Medical & Biological Engineering & Computing https://doi.org/10.1007/s11517-017-1781-0

ORIGINAL ARTICLE

A novel high input impedance front-end for capacitive biopotential measurement Rongrong Wu 1 & Yue Tang 1 & Zhiyu Li 2 & Limin Zhang 1

&

Feng Yan 1

Received: 9 July 2017 / Accepted: 22 December 2017 # International Federation for Medical and Biological Engineering 2018

Abstract For capacitive biopotential measurement, a novel high input impedance front-end is proposed. The input impedance of the frontend can achieve more than 100 GΩ by matching the peripheral parameters. The front-end’s noise model is provided, and noise optimization is given further. The analysis shows the proposed front-end can achieve at least two orders of input impedance more than the non-inverting amplifier circuit with the same peripheral parameters at the cost of only increasing twice input-referred noise. The final experimental results verify the analysis and show the front-end’s feasibility of capacitive sensing electrocardiogram signal. Keywords Biopotential sensing . Capacitive measurement . High input impedance . Input-referred noise

1 Introduction According to the World Health Organization’s report on cardiovascular disease in 2015, deaths from cardiovascular disease account for 31% of all year-round disease deaths, which has been the world’s top cause of death [14]. The cardiovascular disease report of China in 2015 points out that the current number of cardiovascular patients in China is as high as 290 million [1]. In such serious cases, the need of equipment for real-time detection of electrocardiogram (ECG) signal is particularly prominent. As the essential part for sensing biopotential, the detection electrode is divided into three categories: wet electrode, dry electrode, and non-contact electrode [9, 10, 12]. The wet electrode detects bioelectrical signal from human body through the gel electrolyte on skin; similarly, the dry electrode is in contact with the skin and sensing the signal through the metal coating, whereas the non-contact electrode can sense bioelectrical signal with a gap between the skin and the sensor plate,

* Limin Zhang [email protected] 1

School of Electronic Science and Engineering, Nanjing University, Nanjing, Jiangsu Province, China

2

UAV Research Institute, Nanjing University of Aeronautics and Astronautics, Nanjing, Jiangsu Province, China

such as hair, cloth, insulation, or air. The usage of wet electrode in medical devices is mature enough; however, gel electrolyte limits its application on wearable medical equipment. For dry electrode and non-contact electrode, the acknowledged equivalent model between the electrode and human body is a series or parallel model of resistance and capacitance [4]. If the equivalent model between the body and the electrode is simplified as a coupling capacitor, the equivalent input impedance of the post front-end should be at least 100 GΩ to detect 0.1 Hz low frequency signal for the coupling capacitance as low as several pF. Generally, in the absence of an input bias network, the leakage-current of the FET will cause the amplifier to saturate (usually after only a few seconds), which is unacceptable for non-contact detection of biological electrical signals. To solve the problem, Mathews et al. achieved the ultra-high impedance of a circuit via using the bootstrap circuit [5–7, 12], Prance et al. realized the high input impedance by preciously designing the protection circuit ring [8], and Yu et al. made use of the reverse leakage current of silicon diodes whose performance is equivalent to an ultra-high resistance to provide a current bias circuit for the front-end [3]. In a word, the elaborate design for current bias path and shielding technique is significant for achieving an ultra-high input impedance, which is one of the key points for designing wearable detection sensor. Except the ultra-high input impedance, the input referred noise is another one key point for detecting biological

