sensors Article

High-Accuracy, Compact Scanning Method and Circuit for Resistive Sensor Arrays Jong-Seok Kim, Dae-Yong Kwon and Byong-Deok Choi * Department of Electronic Engineering, Hanyang University, 222 Wangsimni-ro, Deongdong-gu, Seoul 04763, Korea; [email protected] (J.-S.K.); [email protected] (D.-Y.K.) * Correspondence: [email protected]; Tel.: +82-22-220-2311 Academic Editor: Vittorio M. N. Passaro Received: 26 November 2015; Accepted: 21 January 2016; Published: 26 January 2016

Abstract: The zero-potential scanning circuit is widely used as read-out circuit for resistive sensor arrays because it removes a well known problem: crosstalk current. The zero-potential scanning circuit can be divided into two groups based on type of row drivers. One type is a row driver using digital buffers. It can be easily implemented because of its simple structure, but we found that it can cause a large read-out error which originates from on-resistance of the digital buffers used in the row driver. The other type is a row driver composed of operational amplifiers. It, very accurately, reads the sensor resistance, but it uses a large number of operational amplifiers to drive rows of the sensor array; therefore, it severely increases the power consumption, cost, and system complexity. To resolve the inaccuracy or high complexity problems founded in those previous circuits, we propose a new row driver which uses only one operational amplifier to drive all rows of a sensor array with high accuracy. The measurement results with the proposed circuit to drive a 4 ˆ 4 resistor array show that the maximum error is only 0.1% which is remarkably reduced from 30.7% of the previous counterpart. Keywords: resistive sensor; sensor array; crosstalk; read-out circuit; row driver

1. Introduction Resistive sensors, such as piezoresistive tactile sensors [1–6], infrared sensors [7–10], and light-dependent resistors [11,12], have been widely used for measurement or instrumentation in a variety of fields. These sensors are usually applied in the form of an array to obtain a required measurement resolution. When we use resistive sensor arrays, we should consider the crosstalk current between each sensor element in the array [13–16]. The crosstalk currents can flow through the unintended current paths when we read the voltage of the resistive sensor element of interest in a selected row of the array. For example, as shown in Figure 1, let us assume that we read the resistance of R22 by applying the driving voltage (VDRIVE ) to the second row and measuring the voltage at Column 2 terminal. Then, other than the current flowing through R22, the unintended currents are added through the dotted path to Column 2 terminal, which pass through all the other resistor elements (R11, R12, R13, R21, R23, R31, R32, and R33 in Figure 1). Therefore, we cannot find a true value of R22 by measuring the voltage at the column terminal. This causes the accuracy problem in the sensor system. The crosstalk current problem in the sensor array has been addressed in the previous literature, and several solutions have been proposed. The crosstalk current can be avoided by inserting a diode [17,18] or a transistor [19,20] for each element in the sensor array, but these methods are seldom used for integrated sensors in VLSI/MEMS chips since the system complexity is drastically increased due to the additional diodes, transistors, and physical wires. Voltage feedback methods have been proposed to remove the crosstalk current [11,21,22], but it is known that those methods still suffer from

Sensors 2016, 16, 155; doi:10.3390/s16020155

www.mdpi.com/journal/sensors

Sensors 2016, 16, 155

2 of 16

Sensors 2016, 16, 155

2 of 16

theSensors measurement 2016, 16, 155 errors which are increased as a sensor array becomes large [23]. The most2 widely of 16 used method to solve the crosstalk current problem is the zero-potential scanning method shown widely used method to solve the crosstalk current problem is the zero-potential scanning method widely used method to solve the crosstalk current problem is the zero-potential scanning method Figure 2 [13–16,21,24–30]. This method has an has operational amplifier at each column terminal. Since shown Figure 2 [13–16,21,24–30]. This method an operational amplifier at each column terminal. shown Figure 2 [13–16,21,24–30]. Thisvoltages method of hasthe anoperational operational amplifier atmust each be column terminal. both the negative and positive input amplifier equal, due to Since both the negative and positive input voltages of the operational amplifier must be equal, duethe Since both the negative and positive input voltages of the operational amplifier must be equal, due virtual characteristics of an operational amplifier, all theallcolumns and and the rows have thethe same to theshort virtual short characteristics of an operational amplifier, the columns the rows have to the virtual short characteristics of an operational amplifier, all the columns and the rows have the voltage 0 V except a selected row whose voltage is VDRIVE . Consequently, thethe currents only flow sameof voltage of 0 Vfor except for a selected row whose voltage is VDRIVE . Consequently, currents only same voltage of 0 V except for a selected row whose voltage is VDRIVE. Consequently, the currents only through the resistive sensor sensor elements in thein selected row without crosstalk currents. flow through the resistive elements the selected row without crosstalk currents. flow through the resistive sensor elements in the selected row without crosstalk currents.

Figure 1. Crosstalk current current paths in resistive sensor array. Figure pathsin inaaaresistive resistivesensor sensor array. Figure1.1.Crosstalk Crosstalk current paths array.

Figure 2. Zero-potential scanning method to eliminate the crosstalk current.

Figure2.2.Zero-potential Zero-potential scanning scanning method current. Figure methodto toeliminate eliminatethe thecrosstalk crosstalk current.

The zero-potential scanning method can be classified into two groups based on type of a row The zero-potential scanning method can be classified into two groups based on type of a row driver: a row driver composed of digital buffers (or multiplexers) [13,15,16,24–27] (which is referred The zero-potential scanning can be(or classified into two groups based on type of a row driver: a row driver composed ofmethod digital buffers multiplexers) [13,15,16,24–27] (which is referred to as Type I in this paper) and a row driver composed of operational amplifiers [14,28–30] (which is driver: a row driver composed of digital buffers (or multiplexers) [13,15,16,24–27] (which is referred to as Type I in this paper) and a row driver composed of operational amplifiers [14,28–30] (which is referred as Type II in this paper). Although both of the types were proposed to remove the crosstalk to as Type to Itoin paper) and a rowAlthough driver composed operational amplifiers [14,28–30] (which is referred asthis Type II in this paper). both of theof types were proposed to remove the crosstalk current, we found that Type I row driver still suffers from the inaccuracy problem and Type II row referred to we as Type in this paper). of the types were proposed to remove theIIcrosstalk current, foundII that Type I rowAlthough driver stillboth suffers from the inaccuracy problem and Type row driver has the drawbacks of high power consumption, high cost, and high system complexity. These driverwe hasfound the drawbacks of Ihigh high cost, and high system complexity. These current, that Type rowpower driverconsumption, still suffers from the inaccuracy problem and Type II row issues will be analyzed in detail in Section 2. issues will analyzed inofdetail Section 2. driver has thebedrawbacks high in power consumption, high cost, and high system complexity. These In this paper, we analyze and reveal the drawbacks of the previous circuits and propose a new In this paper, we in analyze and reveal the drawbacks of the previous circuits and propose a new issues will be analyzed detail in Section 2. zero-potential scanning method with a compact row driver that alleviates the drawbacks of the Type zero-potential scanninganalyze method with areveal compact driver thatofalleviates the drawbacks of the propose Type paper, therow drawbacks the previous circuits and II In rowthis driver whilewe maintainingand the advantages. The paper is organized as follows. In Section 2, we II row driver while maintaining the advantages. The paper is organized as follows. In Section 2, we a new zero-potential scanning compact row thatand alleviates theprinciple drawbacks of the analyze the previous circuits method to revealwith theiralimitations. Thedriver structure operation of the analyze the previous circuits to reveal their limitations. The structure and operation principle of the Type II row driver while maintaining the advantages. The paper is organized as follows. In Section proposed circuit are described in Section 3. In Section 4, the design, simulation, and measurement 2, proposed circuit are described in Section 3. In Section 4, the design, simulation, and measurement weresults analyze previousand circuits to reveal their limitations. The structure operation arethe discussed, performance comparisons are presented. Finally,and conclusions areprinciple given in of results are discussed, and performance comparisons are presented. Finally, conclusions are given in theSection proposed 5. circuit are described in Section 3. In Section 4, the design, simulation, and measurement Section 5.

results are discussed, and performance comparisons are presented. Finally, conclusions are given in Section 5.

