Abstract
A novel touch screen control technique is introduced, which scans each frame in two steps of concurrent multichannel driving and differential sensing. The proposed technique substantially increases the scan rate and reduces the ambient noise effectively. It is also extended to a multichip architecture to support excessively large touch screens with great scan rate improvement. The proposed method has been implemented using 0.18 μm CMOS TowerJazz process and tested with FPGA and AFE board connecting a 23-inch touch screen. Experimental results show a scan rate improvement of up to 23.8 times and an SNR improvement of 24.6 dB over the conventional method.
1. Introduction
Touch screen panels (TSPs) are very convenient computer human interface. They are used to detect human touch or a special pen by changing one of their physical properties. For medium- and small-size TSPs for mobile products, projected mutual capacitive-type TSPs have superior visibility and durability and also provide multitouch functions [1]. They are also increasingly adopted for large PC monitors and TV screens [2, 3]. In mutual capacitance TSPs, touch detection is determined by measuring the change in the mutual capacitance.
Figure 1 shows the conventional touch screen controller (TSC) system scanning a projected mutual capacitance TSP. It works to drive a sequence of pulses to the driving electrodes (TX) which results in a staircase signal after integration of the output signal coming out through the sensing electrodes (RX) [2, 4, 5]. The touch screen panel structure used in Figure 1 uses a crossline structure with large plates to magnify the value of the mutual capacitance to get better touch detection results. There are other different electrode patterns invented to control the values of mutual capacitances, self-capacitances, and electrode resistance and so control the touch detection sensitivity and accuracy [6, 7].
Single-line driving single-line sensing (SDSS) is considered the conventional scan method. It has a frame scan rate [3] given by the following formula:
Here, is the number of TX lines, is the number of RX lines, and is a period of one TX pulse. is the number of integration steps (or the number of TX pulses per sensing point). As indicated by (1), conventional TSC systems sequentially drive each TX line and sequentially sense each RX line, resulting in an unacceptable low frame scan rate for large touch screen panels.
Different research efforts have been done to improve the frame scan rate and signal to noise ratio as they are the most important issues in controller design [8–10]. Mutual capacitance touch screen panels are susceptible to noise, so much of research has been focused on how to reduce the noise and increase the touch signal strength [11, 12] by single-line driving differential sensing (SDDS). While there are software algorithm approaches [13–15], many papers proposed circuit-level approaches to address the noise reduction problem [16, 17]. Frame scan rate has been solved for small- and medium-size screens by different driving methods [12]. However it is still not efficient for large-size TSPs.
In this paper, we propose a new TSC architecture. This architecture can be used in multichip architecture to share scanning of a large TSP [18]. As the size of TSPs grows, in general, the number of driving (TX) lines and sensing (RX) lines tends to increase to keep their high resolution of touch detection. It becomes therefore difficult to control a large TSP because frame scan rate is much lower than required. Our touch screen controller used the proposed concurrent driving technique with differential sensing in order to improve frame scan rate and eliminate noise effects.
2. Concurrent Driving Method
Large touch screen panels usually require a large number of driving (TX) and sensing (RX) lines in order to satisfy their touch detection resolution requirements. This imposes high burden on TSC design. The time required to scan one frame grows along with the number of TX and RX lines in most touch screen control methods based on (1). However most TSC systems have a target frame scan rate that they must satisfy to ensure touch detection quality which is usually in the range of 100 Hz~200 Hz [8, 9].
Our TSC provides an effective solution with a very high scan rate for large touch screen panels. It consists of three components driving a specified number of TX lines, sensing a specified number of RX lines, and running detection algorithm software on the sensed data and then it sends touch position data to the next chip in multichip architecture via serial link. A novel driving and scanning technique is introduced, which reduces the frame scan time and ambient noise effects. The controller can be used for touch screen panels of different sizes, fabricated with various materials and various patterns.
Our touch screen scanning method consists of two steps to scan one frame of touch screen panel. In the first step (concurrent driving differential sensing (CDDS)), all TX lines (rows) are concurrently driven by opposite polarity pulses, which is described by Figure 2(a). This step determines the touched RX lines by differentially sensing the output of each two adjacent RX lines. The idea behind this technique is to cancel the common mode signal while leaving only the differential signal by applying opposite excitation pulses to every adjacent cell in the same RX line. Since the major portion of ambient noise often appears as common mode signals in the adjacent RX lines, such noise can be canceled effectively by differential sensing leading to an increased signal to noise ratio (SNR). This step results in zero output in case of no touch while giving nonzero output if any cell on the two RX lines is touched. The second step (single-line driving differential sensing (SDDS)) is to sequentially scan each pair of two adjacent cells in only the touched RX lines as shown in Figure 2(b).