Med Biol Eng Comput

electrical signals because it determines the sensitivity for weak biological signals. Actually, the input referred noise has been mentioned in many papers about dry contact or non-contact sensors. For the non-contact sensor designed by Sullivanetal et al. [11], the referred input noise in the frequency band of 1 – 100 Hz is 2 μVRMS for the distance between electrode and human body surface being 0.2 mm, and 17 μVRMS for the distance with 3.2 mm. For the wireless non-contact ECG sensing electrodes proposed by Yu et al. [2], the input referred noise in 0.7−100 Hz frequency band is 3.8 μVRMS with the gain of 46 dB. Actually, the experiment results of the input referred noise are usually given and the optimization for input noise is rarely mentioned. In the paper, a novel high input impedance front-end is proposed for capacitive biopotential measurement. The input impedance of the front-end can achieve more than 100 GΩ by matching peripheral parameters and the input referred noise can be optimized. In the second part, design of the proposed front-end is illustrated first, then the corresponding noise model is established and noise optimization is given, and the noise is compared to the noninverting amplifier circuit with the same peripheral parameters further. In the third part, the analysis is verified by experiment results. Finnally, dry contact and non-contact ECG measurement results are given to verify the front-end’s feasibility of capacitive biopotential measurement.

R2 ¼ R1  ð1 þ δþ Þ

ð3Þ

and R3 is expressed by the mismatch coefficient δ− and R4 as R3 ¼ R4  ð1 þ δ‐ Þ

ð4Þ

Then, Eq. (2) can be written as Req ¼

R1  ð1 þ δþ Þ δþ −δ−

ð5Þ

For the resistors R3 and R4, it is recommended to choose 10 kΩ with high accuracy like 0.1% tolerance for easy measurement. Furthermore, the two resistors are selected manually via digital multimeter to ensure the mismatch coefficient δ− lower than 0.01%. For this condition, Eq. (5) can be simplified as
 1 ð6Þ Req ¼ R1  1 þ δþ It is concluded that the larger value of R1 leads to higher mismatch coefficient δ+. If the mismatch coefficient δ+ is 0.1%, then Req ¼ 1001  R1

2 Method 2.1 Front-end design The novel high input impedance front-end is composed by four resistors R1, R2, R3, R4, and an amplifier Amp, as shown in Fig. 1, where Ceq is an equivalent input capacitance between the body and the electrode. If the input impedance of the amplifier is as high as 1011–1013 Ω, the input impedance of the front-end can achieve more than 100 GΩ by matching the resistance parameters. According to the Kirchhoff voltage law, Kirchhoff current law, and the conceptions of virtual short circuit and virtual open circuit for amplifier circuit, the input impedance can be expressed by Eq. (1), Vi 1 R1 R2 R3 1 ¼ þ ¼ þ Req Ii sC eq R2 R3 −R1 R4 sC eq

ð1Þ

Then, the value of the input impedance can be written as R1 R2 R3 Req ¼ ¼ R2 R3 −R1 R4

amplifier’s input impedance as long as the denominator is exactly zero. If R2 is expressed by the mismatch coefficient δ+ and R1 as

R1  R2 R4 R2 −R1  R3

ð2Þ

Considering the amplifier’s input impedance in parallel, the input impedance of the proposed front-end will reach the

ð7Þ

For example, if an expect value of Req is 100 GΩ, the resistor R1 can be chosen as 100 MΩ with the mismatch coefficient δ+ 0.1% of the resistor R2. In a word, by choosing the resistance level of R1 and R2 as MΩ and matching the resistance values R3 and R4, R1 and R2, the input resistance can reach as high as hundred GΩ or several TΩ further.

2.2 Amplitude-frequency response In this part, the amplitude-frequency response of the proposed front-end is given. According to the Kirchhoff voltage law, Kirchhoff current law, and the conceptions of virtual short circuit and virtual open circuit for amplifier circuit, the transfer function of the proposed front-end can be expressed by Eq. (8), H ðsÞ ¼

R1 R2 ðR3 þ R4 Þ R3 R1 R2 þ ðR2 R3 −R1 R4 Þ

1 sCeq

ð8Þ

Dividing both the molecular and denominator with (R2 × R3), Eq. (8) can be written as

Med Biol Eng Comput Fig. 1 Schematic of the proposed front-end

H ðsÞ ¼


 R1 R4
  1þ 1 R4 R1 R3 þ R1 1−  R3 R2 sCeq

ð9Þ

2.3 Noise model

Compared with the gain of the normal non-inverting amplifier circuit, the amplitude-frequency response of the proposed front-end has an additional frequency-dependent coefficient. The parameters of the proposed front-end are shown in Table 1, where the coupling capacitance Ceq between human body and electrode is 470 pF, an additional resistor Radd in series R2 to decrease the resistance difference of R1 and R2. The theoretical calculation of the input impedance using Eq. (2) is 111 GΩ. The corresponding amplitude-frequency response is shown in Fig. 2, which shows first-order high-pass characteristics