Sensors 2016, 16, 155

3 of 16

Sensors 2016, 16, 155

3 of 16

2. Analysis of Previous Row Drivers for Resistive Sensor Arrays 2. Analysis of Previous Row Drivers for Resistive Sensor Arrays

As described in Section 1, the zero-potential scanning method is widely used for resistive sensor As described in Section 1, the zero-potential scanning method is widely used for resistive sensor arrays to eliminate the crosstalk current. There are two types of row drivers for the zero-potential arrays to eliminate the crosstalk current. There are two types of row drivers for the zero-potential scanning method: a row driver composed of digital buffers and a row driver composed of operational scanning method: a row driver composed of digital buffers and a row driver composed of operational amplifiers. In this section, we analyze these two previous types from the perspective of the accuracy amplifiers. In this section, we analyze these two previous types from the perspective of the accuracy andand thethe system complexity. system complexity. 2.1. Row Driver Composed of Digital Buffers (Type I)

2.1. Row Driver Composed of Digital Buffers (Type I)

The row driver in Figure Figure3,3,which whichisisreferred referred Type I here, The row drivercomposed composedof ofdigital digital buffers buffers in to to as as Type I here, hashas been proposed for resistive sensor arrays [13,15,16,21,24–27]. The system is made up of a resistive been proposed for resistive sensor arrays [13,15,16,21,24–27]. The system is made up of a resistive sensor array, column operational includingdigital digitalbuffers. buffers. sensor array, column operationalamplifiers, amplifiers,and and aa row row driver driver including

Figure 3. Zero-potential scanning method with a row driver composed of digital buffers (Type I). Figure 3. Zero-potential scanning method with a row driver composed of digital buffers (Type I).

In order to verify how accurate row driver Type I is, we will calculate the output voltage of a In order to verifyamplifier how accurate row Typecircuit I is, we will of calculate the output voltage column operational by using thedriver equivalent model Type I shown in Figure 4. Asof a column operational amplifier by using the equivalent circuit model of Type I shown in Figure shown in Figure 4a, a digital buffer is usually an inverter composed of PMOS (MP) and NMOS (M4. N) As shown in Figure 4a, abe digital buffer inverter composed of PMOS transistors. It can modeled asisausually simple an switch circuit which has some (M on-resistance (RON(M ). N ) P ) and NMOS transistors. It when can beamodeled as a simple which row, has some on-resistance (RON Therefore, to a selected the Type I circuit can be ).modeled Therefore, digital buffer deliversswitch VDRIVEcircuit as shown in Figure WhenVaDRIVE digitaltobuffer drivesrow, a selected row,I circuit a static can current flows through RON in when a digital buffer 4b. delivers a selected the Type be modeled as shown thus creating a voltage drop (V DROP ) across R ON . The voltage drop across R ON , V DROP , can be expresses Figure 4b. When a digital buffer drives a selected row, a static current flows through RON thus creating as: a voltage drop (VDROP ) across RON . The voltage drop across RON , VDROP , can be expresses as: R

ON R V = (VDRIVE − VREF ) VDROP DROP “ pVDRIVE ´ VREFRqTOT + RONON R TOT ` RON

(1)

(1)

where VDRIVE is the row driving voltage, VREF is the reference voltage tied to the positive inputs of all

where VDRIVE is the row driving voltage, VREF is the reference voltage tied to the positive inputs of all the column operational amplifiers, and RTOT is the total resistance of all the resistive sensor elements total resistance of all the theincolumn operational amplifiers, RTOT is the selected row. Since they are and connected inthe parallel, RTOT is expressed as: resistive sensor elements in the selected row. Since they are connected in parallel, RTOT is expressed as: 1

RTOT

1

N

=

1

1 RSj j =N ÿ

(2)

1 “ (2) R R TOT Sj Using Equations (1) and (2), we can calculate the output voltage of the jth column operational j “1

amplifier, VOUT,j as:

Sensors 2016, 16, 155

4 of 16

Using (1) and (2), we can calculate the output voltage of the jth column operational Sensors 2016,Equations 16, 155 4 of 16 amplifier, VOUT,j as: ` ˘ RF RF (3) VOUT,j “ V V ´− VVROW,i ´ VREF VOUT , j =REF − V ( ) RSj (3) REF ROW , i REF RSj

R R Sj

F = V REF − (DRIVE V DRIVE −´VV − V REF q “ VREF ´ pV ´V ) REF DROP DROP

RF RSj

(4) (4)

RF RTOTR TOTR F R TOT RTOT + RON` RRSjON RSj

(5) (5)

RF 1 1 RF R {R R R R 1 +1R` ON TOT Sj ON TOT Sj

(6) (6)

“ VREF pV(VDRIVE −´VV = V´ ) q REF REF − DRIVE REF “ VREF pV (VDRIVE ´ = V´ − VV )q REF REF − DRIVE REF

where VROW,i RSjSjisisthe theresistance resistanceofofthe the sensor element at the intersection of the theith ith row row voltage, voltage, R sensor element at the intersection of the where VROW,iisisthe selected row and jth column. selected row and jth column.

(a)

(b) Figure 4. (a) Equivalent switch model of a digital buffer; and (b) equivalent circuit model of a resistive

Figure 4. (a) Equivalent switch model of a digital buffer; and (b) equivalent circuit model of a resistive sensor array with the Type I row driver. sensor array with the Type I row driver.

Equation (6) clearly reveals that the jth column voltage, VOUT,j, is dependent upon the sensor Equation (6)other clearly revealsbecause that the jthiscolumn VOUT,j is dependent uponelements the sensor elements of the columns, RTOT the totalvoltage, resistance of all, the resistive sensor in the same row. If Rcolumns, ON = 0, the because equationRisTOT reduced VOUT,j = VREF − (V − VREF ) RF/RSj sensor which shows elements of the other is thetototal resistance ofDRIVE all the resistive elements that the column voltage is determined by the sensor element in the jth column alone independently in the same row. If RON = 0, the equation is reduced to VOUT,j = VREF ´ (VDRIVE ´ VREF ) RF /RSj of the other that sensor the same row. However, digital buffers,inRthe ON cannot be zero, which shows theelements columnin voltage is determined by in thereal sensor element jth column alone because RON isof the on-resistance of elements the transistor used in the digital buffer asinshown in Figure 4a. WeRON independently the other sensor in the same row. However, real digital buffers, measured the on-resistances of the several commercially available digital buffers and foundbuffer that they are cannot be zero, because RON is on-resistance of the transistor used in the digital as shown in the range of tens of ohms [31–34]. Therefore, in Equation (6), if R ON/RTOT is not much smaller than 1, in Figure 4a. We measured the on-resistances of several commercially available digital buffers and VOUT,j is affected by RON/RTOT. As expressed in Equation (2), RTOT becomes smaller as the number of found that they are in the range of tens of ohms [31–34]. Therefore, in Equation (6), if RON /RTOT is not columns (N) increases and/or as RSj decreases. Conclusively, VOUT,j of Type I is affected by the other much smaller than 1, VOUT,j is affected by RON /RTOT . As expressed in Equation (2), RTOT becomes sensor elements in the same row, and this causes inaccuracy in reading the sensor resistance. This smaller as the number of columns (N) increases and/or as RSj decreases. Conclusively, VOUT,j of Type I inaccuracy problem with the Type I row driver will be also presented through simulations and is affected by the other sensor elements in the same row, and this causes inaccuracy in reading the measurements in Sections 4.

