P C pin changes to the "HZ-In" state, and one of the LEDs is activated (DR or DIR steps in Table I).The amount of light that the photodiode receives from the activated LED, either by transmission or by reflection through a finger or earlobe, increases the reverse current that circulates through it (I PD in Fig. 3), varying the discharge rate of the capacitor.As the P C pin has been set to the "HZ-In" state, it does not influence the C discharging process.The time measurements obtained during the DR or DIR steps (T R or T IR , respectively) are taken from the beginning of the discharging to the instant, in which the P C pin detects a change of logic state in the V C voltage of Fig. 3.This situation occurs for V C = V TL , where V TL is the threshold voltage of the P C pin for the transition from "1" to "0."The time evolutions of the voltages on the DP pins V IR , V R , and V C are shown in Fig. 4.
On the other hand, obtaining a stable sampling period, T S , requires the following (see Fig. 4): As stated, this implies that T ch,R and T ch,IR change after each measurement.In any case, the selected value of T S must ensure Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.I.
a minimum value for T ch,R and T ch,IR when the maximum values of T IR and T R are reached, which determines the maximum sampling frequency.The value of a discharge time measurement when activating any of the LEDs, T M , can be evaluated by analyzing the circuit formed by the capacitor and photodiode in Fig. 3. Let us consider that I PD can be written as where I P is the photocurrent due to the incident light on the photodiode, and I D is the dark current given by where I S is the reverse saturation current, q is the electron charge, k B is the Boltzmann constant, and T is the absolute temperature.Since the evolution of V C (t) during capacitor discharge is due solely to I PD , we can write Substituting the value of I PD given by ( 2) and (3) in this expression, the following differential equation is obtained: Finding T M simply requires us to integrate the right-hand term of this expression between the capacitor's initial and final voltage during the discharge process (V DD and V TL , respectively) The integral can be evaluated by performing the change of variable Such that, (6) becomes where If we assume that incident light power is constant during the short duration of the C discharge, then I P is also constant, and from (8), we find This expression can be simplified by taking into account, first, that the natural logarithm in (9) approaches zero, since and second, the term that multiplies the natural logarithm verifies Such that the second term in square brackets is much smaller than the first, and ( 9) is just It is important to note that ( 12) is obtained for a photodiode with a reverse bias voltage that varies during the discharge process.In contrast, reverse bias voltages are constant in the photodiodes of Figs. 1 and 2. T M in (12) can also be expressed more conveniently, considering that I P has a component due to ambient light, I AL , and a component due to reflected/transmitted light, I RT , such that Finally, ( 12) is where I o = I S + I AL is a constant term during the discharging time and satisfies I o ≪ I RT (carefully selecting the photodiode and reducing ambient light).Thus, ( 14) reduces to This equation shows the inverse relationship between T M and I RT .

B. HR and SpO 2 Values
As shown in (15), the information provided by T M is identical to that provided by I RT .Next, it is demonstrated how, from the different T M values, it is possible to obtain HR and SpO 2 similar to how it is carried out in a pulse oximeter based on the circuits in Figs. 1 and 2.
In the simplest cases, HR can be calculated from the difference between the times at which the minimum I IR values occur in a cycle i and in a cycle i + n of the plethysmogram (n is an integer selected by the designer to provide greater stability in the HR value).It should be noted that these instants match those in which the peaks of the plethysmogram formed from the measurements T IR occur, as deduced from (15).If t IR (peak i ) and t IR (peak i+n ) are the times at which the ith and the (i + n)th peaks in T IR are reached, respectively, then, HR, in beats per minute (BPM), can be calculated as There are other more complex methods for estimating HR that considers the existence of motion artifacts [31], [32], but this falls outside the scope of this article.
The ac and dc values of I RT for the red and IR wavelengths are needed to find the SpO 2 values [33].The dc value corresponds to the I RT value for basal blood flow, i.e., the maximum value of I RT in one pulse cycle.In comparison, the ac value is the difference between the maximum and minimum values of I RT during one pulse cycle.
Let I RT,R (trough) and I RT,IR (trough) be the minimum values of I RT in a cycle of the red and IR plethysmogram, respectively, with I RT,R (peak) and I RT,IR (peak) being their maximum values.The ac and dc values of I RT in a cycle of the red and IR plethysmograms I RT,R (ac), I RT,R (dc), I RT,IR (ac), and I RT,IR (dc) are, therefore, given by Meanwhile, the perfusion ratios, P R and P IR , are required to find the SpO 2 value With these definitions, the normalized absorption ratio, R, is defined as I RT,IR (ac) I RT,IR (dc) .
However, the new circuit provides plethysmograms based on time rather than current measurements.Obtaining R as a function of time measurements requires the following values derived from (15): Meaning ( 21) and ( 22) can be written and, with these values, ( 23) is Finally, from R, it is found in [34] that In practice, however, SpO 2 is assumed to be a linear function of R where a and b are constants obtained by calibration.
With the new circuit, the SpO 2 value is therefore ascertained from the plethysmogram maximums and minimums during a pulse wave using (30) and linear equation (32).This procedure is very common for obtaining SpO 2 , although other more complex methods relate R to SpO 2 [35].