Table 1 Parameters of the proposed front-end

Parameter

Value

R1

101.7 MΩ 101.10 MΩ 510 kΩ 9.9986 kΩ 9.9986 kΩ 470 pF 111 GΩ

R2 Radd R3 R4 Ceq Req

with the cut-off frequency at 0.003 Hz and passband gain of 6 dB.

In the theoretical model, several noise sources are considered, including the noise sources caused by the resistors and the noise sources caused by the amplifier. The noise model caused by the resistors is considered as thermal noise for total bandwidth and the noise model caused by the amplifier is consisted of 1/f noise and white noise, the corresponding bandwidth is set according to the adopted amplifier. The noise model for the proposed front-end is given in Fig. 3, including four thermal noise sources Vt1, Vt2, Vt3, and Vt4 caused by four resistors R1, R2, R3, and R4, respectively, an equivalent voltage noise source V1 and two equivalent current noise sources IN1 and IN2 caused by the amplifier. All noise sources are independent. The thermal noise Vtn caused by the resistor Rn can be expressed with Eq. (10), pffiffiffiffiffiffiffiffiffiffiffiffi ð10Þ V tn ¼ 4kTRn where the parameter k is Boltzmann constant 1.38 × 10−23 J/K, T is the absolute temperature under room temperature 298.15 K, Rn presents the resistance calculated. The equivalent voltage noise V1 is composed by white noise and 1/f noise [15]. If the noise voltage density of the white noise is K, and the noise voltage density of 1/f noise is C, then the equivalent voltage noise V1 as function of frequency f can be written as

Med Biol Eng Comput Fig. 2 The magnitude-frequency response of the proposed frontend

Eq. (11), sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1Hz þ K2 V 1 ¼ C2  f

ð11Þ

The value of C and K can be read by the input referred voltage noise density curve in the datasheet of the amplifier. For example, from the input referred voltage noise curve in the datasheet of LMP7701 amplifier [13], the Fig. 3 The noise model for the proposed front-end

value K can be obtained by reading the input referred noise at high frequencies, the equivalent voltage noise V1 at 1 Hz is equal to the input referred noise at 1 Hz. The noise voltage density of 1/f noise C can be calculated by Eq. (11). Related parameters of equivalent voltage noise for LMP7701 amplifier are shown in Table 2. For the amplifier LMP7702, the 1/f noise bandwidth is from 1 Hz to 1 kHz. According to Fig. 3, every source’s contribution to the output noise can be calculated. For example, the contribution of

Med Biol Eng Comput Table 2 Related parameters of equivalent voltage noise for LMP7701 amplifier Voltage noise density @10 kHz

9 nV/√Hz

K Voltage noise density @1 Hz

9 nV/√Hz 118 nV/√Hz

C

117.66 nV/√Hz

equivalent voltage noise V1 can be obtained according to the following equations.   V o − V p −V V p −V i ¼ ð12Þ R1 =ðsCR1 þ 1Þ R2 Vp ¼ Vn ¼ Vo 

R3 R3 þ R4

ð13Þ

From Eq. (12) and Eq. (13), Eq. (14) can be derived, V o1 ¼ V 1 

ðR1 R2 sC þ R1 þ R2 ÞðR3 þ R4 Þ ðR1 R2 sC þ R2 ÞR3 −R1 R4

ð14Þ

By substituting Eq. (8) into Eq. (14), Eq. (15) can be obtained,

V o1 ¼ V 1  H ðsÞ 

1 sC

1 R1 ==R2 == sC

ð15Þ

this shows the contribution of noise source V1 to the output voltage noise. The contributions of other noise sources can be calculated in the same way. Their contributions to the output voltage noise are shown as following, V t1 : V o2 ¼ V t1  H ðsÞ 

