I. Introduction

Capacitive touchscreen technology has been widely adopted in various applications, ranging from mobile devices such as cell phones and tablets to automotive and smart consumer electronics, owing to its intuitive interface. A touch system typically consists of a touchscreen panel (TSP), a touchscreen controller (TSC), and a host processor, as illustrated in Fig. 1. When a touch object, such as a finger or stylus pen, makes contact with the TSP, the touch signal is transferred to an analog front-end (AFE) in the TSC through a flexible PCB (FPCB). Subsequently, the analog touch signal is converted into digital raw data in the AFE. A digital back-end (DBE) processes the raw touch data to determine the touch coordinates. The TSC then sends the touch coordinates to the host processor, which utilizes the touch information as user input.

Fig. 1. Capacitive touch system architecture.

For a better user experience, it is imperative that the touch system maintains a high signal-to-noise ratio (SNR) and operates at a high frame rate, as depicted in Fig. 2. Recent trends have seen the integration of the TSP within the display panel itself for a thinner form factor, improved visibility, and a simplified manufacturing supply chain. This integration brings the TSP closer to the display panel, resulting in increased interference from display noise during touch sensing operations. This is due to the larger coupling capacitance formed between the TSP and the display panel ^[1-6], as indicated by C$_{\mathrm{c,d}}$ and C$_{\mathrm{c,s}}$ in Fig. 3. Fig. 3 shows a cross-sectional view illustrating how display noise interference is coupled to capacitive touch systems. The source of display noise varies with display types: Liquid crystal displays (LCDs) have a common-voltage (VCOM) plane or grid, which introduces capacitively coupled display noise to the touch sensor due to the voltage transition of the source and gate driver lines, referred to as VCOM noise ^[1]. Organic light emitting diode (OLED) displays have a common cathode layer, called ELVSS, which introduces a capacitively coupled display noise to the touch sensor synchronized with H-sync signals ^[3,4]. C$_{\mathrm{c,d}}$ is the parasitic coupling capacitance between a driving channel and the ELVSS or VCOM. Similarly, C$_{\mathrm{c,s}}$ is the parasitic coupling capacitance between a sensing channel and the ELVSS or VCOM. The VCOM noise in LCDs or the zebra pattern (black-white pattern) in OLED can significantly degrade the SNR.

Fig. 2. Key performance metrics of touch system: (a) SNR; (b) frame rate.

Fig. 3. Capacitively coupled display noise interference to capacitive touch system.

While several works have comprehensively reviewed capacitive touch systems ^[7,8], they primarily focus on the system architecture with a broad overview, lacking detailed discussions and evaluations on circuit techniques aimed at improving display noise immunity. The original contributions of this paper are as follows:

1. We provide a review of circuit techniques aimed at enhancing the display noise immunity in the AFE of mutual-capacitive touch systems ^{[1-3, 9-16]}. Specifically, we examine the effectiveness of the following techniques: high-voltage driving scheme at a transmitter (TX) [^{9, 11-14],} fully differential sensing scheme at a receiver (RX) ^{[1, 2, 11, 15, 16],} and touch-display frequency synchronization scheme ^[1].

2. We analyze and validate the effectiveness of each circuit technique using Verilog-A modeling of read-out circuitry with electrical modeling of the TSP. This analytical approach can also be extended to address other noise sources and can aid in the design and optimization of noise-immune AFEs for capacitive touch systems.

Section II of this paper reviews the circuit techniques to enhance the display noise immunity. Section III elaborates on the testbench specifications and circuitry for testing display noise immunity. Section IV evaluates the circuit techniques aimed at enhancing the display noise immunity. Section V concludes this paper.

II. Review of Circuit Techniques to Enhance Display Noise Immunity

Here we review the following three circuit techniques that enhance noise immunity against display noise interference in the AFE of mutual-capacitive touch systems, as shown in Fig. 4:

Fig. 4. Mutual-capacitive touch system architecture with three strategies to enhance display noise immunity at analog front-end analyzed in this work.

