Mutual Capacitive Sensing Touch Screen Controller for Ultrathin Display with Extended Signal Passband Using Negative Capacitance

sensors

Article Mutual Capacitive Sensing Touch Screen Controller for Ultrathin Display with Extended Signal Passband Using Negative Capacitance

Chang-Ju Lee 1,2, Jong Kang Park 1 , Canxing Piao 1, Han-Eol Seo 2, Jaehyuk Choi 1,* and Jung-Hoon Chun 1,*

1 College of Information and Communication Engineering, Sungkyunkwan University, Suwon 16419, Korea; [email protected] (C.-J.L.); [email protected] (J.K.P.); [email protected] (C.P.) 2 Samsung Electronics, Hwaseong 18448, Korea; [email protected] * Correspondence: [email protected] (J.C.); [email protected] (J.-H.C.); Tel.: +82-31-299-4596 (J.-H.C.)

Received: 4 September 2018; Accepted: 23 October 2018; Published: 26 October 2018

Abstract: Flexible and thin displays for smart devices have a large coupling capacitance between the sensor electrode of the touch screen panel (TSP) and the display electrode. This increased coupling capacitance limits the signal passband to less than 100 kHz, resulting in a signiﬁcant reduction in the received signal, with a driving frequency of several hundred kilohertz used for noise avoidance. To overcome this problem, we reduced the effective capacitance at the analog front-end by connecting a circuit with a negative capacitance in parallel with the coupling capacitance of the TSP. In addition, the in-phase and quadrature demodulation scheme was used to address the phase ﬂuctuation between the signal and the clock during demodulation. We fabricated a test chip using the 0.35 µm CMOS process and obtained a signal-to-noise ratio of 43.2 dB for a 6 mm diameter metal pillar touch input.

Keywords: touch screen panel; ﬂexible display; touch screen controller; negative capacitance; IQ demodulation

1. Introduction In recent years, ultra-thin and flexible displays have attracted a great deal of attention. The touch screen panel (TSP) on such a thin display must also be very thin, which makes the distance between the touch sensor electrode and the display electrode extremely short, resulting in a large capacitance between the two electrodes. This increased capacitance means the TSP’s signal transfer characteristics can show those of a low pass filter (LPF) with a very low cut-off frequency. Thus, a touch signal with high frequency can be significantly attenuated while passing through the TSP. We will consider four techniques proposed in previous studies to improve the deterioration of the touch performance by this signal attenuation. Although some of these techniques have not been proposed specifically to address the issue of bandwidth reduction resulting from the increased display coupling capacitance, these potential techniques are considered together as a solution to this problem. The first technique involves sensing the touching of objects using low frequencies where the signal is not attenuated. However, this technique increases the interference with the touch signal from the strong low frequency noises generated by lamps, battery chargers, and home appliances. Of course, using a higher order filter can somewhat compensate for the weakness of the low-frequency driving technique, but if the touch sensing frequency is very close to the noise frequencies, the physical size of the filter would need to be very large. The second method is driving the transmitter (Tx) outputs in touch screen controllers (TSCs) above 10 V to compensate for the signal attenuation [1]. Although this technique is a simple and powerful way to improve the degraded touch performance, it not only

Sensors 2018, 18, 3637; doi:10.3390/s18113637 www.mdpi.com/journal/sensors Sensors 2018, 18, 3637 2 of 13 requires an additional process for fabricating a high-voltage transistor to drive a high voltage, but also increases the power consumed in the TSP. Furthermore, the high-voltage transistors can considerably increase the size of the Tx block itself, depending on the implementation scheme. The third technique was proposed to speed up the circuit operation by charging the display coupling capacitance using a separate circuit with a large charge-supplying capability rather than the main signal processing path for touch sensing [2]. Because the sensing circuit in the main path only reacts to the amount of signal change by touch, it can be very fast. However, in mutual capacitive sensors, the sensor electrodes for the receiver (Rx) are virtually tied to a fixed voltage. Thus, it is difficult to directly apply this technique to supply a non-zero predetermined charge. The last technique involves improving the sensor electrode of the TSP itself using mesh-shaped metal electrodes instead of conventional face-shaped indium-tin-oxide (ITO) transparent electrodes [3]. The mesh-shape of the sensor electrode can reduce the coupling capacitance between the sensor and display electrodes, and its low resistance can compensate for the effect of the increased capacitance in terms of the bandwidth of the LPF. Nevertheless, in TSCs that use a charge amplifier (CA) scheme because of its excellent immunity to low-frequency noises, the frequency characteristics of the amplifier in the CA stage can limit the bandwidth of the touch sensor [2,4–7]. This will be discussed in more detail in Section2. This paper focuses on how to improve the overall bandwidth of a touch sensor that has been decreased by the increase in the display coupling capacitance of mutual capacitive touch sensors with CA circuits, even though the TSP has a sufficiently large passband using metal-mesh electrodes. As a solution, we propose a scheme to dynamically extend the bandwidth of the signal passband toward higher frequencies by connecting circuits with negative capacitance (NC) to the Rx sensor electrodes to offset the display coupling capacitance. In addition, we applied an in-phase and quadrature (IQ) demodulation scheme to our design to desensitize the touch sensor to external temperature changes, because numerous metals used as TSP electrodes in ultrathin displays have large temperature coefficient of resistivity (TCR) values [8,9]. The remainder of this paper is organized as follows. In Section2, we examine the bandwidth reduction phenomenon of the signal due to the increase in the display coupling capacitance of an ultra-thin display, and propose a circuit with NC to overcome this problem. Section3 describes the implementation of a TSC with an NC circuit in detail. The final section summarizes and discusses the measurement results for a test chip.

2. Circuit Analysis of TSP Sensor and Proposed NC Circuit

2.1. Passband Lowering Phenomenon by Large Display Coupling Capacitance Capacitive TSPs have various shapes and structures according to the method used to form the sensor electrode [10]. Among these, on-cell TSPs, where sensor electrodes are formed on the display’s top substrate, and in-cell TSPs, which reuse the display electrodes, are extensively used in ultra-thin displays because they do not substantially increase the thickness of the display. In particular, because on-cell TSPs are driven independently of the display operation, the touch performance cannot be degraded even in a display requiring a high resolution and high frame rate. Therefore, it is very important for the on-cell TSP in an ultra-thin display to maintain the touch performance obtained in thick TSPs. Figure1 shows an on-cell TSP with X-axis and Y-axis sensor electrodes on the same plane, along with the vertical structure of a display module. In the vertical structure, the touch sensor electrodes are usually formed on the opposite side of the display’s common electrode in the top substrate of the display, which corresponds to the encapsulation glass in an OLED display and a color ﬁlter substrate in an LCD. At the top of the display, a cover window made of reinforced glass is used to protect the display from scratches or external impact. In conventional thick TSPs, the display’s top substrate and cover glass generally have thicknesses of a few hundred micrometers. Sensors 2018, 18, 3637 3 of 13 Sensors 2018, 18, x FOR PEER REVIEW 3 of 14

FigureFigure 1. 1.( a()a) Touch Touch screen screen panelpanel (TSP)(TSP) withwith co coplanarplanar sensor sensor electrodes electrodes and and (b ()b vertical) vertical structure structure of of displaydisplay module module embedded embedded withwith TSP.TSP.