1 sCR1

ð16Þ

V t2 : V o3 ¼ V t2  H ðsÞ 

1 sCR2

ð17Þ 1 sC

R4  1 R3 þ R4 R1 ==R2 == sC 1 R3 sC V t4 : V o5 ¼ −V t4  H ðsÞ   1 R3 þ R4 R1 ==R2 == sC 1 I N 1 : V o6 ¼ I N 1  H ðsÞ  sC 1 R3 R4 sC I N 2 : V o7 ¼ −I N 2  H ðsÞ   1 R3 þ R4 R1 ==R2 == sC V t3 : V o4 ¼ −V t3  H ðsÞ 

ð18Þ

ð19Þ

ð20Þ

ð21Þ

Then, the total output voltage can be calculated by Eq. (22). qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi V o ¼ V 2o1 þ V 2o2 þ V 2o3 þ V 2o4 þ V 2o5 þ V 2o6 þ V 2o7 ð22Þ The input referred voltage noise can be written as Eq. (23). Vi ¼

Vo : H ðsÞ

ð23Þ

According to the parameters shown in Table 1, the input referred voltage noise as function of the frequency can be calculated, as shown in Fig. 4 (dotted line). The theoretical input referred voltage noise in bandwidth of 1−100 Hz is 6.1169 μVRMS. Based on widely used for circuit analysis simulation software Multisim, the simulation result for the input referred voltage noise as function of the frequency is also shown in Fig. 4 (solid line) and the input equivalent voltage noise is 6.1227 μVRMS for 1–100 Hz bandwidth. The simulation result agrees well with the theoretical result. Usually, the input coupling capacitance varies for different applications. For example, the coupling capacitance is about several nF for dry contact measurement and only tens of pF for non-contact measurement. The input referred voltage noise with 1–100 Hz bandwidth as function of the input coupling capacitance is shown in Fig. 5. The input referred voltage noise decreases with the increase of the input coupling capacitance. For the input coupling capacitance from 10 pF to 2 nF, the input referred voltage noise degrades from 0.28 mV to 1.47 μV. The input referred voltage noise with 1–100 Hz bandwidth as function of the input impedance is shown in Fig. 6. The input referred voltage noise decreases no more than 0.02 μV with the increase of the input impedance from 10 GΩ to 1 TΩ. The analysis above all is based on the condition that the resistances R1 and R2 in positive feedback loop are both about 100 MΩ. Actually, the resistances R1 and R2 affect the input referred voltage noise. For the same input impedance, the input referred voltage noise with 1–100 Hz bandwidth as function of the resistances R1 and R2 is shown in Fig. 7. The input referred voltage noise decreases with the increase of the resistances R1 and R2. For the resistances R1 and R2 from 20 kΩ to 100 MΩ, the input referred voltage noise decreases from 2.9 mV to 6.13 μV. To realize the input referred voltage noise as small as possible, the resistances R1 and R2 should be chosen as large as possible.

2.4 The comparison with non-inverting amplifier circuit As shown in Fig. 1, the proposed front-end can be seen as a modified circuit based on the non-inverting amplifier circuit by adding the positive feedback loop with the resistors R 1 and R2 . It is necessary to give the

Med Biol Eng Comput Fig. 4 The input referred voltage noise as function of the frequency

voltage noise comparison of the proposed front-end and the non-inverting amplifier. The noise model of the noninverting amplifier circuit is shown in Fig. 8, and the parameters are as same as those in Fig. 3. According to the Kirchhoff voltage law, Kirchhoff current law, and the conceptions of virtual short circuit and virtual open circuit for amplifier circuit, the transfer

Fig. 5 The input referred voltage noise as function of the input coupling capacitance

function of the non-inverting amplifier circuit can be expressed by Eq. (24), 0

H ðsÞ ¼

R3 þ R4  R3

s 1 sþ R1 C eq

ð24Þ

Med Biol Eng Comput Fig. 6 The input referred voltage noise as function of the input impedance