sensor resistance. This inaccuracy problem with the Type I row driver will be also presented through simulations and measurements in Section 4. 2.2. Row Driver Composed of Operational Amplifiers (Type II) Figure 5 shows the zero-potential scanning method using a row driver composed of operational amplifiers (referred to as Type II here) [14,28–30]. Both Type I and Type II commonly use operational

Sensors 2016, 16, 155

5 of 16

2.2. Row Driver Composed of Operational Amplifiers (Type II) Figure 5 shows the zero-potential scanning method using a row driver composed of operational Sensors 2016, 16, 155 of 16 amplifiers (referred to as Type II here) [14,28–30]. Both Type I and Type II commonly use5operational amplifiers at the atcolumn terminals, but usesoperational operational amplifiers instead of digital amplifiers the column terminals, butType TypeIIII also also uses amplifiers instead of digital buffers used the in Type I rowI row driver. buffersinused the Type driver. Sensors 2016, 16, 155

5 of 16

amplifiers at the column terminals, but Type II also uses operational amplifiers instead of digital buffers used in the Type I row driver.

Figure 5. Zero-potential scanning scanning method withwith a rowadriver composed of operational amplifiers (Type II). Figure 5. Zero-potential method row driver composed of operational amplifiers (Type II).

Whereas Type I suffers from inaccuracy in reading a sensor resistance, as analyzed in Section 2.1, Type II is effective in achieving sufficient accuracy. Figure 6 shows the equivalent circuit model of Whereas I suffers from inaccuracy in reading a sensor resistance, as. However, analyzedbecause in Section 2.1, Type II.Type The on-resistance of the selection switch in Figure 5 is represented as RON Figure 5. Zero-potential scanning method with a row driver composed of operational amplifiers (Type Type II no is effective in achieving sufficient shows the positive equivalent static current flows through RON, no accuracy. voltage dropFigure occurs 6across it, the inputcircuit ofII).the model amplifiers is Vof DRIVE , and the rowswitch voltage in VROW,i is virtually equal to VDRIVE of Typeoperational II. The on-resistance the selection Figure 5 is represented as due RONto. the However, Whereas Type I suffersoffrom inaccuracy inamplifiers. reading a Consequently, sensor resistance, asoutput analyzed in Section 2.1, virtual short characteristic the operational the voltage of the jth becauseType no static current in flows through RON ,accuracy. no voltage drop occurstheacross it, thecircuit positive input II is effective achieving sufficient Figure 6 shows equivalent model of of the column amplifier of Type II can be expressed as: operational is VDRIVE , and the row voltage VROW,i is virtually to VDRIVE due to the Type amplifiers II. The on-resistance of the selection switch in Figure 5 is represented as Requal ON. However, because RFacross it, the no static current flowsofthrough RV ON, no= voltage occurs of (7) the virtual short characteristic the operational amplifiers. Consequently, thepositive outputinput voltage of the jth V REF − (Vdrop OUT , j DRIVE − V REF ) R Sj operational amplifiers is V DRIVE , and the row voltage V ROW,i is virtually equal to V DRIVE due to the column amplifier of Type II can be expressed as: virtual short characteristic of the operational amplifiers. Consequently, the output voltage of the jth In Equation (7), it is clear that the column voltage is determined by the resistive sensor element column amplifier of Type II can be expressed as: R in the jth column alone independently the´other sensor´elements VOUT,j “ VofREF pVDRIVE VREF q inF the same row. Therefore,the (7) RSj Type I. Type II row driver can avoid the inaccuracy problem, unlike row R driver VOUT , j = V REF − (V DRIVE − V REF ) F (7) R Sj

In Equation (7), it is clear that the column voltage is determined by the resistive sensor element In Equation (7), it is clear that the column voltage is determined by the resistive sensor element in the jth column alone independently of the other sensor elements in the same row. Therefore, the in the jth column alone independently of the other sensor elements in the same row. Therefore,the Type II row can avoid the the inaccuracy rowdriver driver Type Typedriver II row driver can avoid inaccuracyproblem, problem, unlike unlike row Type I. I.

Figure 6. Equivalent circuit model of a resistive sensor array with row driver Type II.

However, this requires a large number of operational amplifiers. Since an operational amplifier consumes a considerable static power, it can significantly increase the power consumption, especially Figure 6. Equivalent circuit model of a resistive sensor array with row driver Type II.

Figure 6. Equivalent circuit model of a resistive sensor array with row driver Type II. However, this requires a large number of operational amplifiers. Since an operational amplifier consumes a considerable static power, it can significantly increase the power consumption, especially

Sensors 2016, 16, 155

6 of 16

Sensors 2016, 16,this 155 However,

6 of 16 requires a large number of operational amplifiers. Since an operational amplifier consumes a considerable static power, it can significantly increase the power consumption, especially for a large sensor array. For example, a sensor array with N columns and M rows, the Type row for a large sensor array. For example, a sensor array with N columns and M rows, the Type row driver driver II requires N + M operational amplifiers, while the Type I row driver uses M operational II requires N + M operational amplifiers, while the Type I row driver uses M operational amplifiers. amplifiers. A large number of operational amplifiers also increases the system complexity, volume, A large number of operational amplifiers also increases the system complexity, volume, and cost.

and cost.

3. Proposed Row Driver for Zero-Potential Scanning Method (Type III) 3. Proposed Row Driver for Zero-Potential Scanning Method (Type III)

To resolve the problems of inaccuracy of Type I and high circuit complexity of Type II mentioned in To resolve the problems of inaccuracy of Type I and high circuit complexity of Type II mentioned Section 2, we propose a row driver using ausing singlea operational amplifier for the for zero-potential scanning in Section 2, we propose a row driver single operational amplifier the zero-potential method (referred to as Type III) as shown in Figure 7. The Type II row driver uses an operational scanning method (referred to as Type III) as shown in Figure 7. The Type II row driver uses an amplifier for each row, but the proposed rowproposed driver uses singleuses operational amplifier for the entire operational amplifier for each row, but the rowadriver a single operational amplifier rows, maintaining themaintaining merit of Type for while the entire rows, while the II. merit of Type II.

Figure 7. Proposed zero-potential scanning method with the proposed compact row driver using a Figure 7. Proposed zero-potential scanning method with the proposed compact row driver using a single operational amplifier (Type III). single operational amplifier (Type III).