ForFor ultra-thin ultra-thin displays, displays, the the thickness thickness of of the the display’s display’s top top substrate, substrate, whichwhich isis denoteddenoted byby dd,, isis the firstthe to first be deceasedto be deceased among among the stacked the stacked layers inlayers Figure in1 b.Figure This is1b. because This is it because does not itsignificantly does not affectsignificantly the durability affect the of the durability display of module. the display It has module. been reported It has been that reported the thickness that the has thickness been reduced has frombeen tens reduced of micrometers from tens toof a micrometers few micrometers to a few by replacing micrometers the thickby replacing glass with the a thick thin plasticglass with film a [11 ]. Thisthin decrease plastic film in thickness [11]. This greatly decrease increases in thicknes thes coupling greatly increases capacitance the coupling between capacitance the TSP electrodes between and thethe display’s TSP electrodes common and electrodes, the display’s and common Table1 lists electrod the calculatedes, and Table values 1 lists of the the calculated parallel capacitancevalues of betweenthe parallel two capacitance electrodes between at a 4 mm two× electrodes4 mm TSP at a sensor 4 mm × node. 4 mm Here, TSP sensor 4 mm node.× 4 mmHere, is 4 themm size × 4 of themm most is the common size of sensorthe most node common for smartphones sensor node [for12 ],smartphones and the capacitance [12], and valuesthe capacitance are calculated values for relativeare calculated permittivity for relative values permittivity of 4 and 10, consideringvalues of 4 and the 10, possibility considering of using the possibility TSP substrates of using of variousTSP substrates of various materials. From this calculation, we can deduce that a smartphone or tablet materials. From this calculation, we can deduce that a smartphone or tablet with 20–30 channels will with 20–30 channels will have a coupling capacitance ranging from a few hundred picofarads to have a coupling capacitance ranging from a few hundred picofarads to even a few nanofarads. even a few nanofarads.

Table 1. Capacitance between sensor electrode and display’s common electrode at sensor node. Table 1. Capacitance between sensor electrode and display’s common electrode at sensor node.

d εr = 4 εr = 10 d εr = 4 εr = 10 50 [[µm]µm] 5.65 5.65 [pF] 14.2 14.2 [pF] [pF] 10 [µm] 28.3 [pF] 71.0 [pF] 10 [µm] 28.3 [pF] 71.0 [pF]

NumerousNumerous touch touch sensingsensing integratedintegrated circuits (ICs) (ICs) use use a acharge charge amplifier amplifier (CA) (CA) in inthe the first first stage stage ofof the the analog analog front-end front-end (AFE)(AFE) becausebecause of its excell excellentent noise noise immunity immunity [4,5]. [4,5 ].They They are are designed designed to to havehave the the transfer transfer characteristics characteristics ofof aa band-passband-pass filt filterer (BPF) (BPF) with with a amaximum maximum transfer transfer gain gain at several at several hundredhundred kilohertz kilohertz to blockto block low-frequency low-frequency noises noises from from power power supplies, supplies, lamps, lamps, and chargers. and chargers. However, anHowever, increase inan theincrease display’s in the coupling display’s capacitancecoupling capacitance decreases decreases the upper the cut-offupper cut-off frequency frequency of the of BPF, whichthe BPF, moves which the moves passband the passband of the circuit of the to circuit low frequencies. to low frequencies. ThereThere were were various various technical technical challenges challenges to to overcome overcome this this passband passband lowering lowering problem. problem. OneOne ofof the the most notable achievements was replacing the ITO with a low-resistance metal as a sensor most notable achievements was replacing the ITO with a low-resistance metal as a sensor electrode electrode material. Recently, mesh-shaped metal electrodes have become widely used in many material. Recently, mesh-shaped metal electrodes have become widely used in many TSPs. TSPs. Nevertheless, if the display coupling capacitance is overwhelmingly large, the passband may be Nevertheless, if the display coupling capacitance is overwhelmingly large, the passband may decreased because of the CA circuit characteristics of the AFE, even if the bandwidth of the TSP is be decreased because of the CA circuit characteristics of the AFE, even if the bandwidth of the TSP maintainedis maintained at the at level the level that existedthat existed prior toprior using to low-resistanceusing low-resistance electrodes. electrodes. Regardless Regardless of the thicknessof the reductionthickness of reduction the TSP, theof the feedback TSP, the capacitance feedback capaci insidetance the CAinside circuit the CA is not circ changeduit is not muchchanged because much the mutualbecause capacitance the mutual sensed capaci usingtance thesensed feedback using capacitancethe feedback is capacitance not significantly is not affected significantly by the affected thickness ofby the the TSP. thickness Therefore, of the increaseTSP. Therefore, in the aforementioned the increase in couplingthe aforementioned capacitance coupling decreases capacitance the feedback gaindecreases and finally the feedback decreases gain the and bandwidth finally decreases of the closed-loop the bandwidth circuit, of whichthe closed-loop has an amplifier circuit, which with one polehas inan the amplifier CA stage. with Ifone this pole bandwidth in the CA isstage. smaller If this than bandwidth the first is polesmaller of thethan TSP, the first the upperpole of cut-offthe frequencyTSP, the upper of the BPFcut-off of frequency the CA stage of the is pulledBPF of downward.the CA stage Figureis pulled2 shows downward. a typical Figure circuit 2 shows model a of touchtypical sensors circuit which model sense of mutualtouch sensors capacitance which using sense the mutual CA and capacitance the simulation using results the CA of theand transfer the

SensorsSensors 20182018,, 1818,, 3637x FOR PEER REVIEW 4 of 1314

simulation results of the transfer function of a CA stage connected to a TSP for two cases with large Sensors 2018, 18, x FOR PEER REVIEW 4 of 14 functionand small of values a CA stage for the connected coupling to capacitanc a TSP for twoe between cases with the largedisplay and and small TSP values electrodes for the (C couplingS). Here, capacitance between the display and TSP electrodes (C ). Here, C denotes a mutual capacitance CM denotessimulation a mutualresults of capacitance the transfer betweenfunction of a acertain CA stage driving connectedS electrode toM a TSP and for a twosensing cases electrodewith large in a between a certain driving electrode and a sensing electrode in a TSP, and RS denotes a resistance of TSP,and and small RS denotesvalues for a theresistance coupling of capacitanc a sensinge between electrode. the displayAdditionally, and TSP in electrodes these simulations, (CS). Here, we a sensing electrode. Additionally, in these simulations, we assumed that the amplifier used at the CA assumedCM denotes that the a mutual amplifier capacitance used at betweenthe CA astag certaine had driving a unity electrode gain bandwidth and a sensing (UGBW) electrode of 10in MHza stageand TSP,that had andthe a unity feedbackRS denotes gain capacitance bandwidth a resistance (UGBW)of of the a CAsensing of stage 10 electrode. MHz (CFB) and was Additionally, that 2 pF, the reflecting feedback in these the capacitance designsimulations, values of thewe used CA stagein commonassumed (CFB) was thattouch 2 the pF, sensors. amplifier reflecting When used the design atC Sthe is CA valueschanged stag usede hadfrom in a commonunity10 pF gain to touch bandwidth1 nF, sensors. the peak(UGBW) When frequency ofC S10is MHz changed of the frompassbandand 10 that pF is tosignificantlythe 1 feedback nF, the peak capacitance lowered. frequency If of the the ofpassband CA the stage passband is(C loweredFB) was is significantly 2 bypF, thisreflecting large lowered. CtheS, itdesign is Ifnecessary the values passband used to drop is S loweredthe Txin commondriving by this frequency,touch large sensors.CS, itwhich is When necessary is approximatelyC is tochanged drop the fromseveral Tx 10 driving pFhundred to frequency,1 nF, kilohertz the peak which with frequency is a approximatelysmall of C theS, even severaldownpassband to hundred 100 kHz.is significantly kilohertz In this case, withlowered. the a smallinterference If theC passbandS, even of downlow-frequency is lowered to 100 by kHz.this noises large In thissignificantly CS, it case, is necessary the deteriorates interference to drop the of the Tx driving frequency, which is approximately several hundred kilohertz with a small CS, even low-frequencytouch sensing performance. noises significantly deteriorates the touch sensing performance. down to 100 kHz. In this case, the interference of low-frequency noises significantly deteriorates the touch sensing performance.