C. Errors and Uncertainties
The main source of error when using (15) to form the plethysmograms is the photodiode capacitor C PD , which in the circuit in Fig. 3 is located in parallel with C. Considering this capacitor C should be replaced by C PD + C in (4).If C PD is a constant, this element would not be a problem for determining R or SpO 2 since (30) does not show any dependence on the capacitance of the circuit.However, C PD varies slightly with the voltage applied to the photodiode and cannot be handled as a constant in (6).Fortunately, C PD is in the range of just a few picofarads, and its variations, C PD , maybe even smaller.It is, therefore, possible to choose C in such a way that C ≫ C PD , while simultaneously keeping T R and T IR small enough to consider a constant value of I PD during discharge.
Meanwhile, the presence of C PD does not modify the HR value since it does not vary the instants of time, in which the peaks and troughs are reached in the plethysmograms.
On the other hand, errors in the precise determination of R R and R IR have no practical importance since their only effect is a slight modification of I RT,R and I RT,IR and, therefore, of the temporal measurements.However, the modification occurs in the same way in all these measurements, as indicated by (15), and consequently, it does not affect the HR and SpO 2 calculations.For the same reason, it is not necessary to consider the output resistors of the FPGA pins since their values are generally low.
Since R and SpO 2 depend on time measurements, the uncertainty in their estimates, u(R) and u(SpO 2 ), comes from the uncertainty in their time measurements, u(T M ).There are two causes for u(T M ) [36].The first one is quantization in clock cycles of the time measurements, which produces quantization uncertainty.However, this uncertainty can be neglected if the values of T M (expressed in DP clock cycles) are large enough.Thus, the most important cause of uncertainty is the electronic noise in V C at the trigger instant.This uncertainty, u trigger , depends inversely on the slope of the discharge curve of V C when it reaches voltage V TL [37] Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.
being α, a constant related to the noise level in the circuit's node.Meanwhile, whenever the same conditions that led to ( 15) are verified, it is trivial to obtain from (4) and (33) becomes This expression allows us to find a quality measurement in the estimation of T M , u(T M )/T M using ( 14) Thus, u(T M )/T M is constant, regardless of C, for a given noise level in the circuit.It, therefore, makes no sense to increase C, to achieve higher quality in T M , beyond a value that allows us to neglect quantization uncertainty and which verifies C ≫ C PD .
On the other hand, to find the quality in the estimation of R, u(R)/R, it must be remembered that R in ( 30) is a function of four measurements: T R (peak), T R (trough), T IR (peak), and T IR (trough).Since T M can be any of these time measurements, we can write Using ( 28)-( 30) and ( 36), after some simple algebraic operations, (37) becomes Thus, quality in estimating R depends on physiological parameters, P R and P IR (which the designer has no influence over), and electrical noise in V C , which the designer must try to minimize.Nevertheless, P R and P IR are much less than one and ( 38) can be approximated by a constant value Finally, using (32), quality in estimating SpO 2 is given by This result shows that quality in estimating SpO 2 depends on quality in estimating R and on the SpO 2 value itself.This phenomenon also occurs in pulse oximeters based on the circuits in Figs. 1 and 2. However, for usual values of a and SpO 2 a SpO 2