The scanning operation of the proposed circuit is illustrated in Figure 8. When the first row is The scanning of the proposed circuit is illustrated in other Figureswitches 8. When first row driven, a pair of operation switches, SW 1A and SW1B, are turned on, while the arethe turned off. is driven, pair the of switches, and SWinput on, while the other Then,aboth output andSW the of turned the operational amplifier areswitches connectedare to turned the firstoff. 1Anegative 1B , are Then, both thethe output androw theisnegative input of the operational are connected toare theturned first row. row. When second driven, SW 2A and SW2B are turned amplifier on, and the other switches off. the With this driving the2B remaining row, proposed Type III can drive all theoff. When second row isrepeated driven, throughout SW2A and SW are turned on,the and the other switches are turned rows a single operational amplifier. With thisusing driving repeated throughout the remaining row, the proposed Type III can drive all the rows selected row is driven by a row driver, it can be modeled by the equivalent circuit in using a When single aoperational amplifier. Figure 9, where R ON,A and ON,B represent thedriver, on-resistnace row selection SWcircuit 1A and in When a selected row isRdriven by a row it can of be the modeled by the switches equivalent SW 1B in Figure 8, respectively. As shown in the figure, the row driving current flows through RON,A, Figure 9, where RON,A and RON,B represent the on-resistnace of the row selection switches SW1A and no current flows through RON,B, because the input impedance of the operational amplifier is SWbut 1B in Figure 8, respectively. As shown in the figure, the row driving current flows through RON,A , almost infinite. Consequently, notwithstanding the fact that the row driving current makes the but no current flows through RON,B , because the input impedance of the operational amplifier is almost voltage drop across RON,A, the row voltage VROW,j is regulated to VDRIVE by the virtual short infinite. Consequently, notwithstanding the fact that the row driving current makes the voltage drop characteristic of the operational amplifier. Therefore, the output voltage of the jth column operational across RON,A , the row voltage VROW,j is regulated to VDRIVE by the virtual short characteristic of the amplifier of Type III can be expressed as:

operational amplifier. Therefore, the output voltage of the jth column operational amplifier of Type III R can be expressed as: VOUT , j = V REF − (V DRIVE − V REF ) F (8) RSj R F VOUT,j “ VREF ´ pVDRIVE ´ VREF q (8) RSj This equation is exactly the same with Equation (7), and VOUT,j is determined by the sensor element in the jth column alone and is independent of the other sensor elements; therefore, the

Sensors 2016, 16, 155

7 of 16

This equation is exactly the same with Equation (7), and VOUT,j is determined by the sensor element in2016, the16,jth Sensors 155column alone and is independent of the other sensor elements; therefore, 7 of 16 the Sensors 2016, 16, 7 of 16 proposed Type III 155 driver is as accurate as the Type II in measuring the sensor resistance. Additionally, proposed Typethat III driver is as accurate asrows the Type in measuring thefloating sensor resistance. Additionally, it should be noted all the unselected areIInot electrically but also regulated to VREF Type III driver as accurate as the Type in electrically measuring the sensor resistance. Additionally, itproposed should noted that allis the unselected rows are IInot floating but also regulated tothe VREF by the columnbe drivers (not the row drivers). This means that no current flows through sensor it should be noted that (not all the rowsThis are means not electrically floatingflows but also regulated to VREF by the column drivers theunselected row drivers). that no current through the sensor elements of the unselected rows. Consequently, the accuracy of the Type III row driver is expected to by the column drivers (not the row drivers). This means that no Type current through the sensor elements of the unselected rows. Consequently, the accuracy of the III flows row driver is expected to be identical to of that ofunselected the Typerows. II row driver, because no row or column is eletrcially floating in both elements the Consequently, the accuracy of the Type III row driver is expected to of be identical to that of the Type II row driver, because no row or column is eletrcially floating in both the row drivers. The simulation measurement results will in Section Section bethe identical to that of the Typeand II row driver, because no row orbe column is eletrcially floating of row drivers. The simulation and measurement results will bepresented presented in 4.4. in both of the row drivers. The simulation and measurement results will be presented in Section 4.

(a) (a)

(b) (b) Figure 8. Scanning operation of the proposed row driver Type III. (a) 1st and (b) 2nd row is driven Figure 8. Scanning operation of the proposed row driver Type III. (a) 1st and (b) 2nd row is driven. Figure 8. Scanning operation of the proposed row driver Type III. (a) 1st and (b) 2nd row is driven

Figure 9. Equivalent circuit model of a resistive sensor array with the proposed Type III row driver. Figure 9. Equivalent circuit model of a resistive sensor array with the proposed Type III row driver.

Figure 9. Equivalent circuit model of a resistive sensor array with the proposed Type III row driver.

Sensors 2016, 16, 155

8 of 16

From the operation described above, we can find that the proposed Type III driver can drive all of the rows using a single operational amplifier without the inaccuracy problem. That is to say, the proposed Type III driver can drastically reduce the number of operational amplifiers. For example, to drive N ˆ M resistive sensor array, the proposed Type III needs only N + 1 operational amplifiers, which is a markedly reduced number compared with that of Type II, which uses N + M operational amplifiers. With the reduced number of operational amplifiers, the proposed Type III can considerably reduce the power consumption, system complexity, and cost compared with Type II. Although each row has two additional switches, they are much simpler than the operational amplifier. The characteristics of three types of row drivers are summarized in Table 1. All types of the row drivers can eliminate crosstalk current because they are all based on the zero-potential scanning method. However, Type I significantly degrades the accuracy because the voltage measurement of a sensor in a row is affected by the other sensors in the same row. We reveal that this inaccuarcy is not caused by the crosstalk current, but comes from the row driver circuit itself. Type II successfully eliminates both inaccuracy and crosstalk current, but leads to high power consumption, high cost, and high complexity, since it requires a large number of amplifiers. On the contrary, the proposed Type III not only removes both crosstalk and inaccuracy, but requires only one amplifier for all rows. Therefore, we can conclude that the proposed Type III simulataneously takes the advantages of both Type I and II. Table 1. Comparison of three types of row drivers. Type I Crosstalk current Inaccuracy Number of operational amplifiers (for N ˆ M array)

eliminated exists N

Type II eliminated eliminated N+M

Type III (Proposed) eliminated eliminated N+1

4. Results and Discussion To compare the characteristics of the previous Type I, II, and the proposed Type III row driver circuits, before real implementation and measurement, we performed PSPICE simulations [35]. The resistance of the sensor elements are assumed to be in the range from 100 Ω to 1 MΩ, which can cover most of the previously reported resistive sensors [1–4,6–9,11,12,14]. For accurate read-out of the sensor resistance, we should be careful to choose an operational amplifier since an operational amplifier has non-ideal properties affecting the accuracy. First, an operational amplifier should provide a low-offset voltage which is a source of output voltage error. Second, the open-loop gain should be large enough to reduce gain error. Third, it should provide a sufficient current driving capability. In addition, an input bias current flowing into both inputs of an operational amplifier should be very small. Considering these requirements, we use the high precision amplifier, LMP7701 for both column and row amplifiers [36]. The operational amplifier has the specifications as follows: typical offset voltage = ˘37 µV, open-loop gain = 121 dB, maximum driving current = 86 mA, and maximum input bias current = 1 pA. With these specifications, errors caused by the non-ideal properties of the operational amplifier can be neglected. If inccuracy exists in reading a sensor resistance, it causes the output voltage to deviate from the true value as discussed in Section 2; thus, we can define the percentage error of the output voltage as: ˇ ˇ ˇV ´ VIDEAL ˇˇ e p%q “ ˇˇ OUT ˇ ˆ 100 VIDEAL

(9)

where VOUT is measured (or simulated) the output voltage of the column amplifier and VIDEAL is the ideal output voltage which should be obtained without any inaccuracy as expressed in Equations (7) or (8). Figure 10a shows the schematic for simulations of row driver Type I with N ˆ M resistive sensor array. As described in Section 2, the digital buffer in row driver Type I can be modeled as a

Sensors 2016, 16, 155

9 of 16

Sensors 2016, 16, 155

9 of 16

resistor, RON . Here, RON is set at 10 Ω which is a typical resistance found in many commercial digital buffers [31,32,34]. we observe will observe the output voltage at Column 1 (V ), where the resistor [31,32,34]. Now Now we will the output voltage at Column 1 (VOUT1 ), OUT1 where the resistor to be to be sensed, RS11 andRRS21S21 , are assumed 1 kΩ and respectively. of the other resistors S11and , are assumed to to bebe 1 kΩ and 100100 Ω,Ω, respectively. AllAll of the other resistors areare sensed, R assumed to to bebe 100 Ω.Ω.InInthis therow rowisisdriven driventotoVV (0.5 DRIVE (0.5 assumed 100 thissimulation, simulation,when when aa row row is selected, selected, the DRIVE V) V) andand all all thethe other rows are biased to V (0.6 V). other rows are biased to VREF REF (0.6 V).