Similar to the noise derivation of the proposed front-end, the noise source’s contribution to the output voltage noise is listed as following,

0

0

V 1 : V o1

1 sC eq 0 0 ¼ V 1  H ðsÞ  1 R1 == sC eq

Fig. 7 The input referred voltage noise as function of the resistances R1 and R2

0

0

0

0

V t1 : V o2

ð25Þ V t3 : V o4

1 sCeq ¼ V t1  H ðsÞ  1 R1 == sCeq 0

0

R4 ¼ −V t3  H ðsÞ   R3 þ R4 0

0

ð26Þ

1 sC eq 1 R1 == sC eq

ð27Þ

Med Biol Eng Comput Fig. 8 The noise model of the non-inverting amplifier circuit

0

0

V t4 : V o5

0

0

I N 1 : V o6

1 sC eq R3 0 0 ¼ −V t4  H ðsÞ   1 R3 þ R4 R1 == sC eq 1 ¼ I N 1  H ðsÞ  R1 sC eq 0

0

0

ð28Þ

0

I N 2 : V o7

0

1 sCeq 1 R1 == sC eq

ð30Þ

The total output voltage noise is expressed by Eq. (31). ð29Þ 0

Vo ¼

Fig. 9 The comparison of the two input referred voltage noise curves

R3 R4 ¼ −I N 2  H ðsÞ   R3 þ R4 0

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 0 2 0 2 0 2 0 2 0 2 0 2 V o1 þ V o2 þ V o4 þ V o5 þ V o6 þ V o7

ð31Þ

Med Biol Eng Comput Table 3 circuits

The comparison of the input referred voltage noise for two The input coupling capacitance and input impedance

Non-inverting 470 pF–100 MΩ amplifier circuit Proposed sensor 470 pF–100 GΩ

Simulated (VRMS)

Theoretical (VRMS)

4.3332 μV 4.3340 μV 6.1273 μV 6.1213 μV

The input referred noise is expressed by Eq. (32). 0

Vi ¼

0

Vo 0 H ðsÞ

ð32Þ

For the same parameters, the input referred voltage noise as function of the frequency is shown in Fig. 9 (dashed line), which is slightly lower than that of the proposed front-end shown by solid line in Fig. 9. The input referred voltage noise results for 1–100 Hz bandwidth of two circuits are compared in Table 3, which shows that the input referred voltage noise of the proposed front-end is no more than twice larger than that of the non-inverting amplifier circuit, while the input impedance of the proposed front-end is about 1000 times larger than that.

parameters in Table 1 are given. The applications on dry contact and non-contact ECG measurement are also illustrated.

3.1 Amplitude-frequency response As mentioned above, the proposed front-end can be seen as a first-order high pass filter, whose cut-off frequency is determined by the input impedance and the input coupling capacitance. The peripheral parameters of the proposed front-end are selected according to Table 1, except the input coupling capacitance 47 pF instead of 470 pF for increasing the cut-off frequency to 0.028 Hz to be measured easily. As cut-off frequency of front-end circuit is about 0.028 Hz, which is too low to measure automatically for most instruments, the amplitudefrequency response is measured manually with the signal generator AFG1002 and oscilloscope TBS1102B by scanning the frequency from 0.01 to 200 Hz. Considering the signal’s period is variable, 2–3 complete waveforms are kept on the oscilloscope’s screen to ensure the measurement accuracy by adjusting the time scale of the oscilloscope. The measured amplitude-frequency response is shown in Fig. 10. The cut-off frequency is about 0.028 Hz, which means the input impedance is about 113.9 GΩ, agreeing well with the theoretical result 111 GΩ.