FigureFigure 2.2. Pass-band lowering phenomenon due to large CCS. Figure 2.Pass-band Pass-band lowering lowering phenomenonphenomenon due due to to large large CS. S. 2.2. Design of NC 2.2. 2.2.Design Design of NC of NC To improve the degraded signal transfer characteristics resulting from the large C , we can simply To Toimprove improve the the degraded degraded signal signal transfertransfer characteristics resulting resulting from from the the large largeS CS, CweS, wecan can increasesimplysimply increase the increase bandwidth the the bandwidth ofbandwidth the amplifier of of the the at amplifieramplifier the CA stage. at thethe However, CACA stage. stage. this However, However, method this causes this method method a serious causes causes increase a a inserious powerserious increase consumption increase in inpower power [13 ].consumption Inconsumption our work, [13]. we[13]. proposed InIn ourr work,work, a technique we we proposed proposed to enlarge a techniquea technique the signal to enlargeto passband enlarge the the by arrangingsignalsignal passband passband an NC by generating byarranging arranging circuit an an NC NC in generating parallelgenerating with circuitCS, as inin describedparallel parallel with with in FigureC SC, asS, asdescribed3 .described The NC in isFigure in generated Figure 3. 3. usingThe TheNC a Miller NCis generatedis capacitorgenerated using connectedusing a aMiller Miller in parallel capacitorcapacitor with connected the amplifier inin parallel parallel with awith positivewith the the amplifier gain, amplifier as depictedwith with a ina positive gain, as depicted in Figure 3b. Here, the NC value is approximately −CNC0 × (R2/R1) where Figurepositive3b. gain, Here, as thedepicted NC value in Figure is approximately 3b. Here, the− NCCNC value0 × (R is2/ approximatelyR1) where CNC 0−CisNC the0 × base(R2/R feedback1) where CNC0 is the base feedback capacitance of the proposed NC circuit, and R1 and R2 correspond to the capacitanceCNC0 is the base of the feedback proposed capacitance NC circuit, of and the R proposed1 and R2 correspond NC circuit, to and the R resistors1 and R2 shown correspond in Figure to 3theb. resistors shown in Figure 3b. As a result, CS is compensated by the inserted NC, and the passband Asresistors a result, shownCS is in compensated Figure 3b. As by thea result, inserted CS is NC, compensated and the passband by the ofinserted the signal NC, transfer and the function passband is of the signal transfer function is effectively extended. Figure 4 shows the simulation results for the effectivelyof the signal extended. transfer Figurefunction4 shows is effectively the simulation extended. results Figure for the 4 shows signal the transfer simulation characteristics results for of thethe signal transfer characteristics of the CA stage with and without NC. It demonstrates that the CA stage with and without NC. It demonstrates that the proposed NC circuit can increase the driving signalproposed transfer NC characteristics circuit can increase of the the CAdriving stage frequency with and of the without touch sensor NC. fromIt demonstrates below 100 kHz that to the frequency of the touch sensor from below 100 kHz to several hundred kilohertz. The increased frequency proposedseveral NC hundred circuit kilohertz. can increase The increased the driving frequency frequency of the of touchthe touch driving sensor signal from is very below effective 100 kHz in to ofseveral thesuppressing touch hundred driving the kilohertz. strong signal exte is The veryrnal increased low-frequency effective frequency in suppressing noises. of the the touch strong driving external signal low-frequency is very effective noises. in suppressing the strong external low-frequency noises.

Figure 3. (a) Charge ampliﬁer with negative capacitance circuit and (b) circuit implementation of negative capacitance (NC).

Sensors 2018, 18, x FOR PEER REVIEW 5 of 14

Sensors 2018Figure, 18, 3. 3637 (a) Charge amplifier with negative capacitance circuit and (b) circuit implementation of 5 of 13 negative capacitance (NC).

FigureFigure 4. 4.Signal Signal transfer transfer functionfunction (STF)(STF) at CA stage stage with with and and wi withoutthout proposed proposed NC NC circuit. circuit.

OnOn the the other other hand, hand, becausebecause thethe proposed NCNC circuitcircuit includes includes an an amplifier, amplifier, the the frequency frequency characteristicscharacteristics of of the the NC NC are affectedare affected by the by frequency the frequency characteristics characteristics of the amplifier.of the amplifier. The impedance The ofimpedance the NC circuit, of the including NC circuit, an including amplifier withan amplifier one pole, with is derivedone pole, by isEquation derived by (1), Equation and the (1), frequency and characteristicsthe frequency are characteristics shown in Figure are shown5. in Figure 5. 1+ As()β ()=⋅1 NC ,amp 2 ! ZsNC 1 1 + ANC,amp(s)β2 (1) ( ) = sC · +−()( ) ZNC s NC0 11AsNC ,amp β2 (1) sCNC0 1 + ANC,amp(s)(β2 − 1) A R where As()==DC 2 ,β 1 , NC ,amp = ADC2 2 =+ R1 where ANC,amp(s) s s , β2 RR12+ , ADC2 and p2 stand for the DC gain and the 3-dB pole 1+1+ p R1 R2 p 2 frequency of the amplifier2 inside the proposed NC circuit, respectively. Equation (1) confirms that in the low-frequencyADC2 and p2 region,stand for the theNC DCcircuit gain has and a negative the 3-dB capacitance pole frequency of −C ofNC the0 × amplifier(R2/R1). Forinside reference, the proposed NC circuit, respectively. Equation (1) confirms that in the low-frequency region, the NC this value is equal to the negative value of the product of the Miller gain (AM = R2/R1) and the base circuit has a negative capacitance of −CNC0 × (R2/R1). For reference, this value is equal to the negative◦ feedback capacitance (CNC0). However, the phase of the NC circuit moves away from 270 above value of the product of the Miller gain (AM = R2/R1) and the base feedback capacitance (CNC0). Sensorsa certain 2018 frequency,, 18, x FOR PEER as shown REVIEW in Figure5. 6 of 14 However, the phase of the NC circuit moves away from 270° above a certain frequency, as shown in Figure 5.

FigureFigure 5. 5. MagnitudeMagnitude and and phase phase characteristics characteristics of of impedance impedance of of proposed NC circuit.

To analyze the design factors affecting the characteristics of the NC circuit in more detail, we conducted a pole-zero plot analysis of the impedance of the circuit, as depicted by Figure 6.

Figure 6. Pole-zero plot of impedance of proposed NC circuit.