(a)

(b)

(c)

(d)

(e)

Figure 10. (a) Schematic for the simulations of the Type I row driver; (b) transient simulation results;

Figure 10. (a) Schematic for the simulations of the Type I row driver; (b) transient simulation results; (c) output errors as a function of the sensor resistance with different column numbers; (d) output (c) output errors as a function of the sensor resistance with different column numbers; (d) output errors errors as a function of the sensor resistance with different adjacent sensor resistance (RADJ), and as a function of the sensor resistance with different adjacent sensor resistance (RADJ ), and (e) output (e) output errors as a function of the sensor resistance with different row numbers. errors as a function of the sensor resistance with different row numbers.

The transient simulation results of row driver Type I are shown in Figure 10b. By selecting Row (assumed 1 kΩ here). 1 and thesimulation output voltage at Column 1, we obtain ofin RS11 Thereading transient results of row driver Type I the are value shown Figure 10b. to Bybe selecting Row 1 should 0.85 V voltage if RON is ideally 0 Ω in (6).the However, in the figure, the error andVOUT1 reading thebeoutput at Column 1,Equation we obtain value ofasRshown (assumed to be 1 kΩ here). S11 of 7.0% is observed, which agrees very well with the error predicted in Equations (6) and (9). Then, VOUT1 should be 0.85 V if RON is ideally 0 Ω in Equation (6). However, as shown in the figure, the

Sensors 2016, 16, 155

10 of 16

error of 7.0% is observed, which agrees very well with the error predicted in Equations (6) and (9). Then, we select Row 2 and read the output voltage at Column 1 to find out the value of RS21 (assumed to be 100 Ω here). Ideally, VOUT1 should be 3.1 V, but the output error reaches 23.0% as also calculated in Equations (6) and (9). The simulated 99% settling time is 13 µs. To also find the impact of the number of columns, we performed the simulations by changing the number of columns while fixing the number of rows at ten. Figure 10c shows the output voltage error as the sensor resistance to be measured (which is referred to as target sensor here) is swept from 100 Ω to 1 MΩ, while all of the other sensor resistances in the same row stay at 100 Ω. If the target sensor resistance is 1 MΩ, which is 10,000 times larger than the other sensor elements, the error is very small. However, if the target sensor is 100 Ω, the error is drastically increased. If the target sensor is 100 Ω, the simulation result with a single column shows that the error is under 10%. However, with 10 columns, the error reaches almost 40%, and with 30 columns, the error is increased to over 60%. These results agree very well with the expectation from Equation (6). Figure 10d shows the output voltage error for different resitance of all the other sensor elements (RADJ ) in a 10 ˆ 10 array. RADJ is swept from 100 Ω to 1000 kΩ. As expected in Equations (2) and (6), the error is increased as RADJ becomes similar to RON (100 Ω here). When RADJ is 1000 kΩ, the error is 7.3%, but when RADJ is decreased to 100 Ω, the error is increased to 40.2%. From the simulation results in Figure 10c,d, we can find that the Type I row driver suffers from the reading inaccuracy when the array becomes larger and RADJ becomes similar to RON ; therefore, it is not suitable for large sensor array applications. Figure 10e shows the output voltage error for different numbers of rows with ten columns. There is no difference in the error for various numbes of rows, because the error is caused by RON and the currents flows through RON of the selected row only. In other words, all the other rows, except the selected row, has no current, because their row and column voltages are tied to VREF as explained in Section 2. Figure 11a shows the schematic for simulations of row driver Type II. As mentioned in Section 2, the selection switch is modeled as a resistor, RON . All the simulation conditions are the same with the previous simulation with the Type I row driver. The transient simulation results of Type II are shown in Figure 11b. Unlike Type I, the Type II row driver creates a very small error even when the target sensor resistance is small (100 Ω). The simulated 99% settling time is 14 µs when a driven row is swiched from Row1 to Row 2. Figure 11c shows the error of the Type II circuit as we sweep the target sensor resistance for a variety number of columns while fixing the number of rows at ten. As revealed in Equation (7) where the output voltage is only determined by a target sensor, not being affected by the other sensors in the same row, the simulation results show that the errors of the Type II circuit are very small: under 0.1%. In addition, as shown in Figure 11d, for different resitance of all the other sensor elements (RADJ ) from 100 Ω to 1000 kΩ, the Type II circuit still provides the small error under 0.1%. This is also evident in Equation (7) because there is no parameter from the other sensor elements, unlike Equations (2) and (6) for the Type I circuit. Therefore, we can conclude that the Type II row driver circuit can successfully resolve the inaccuracy problem, which is the main advantage of Type II. However, as described in Section 2, row driver Type II needs the same number of operational amplifiers as the rows; consequently, the complexity, fabrication cost, and power consumption are severely increased. Figure 11e shows the output voltage error for different numbers of rows with ten columns. For all different numbers of rows, the error remains very small under 0.1%. Figure 12a shows the schematic for simulations of the proposed Type III row driver. As described in Section 3, the selected row is connected to both the output and negative input of the operational amplifier, but the other row is disconnected (see Figure 8). All simulation conditions are the same with the previous simulations above. Figure 12b shows the transient simulation result of the proposed Type III row driver. Very similarly with Type II, the proposed Type III row driver also shows a very small error under 0.1% even when the target sensor resistance is

Sensors 2016, 16, 155

11 of 16

small (100 Ω). The simulated 99% settling time is 14 µs which is equal to that of Type II row driver. ThisSensors result agrees 2016, 16, 155well with the expectation described in Section 3. 11 of 16

(a)

(b)

(c)

(d)

(e)

Figure 11. (a) Schematic for the simulations of the Type II row driver, (b) transient simulation results,

Figure 11. (a) Schematic for the simulations of the Type II row driver, (b) transient simulation results, (c) output errors as a function of the sensor resistance with different column numbers, (d) output (c) output errors as a function of the sensor resistance with different column numbers, (d) output errors errors as a function of the sensor resistance with different adjacent sensor resistance (RADJ), and (e) as a function of the sensor resistance with different adjacent sensor resistance (RADJ ), and (e) output output errors as a function of the sensor resistance with different row numbers. errors as a function of the sensor resistance with different row numbers.