3.2 Noise performance

3 Results To verify the feasibility of the proposed front-end, the experiment results for the front-end using the amplifier LMP7701 and Fig. 10 The measured amplitudefrequency response

The noise performance is measured by the noise analysis equipment B&K-3160-A-042, and the result is shown in Fig. 11 (dotted line). The collecting hardware named LANXI has the frequency range from DC to 51.2 kHz and works along with software named PULSE, which are both produced

Med Biol Eng Comput Fig. 11 The input referred voltage noise density of experiment and theoretical results

by Brüel & Kjær, a company in Denmark. This system has exceptional flexibility, stability and accuracy with high quality levels from 1 μV to 316 mVor 10 Vobtained. The experiment result is coincident with the theoretical result (solid line). Table 4 lists the theoretical result, simulation result, and experiment result of the input referred voltage noise for 1– 100 Hz bandwidth of the proposed front-end. The theoretical result agrees well with the simulation result, but slightly larger than the experiment result. The reason might come from that the theoretical and simulation results are calculated according to the typical parameter in datasheet, which may be different from that of the actual used amplifier in the experiment.

top layer of the printed circuit board is designed to sensing signals with tin-free solder, while the bottom layer is designed to place the proposed front-end and the connection port. The whole bottom layer is covered by GND net and the guarding

3.3 Application results The two proposed front-ends with the parameters shown in Table 1 are used to detect ECG signal. The system is one channel ECG detection system including two electrodes and a development kit ADS1292RECG-FE of Texas Instruments. The electrode is a round printed circuit board with the diameter of 4 cm and thickness of 2 mm as shown in Fig. 12, where the Table 4

Different results comparison of input referred voltage noise

Type

The input referred voltage noise (VRMS)

Theoretical result Simulated result Tested result

6.1213 μV 6.1273 μV 1.9108 μV Fig. 12 PCB layout of the proposed front-end

Med Biol Eng Comput 10

Fig. 13 The ECG signal with dry contact detection

Voltage (mV)

5

0

−5

−10 0

1

2

3

4

5 Time (s)

6

7

8

9

10

notch and a 0.08 Hz digital high-pass filter, where the QRS wave and T wave are clearly observed.

technique is not used around the input ports. The connection port is designed with SMA connection, where the GND guarding is applied to reduce the external noise. For the proposed front-end, the resistors are SMD resistors with 0402 footprints and the amplifier is SOIC footprint. The frond-end circuit is powered by two batteries with voltage 3.7 V. Coaxiallines are used to contact the electrodes and the development kit. After the signal detected by the front-ends, the development kit ADS1292RECG-FE translates the analogue signal into digital signal at 250 sample rate, and sends to MCU MSP430F5969, which contacts with computer by serial communication to record data in real time. The recorded data is saved as txt or excel without any noise suppression. To obtain the ECG signal clearly, a 50 Hz digital notch and a 0.08 Hz digital high-pass filter are used to remove the power line interference and the baseline drift in MATLAB. For the one channel ECG detection system, the two electrodes’ places are close as the position of v1 and v4 in the 12 leads to ensure the experiment results similar to the traditional measurements. The human test subject is the volunteer from our group, one male student aged 26 without known cardiac pathology. In dry contact detection, the electrodes contact the body skin directly. Figure 13 shows the dry contact ECG detection result with a 50 Hz digital notch and a 0.08 Hz digital high-pass filter, where the P wave, QRS wave, and T wave are clearly observed. In non-contact detection, the electrodes and the skin are separated by a modal cotton t-shirt which is made up by 93% recycled cellulose fiber and 7% spandex. Figure 14 shows the non-contact detection result with a 50 Hz digital

4 Discussion As mentioned above, the ultra-high input impedance of the proposed front-end is implemented by matching the peripheral parameters. According to the theoretical derivation, the input impedance value relays on not only the equivalence of R3 and R4, but also the difference and the resistances of R1 and R2. Considering every possible noise source, the noise model is established to estimate the input referred noise level of the proposed front-end. To verify the correction of the model, the simulation results are compared with the theoretical results, showing well agreement as shown in Fig. 4. When the input coupling capacitance becomes larger for the medium’s variation, the input referred noise will become smaller under the condition of constant input impedance, as shown in Fig. 5. Under the condition of the constant input coupling capacitance, it is wonderful to find that the increase of the input referred impedance can decrease the input referred noise slightly and the corresponding relationship is shown in Fig. 6. Considering that the input referred impedance mostly depends on the value of R1, R2, it is concluded that R1, R2 should be selected as large as possible to achieve small input referred noise from Fig. 7. Therefore, the resistances R1 and R2 should be chosen as large as possible to do benefit to the input referred impedance with slight decrease of the input referred noise.