The differential integrator input integrates the difference between two adjacent lines. Through concurrent driving, each TX signal propagates through one TX line and two RX lines connected by two mutual capacitances as shown in Figure 3. Hence the differential RX output, in case of single-line driving, can be expressed by
Here and are the transfer functions (TFs) for TX line and RX line, respectively. and are TFs for the mutual capacitance of the first and the second cells, respectively. is TF for the RC circuit connecting the first and the second cells. Considering that the resistance and self-capacitance connecting two adjacent cells are very small, we can ignore their transfer function. Therefore the differential output of two RX lines equals
In case of no touch, the output is zero since the two mutual capacitances are equal and so are their transfer functions. On the contrary when there is a touch point affecting the two cells in different amount, the output is not zero. The output has positive or negative value based on which cell is touched and the polarity of the input signal as shown in Figure 2.
Applying the superposition theorem on the system of concurrent differential driving case, the system output is the summation of the effect of all supplies which results in zero output in case of no touch, positive in case of touch affecting the even ordered cells (marked in purple in Figure 2), and negative in case of touch affecting the odd ordered cells (marked in green in Figure 2).
The timing diagram of Figure 4 illustrates the two steps. The first step (CDDS) drives all TX lines concurrently with opposite polarity pulses and sequentially reads out every pair of two RX lines. During this step, each pair of RX lines is marked as touched or untouched. Once all the touched pairs are determined, the second step (SDDS) starts by selecting the first touched pair of RX lines and sequentially drives the TX lines starting with the first TX line. Each cell on the selected RX line is marked based on the differential integrator output. A positive integrator output indicates that the first cell of the RX pair is touched, while a negative integrator output indicates that the second cell is touched. The above process is repeated for all TX lines to scan all touched RX lines. Once all the touched RX lines are scanned, all touched cells (touch positions) can be determined.
Our method eliminates the needs for scanning all cells every frame which is the major drawback of conventional methods, and so it can improve the frame scan rate. For an example touch screen of lines, a conventional sensing method requires seconds where is the time for scanning one cell. The proposed method, on the other hand, needs only seconds, where is the number of RX pairs that are determined as touched pairs. Here, the first term is the time consumed by the concurrent scan step which scans each of the RX lines and finds touched lines. The second term is the time consumed by the selective scan step which scans only the touched RX lines.
Next, consider an example where five fingers are touching different RX lines. Assuming that one touch point affects three RX lines on average [5], the number of affected RX lines is 15 and the number of RX pairs is 10. Then the proposed method has a total scanning time of . This is an improvement of 6.63 times compared with the conventional method. However in case of no touch event for long time, our scanning method keeps repeating the first step until touch event occurs, so that our TSC keeps scanning the touch screen panel every seconds which equals 0.5 milliseconds for 1.25 MHz excitation signal and eight integration steps. Therefore it increases the touch screen response because the first step takes very short time to finish.
For our proposed method, a generalized form of the frame scan rate is given by
Our proposed method employs a differential sensing circuit to cancel ambient noise, which is the major cause of poor touch detection performance in most cases.
It can substantially speed up the overall sensing process allowing a very fast frame scan rate. It, therefore, is well suited for large touch screens.
3. Multichip Architecture
To support large touch screen panels, we may use one of the two conventional methods. The first is fabrication of another TSC that has a larger number of TX and RX lines. This, however, is time consuming and costly and cannot achieve a high frame scan rate. The second is to use multichip architecture with each chip employing a conventional scanning technique. For the first method, the scan time is proportional to the square of the screen size. The second method can allow sequential TX driving and simultaneous RX sensing. Thus its scan time linearly increases in proportion to the number of TX and RX lines.
In this section, we propose a novel scanning method which can efficiently reduce the overall scan time in multichip architecture. Figure 5 shows a TSC designed for this multichip architecture. This TSC conducts three main functions: drive a set of TX lines, sense a set of RX lines, and run detection algorithm. The chip consists of two main blocks: CPU and TX/RX controller. The CPU is connected with other functional blocks through an internal system bus called AHB bus. The CPU and TX/RX controller communicate through a set of control registers. While the TX/RX controller controls the TX and RX switches, the CPU reads sensing data from its RX memories and runs an algorithm for touch position calculation. Table 1 summarizes the function of major blocks in the proposed TSC chip of Figure 5.