In Equation (1), the NC circuit has two poles at the origin and right-half-plane (RHP) and one zero at the left-half-plane (LHP) in the s-domain, and the poles and zero have the values found using Equation (2). In this equation, ωu,NC stands for the product of the DC gain of the amplifier inside the proposed NC circuit (ADC2) and 3 dB bandwidth (p2), which means the unity gain bandwidth of the amplifier: =− − ⋅  ωωβZu,NC12p, 2  =− + ⋅() −  ωωPu,NC12p,1 β 2 (2)  =  ωP2 0

Then, the phase of the impedance from the pole-zero plot can be obtained by Equation (3). ∠=−−=−−°() Zs θθθ123 θθ 1290 (3)

If the amplifier has a sufficiently large DC gain and the negative feedback gain (β2) is not very close to 1 or 0, the RHP pole and LHP zero have values near ωu,NC × (1 − β2) and −ωu,NC × β2, respectively. Meanwhile, θ1 and θ2 are almost 0° and 180° at low frequencies, and as the frequency increases, both converge to 90°. This observation explains why the impedance of the proposed circuit with NC characteristics at low frequencies gradually changes to a positive capacitance as it goes to high frequencies.

Sensors 2018, 18, x FOR PEER REVIEW 6 of 14

Sensors 2018, 18, 3637 6 of 13 Figure 5. Magnitude and phase characteristics of impedance of proposed NC circuit.

To analyze the the design design factors factors affecting affecting the the charac characteristicsteristics of ofthe the NC NC circuit circuit in more in more detail, detail, we weconducted conducted a pole-zero a pole-zero plot plot analysis analysis of the of theimpeda impedancence of the of thecircuit, circuit, as depicted as depicted by Figure by Figure 6. 6.

Figure 6. Pole-zeroPole-zero plot of impedance of proposed NC circuit.

In Equation (1), the NC circuit has two poles at the origin and right-half-planeright-half-plane (RHP) and one zero atat thethe left-half-planeleft-half-plane (LHP) (LHP) in in the the s-domain, s-domain and, and the the poles poles and and zero zero have have the values the values found found using Equation (2). In this equation, ω stands for the product of the DC gain of the ampliﬁer inside using Equation (2). In this equation,u,NC ωu,NC stands for the product of the DC gain of the amplifier the proposed NC circuit (A ) and 3 dB bandwidth (p ), which means the unity gain bandwidth of inside the proposed NC circuitDC2 (ADC2) and 3 dB bandwidth2 (p2), which means the unity gain the ampliﬁer: bandwidth of the amplifier:  ω = −p − ω · β ,  Z1 2 u,NC 2  ωωβ=−p, − ⋅ ωP1 Zu,NC=12−p2 + ωu,NC · ( 21 − β2), (2)   =− + ⋅() −  ωPωω2 Pu,NC=120 p,1 β 2 (2)  = Then, the phase of the impedance ω fromP2 the0 pole-zero plot can be obtained by Equation (3).

Then, the phase of the impedance from the pole-zero plot can be◦ obtained by Equation (3). ∠Z(s) = θ1 − θ2 − θ3 = θ1 − θ2 − 90 (3) ∠=−−=−−°() Zs θθθ123 θθ 1290 (3) If the amplifier has a sufficiently large DC gain and the negative feedback gain (β2) is not very closeIf to the 1 oramplifier 0, the RHP has a pole sufficiently and LHP large zero DC have gain values and the near negativeωu,NC ×feedback(1 − β2 gain) and (β−2)ω isu,NC not ×veryβ2, ◦ ◦ closerespectively. to 1 or Meanwhile,0, the RHP θpole1 and andθ2 are LHP almost zero 0 haveand values 180 at near low frequencies,ωu,NC × (1 − andβ2) asand the −ω frequencyu,NC × β2, ◦ respectively.increases, both Meanwhile, converge to θ1 90 and. This θ2 are observation almost 0° explains and 180° why at low the frequencies, impedance of and the as proposed the frequency circuit withincreases,NC characteristics both converge at to low 90°. frequencies This observat graduallyion explains changes why to a positivethe impedance capacitance of the as proposed it goes to circuithigh frequencies. with NC characteristics at low frequencies gradually changes to a positive capacitance as it goes Into orderhigh frequencies. to use the high-frequency touch-driving signal in the TSP, the valid region of NC should be sufficiently wide. We can consider two methods to attain the necessary width: using an amplifier with a large UGBW and optimizing the feedback gain setting. When the UGBW of the amplifier becomes larger, the impedance of the proposed circuit can maintain a negative capacitance characteristic up to a higher frequency region because the RHP pole and LHP zero move away from each other. However, the increase in the UGBW necessarily accompanies an increase in the power consumption of the amplifier. Therefore, it is necessary to design the UGBW of the amplifier to ensure that the NC characteristic is maintained just until the driving frequency of the touch sensor to attain a low-power design. As for the selection of the feedback gain, because the RHP pole and LHP zero move in the same direction based on the change in the value of β2, we simulated a change in the valid region of the NC. When setting the NC value, because the Miller gain rather than the feedback gain provides direct intuition in the circuit analysis and β2 is equal to 1/(1 − AM), the Miller gain is used as an independent variable in the result plot. Figure7 plots the relationship between the frequency with 15 ◦, 30◦, and 45◦ phase shifts from the ideal negative capacitance and the Miller gain of the NC circuit. In this figure, we can find that the frequency at which the phase shift occurs decreases as the Miller gain increases or the β2 value of the NC circuit decreases. This is because the closed loop bandwidth of the NC circuit decreases as the Miller gain increases. Therefore, using a small CNC0 and an excessively Sensors 2018, 18, x FOR PEER REVIEW 7 of 14

In order to use the high-frequency touch-driving signal in the TSP, the valid region of NC should be sufficiently wide. We can consider two methods to attain the necessary width: using an amplifier with a large UGBW and optimizing the feedback gain setting. When the UGBW of the amplifier becomes larger, the impedance of the proposed circuit can maintain a negative capacitance characteristic up to a higher frequency region because the RHP pole and LHP zero move away from each other. However, the increase in the UGBW necessarily accompanies an increase in the power consumption of the amplifier. Therefore, it is necessary to design the UGBW of the amplifier to ensure that the NC characteristic is maintained just until the driving frequency of the touch sensor to attain a low-power design. As for the selection of the feedback gain, because the RHP pole and LHP zero move in the same direction based on the change in the value of β2, we simulated a change in the valid region of the NC. When setting the NC value, because the Miller gain rather than the feedback gain provides direct intuition in the circuit analysis and β2 is equal to 1/(1 − AM), the Miller gain is used as an independent variable in the result plot. Figure 7 plots the relationship between the frequency with 15°, 30°, and 45° phase shifts from the ideal negative capacitance and the Miller gain of the NC circuit. In this figure, we can find that the frequency at which the phase shift occurs decreases as the Miller gain increases or the β2 value of the NC circuit decreases. This is because the closed loop Sensors 2018, 18, 3637 7 of 13 bandwidth of the NC circuit decreases as the Miller gain increases. Therefore, using a small CNC0 and an excessively large Miller gain to obtain a certain NC value is undesirable in terms of the largedecrease Miller in gainthe tovalid obtain region a certain of the NC NC. value In iscontrast, undesirable using in termsa large of theCNC0 decrease and small in the Miller valid gain region is ofadvantageous the NC. In contrast,for a low-power using a large circuitCNC design,0 and smallbut this Miller combination gain is advantageous can increase for the a low-powerchip size. circuitTherefore, design, it is butimportant this combination to select the can approp increaseriate the Miller chip gain size. considering Therefore, it the is importantproduct cost. to select the appropriateConsequently, Miller gain when considering the proposed the productcircuit is cost. used for an NC, the UGBW of the amplifier, Miller gain Consequently,of the NC circuit, when and the C proposedNC0 value should circuit is be used optimized for an NC, considering the UGBW the of driving the ampliﬁer, frequency Miller of gain the oftouch the sensor. NC circuit, and CNC0 value should be optimized considering the driving frequency of the touch sensor.