6

Fig. 14 The ECG signal with non-contact detection

5

Voltage (mV)

4 3 2 1 0 −1 −2 0

1

2

3

4

5 Time (s)

6

7

8

9

10

Med Biol Eng Comput

Furthermore, the temperature drift of the resistance will influence the matching degree. The temperature drift coefficient of the resistor is expressed with parts per million (ppm), such as ± 5, ± 10, ± 25 ppm, and so on. As supposed, there would be a tolerance range for these four resistances’ value with considering the temperature drift coefficient of each resistance. In order to decrease the analysis’s complex, we choose the four resistors R1, R2, R3, and R4 with lower temperature drift coefficient. Anyway, the limits caused by the temperature drift coefficient should be analyzed and the solution to reduce the influence of resistance’s change due to temperature drift should be studied in the future. Moreover, the performance of the proposed front-end is compared with the basic circuit of the non-inverting amplifier. From Fig. 9, it can be concluded that the proposed front-end can increase the input impedance at least two or three magnitudes more than non-inverting amplifier circuit with same parameters but keep the input referred noise no more than twice larger than that of the latter.

2.

3.

4.

5.

6. 7. 8.

5 Conclusions The paper proposes a bio-front-end circuit which impedance can achieve more than 100 GΩ via precisely matching peripheral parameters with normal values. The noise model and noise optimization method for the front-end are given, and the correction is verified by simulation and experiment results. As analyzed, it is recommended to use larger resistors for R1 and R2 which is benefical to increase the input impedance and decrease the input referred noise. Compared with the non-inverting amplifier circuit, the proposed front-end can achieve the input impedance at least two orders of magnitude larger than that of the conventional non-inverting amplifier circuit, but keep the input-referred noise no more than twice larger than that of the latter. Its feasibility for dry-contact and non-contact biopotential measurement has been illustrated with ECG signal detection. In the future work, the limits caused by the temperature drift coefficient should be analyzed, and the solution to reduce the influence of resistance’s change due to temperature drift should be studied. Funding information This work was supported by the National Key R&D Program of China (No. 2016YFA0202102), the Key Technologies R&D Program of Jiangsu Province (No. BE2014712 and BE2010190), and the National Nature Science Foundation Program of China (No. 61571376 and 11304152).

Compliance with ethical standards Conflict of interest The authors declare that they have no conflict of interest.

References 1.

Chen WW, Du WL (2016) Report on cardiovascular disease in China. National Center for Cardiovascular Diseases, Beijing, pp

9.

10.

11.

12.

13. 14. 15.