First, a higher driving voltage (V$_{\mathrm{TX}}$) at the AFE TX improves the SNR under display noise by directly enhancing the signal power ^{[9, 11-14]}. However, hardware overhead is inevitable, such as utilizing high-voltage devices and a charge pump to generate a higher supply voltage on the chip or additional pads allocated for higher supply voltages off-chip.

Second, a fully differential sensing scheme at the AFE RX is effective in rejecting the common-mode display noise ^{[1, 2, 11, 15, 16]}. However, due to manufacturing mismatches in the TSP, the parasitic capacitance and resistance of the touch sensor electrodes and the coupling capacitance between the touch sensor electrode and the ELVSS or VCOM differ channel by channel. In addition, a parasitic capacitance at the routing path on a flexible printed circuit board (FPCB) from the TSP to the AFE also differs channel by channel. Therefore, the display noise is no longer an ideal common noise source, making residual noise even with the differential sensing scheme.

Third, the effect of display noise can be reduced by synthesizing the touch signal frequency with a phase-locked loop (PLL) using periodic display control signals so that the touch signal avoids display noise interference in the frequency domain. In ^[1], the touch signal frequency is set in the dips of the display noise spectrum to minimize the impact of display noise. On the other hand, the touch operation can be separated from the display operation in the time domain. Since the display noise interference mainly occurs when a display driver integrated circuits (IC) starts to drive the display panel, the touch operation can avoid it using H-blank or V-blank time, yet within a limited time ^[12]. In the case of asynchronous touch-display operation, the duty cycle of touch signals can be adaptively controlled to avoid display noise interference ^[3].

In the following sections, the effectiveness of each circuit technique is evaluated with Verilog-A modeling of the readout circuitry and electrical modeling of the TSP. In this work, a lock-in sensing capacitance readout scheme shown in Fig. 5 is employed for an AFE of capacitive touch systems. Since the capacitance signal is DC, the TX drives the TSP channels to modulate the mutual capacitance signals to the signal frequency (f$_{\mathrm{sig}}$). The modulated capacitance signal is amplified at the charge amplifier (CA) and down-converted into DC again by a mixer (chopper). The out-of-band noise and error are attenuated by the low-pass filter (LPF), of which output is proportional to the touch signal (${\Delta}$C$_{\mathrm{M}}$).

Fig. 5. Effect of display noise in capacitive touch system based on lock-in sensing capacitance readout scheme.

III. Testbench Specifications and Circuitry for Testing Display Noise Immunity

To investigate the effectiveness of each circuit technique regarding display noise immunity, the TSP and read-out circuitry are modeled. Fig. 6 shows the electrical modeling of the mutual-capacitive TSP based on a unit cell that models a cross node between the driving and sensing channels with equivalent resistances and capacitances. R$_{\mathrm{drv}}$ and R$_{\mathrm{sen}}$ are the unit-cell parasitic resistances of a driving and sensing channel, respectively. C$_{\mathrm{p,drv}}$ and C$_{\mathrm{p,sen}}$ are the unit-cell parasitic self (grounded) capacitances of a driving and sensing channel, respectively. C$_{\mathrm{c,drv}}$ is the unit-cell parasitic coupling capacitance between a driving channel and the ELVSS. Similarly, C$_{\mathrm{c,sen}}$ is the unit-cell parasitic coupling capacitance between a sensing channel and the ELVSS. C$_{\mathrm{M}}$ is the mutual capacitance between the driving and sensing channels in the unit cell. ${\Delta}$C$_{\mathrm{M}}$ is the C$_{\mathrm{M}}$ change by the touch object. V$_{\mathrm{DN}}$ represents display noise, capacitively coupled from the ELVSS through C$_{\mathrm{c,drv}}$ and C$_{\mathrm{c,sen}}$. Although the display noise may vary slightly across the EVLSS, the display noise injected into each cell is assumed to be the same in this work.

Fig. 6. Electrical modeling of capacitive TSP based on unit cell model.