Figure 7. Phase-shifted frequency according to variousvarious Miller gains of proposed NC circuit. 3. Full Chip Implementation 3. Full Chip Implementation 3.1. Block Diagram of Touch Screen Controller 3.1. Block Diagram of Touch Screen Controller Figure8 shows a block diagram of a full-chip TSC used for testing. In order to prevent noises Figure 8 shows a block diagram of a full-chip TSC used for testing. In order to prevent noises generated in other blocks from interfering with the noise-sensitive analog blocks, three separate power generated in other blocks from interfering with the noise-sensitive analog blocks, three separate sources are supplied to the chip: VDDA for the analog block, VDD for the digital block, and VDDT power sources are supplied to the chip: VDDA for the analog block, VDD for the digital block, and for the Tx driver in the figure. Furthermore, the voltage regulated by a low drop-out (LDO) circuit is VDDT for the Tx driver in the figure. Furthermore, the voltage regulated by a low drop-out (LDO) used in the analog circuits located in the touch signal processing path for superior noise performance. circuit is used in the analog circuits located in the touch signal processing path for superior noise The chip communicates with a field programmable gate array (FPGA) chip through inter-integrated performance. The chip communicates with a field programmable gate array (FPGA) chip through circuit (I2C) and general purpose input/output (GPIO) interfaces for the digital processing of touch inter-integrated circuit (I2C) and general purpose input/output (GPIO) interfaces for the digital data to enhance the touching performance and calculation of the coordinates of the touched object. In the block diagram, the Rx channels of the TSP are connected to the negative inputs of single-ended CAs that include NC circuits. A pair of adjacent CA outputs passing through a 3:2 multiplexer (MUX) is identically input to two differential programmable gain amplifiers (DPGAs) for IQ demodulation. By repeatedly alternating the CA output pairs using the 3:2 MUX, the number of DPGAs can be reduced by nearly half [14]. The differential sensing scheme can not only eliminate the noise common to Rx channels, such as the display and internal bias noise, but also alleviate the newly introduced noise by the NC circuit [6]. The outputs of the upper DPGA in Figure8 are demodulated with an in-phase clock that has the same triggering edge as the Tx driving pulse, and the outputs of the lower DPGA are demodulated with a clock that is shifted 90◦ compared to the in-phase clock. Each demodulated output is converted to a digital code by a ∆-Σ analog-to-digital converter (ADC) and then transmitted to the FPGA chip through an I2C communication interface to calculate the touch coordinates. The digital block is composed of decimation filters for the ∆-Σ ADC, a timing controller, parameter setting registers for the analog IPs, and an orthogonal code table for simultaneous multiple Tx driving operations [15]. The orthogonal code used is a modified Hadamard matrix with the maximum column sum balanced from 4 to 16 Tx lines [16]. The Tx driver consists of a digital-to-analog converter (DAC) and numerous buffers. The DAC generates a sine-wave with almost no harmonics at high frequencies to make the touch data insensitive to environmental changes in the TSP, such as temperature change. Sensors 2018, 18, x FOR PEER REVIEW 8 of 14

processing of touch data to enhance the touching performance and calculation of the coordinates of the touched object. In the block diagram, the Rx channels of the TSP are connected to the negative inputs of single-ended CAs that include NC circuits. A pair of adjacent CA outputs passing through a 3:2 multiplexer (MUX) is identically input to two differential programmable gain amplifiers (DPGAs) for IQ demodulation. By repeatedly alternating the CA output pairs using the 3:2 MUX, the number of DPGAs can be reduced by nearly half [14]. The differential sensing scheme can not only eliminate the noise common to Rx channels, such as the display and internal bias noise, but also alleviate the newly introduced noise by the NC circuit [6]. The outputs of the upper DPGA in Figure 8 are demodulated with an in-phase clock that has the same triggering edge as the Tx driving pulse, and the outputs of the lower DPGA are demodulated with a clock that is shifted 90° compared to the in-phase clock. Each demodulated output is converted to a digital code by a Δ-Σ analog-to-digital converter (ADC) and then transmitted to the FPGA chip through an I2C communication interface to calculate the touch coordinates. The digital block is composed of decimation filters for the Δ-Σ ADC, a timing controller, parameter setting registers for the analog IPs, and an orthogonal code table for simultaneous multiple Tx driving operations [15]. The orthogonal code used is a modified Hadamard matrix with Sensorsthe2018 maximum, 18, 3637 column sum balanced from 4 to 16 Tx lines [16]. The Tx driver consists of 8a of 13 digital-to-analog converter (DAC) and numerous buffers. The DAC generates a sine-wave with almost no harmonics at high frequencies to make the touch data insensitive to environmental Moreover,changes because in the TSP, the DACsuch as is temperature implemented change to simultaneously. Moreover, because generate the DAC two wavesis implemented with opposite to phases,simultaneously multiple driving generate pulses two canwaves be simultaneously with opposite outputphases, accordingmultiple driving to the orthogonal pulses can code be by controllingsimultaneously the output output switches. according to the orthogonal code by controlling the output switches.

FigureFigure 8. Block8. Block diagram diagram of of full fullchip chip includingincluding interfaces interfaces with with a aFPGA FPGA chip. chip.

3.2. IQ3.2. Demodulation IQ Demodulation Multiple AFEs with a lock-in architecture for enhancing the signal-to-noise ratio (SNR) are used to demodulate the outputs of the front-end using a clock with a predetermined phase so that the phase of the clock can be locked to the touch signal received for each touch sensor node [4,5]. However, this scheme has the problem that the strength of the demodulated signal is reduced by a temporal phase mismatch between two signals, which can be caused by changes in the device temperature, changes in the parasitic components of the TSP as a result of touch events, and the demodulation clock’s jitter. In particular, ultra-thin displays such as flexible displays are more sensitive to heat. As previously mentioned, metals with low resistivity are used in ultra-thin displays instead of ITO to ensure a wide bandwidth performance for the TSP. However, most of the metals that can be used in ultra-thin displays have relatively large TCR values of approximately 0.004, which are 20 times greater than that of ITO (0.0002) [8]. This means that thin TSPs using metal electrodes can be more temperature sensitive than previous TSPs. Because the resistance of the sensor electrode determines the cut-off frequency of the passband of the TSP, a large change in the sensor resistance with a change in temperature can have a profound effect on the touch performance. To evaluate the sensitivity of touch signals to temperature changes in TSPs using metal electrodes, we consider the following example that reflects the characteristics of a typical touch sensor. The TSP forms a second-order LPF with two poles at a cut-off frequency, and the feedback impedance in the CA stage has a pole at the same frequency for low-frequency noise avoidance. If the Tx signal is driven with the cut-off frequency and the temperature of the TSP changes by 25 ◦C, the phase of the received signal is shifted by 9.04◦ as a result of the variation in the bandwidth of the passband of the TSP. Then, Sensors 2018, 18, 3637 9 of 13 this phase difference of 9.04◦ compared to the demodulation clock causes a signal loss of 1.24% in the demodulation process using a square wave. In conventional mutual-capacitive TSPs, because the maximum variation of the mutual capacitance by touch is approximately 10% of the capacitance sensed at no touch, this 1.24% signal reduction after demodulation actually implies a reduction of 12.4% of the meaningful signal for touch sensing, which is a value that can seriously deteriorate the touch performance. To address the dynamic phase mismatch caused by environmental variations like external temperature changes, conventional demodulation circuits periodically update the phase of the demodulation clock stored in the memory or use a temperature sensor. However, the periodic updating method for the clock phase has an interval in which the touch sensor cannot cope with a sudden temperature change before the next update time. During this interval, the response to a touch may slow down or malfunction. Although the temperature sensor built into the TSC can compensate for the above interval problem, because each sensor node has a different phase delay for the signal based on the temperature change, this method is also limited to precisely compensating for the phase deviation of all the sensor nodes. Furthermore, in the case of displays with self-luminous elements such as OLED displays, a spatial temperature distribution may occur on the TSP depending on the display image, resulting in the failure of the temperature compensation. In this work, we used an IQ demodulation technique to make the touch sensor insensitive to a sudden temperature change. In the IQ demodulation scheme, even if a phase mismatch occurs between two signals, the final value of the touch signal is determined regardless of the phase mismatch using the trigonometric formula of Equation (4).