Figure 12c shows the error of the proposed Type III circuit as we sweep the target sensor Figure 12c the error of the proposedwhile Type fixing III circuit we sweep the target resistance resistance forshows a variety number of columns the as number of rows at ten.sensor The proposed III circuit alsoofshows the low error under In addition, in Figure 12d,Type for different for aType variety number columns while fixing the0.1%. number of rowsasatshown ten. The proposed III circuit 100 Ω in to Figure 1000 kΩ, thefor proposed Type III circuit of all other sensor elements (RADJ) from alsoresitance shows the lowthe error under 0.1%. In addition, as shown 12d, different resitance of all the small error under like theΩType II circuit. Figure 12e shows theIII output voltage errorthe the provides other sensor elements (RADJ )0.1% from 100 to 1000 kΩ, the proposed Type circuit provides forerror different numbers of rows with II tencircuit. columns. For all numbers rows, the error small under 0.1% like the Type Figure 12edifferent shows the outputofvoltage error forremains different very small, under 0.1% like the Type II circuit. From the results in Figure 12c–e, it is found that the numbers of rows with ten columns. For all different numbers of rows, the error remains very small, Type III circuit can resolve the inaccuracy problem, like the Type II row driver, but with a single row amplifier.

Sensors 2016, 16, 155

12 of 16

under 0.1% like the Type II circuit. From the results in Figure 12c–e, it is found that the Type III circuit canSensors resolve inaccuracy problem, like the Type II row driver, but with a single row amplifier. 2016,the 16, 155 12 of 16

(a)

(b)

(c)

(d)

(e)

Figure 12. (a) Schematic for the simulations of the proposed row driver Type III; (b) transient

Figure 12. (a) Schematic for the simulations of the proposed row driver Type III; (b) transient simulation results; (c) output errors as a function of the sensor resistance with different column simulation results; (c) output errors as a function of the sensor resistance with different column numbers; numbers; (d) output errors as a function of the sensor resistance with different adjacent sensor (d) output errors as a function of the sensor resistance with different adjacent sensor resistance(RADJ ); resistance(RADJ); and (e) output errors as a function of the sensor resistance with different row numbers. and (e) output errors as a function of the sensor resistance with different row numbers.

In addition to simulations, we also evaluated the characteristics of the previous row driver Type I, addition to simulations, we alsoby evaluated the characteristics the with previous driver II, In and the proposed Type III circuits, establishing a 4 × 4 resistor of array thoserow three typesType of I, row the drivers on a printed board (PCB) as shown Figure 13. Thewith high-precision II, and proposed Type IIIcircuit circuits, by establishing a 4 ˆin4 resistor array those threeamplifier, types of row LMP7701 used for simulations used measurements. The high-accuracy drivers on a which printediscircuit board (PCB) as above shownisinalso Figure 13.for The high-precision amplifier, LMP7701 resistors with the tolerance of 0.1% are used for the resistor array. A digital buffer (SN54AHCT126) which is used for simulations above is also used for measurements. The high-accuracy resistors with 15 are Ω isused used for for row driver Type I [34]. The row-selection switch of the Type II row RON of about thewith tolerance of 0.1% the resistor array. A digital buffer (SN54AHCT126) with RON of ON is typically 2.5 Ω [37]. The proposed driver is realized with SPDT switches (ADG733) of which R about 15 Ω is used for row driver Type I [34]. The row-selection switch of the Type II row driver is Type III row driver uses an eight-channel multiplexer (CD4051B) of which RON is typically 125 Ω as realized with SPDT switches (ADG733) of which RON is typically 2.5 Ω [37]. The proposed Type III a row selection switch [38]. As shown in the figure, the Type II row driver uses four operational row driver uses an eight-channel multiplexer (CD4051B) of which RON is typically 125 Ω as a row amplifiers to drive four rows, but the proposed Type III uses only one operational amplifier. Figure 14 shows the measured waveforms of the previous Type I, II, and the proposed Type III row drivers with the 4 × 4 resistor array shown in Figure 13. Similar to the simulations above, when

Sensors 2016, 16, 155

13 of 16

selection switch [38]. As shown in the figure, the Type II row driver uses four operational amplifiers to drive four rows, but the proposed Type III uses only one operational amplifier. Figure 14 shows the measured waveforms of the previous Type I, II, and the proposed Type III Sensors 2016, 16, 155the 4 ˆ 4 resistor array shown in Figure 13. Similar to the simulations above, 13 ofwhen 16 row drivers Sensors 2016,with 16, 155 13 of 16 we select Row 1 we measure the value of RS11 by reading the output voltage at Column 1 (VOUT1 ). we select Row 1 we measure the value of RS11 by reading the output voltage at Column 1 (VOUT1). Since weRselect 1 we measure theinvalue RS11 by reading theshould outputbe voltage at ideally, Columnif1there (VOUT1is). no Since Since isRow assumed to be 1 kΩ this of experiment, VOUT1 0.85 V, error. assumed to be 1 kΩ in this experiment, VOUT1 should be 0.85 V, ideally, if there is no error. As RS11 isS11 to be 1the kΩerror in this VOUT1 should be 0.85 V, ideally, if there is III no row error. As RS11 is assumed As shown in the figure, of experiment, the Type I row driver reaches 9.5%, but Type II and drivers shown in the figure, the error of the Type I row driver reaches 9.5%, but Type II and III row drivers shown insmall the figure, the error ofThen, the Type select I row driver 9.5%, but Type voltage II and IIIatrow drivers1 to show very Row22reaches andread read theoutput output Column show very smallerrors errorsofof0.1%. 0.1%. Then, we we select Row and the voltage at Column 1 to show very small errors of 0.1%. Then, we select Row 2 and read the output voltage at Column 1 toI is find outout RS21 Asexpected expectedininEquation Equation (6), output error of Type find RS21(assumed (assumedto tobe be 100 100 Ω Ω here). here). As (6), thethe output error of Type I is find increased out RS21 (assumed tobut be 100 ΩIIhere). As expected in Equation (6), the output error errors of Type is further to to 30.7%, Type and Type III III row drivers stillstill provide very small ofIof 0.1%. further increased 30.7%, but Type II and Type row drivers provide very small errors further increased to 30.7%, but Type II and Type III row drivers still provide very small errors of Additionally, both Type II and Type III show exactly transienttransient responseresponse (rising time); thus, 0.1%. Additionally, both Type II and Type the III show theidentical exactly identical (rising 0.1%. Additionally, both Type II and Type III show the exactly identical transient response (rising their dynamic characteristics are identicalare when the driven are switched. time); thus, their dynamic characteristics identical when rows the driven rows are switched. time); thus, their dynamic characteristics are identical when the driven rows are switched.

Figure13. 13.Test Testboard board with row row drivers Type I, II, III, 4 resistor array. Figure III,and and444×׈ resistor array. Figure 13. Test boardwith with rowdrivers drivers Type Type I, I, II, II, III, and 4 4resistor array.

Figure 14. Measured waveforms of row drivers Type I, II, and the proposed Type III.

Figure 14. Measured waveforms of row drivers Type I, II, and the proposed Type III. Figure 14. Measured waveforms of row drivers Type I, II, and the proposed Type III.