2–14 Available: http://www.nccd.org.cn/UploadFile/201607/ 20160718113619135135.pdf Chi YM, Cauwenberghs G (2010) Wireless non-contact EEG/ECG electrodes for body sensor networks. In International Conference on Body Sensor Networks, IEEE,pp. 297–301 Chi YM, Deiss SR, Cauwenberghs G (2009) Non-contact low power EEG/ECG electrode for high density wearable biopotential sensor networks. In: Sixth International Workshop on Wearable and ImplanTable Body Sensor Networks. Berkeley, CA, USA, June, IEEE Computer Society, pp. 246–25 Chi YM, Jung TP, Cauwenberghs G (2010) Dry-contact and noncontact biopotential electrodes: methodological review. IEEE Rev Biomed Eng 3:106–119. https://doi.org/10.1109/RBME.2010. 2084078 Harland CJ, Clark TD, Prance RJ (2003) High resolution ambulatory electrocardiographic monitoring using wrist-mounted electric potential sensors. Meas Sci Technol 14(7):923–928. https://doi.org/ 10.1088/0957-0233/14/7/305 Mathews R, Krupka MA, Hibbs AD (2004) Sensor system for measuring biopotentials. U. S. Patent 6 961 601 Prance H (2011) Sensor developments for electrophysiological monitoring in healthcare. Appl Biomed Eng. InTech Prance RJ, Debray A, Clark TD, Prance H, Nock M, Harland CJ, Clippingdale A (2000) An ultra-low-noise electrical-potential probe for human-body scanning. Meas Sci Technol 11:291–297, Available: http://iopscience.iop.org/0957-0233/11/3/318. https:// doi.org/10.1088/0957-0233/11/3/318 Spinelli E, Haberman M (2010) Insulating electrodes: a review on biopotential front ends for dielectric skin–electrode interfaces. Physiol Meas 31(10):S183–S198. https://doi.org/10.1088/09673334/31/10/S03 Searle A, Kirkup L (2000) A direct comparison of wet, dry and insulating bioelectric recording electrodes. Physiol Meas 21(2): 271–283. https://doi.org/10.1088/0967-3334/21/2/307 Sullivan TJ, Deiss SR, Cauwenberghs G (2007). A low-noise, noncontact EEG/ECG sensor. In: Biomedical Circuits and Systems Conference, IEEE, pp. 154–157 Sun Y, XB Y (2016) Capacitive biopotential measurement for electrophysiological signal acquisition: a review. IEEE Sensors J 16(9): 2832–2853. https://doi.org/10.1109/JSEN.2016.2519392 LMP7701 data sheet, Available: http://www.ti.com.cn/cn/lit/ds/ symlink/lmp7701.pdf WHO Media Center, BCardiovascular diseases^. Available: http:// www.who.int/mediacentre/factsheets/fs317/zh/ Yang JG (2015) Noise index. In: Hello, the amplifier, 1st edn. Beijing Publishing House Group, Beijing, pp. 29–60

Rongrong Wu received the Bachelor’s degree in Shool of Electronic Science and E n g i n e e r i ng f r o m N a n j i n g University, China, in 2016. She is currently a postgraduate student and pursuing for a master degree in school of Electronic Science and Engineering, Nanjing University. Her main research interests include biomedical signal processing, analog circuit design, and biomedical noise control.

Med Biol Eng Comput Yue Tang received the Bachelor’s degree in Lightning Protecting Science and Te c h n o l o g y f r o m N a n j i n g University of Information Science & Technology, China, in 2014. He is currently pursuing the P h. D . d e g r e e in s c ho o l o f Electronic Science and Engineering, Nanjing University, Nanjing, China. His main research interests include biosensor design and biomedical signal processing, as well as their applications for wearable devices and

Limin Zhan g received the Bachelor’s degree in Mechanical Manufacture and Automation from Nanjing University of Aeronautics and Astronautics, China, in 2000, the Master’s degree in electromechanical engineering from Nanjing University of Aeronautics and Astronautics, China, in 2003, and the Ph.D. degree in signal and information processing from Nanjing University, China, in 2012. She went to University of Western Australia as a visiting scholar in 2013. She is currently an Associate Professor in the School of Electronic Science and Engineering, Nanjing University, China. Her main research interests include analog sensor design and analog integrated circuit design, as well as their applications for biosensor design, biomedical signal processing, and active noise control.

Zhiyu Li was born in China in 1976. He received the M.S. degree in measuring and testing technologies and instruments from Nanjing University of Aeronautics and Astronautics in 2008. He is currently a master tutor and researcher in the UAV Research Institute of Nanjing University of Aeronautics and Astronautics. His major areas of research are flight control of UAV and automatic measurement technique.

Feng Yan received the Bachelor’s degree and Master’s degree in Physics from Nanjing University, China, in 1990 and 1993, respectively, and the Ph.D. degree in Electrical and Computer Engineering from Rutgers, the State University of New Jersey, USA, in 2002. He is currently a Professor in the School of Electronic Science and Engineering, Nanjing University, China. His main research interests include device design, sensor design, as well as their applications for biosensor design, and optical imaging.

low power consumption.