π 2 (Asinθ)2 + Asin + θ = (Asinθ)2 + (Acosθ)2 = A2 (4) 2 where θ is the amount of phase mismatch between the demodulation clock and the signal. In addition, because this technique does not require special chip operation scenarios to lock the phase of the demodulation clock, the demodulation process and circuit implementation can be very simple. Figure9 illustrates a pair of AFE circuits designed for IQ demodulation, where the squaring and summing operationsSensors 2018 are, 18 performed, x FOR PEER REVIEW in the FPGA. 10 of 14 On the other hand, because the IQ demodulation technique requires two signal paths, which are an I-phaseOn signalthe other path hand, and because a Q-phase the IQ signal demodulation path, the technique demodulation requires circuits two signal are paths, required which twice. are an I-phase signal path and a Q-phase signal path, the demodulation circuits are required twice. However, considering the overhead of the conventional demodulation scheme, which includes the However, considering the overhead of the conventional demodulation scheme, which includes the memories of the sensor nodes for storing phase information, multiple clock lines, temperature sensing memories of the sensor nodes for storing phase information, multiple clock lines, temperature circuits,sensing complex circuits, control complex logic control circuits, logic andcircuits, complex and complex chip driving chip driving scenarios, scenarios, the sizethe size penalty penalty of the additionalof the additional blocks for blocks IQ demodulation for IQ demodulation may not may be not large. be large.

FigureFigure 9. 9.Pair Pair of of analog analog front-endfront-end circuits for for IQ IQ demodulation. demodulation.

3.3. Remaining Blocks for Analog Signal Processing The demodulated signal is passed through a 2nd-order LPF to remove the high-frequency components of the demodulated signal and other high-frequency noise components. It is then amplified by a variable gain amplifier (VGA) to minimize the SNR reduction due to the quantization noise in the ADC. Finally, the analog signal is converted to digital code by a 2nd-order incremental Δ-Σ ADC [5,17].

4. Measurement Results and Discussion The test chip shown in Figure 10 was implemented using a 0.35 µm 1P4M-CMOS process. It consisted of analog signal processing circuits and Δ-Σ ADCs that could support up to 10 Rx channels, along with various digital circuits, such as those for the data memory, timing controller, control registers, chip-to-chip communication interface circuits, and 15 Tx drivers. The entire Tx and Rx occupy 1.12 mm2 and 3.78 mm2 respectively. The area of the ADC block is 0.9 mm2, and the digital blocks with data memory, timing controller, and registers occupy 5.62 mm2.

Figure 10. Microphotograph of test chip.

Sensors 2018, 18, x FOR PEER REVIEW 10 of 14

On the other hand, because the IQ demodulation technique requires two signal paths, which are an I-phase signal path and a Q-phase signal path, the demodulation circuits are required twice. However, considering the overhead of the conventional demodulation scheme, which includes the memories of the sensor nodes for storing phase information, multiple clock lines, temperature sensing circuits, complex control logic circuits, and complex chip driving scenarios, the size penalty of the additional blocks for IQ demodulation may not be large.

Sensors 2018, 18, 3637 Figure 9. Pair of analog front-end circuits for IQ demodulation. 10 of 13

3.3. Remaining Blocks for Analog Signal Processing 3.3. Remaining Blocks for Analog Signal Processing The demodulated signal is passed through a 2nd-order LPF to remove the high-frequency componentsThe demodulated of the demodulated signal is passed signal throughand other a 2nd-orderhigh-frequency LPF tonoise remove components. the high-frequency It is then componentsamplified by of a thevariable demodulated gain amplifier signal (VGA) and other to minimize high-frequency the SNR noise reduction components. due to It the is then quantization ampliﬁed bynoise a variable in the ADC. gain ampliﬁerFinally, the (VGA) analog to minimize signal is theconverted SNR reduction to digital due code to theby a quantization 2nd-order incremental noise in the ADC.Δ-Σ ADC Finally, [5,17]. the analog signal is converted to digital code by a 2nd-order incremental ∆-Σ ADC [5,17].

4. Measurement Results and Discussion µ The test chip chip shown shown in in Figure Figure 10 10 was was implem implementedented using using a a0.35 0.35 µm m1P4M-CMOS 1P4M-CMOS process. process. It Itconsisted consisted of ofanalog analog signal signal processing processing circuits circuits and and Δ∆-Σ-Σ ADCsADCs that that could could support support up up to to 10 Rx channels, along with various digital circuits, circuits, such such as as those those for for the the data data memory, memory, timing controller, controller, control registers,registers, chip-to-chipchip-to-chip communication communication interface interface circuits, circuits, and and 15 15 Tx Tx drivers. drivers. The The entire entire Tx and Tx 2 2 2 Rxand occupy Rx occupy 1.12 mm1.12 mmand2 3.78and mm3.78 mmrespectively.2 respectively. The area The of area the of ADC the blockADC isblock 0.9 mm is 0.9, and mm the2, and digital the 2 blocksdigital withblocks data with memory, data memory, timing controller,timing controller, and registers and registers occupy occupy 5.62 mm 5.62. mm2.

Figure 10. Microphotograph of test chip.

We examined the effect of the proposed NC compensation by comparing the outputs with and without the compensation at the CA stage. For comparison, we constructed a test circuit with the same conﬁguration as the equivalent circuit model of the TSP shown in Figure2. The Tx blocks drove one node of the mutual capacitance with a sine wave that had a peak-to-peak voltage of 3 V and frequency of 200 kHz, and the CA output signals were measured while changing the capacitance corresponding to the display coupling capacitance to various values. As shown in the two measurement results on the left side of Figure 11, when the NC compensation was absent, the CA output had a peak-to-peak voltage of 220 mV, whereas when it was compensated using NC, the output increased approximately three-fold to 642 mV. In addition, Figure 11c shows that the gain of the CA is not substantially reduced even though Cs increases. This result conﬁrmed that the touch performance could be improved using a high-frequency touch signal even in a TSP with a large display coupling capacitance if an NC corresponding to the display coupling capacitance was generated using the proposed NC circuit. Sensors 2018, 18, x FOR PEER REVIEW 11 of 14

We examined the effect of the proposed NC compensation by comparing the outputs with and without the compensation at the CA stage. For comparison, we constructed a test circuit with the same configuration as the equivalent circuit model of the TSP shown in Figure 2. The Tx blocks drove one node of the mutual capacitance with a sine wave that had a peak-to-peak voltage of 3 V and frequency of 200 kHz, and the CA output signals were measured while changing the capacitance corresponding to the display coupling capacitance to various values. As shown in the two measurement results on the left side of Figure 11, when the NC compensation was absent, the CA output had a peak-to-peak voltage of 220 mV, whereas when it was compensated using NC, the output increased approximately three-fold to 642 mV. In addition, Figure 11c shows that the gain of the CA is not substantially reduced even though Cs increases. This result confirmed that the touch performance could be improved using a high-frequency touch signal even in a TSP with a large

displaySensors coupling2018, 18, 3637 capacitance if an NC corresponding to the display coupling capacitance11 of was 13 generated using the proposed NC circuit.