With the voltage stimulation from the TX, the mutual capacitance change ${\Delta}$C$_{\mathrm{M}}$ is converted into a voltage signal by a charge amplifier, which is based on an operational transconductance amplifier (OTA) with capacitive feedback. Fig. 7(a) shows the schematic of a differential charge amplifier (DCA). The differential output of the DCA V$_{\mathrm{DCA}}$ (=V$_{\mathrm{outp}}$-V$_{\mathrm{outm}}$) is expressed by (1), where R$_{\mathrm{FB}}$ and C$_{\mathrm{FB}}$ are the feedback resistance and capacitance, respectively.

V$_{\mathrm{in,CM}}$ denotes the input common-mode voltage of the DCA. The detailed derivation process of (1) is described in Appendix.

Fig. 7(b) shows the schematic of a chopper, which converts the DCA output to the DC value with the chopping clocks. The chopper consists of four CMOS switches, controlled by the en and enb signals which have the same frequency as the signal frequency of capacitance modulation. In this work, a multi-feedback low-pass filter (MFB LPF) is chosen because it can achieve second-order low-pass filtering using only one OTA. Fig. 7(c) shows the schematic of an MFB LPF. The differential output of the MFB LPF V$_{\mathrm{LPF,OUT}}$(=V$_{\mathrm{outp}}$-V$_{\mathrm{outm}}$) is expressed by (2) where V$_{\mathrm{LPF,IN}}$ denotes the differential input of V$_{\mathrm{inp}}$-V$_{\mathrm{inm}}$. The detailed derivation process of (2) is described in Appendix.

Fig. 7. Schematic of capacitive touch readout circuitry: (a) differential charge amplifier; (b) chopper; (c) multi-feedback low-pass filter.

Table 1 shows the testbench specification of the capacitive TSP and the readout circuitry. One sensing channel of the TSP consists of 20 unit cells. Each unit cell has R$_{\mathrm{drv}}$ and R$_{\mathrm{sen}}$ of 10 ${\Omega}$, C$_{\mathrm{p,drv}}$ and C$_{\mathrm{p,sen}}$ of 2 pF, C$_{\mathrm{c,drv}}$ and C$_{\mathrm{c,sen}}$ of 8 pF. C$_{\mathrm{M}}$ is 1 pF, while assuming ${\Delta}$C$_{\mathrm{M}}$ is 0.1 pF in this testbench. The DCA and MFB LPF are modeled using Verilog-A with an amplifier that has a DC gain of 60 dB and a bandwidth of 10 MHz. For display noise, the H-sync frequency is set to 167.8 kHz considering recent specifications of mobile displays ^[17-19], as shown in Table 2. The peak-to-peak swing of the display noise voltage at ELVSS is assumed as 1 V.

Table 1. Testbench specifications of (a) touchscreen panel (TSP); (b) analog front-end (AFE)

Table 2. Recent specifications of mobile displays ^[17-19].

Fig. 8 shows the AC response of the DCA with the TSP. With C$_{\mathrm{M}}$ only considered, the DCA achieves a flat band gain of -22 dB as expected from (1) and a 3-dB bandpass from 14.3 kHz to 2.19 MHz. With the whole TSP, however, the frequency range over 60 kHz is low-pass filtered by the parasitic RC inside the TSP. Given this, the suitable signal frequency range is 100-500 kHz. The MFB LPF achieves a DC gain of 18 dB with a 3-dB bandwidth of 4.81 kHz. The RX supply voltage is set to 3.3 V.

Fig. 8. Simulated AC response of DCA with TSP.