Figure 15 shows the measured output errors of the three types of row drivers with 4 × 4 resistor Figure 15 shows the measured output errors of the three types of row drivers with 4 × 4 resistor array as a 15 function themeasured target sensor resistance the other resistors have 100 Ω. 4 The Figure showsof the output errors while of theall three row drivers with ˆ 4target resistor array as a function of the target sensor resistance while all thetypes other of resistors have 100 Ω. The target sensor resistance is swept fromsensor 100 Ωresistance to 1 MΩ. As shown in the figure, row driver Type IThe shows array as a function of the target while all the other resistors have 100 Ω. target sensor resistance is swept from 100 Ω to 1 MΩ. As shown in the figure, row driver Type I shows severely large error over 30.7% as the target sensor resistance is decreased to a comparable value with sensor resistance is swept from as 100 to 1 MΩ. shown isindecreased the figure, driver Type shows severely large error over 30.7% theΩtarget sensorAs resistance to arow comparable valueI with all the other resistors (100 Ω), but the Type II and the proposed Type III row drivers provide small all the other resistors (100 Ω), but the Type II and the proposed Type III row drivers provide small errors under 0.1%. From these measurement results, which agree very well with both the calculations errors under 0.1%. From these measurement results, which agree very well with both the calculations and simulations in Sections 2 and 3, we confirm that the Type I row driver suffers from the reading and simulations in Sections 2 and 3, we confirm that the Type I row driver suffers from the reading

Sensors 2016, 16, 155

14 of 16

severely large error over 30.7% as the target sensor resistance is decreased to a comparable value with all the other resistors (100 Ω), but the Type II and the proposed Type III row drivers provide small errors under 0.1%. From these measurement results, which agree very well with both the calculations and simulations in Sections 2 and 3 we confirm that the Type I row driver suffers from 2016, 16, 155 14 of 16 the Sensors reading inaccuracy, but the Type II and the proposed Type III row drivers effectively solve the inaccuracy problem. inaccuracy, but the Type II and the proposed Type III row drivers effectively solve the inaccuracy The performance comparison of the Type I, II, and the proposed Type III row driver circuits problem. is summarized in Table comparison 2. The proposed Type III and rowthe driver provides requires The performance of the Type I, II, proposed Type small III rowerror, driverbut circuits is only one more operational amplifier than row driver Type I. Therefore, we can conclude that summarized in Table 2. The proposed Type III row driver provides small error, but requires only onethe proposed row driver amplifier Type III circuit is very efficient remove thewe inaccuracy withthat littlethe increase in the more operational than row driver Type to I. Therefore, can conclude proposed circuit rowcomplexity. driver Type III circuit is very efficient to remove the inaccuracy with little increase in the circuit complexity.

Figure Measured outputerrors errorsofofthree threetypes types of of row row driver driver circuits 4 resistor array. Figure 15.15. Measured output circuitswith with44׈ 4 resistor array. Table 2. Performance comparison of row drivers.

Table 2. Performance comparison of row drivers. Measured maximum error 1 Number of operational amplifiers (for 1 N × M array)

Type I 30.7%I Type N

Type II Type III (Proposed) ≤0.1% ≤0.1% Type II Type III (Proposed) N+M N+1

30.7% ď0.1% ď0.1% Measured maximum error 1 Measurement conditions: 4 × 4 resistor array, target sensor resistance = 100 Ω, and all the other Number of operational amplifiers (for N ˆ M array) N N+M N+1 sensor resistance = 100 Ω

Measurement conditions: 4 ˆ 4 resistor array, target sensor resistance = 100 Ω, and all the other sensor resistance = 100 Ω. 1

5. Conclusions

5. Conclusions In order to resolve the inaccuracy and high complexity problems found in the previous row drivers, we propose a new row driver circuit which uses only one operational amplifier to drive all In order to resolve the inaccuracy and high complexity problems found in the previous row rows of a sensor array. Although the proposed Type III row driver has only one operational amplifier drivers, we propose a new row driver circuit which uses only one operational amplifier to drive all to drive all the rows, it successfully removes the inaccuracy problem by using a proper feedback operation. rows of athe sensor array. Although the proposed Type IIIthe row driver has only one operational amplifier With reduced number of operational amplifiers, proposed Type III circuit can achieve the low to drive all the rows, successfully removes the inaccuracy proper feedback operation. power, low cost,itand low complexity which are stronglyproblem requiredby forusing large asensor arrays. The PSPICE With the reduced number of operational amplifiers, the proposed Type III circuit can achieve simulation results show that the proposed circuit provides a small error of under 0.1% error the evenlow power, low cost, and low complexity which are strongly required for large sensor arrays. The PSPICE with 30 columns of a sensor array. The measurement results from the circuit implementation show simulation show thatwith the aproposed circuit provides small iserror of underreduced 0.1% error with that the results maximum error 4 × 4 resistor array is 0.1%awhich remarkably fromeven 30.7% 30 columns of a sensor array. The measurement results from the circuit show that the of the previous counterpart. From these improved performances, we canimplementation conclude that the proposed Type III error row driver very well suited resistive array applications. maximum with circuit a 4 ˆ 4isresistor array is for 0.1% whichsensor is remarkably reduced from 30.7% of the

previous counterpart.This From thesewas improved performances, weTechnology can conclude that the Program proposed Type III Acknowledgments: research supported by Nano Material Development through rowthe driver circuit is very well suited for resistive sensor array applications. National Research Foundation of Korea (NRF) funded by the Ministry of Education, Science and Technology (2012M3A7B4035201). Author Contributions: Jong-Seok Kim provided the initial idea of this research, performed the experiments, and wrote the paper. Dae-Yong Kwon designed the test board and performed the experiments. Byong-Deok Choi contributed to the data analysis and the verification of the experiments, and wrote the paper. Conflicts of Interest: The authors declare no conflict of interest.

Sensors 2016, 16, 155

15 of 16

Acknowledgments: This research was supported by Nano Material Technology Development Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education, Science and Technology (2012M3A7B4035201). Author Contributions: Jong-Seok Kim provided the initial idea of this research, performed the experiments, and wrote the paper. Dae-Yong Kwon designed the test board and performed the experiments. Byong-Deok Choi contributed to the data analysis and the verification of the experiments, and wrote the paper. Conflicts of Interest: The authors declare no conflict of interest.

References 1. 2. 3. 4.

5.

6. 7. 8. 9.

10. 11. 12. 13.

14. 15. 16.

17. 18.