FigureFigure 11. 11. MeasuredMeasured outputs outputs of of charge charge amplifier: ampliﬁer: ((aa)) withoutwithout NC NC circuit; circuit; (b )(b with) with NC; NC; and and (c) gain (c) gain of

of CACA stage for variousvariousC CSSvalues. values.

Next,Next, we we compared compared the the ADC ADC outputs outputs to to evaluate evaluate the the SNR SNR of the touchtouch signalsignalon on the the test test chip. chip. In this test, the test chip was connected to a TSP that had 15 Tx and 10 Rx sensor electrodes, and its In this test, the test chip was connected to a TSP that had 15 Tx and 10 Rx sensor electrodes, and its display coupling capacitance per Rx channel was approximately 660 pF. Figure 12 shows the 2D touch display coupling capacitance per Rx channel was approximately 660 pF. Figure 12 shows the 2D data measured with and without the NC compensation. The VGA outputs were adjusted to be the touch data measured with and without the NC compensation. The VGA outputs were adjusted to same in both cases to ensure the same quantization noise for the ADC. We calculated the signal-to-noise be the same in both cases to ensure the same quantization noise for the ADC. We calculated the ratio (SNR) using Equation (5) [18]: signal-to-noise ratio (SNR) using Equation (5) [18]: = − STouchSTouchSignal=− SignalTouch,AVG100 SignalSignalUntouch,AVG ,100, Touch,AVG100 Untouch,AVG 100 q n=99 2 ∑n=0 (Signal[n]−SignalTouch) NTouch = = , (5) RMS100 n 99 100− 2 ()Signal[] n SignalTouch NTouchSNR= [dB] =n=020log STouch , RMS100 NTouchRMS100 (5) 100 where AVG100 and RMS100 denote the average and theSTouch root-mean-square (RMS) of 100 data obtained sequentially, respectively, and SignalSNR[] dBis a= reference20 log value for measuring the noise RMS value as touch NTouch a longer-term average. In Table2, the SNR obtained from thisRMS study100 is the average of the SNR values wheremeasured AVG100 at 20 and different RMS100 locations, denote which the areaverage distributed and the evenly root-mean-square over the TSP. When (RMS) the of NC 100 circuit data was used, the SNR was improved by 19.1 dB from 24.1 dB to 43.2 dB. The improvement in the signal obtained sequentially, respectively, and Signaltouch is a reference value for measuring the noise RMS Sensorsvalue 2018transfer, 18as, xa FOR characteristicslonger-term PEER REVIEW average. from the In C STablecompensation 2, the SN deﬁnitelyR obtained contributed from this to study this SNR is the enhancement. average of the 12 of 14 SNR values measured at 20 different locations, which are distributed evenly over the TSP. When the NC circuit was used, the SNR was improved by 19.1 dB from 24.1 dB to 43.2 dB. The improvement in the signal transfer characteristics from the CS compensation definitely contributed to this SNR enhancement.

Figure 12.Figure Comparison 12. Comparison of oftouch touch sensingsensing 2D data2D withoutdata without negative capacitancenegative (capacitanceleft) and with proposed(left) and with circuit (right). proposed circuit (right).

Table 2. Comparison of proposed touch sensor with state-of-the-art devices.

ISSCC 10’ [2] ISSCC 13’ [4] JSSC 14’ [5] TCAS-I 16’ [19] This Work Process 90 nm 0.35 µm 0.18 µm 0.18 µm 0.35 µm TSP Type Self Mutual Mutual Mutual Mutual Tx: 27 Tx: 12 Tx: 32 Tx: 15 Channel 1 TRX: 24 Rx: 43 Rx: 8 Rx: 10 Rx: 10 SNR 36 dB 39 dB 60 dB 72 dB 43 dB Frame Rate 120 Hz 120 Hz 200 Hz 240 Hz 250 Hz Supply 3 V 3.3 V 2.1–3.3 V 3.3 V 3.3 V Power 12 mW 18.7 mW 6.26 mW 42.6 mW 12.8 mW Area 3.65 mm2 10.4 mm2 2.2 mm2 1.25 mm2 2 4.89 mm2 1 TRX = Tx + Rx (Self-capacitive TSP); 2 Area of Tx and Rx without the digital block for comparison with other touch sensing ICs.

On the other hand, when the NC compensation was applied, the SNR of the 2D touch data was improved by 19 dB, while the touch signal that passed through the CA stage in Figure 11 was increased by 10 dB. This can be explained by the fact that the serious amplification of the device noise and reference noise of the CA-stage amplifier with a large CS was mitigated. Namely, because the equivalent CS value was decreased by the NC circuit, the amplification gain of the noise at the CA stage, including the TSP, was reduced, which enhanced the overall SNR of the sensor [5,19]. In addition, because the AFE that used the NC circuit could acquire a sufficiently large signal even with a small signal gain for the CA stage, the gains of the signal amplification stages after the CA stage could be set to small values, and the noise figure of the total chains of the amplifying stages could be improved. As a result, the reduced amplification of the noise in the AFE block also contributed to the enhancement of the overall SNR of the touch sensor. The main performance of the test chip is summarized in Table 2. It consumed 12.8 mW at a scan rate of 250 Hz and achieved a 43.2 dB SNR on a 6 mm-diameter metal pillar touch. However, there was no information on the display coupling capacitance or thickness of the display top substrate in the state-of-the-art papers, which made it difficult to make a direct comparison with this work. Nevertheless, since the previous experiments were performed on add-on type or thick on-cell type TSPs, the display coupling capacitance is expected to be less than a few tens of picofarads, which may result in better SNR measurement results. Taking this into account, although the device reported here had a slightly lower SNR than those in the two latest studies, the table shows that it had a competitive touch-sensing performance even in a TSP with a large display coupling capacitance of several hundred picofarads.

Author Contributions: C.L., J.P., C.P., H.S., and J.-H.C. conceived and designed the experiments; C.L., J.P. and C.P. performed the experiments; C.L., J.C. and J.-H.C. analyzed the data; C.L. and J.-H.C. wrote the paper.

Funding: This work was supported by the Ministry of Trade, Industry & Energy (project number: 10008040) and Korea Semiconductor Research Consortium support program for the development of the future semiconductor device.

Sensors 2018, 18, 3637 12 of 13

Table 2. Comparison of proposed touch sensor with state-of-the-art devices.