IV. Evaluation of Circuit Techniques Enhancing Display Noise Immunity

1. Differential Sensing Scheme at the RX

Fig. 9 shows the transient simulation results with and without the display noise. When there is no display noise, the DCA outputs show a clear sinusoidal wave which is the touch signal (${\Delta}$C$_{\mathrm{M}}$) modulated by a 300-kHz signal, as shown in Fig. 9(a). With the display noise of 1 V$_{\mathrm{PP}}$ at the ELVSS in Fig. 9(b), the DCA outputs are strongly coupled by the display interference noise, as shown in Fig. 9(c). Because the coupling capacitance of 160 pF (= 8pF ${\times}$ 20 cells) is much larger than the touch signal of 0.1 pF. Nevertheless, the common-mode component of the display noise in the differential output voltage of the DCA is attenuated by employing a differential sensing scheme, as shown in Fig. 9(d). Fig. 9(e) shows that the differential output voltage of the LPF shows the mean touch signal voltage of 33.6 mV and 33.5 mV with and without the display noise, respectively. The remaining display noise is 0.9 mV$_{\mathrm{pp}}$, degrading the SNR from 40.5~dB to 39.6 dB.

Fig. 9. Transient simulation results of testbench: (a) DCA outputs without display noise; (b) display noise at ELVSS; (c) DCA outputs with display noise; (d) differential output of DCA; (e) differential output of LPF.

The effectiveness of a differential sensing scheme can be degraded by mismatches between differential paths. Fig. 10 shows the SNR with the mismatch of -5% to +5% in the capacitances (C$_{\mathrm{p}}$, C$_{\mathrm{c}}$, C$_{\mathrm{M}}$) and resistance (R$_{\mathrm{p}}$) of the TSP in Table 1 and the parasitic capacitance of the FPCB routing (C$_{\mathrm{FPCB}}$) of 1 pF. C$_{\mathrm{p}}$ and C$_{\mathrm{M}}$ are the unit-cell self capacitance and mutual capacitance of the TSP, respectively. C$_{\mathrm{c}}$ is the unit-cell coupling capacitance between the touch sensor and ELVSS. R$_{\mathrm{p}}$ is the unit-cell parasitic resistance of the TSP. For Fig. 10, the TX driving voltage and frequency are 1.8 V and 300 kHz, respectively, with an H-sync frequency of 167.8 kHz shown in Fig. 9(b). In consideration of circuit noise, the total integrated output-referred noise voltage of 330~${\mu}$V$_{\mathrm{rms}}$ is considered at the MFB LPF output. The baseline SNR is 40.5 dB when there is no display noise and no mismatch. Although the SNR is degraded to 39.6 dB with display noise, the R$_{\mathrm{p}}$ mismatch doesn’t further degrade the SNR. The C$_{\mathrm{FPCB}}$ mismatch degrades the SNR to 38.9 dB in the worst case. The mismatches of C$_{\mathrm{p}}$, C$_{\mathrm{c}}$, and C$_{\mathrm{M}}$ significantly degrade the SNR, and higher mismatch ratios result in worse SNR. The worst SNR degradation occurs due to C$_{\mathrm{c}}$ mismatch because the display noise is coupled through C$_{\mathrm{c}}$ which is the largest capacitance.

Fig. 10. Simulated SNR with respect to mismatch in TSP and FPCB.

In summary, the differential sensing scheme is effective in rejecting the display noise despite the C$_{\mathrm{FPCB}}$ and R$_{\mathrm{p}}$ mismatches. However, mismatches in the TSP capacitances, such as C$_{\mathrm{p}}$, C$_{\mathrm{c}}$, and C$_{\mathrm{M}}$, could significantly degrade the SNR. Note that the impact of each mismatch on SNR may vary with the values of the capacitances and resistances.

2. High-voltage Driving Scheme at the TX

Utilizing a k-time high driving voltage at the TX can increase the touch signal power and thus directly enhance the SNR by k$^{2}$ when the display noise is a dominant noise source, as expressed by (3).