III. METHODS, EXPERIMENTAL RESULTS, AND DISCUSSION
As a proof of concept of the new pulse oximeter, the circuit has been designed using a commercial board, namely, the Digilent CMOD A7 (Pullman, Washington).This board uses an FPGA (Xilinx Artix 7 XC7A35T) that will perform the DP functions in the pulse oximeter, with 3.3 V as the V DD value for the pins.Using an FPGA has the advantage of allowing rapid prototyping of the circuit in Fig. 3, while also helping to assess the complexity of a future ASIC that includes the entire circuit.The characteristics of the pins (configured as output pins) of this FPGA ensure (together with resistors R R and R IR ) the current required for ON/OFF control of the LEDs.When configured as an input, a pin presents a very high input impedance, as occurs with the P C pin in Fig. 3.In the board, the FPGA is connected to a UART port interface integrated circuit, FT2232HQ.This port connects the circuit to a laptop to transmit the T R and T IR data obtained by the FPGA and the HR and SpO 2 values.It is, therefore, possible to check the accuracy of the estimates, comparing those made internally in the FPGA with those found using the MATLAB software.The FPGA's internal clock frequency is set at 100 MHz.The time measurements count is therefore incremented every 10 ns.The threshold voltage, V TL , measured experimentally in the P C pin is approximately 1.26 V (this is the minimum value of the reverse bias voltage of the photodiode when obtaining T M ).
The red and IR LEDs and the photodiode in the circuit of Fig. 3 are in a single device, the SFH7072 by OSRAM (Berkshire, U.K.).A value of 350 has been selected for R R and R IR , limiting the turn-on current of the LEDs to approximately 4 mA (choosing the optimal values of resistors R R and R IR shall be a topic to return to in future works).A capacitance of 1 nF has been selected considering all set out above on the value of C. All these elements have been connected to the CMOD A7 on a solderable breadboard for rapid prototyping.Fig. 5 shows the implemented system.
The internal structure programmed into the FPGA is shown in Fig. 6.It displays a 16-bit ring counter with an end of count (EOC) output to indicate the start of a discharging process to the control unit.The count is incremented whenever a "1" is detected during discharging or in the Ch steps.The values generated by the counter, T R and T IR , are taken directly to the control unit and the arithmetic unit.The operations in the arithmetic unit are simple since only the search for the maximums and minimums of T R and T IR is carried out, using these values together with ( 16), (30), and (32) to obtain HR and SpO 2 .Finally, all these data are input into the USB port unit, which communicates with the FT2232HQ for data transmission to the laptop.
The T R and T IR values, as received in the laptop, allow us to obtain plethysmograms, such as the two shown in Fig. 7.In these plethysmograms, the vertical axis shows the T R and T IR values in microseconds; however, as mentioned, the FPGA obtains the values in number of clock cycles.The count varies between 32 500 and 35 000 for the IR plethysmogram, while the variation for the red plethysmogram is between 41 500 and 43 500.The 16-bit ring counter provides these values in Fig. 6, whose maximum count value, 2 16  − 1 = 65 535, has proven to be sufficient in all tests performed.
As discussed in Section II, (1) must be verified to maintain a constant sampling period.To do this and ensure full charge of C, the FPGA state changes to Ch once T R or T IR is detected and continues in this state until the counter (which has not been stopped) reaches its maximum value.At this point, the process of obtaining a measurement on the other plethysmogram begins.Thus, the minimum duration of the Ch step is 65 535 − 43 500 = 22 035 clock cycles.
Proceeding in this way, T S = 2•65535•10 ns.That is, the sampling frequency with which each plethysmogram is obtained is 763 Hz.
It is worth mentioning that the proposed circuit allows the adoption of sampling techniques that reduce the number of samples to generate the characteristic points of a plethysmogram, such as compressive sampling or event-driven sampling [25].It is enough to modify the control unit in Fig. 6 to use these techniques, introducing variable periods of time with the LEDs off.The two plethysmograms in Fig. 7 show how, with this sampling frequency, the details of the pulse wave can be observed quite accurately, in addition to the slow evolution of the basal level.Meanwhile, Fig. 7(b) also shows a slightly higher signal noise level for the red wavelength than for the IR wavelength of Fig. 7(a), as is typical of all pulse oximeters.The perfusion ratios obtained from these plethysmograms, P R and P IR , present normal values, around 2% and 3%, respectively [note that these values make the approximation (39) possible].
The different pulse waves and their identifying characteristics can be observed in time and amplitude.Hence, in Fig. 7, it is easy to distinguish the appearance of two dicrotic pulses [38], highlighted in red.The dicrotic pulse located furthest to the right has been amplified in Fig. 8, for the IR wavelength, clearly showing the resolution obtained in determining its characteristics.Pulse waves are characterized by their systolic amplitude, diastolic amplitude, and dicrotic notch; all these parameters are shown in Fig. 8.
It is important to note that Figs. 7 and 8 show the raw data obtained by the FPGA without any type of digital processing (such as filtering).Obtaining this information allows us to establish physiological parameters of clinical interest, e.g., augmentation, crest time, large artery stiffness index (LASI), or inflection point area ratio (IPA) [39].