On the other hand, when the NC compensation was applied, the SNR of the 2D touch data was improved by 19 dB, while the touch signal that passed through the CA stage in Figure 11 was increased by 10 dB. This can be explained by the fact that the serious amplification of the device noise and reference noise of the CA-stage amplifier with a large CS was mitigated. Namely, because the equivalent CS value was decreased by the NC circuit, the amplification gain of the noise at the CA stage, including the TSP, was reduced, which enhanced the overall SNR of the sensor [5,19]. In addition, because the AFE that used the NC circuit could acquire a sufficiently large signal even with a small signal gain for the CA stage, the gains of the signal amplification stages after the CA stage could be set to small values, and the noise figure of the total chains of the amplifying stages could be improved. As a result, the reduced amplification of the noise in the AFE block also contributed to the enhancement of the overall SNR of the touch sensor. The main performance of the test chip is summarized in Table2. It consumed 12.8 mW at a scan rate of 250 Hz and achieved a 43.2 dB SNR on a 6 mm-diameter metal pillar touch. However, there was no information on the display coupling capacitance or thickness of the display top substrate in the state-of-the-art papers, which made it difficult to make a direct comparison with this work. Nevertheless, since the previous experiments were performed on add-on type or thick on-cell type TSPs, the display coupling capacitance is expected to be less than a few tens of picofarads, which may result in better SNR measurement results. Taking this into account, although the device reported here had a slightly lower SNR than those in the two latest studies, the table shows that it had a competitive touch-sensing performance even in a TSP with a large display coupling capacitance of several hundred picofarads.

Author Contributions: C.-J.L., J.K.P., C.P., H.-E.S., and J.-H.C. conceived and designed the experiments; C.-J.L., J.K.P. and C.P. performed the experiments; C.-J.L., J.C. and J.-H.C. analyzed the data; C.-J.L. and J.-H.C. wrote the paper. Funding: This work was supported by the Ministry of Trade, Industry & Energy (project number: 10008040) and Korea Semiconductor Research Consortium support program for the development of the future semiconductor device. Acknowledgments: The CAD tools’ licenses were supported by the IC Design Education Center (IDEC) of Korea. Conﬂicts of Interest: The authors declare no conﬂict of interest.

References

1. Carey, J. Noise Wars: Projected Capacitance Strikes Back; Cypress Semiconductor Corporation: San Jose, CA, USA, 2012. 2. Kim, H.-R.; Choi, Y.-K.; Byun, S.-H.; Kim, S.-W.; Choi, K.-H.; Ahn, H.-Y.; Park, J.; Lee, D.-Y.; Wu, Z.-Y.; Kwon, H.-D.; et al. A mobile-display-driver IC embedding a capacitive-touch-screen controller system. In Proceedings of the 2010 IEEE International Solid-State Circuits Conference, San Francisco, CA, USA, 7–11 February 2010; pp. 114–115. Sensors 2018, 18, 3637 13 of 13

3. Xu, J.; Zhang, Q.; Li, H.; Liu, T.; Chen, S.; Shang, F. Simulation of color moire pattern in LCD-based metal mesh touch screen. SID 2017, 48, 2087–2090. [CrossRef] 4. Yang, J.-H.; Park, S.-H.; Choi, J.-M.; Kim, H.-S.; Park, C.-B.; Ryu, S.-T.; Cho, G.-H. A highly noise-immune touch controller using filtered-delta-integration and a charge-interpolation technique for 10.1-inch capacitive touch-screen panels. In Proceedings of the 2013 IEEE International Solid-State Circuits Conference, San Francisco, CA, USA, 17–21 February 2013; pp. 390–391. 5. Park, J.-E.; Lim, D.-H.; Jeong, D.-K. A reconfigurable 40-to-67 dB SNR, 50-to-6400 Hz frame-rate, column-parallel readout IC for capacitive touch-screen panels. IEEE J. Solid State Circuits 2014, 49, 2305–2318. [CrossRef] 6. Park, J.-E.; Park, J.; Hwang, Y.-H.; Oh, J.; Jeong, D.-K. A 100-TRX-channel configurable 85-to-385Hz-frame-rate

analog front-end for touch controller with highly enhanced noise immunity of 20VPP. In Proceedings of the 2016 IEEE International Solid-State Circuits Conference, San Francisco, CA, USA, 31 January–4 February 2016; pp. 210–211. 7. Kim, K.-D.; Byun, S.-H.; Choi, Y.-K.; Baek, J.-H.; Cho, H.-H.; Ahn, H.-Y.; Lee, C.-J.; Cho, M.-S.; Lee, J.-H.; Kim, S.-W.; et al. A capacitive touch controller robust to display noise for ultrathin touch screen displays. In Proceedings of the 2012 IEEE International Solid-State Circuits Conference, San Francisco, CA, USA, 19–23 February 2012; pp. 116–117. 8. Wikipedia. Available online: https://en.wikipedia.org/wiki/Electrical_resistivity_and_conductivity (accessed on 27 July 2018). 9. Sklar, B. Digital Communication, 2nd ed.; Prentice-Hall: Upper Saddle River, NJ, USA, 2001. 10. Walker, G.; Fihn, M. LCD In-cell touch. Inf. Disp. 2010, 26, 8–14. 11. Chen, H.-Y.; Lu, Y.-C.; Chang, K.-M.; Yeh, Y.-H.; Chen, K.-J. Highly sensitive touch panel technology for foldable AMOLED. SID 2017, 48, 2052–2055. [CrossRef] 12. GSMARENA. Available online: https://www.gsmarena.com (accessed on 27 July 2018). 13. Allen, P.E.; Holberg, D.R. CMOS Analog Circuit Design, 2nd ed.; Oxford University Press: New York, NY, USA, 2002. 14. Kim, K.-D.; Kang, S.; Choi, Y.-K.; Lee, K.-H.; Lee, C.-H.; Lee, J.-C.; Choi, M.; Ko, K.; Jung, J.; Park, N.; et al. A fully-differential capacitive touch controller with input common-mode feedback for symmetric display noise cancellation. In Proceedings of the 2014 Symposium on VLSI Circuits Digest of Technical Papers, Honolulu, HI, USA, 10–13 June 2014; pp. 1–2. 15. Shin, H.; Ko, S.; Jang, H.; Yun, I.; Lee, K. A 55 dB SNR with 240 Hz frame rate mutual capacitor 30 × 24 touch-screen panel read-out IC using code-division multiple sensing technique. In Proceedings of the 2013 IEEE International Solid-State Circuits Conference, San Francisco, CA, USA, 17–21 February 2013; pp. 388–389. 16. Park, J.; Lee, C.-J.; Kim, D.-Y.; Chun, J.-H.; Kim, J. Application of weighing matrices to simultaneous driving technique for capacitive touch sensors. IEEE Trans. Consum. Electron. 2015, 61, 261–269. [CrossRef] 17. Markus, J.; Silva, J.; Temes, G. Theory and applications of incremental ∆∑ converters. IEEE Trans. Circuits Syst. I 2004, 51, 678–690. [CrossRef] 18. Ko, S.; Shin, H.; Lee, J.; Jang, H.; So, B.-C.; Yun, I.; Lee, K. Low noise capacitive sensor for multi-touch mobile handset’s applications. In Proceedings of the 2010 IEEE Asian Solid-State Circuits Conference, Beijing, China, 8–10 November 2010; pp. 1–4. 19. Heo, S.; Ma, H.; Song, J.; Park, K.; Choi, E.-H.; Kim, J.; Bien, F. 72 dB SNR, 240 Hz frame rate readout IC with differential continuous-mode parallel architecture for larger touch-screen panel applications. IEEE Trans. Circuits Syst. I Regul. Pap. 2016, 63, 960–971. [CrossRef]

© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).