(3)

$ SNR=\left(\frac{S_{touch}}{N_{touch,rms}}\right)^{2}\approx \left(\frac{S_{touch}}{N_{display,rms}}\right)^{2} $

However, the SNR improvement diminishes when there is a chopping timing error because the residual ripple voltage due to the chopping timing error at the LPF output also increases proportional to the driving voltage. Fig. 11(a) shows the simulated SNR enhanced by the high-voltage driving of 1.8 V, 3.3 V, and 8 V for the signal frequency swept from 100 kHz to 500 kHz. In consideration of circuit noise, the total integrated output-referred noise voltage of 330 ${\mu}$V$_{\mathrm{rms}}$ is considered at the MFB LPF output. With 1.8-V, 3.3-V, and 8-V driving, the SNR at 500 kHz is 36.1 dB, 41.3 dB, and 49 dB, respectively, showing a clear improvement of 5.2 dB (~20log(3.3/1.8)) and 7.7 dB (~20log(8/3.3)) well matched with (3). However, for lower signal frequencies, the SNR improvement diminishes because the ripple voltage at the LPF output is less filtered, which is proportional to the driving voltage. With 8-V driving, the SNR reaches 54.8 dB at 190 kHz. With 3.3-V and 1.8-V driving, the best SNR is 50.4 dB at 140 kHz and 47.1 dB at 110 kHz, respectively. Although the peak SNR appears in different frequencies, the SNR also diminishes for higher frequencies due to the low-pass filtering characteristic of the TSP shown in Fig. 8(a).

Fig. 11. Simulated SNR: (a) enhanced by high-voltage driving; (b) degraded by display noise.

3. Touch-display Frequency Synchronization Scheme

Fig. 11(b) shows how display noise degrades the SNR in the frequency domain, with the display noise waveform shown in Fig. 9(b). In this case, the SNR dip predominantly occurs around 83.9 kHz, 251.7 kHz, and 419.5 kHz, which correspond to the odd harmonics of the display noise interference coupled with the H-sync frequency of 167.8 kHz (83.9 kHz = 167.8 kHz/2, 251.7~kHz = 167.8 kHz·3/2, 419.5 kHz = 167.8 kHz·5/2). Although the effectiveness of high-voltage driving is verified as expected by (3), the SNR dip occurs in a frequency range of about +/-25 kHz at the display noise frequencies, preventing the use of signal frequency near the display noise frequencies. Therefore, it is demonstrated that not only the high-voltage driving but also the frequency selection significantly affects the SNR enhancement in the existence of display noise.

Fig. 12(a) displays three different display noise patterns which represent different display data transitions, while Fig. 12(b) shows the power spectrum density (PSD) of each pattern obtained through FFT analysis. The simulated PSD reveals that the display noise varies with each pattern; pattern #1 and #3 exhibit a fundamental frequency tone at 83.9 kHz (=167.8 kHz/2), whereas pattern #2 does not due to its period of 1-H time, as shown in Fig. 12(a), resulting in a fundamental frequency of 167.8 kHz. Pattern #1 and #3, with a period of two times 1-H time, have a fundamental frequency of 83.9 kHz (=167.8 kHz/2). Fig. 12(c) depicts the simulated SNR of the AFE corresponding to each display noise pattern. The SNR dip coincides with the frequencies of the display noise tones present in each pattern, such as 83.9 kHz, 167.8 kHz, 251.7 kHz, 335.6~kHz, and 419.5 kHz. It is noteworthy that the display noise interference occurs at harmonic frequencies half of the H-sync frequency (= n·f$_{\mathrm{sync}}$/2).

Fig. 12. Simulated SNR dependent on display noise patterns: (a) three different display noise patterns; (b) their FFT results; (c) simulated SNR.

V. Conclusions

The circuit techniques enhancing display noise immunity for the AFE of mutual-capacitive touch systems are reviewed, and their effectiveness is evaluated. Demonstrated through behavioral-level circuit simulations based on Verilog-A, the SNR can be effectively enhanced using a high-voltage driving scheme with proper signal frequency selection, avoiding the display-data-dependent fundamental and harmonic tones of display noise interference at the TX. In addition, a fully differential sensing scheme at the RX has limitations in rejecting display noise due to the channel-to-channel mismatch of the TSP, especially capacitance mismatches. These results can be used for designing high-SNR AFE in mutual-capacitive touch systems robust to display noise.

Appendix