TABLE II COMPARISON
For the HR and SpO 2 values obtained by the new pulse oximeter: 1) Value n = 10 was chosen to estimate HR using (16).Thus, each pulse wave generates a new HR value considering the last ten pulse waves for calculation.2) Obtaining the SpO 2 value using (32) requires us to know a and b in this equation.The commonly used empirical values for these parameters are [29] a = 110 and b = 25.For a simple and functional comparison, eight apparently healthy subjects (for whom informed consent was obtained) were selected to estimate HR and SpO 2 values.The tests were performed in a fixed state without any movement.The estimates were obtained using the new circuit and two commercial pulse oximeters.The first of the commercial pulse oximeters, labeled Pulse Oximeter 1 in Table II, is the Vitalcontrol model by Hans Dinslage GmbH (Uttenweiler, Germany) with ±2% accuracy for SpO 2 ranging from 70% to 100% and ±2 beats/minute for HR ranging from 30 to 250.The second, labeled Pulse Oximeter 2, is the FarmaqO from Farmaqo (Fresnes, France) with ±2% accuracy for SpO 2 ranging from 70% to 100% and ±3 beats/minute for HR ranging from 25 to 250.The results are shown in Table II, where the HR values are expressed in BPM, and the SpO 2 values are expressed as a percentage.The results were taken simultaneously for the three pulse oximeters placed on different fingers for 20 s.For commercial pulse oximeters, the values in Table II are those that were most prolonged during the 20 s, while for the new circuit, they are the average values during this time.
Table II shows an acceptable agreement between the values of the three pulse oximeters.The SpO 2 values estimated by the new circuit are generally closer to those of Pulse Oximeter 1, while the HR values are closer to those of Pulse Oximeter 2. Table II also shows the mean absolute errors (MAEs) of the HR estimates obtained from the manual examination of the data recorded with the experimental setup shown in Fig. 5.

IV. CONCLUSION
Current implementations of pulse oximeters require complex circuits to acquire and digitalize the pulse wave signal.Apart from the photodiodes and LEDs required to obtain the plethysmograms used by pulse oximeters, these circuits need a variable number of analog modules, together with an ADC or an SI.Furthermore, all implementations require a DP, which, operating on the acquired data, provides at least the HR and SpO 2 values.
This article presents a new circuit for a pulse oximeter that eliminates the need for such analog modules, ADCs, and SIs.Logically, the circuit needs the same photodiodes and LEDs as other implementations but uses only two resistors to control the LEDs, a capacitor, and the DP.The circuit reduces the hardware required by means of a light-to-time-to-digital conversion, which takes advantage of the characteristics of the pins of the DPs.The new proposal does not impose any special restrictions on the DP used in the circuit, giving the designer the freedom to select microcontrollers, FPGAs, or ASICs as the pulse oximeter's digital core.
As proof of concept, the circuit has been implemented using an FPGA as the DP.The results show plethysmograms with resolution levels (both in amplitude and time) that allow the characteristics of the pulse waves to be clearly distinguished, allowing the pulse oximeter to be used for clinical purposes.
A simple comparison of the new design's performance with two commercial pulse oximeters has found no significant differences between the HR or SpO 2 values provided by the commercial pulse oximeters and those estimated with the new circuit.

Fig. 4 .
Fig.4.Voltage waveforms for V R , V IR , and V C in the circuit of Fig.3and for the different steps of TableI.

Fig. 5 .
Fig. 5. Implemented circuit.(a) With a cavity for finger insertion and (b) without the cover on the SFH7072.

7 .
Plethysmograms obtained during a time interval of 20 s with the new proposed circuit.(a) For the IR wavelength, with T IR measurements in µs.(b) For the red wavelength, with T R measurements in µs.

Fig. 8 .
Fig. 8. Magnification of the second dicrotic pulse marked in red in Fig. 7(a), where the most important features of the dicrotic pulse wave in the IR plethysmogram are presented in more detail.