Shikida, M.; Shimizu, T.; Sato, K.; Itoigawa, K. Active tactile sensor for detecting contact force and hardness of an object. Sens. Actuators A Phys. 2003, 103, 213–218. [CrossRef] Noda, K.; Hoshino, K.; Matsumoto, K.; Shimoyama, I. A shear stress sensor for tactile sensing with the piezoresistive cantilever standing in elastic material. Sens. Actuators A Phys. 2006, 127, 295–301. [CrossRef] Huang, Y.-M.; Tsai, N.-C.; Lai, J.-Y. Development of tactile sensors for simultaneous, detection of normal and shear stresses. Sens. Actuators A Phys. 2010, 159, 189–195. [CrossRef] Wang, Y.; Wang, A.X.; Wang, Y.; Chyu, M.K.; Wang, Q.-M. Fabrication and characterization of carbon nanotube–polyimide composite based high temperature flexible thin film piezoresistive strain sensor. Sens. Actuators A Phys. 2013, 199, 265–271. [CrossRef] Thanh-Vinh, N.; Binh-Khiem, N.; Takahashi, H.; Matsumoto, K.; Shimoyama, I. High-sensitivity triaxial tactile sensor with elastic microstructures pressing on piezoresistive cantilevers. Sens. Actuators A Phys. 2014, 215, 167–175. [CrossRef] Singh, K.; Joyce, R.; Varghese, S.; Akhtar, J. Fabrication of electron beam physical vapor deposited polysilicon piezoresistive MEMS pressure sensor. Sens. Actuators A Phys. 2015, 223, 151–158. [CrossRef] Eminoglu, S.; Tezcan, D.S.; Tanrikulu, M.Y.; Akin, T. Low-cost uncooled infrared detectors in CMOS process. Sens. Actuators A Phys. 2003, 109, 102–113. [CrossRef] Saxena, R.S.; Panwar, A.; Lamba, S.S.; Bhan, R.K. A sub-circuit model of a microbolometer IR detector and its experimental validation. Sens. Actuators A Phys. 2011, 171, 138–145. [CrossRef] Sumesh, M.A.; Karanth, S.; Prakash, S.; Laxmiprasad, A.S.; Nagendra, C.L. Ion beam sputtered Ge–Si–O amorphous thin films for microbolometer infrared detectors and their application in earth sensors. Sens. Actuators A Phys. 2013, 192, 81–91. [CrossRef] Sun, X.Y.; Luo, W.B.; Meng, J.; Qing, X.; Fu, W.Y.; Shuai, Y.; Wu, C.G. Monolithic pyroelectric infrared detectors using SiO2 aerogel thin films. Sens. Actuators A Phys. 2015, 228, 69–74. [CrossRef] Saxena, R.S.; Bhan, R.K.; Aggrawal, A. A new discrete circuit for readout of resistive sensor arrays. Sens. Actuators A Phys. 2009, 149, 93–99. [CrossRef] Wang, X.; Fitch, A.L.; Lu, H.H.; Sreeram, V.; Qi, W. Implementation of an analogue model of a memristor based on a light-dependent resistor. Chin. Phys. B 2012, 21. [CrossRef] Shimojo, M.; Ishikawa, M.; Kanaya, K. A flexible high resolution tactile imager with video signal output. In Proceedings of the IEEE International Conference on Robotics and Automation, Sacramento, CA, USA, 9–11 April 1991; pp. 384–389. D’Alessio, T. Measurement errors in the scanning of piezoresistive sensors arrays. Sens. Actuators A Phys. 1999, 72, 71–76. [CrossRef] Shimojo, M.; Namiki, A.; Ishikawa, M.; Makino, R.; Mabuchi, K. A tactile sensor sheet using pressure conductive rubber with electrical-wires stitched method. IEEE Sens. J. 2004, 4, 589–596. [CrossRef] Yang, Y.-J.; Cheng, M.-Y.; Chang, W.-Y.; Tsao, L.-C.; Yang, S.-A.; Shih, W.-P.; Chang, F.-Y.; Chang, S.-H.; Fan, K.-C. An integrated flexible temperature and tactile sensing array using PI-copper films. Sens. Actuators A Phys. 2008, 143, 143–153. [CrossRef] Snyder, W.E.; St. Clair, J. Conductive elastomers as sensor for industrial parts handling equipment. IEEE Trans. Instrum. Meas. 1978, 27, 94–99. [CrossRef] Prutchi, D.; Arcan, M. Dynamic contact stress analysis using a compliant sensor array. Measurement 1993, 11, 197–210. [CrossRef]

Sensors 2016, 16, 155

19.

20.

21. 22. 23. 24. 25. 26.

27. 28.

29.

30. 31. 32. 33. 34. 35. 36. 37. 38.

16 of 16

Takei, K.; Takahashi, T.; Ho, J.C.; Ko, H.; Gillies, A.G.; Leu, P.W.; Fearing, R.S.; Javey, A. Nanowire active-matrix circuitry for low-voltage macroscale artificial skin. Nat. Mater. 2010, 9, 821–826. [CrossRef] [PubMed] Kaltenbrunner, M.; Sekitani, T.; Reeder, J.; Yokota, T.; Kuribara, K.; Tokuhara, T.; Drack, M.; Schwödiauer, R.; Graz, I.; Bauer-Gogonea, S.; et al. An ultra-lightweight design for imperceptible plastic electronics. Nature 2013, 499, 458–463. [CrossRef] [PubMed] Liu, H.; Zhang, Y.-F.; Liu, Y.-W.; Jin, M.-H. Measurement errors in the scanning of resistive sensor arrays. Sens. Actuators A Phys. 2010, 163, 198–204. [CrossRef] Wu, J.; Wang, L.; Li, J. VF-NSE method measurement error analysis of networked resistive sensor array. Sens. Actuators A Phys. 2014, 211, 45–50. [CrossRef] Wu, J.; Wang, L.; Li, J. Design and crosstalk error analysis of the circuit for the 2-D networked resistive sensor array. IEEE Sens. J. 2015, 15, 1020–1026. [CrossRef] Vidal-Verdú, F.; Bravo-Gálvez, D.; Oballe-Peinado, O. Tactile sensor hardware based on programmable SoC. Electron. Lett. 2010, 46, 1117–1118. [CrossRef] Cheng, M.-Y.; Tsao, C.-M.; Lai, Y.-Z.; Yang, Y.-J. The development of a highly twistable tactile sensing array with stretchable helical electrodes. Sens. Actuators A Phys. 2011, 166, 226–233. [CrossRef] Vidal-Verdú, F.; Oballe-Peinado, Ó.; Sánchez-Durán, J.A.; Castellanos-Ramos, J.; Navas-González, R. Three realizations and comparison of hardware for piezoresistive tactile sensors. Sensors 2011, 11, 3249–3266. [CrossRef] [PubMed] Wang, H.; Zhou, D.; Cao, J. Development of a skin-like tactile sensor array for curved surface. IEEE Sens. J. 2014, 14, 55–61. [CrossRef] Lazzarini, R.; Magni, R.; Dario, P. A tactile array sensor layered in an artificial skin. In Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems, Pittsburgh, PA, USA, 5–9 August 1995; pp. 114–119. Iwashita, A.; Shimojo, M. Development of a mixed signal LSI for tactile data processing. In Proceedings of the 2004 IEEE International Conference on Systems, Man and Cybernetics, The Hague, The Netherlands, 10–13 October 2004; Volume 5, pp. 4408–4413. Castellanos-Ramos, J.; Navas-González, R.; Macicior, H.; Sikora, T.; Ochoteco, E.; Vidal-Verdú, F. Tactile sensors based on conductive polymers. Microsyst. Technol. 2010, 16, 765–776. [CrossRef] Fairchild Inc. NC7NZ34, TinyLogic® UHS Triple Buffer. Available online: www.fairchild.com (accessed on 23 November 2015). Intersil Inc. ISL55100B, Quad 18V Pin Electronics Driver/Window Comparator. Available online: www.intersil.com (accessed on 23 November 2015). Texas Instrument Inc. SN54HCT244, SN74HCT244 Octal Buffers and Line Drivers with 3-State Outputs. Available online: www.ti.com (accessed on 13 January 2016). Texas Instrument Inc. SNx4AHCT126 Quadruple Bus Buffer Gates with 3-State Outputs. Available online: www.ti.com (accessed on 23 November 2015). Cadence Design Systems, Inc. PSpice Designer. Available online: http://www.orcad.com/products/ orcad-pspice-designer/overview (accessed on 23 November 2015). Texas Instrument Inc. LMP7701/LMP7702/LMP7704 Precision, CMOS Input, RRIO, Wide Supply Range Amplifiers. Available online: www.ti.com (accessed on 24 November 2015). Analog Device Inc. CMOS, 2.5 Ohm Low Voltage, Triple/Quad SPDT Switches, ADG733/ADG734. Available online: www.analog.com (accessed on 23 November 2015). Texas Instrument Inc. CD405xB CMOS Single 8-Channel Analog Multiplexer/Demultiplexer with Logic-Level Conversion. Available online: www.ti.com (accessed on 23 November 2015). © 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons by Attribution (CC-BY) license (http://creativecommons.org/licenses/by/4.0/).