Even for extremely large touch screens, only a small number of TSCs are needed as shown in Figure 6. The number of TSCs () needed for a given touch screen size is given bywhere is a function that rounds up to the next integer.
Our chip supports a commercial 23-inch touch screen of projected mutual capacitance type, which has electrode lines of channels. The proposed multi-TSC scanning technique can support larger touch screens by using multiple chips. Let be the number of TSCs. We partition the TX and RX lines of a touch screen into segments of TX and segments of RX lines, respectively. Then the TSCs cooperate in 2 steps as given in Figure 6.
Each TSC generates its TX signals in parallel and receives RX signals directly to/from TSP, and it also calculates the touch positions individually. For example, suppose we control a very large touch screen (132 × 234 lines), nine times larger in size than the touch screen size supported by one TSC introduced in Section 2. The proposed architecture uses three TSCs (TSC0~TSC2) as illustrated by Figure 7. Step 1 (CDDS): each of the TSCs drives its segment of TX lines concurrently, while each of the TSCs determines which RX lines are touched (stored in ); Step 2 (SDDS): each of the TSCs drives entire TX lines sequentially for only the RX lines that have been selected as touched lines in Step 1. The reading operation for RX lines is always conducted sequentially. In Step 1, the reading is done once per every RX pair. On the other hand, in Step 2, the reading is repeated as many times as the number of TX lines per each RX pair that was selected in Step 1.
(a)
(b)
Our proposed architecture, therefore, can further reduce the latency of detection. It also can reduce power consumption and SoC design costs. The following equation gives the frame scan rate of the multichip architecture for large TSP supported:
Here is the number of RX pairs determined as touched pairs by every TSC, is the number of RX lines supported by one chip, and is the total number of TX lines.
4. Experimental Results
4.1. System Implementation
Our proposed method has been implemented with FPGA board connected with analog front end (AFE) board. The AFE board has our chip fabricated in 0.18 µm CMOS TowerJazz process. Figure 8 shows our system implementation and our chip layout. Fabricated chip has all analog switches, sensing circuits, and digital logic for controlling switches. However FPGA board has ADC, DAC, and all control and processing logic. In this system, embedded CPU on FPGA is used to run touch detection algorithm and extract touch points in order to send them to PC through UART cable.
4.2. Scan Rate Improvement
As described in the previous section, our scanning technique (CDDS + SDDS) uses two steps to scan one frame. Conventional touch screen detection methods (SDSS or SDDS) need a total scan time of seconds to scan one frame. Here represents the time required to scan one cell. is determined by the frequency of driving signal and the number of charge integrations required to meet the target performance. On the other hand, for our proposed technique, it takes seconds to finish the first step and seconds to finish the second step. Therefore, the scan complexity of our proposed method is when a single touch is considered, while it slightly increases for multitouch. In contrast, the scan complexity of conventional methods is regardless of the number of touches.
For the multichip TSC architecture, the required scan time does not increase with the number of TSCs employed, since all TSCs operate in parallel. In particular, the parallel processing keeps the time taken by the first step always the same, no matter how many TSCs are used. On the other hand, the required time to finish the second step of our multichip scheme is given by . For example, consider a multichip scheme with 3 TSCs. If the number of touched RX pairs for each TSC is (3, 5, 1), then is 5, which is the largest number of RX pairs of all TSCs. Thus for the above example TSP of lines, our method gives a scan time of 78 × 1 us + 5 × 132 × 1 us = 738 us (assuming ). In contrast, the conventional method requires 123 × 234 × 1 us = 30888 us. This example thus shows that our multichip architecture can substantially improve the frame scan rate compared to conventional schemes.
Figure 9 compares the frame scan time for 3 different schemes simulated with touch screens of various sizes: (1) the proposed multichip architecture with concurrent scan, (2) multichip architecture with a conventional scan, and (3) conventional single TSC. In this comparison, the numbers of TX and RX lines of each touch screen size are scaled in a way that makes the pitch between the lines the same as the pitch of the 23-inch touch screen (a commercial touch screen used for experiment). This comparison is done with the following parameters: 1.25 MHz excitation TX pulses and 8 TX pulses for each charge integration. We also assume that every TSC has disjoint five touch points which leads to ten pairs of RX lines.
As shown in Figure 8, the frame scan rate for our technique is 22 times higher than the conventional scheme for the 70-inch touch screen and 14.3 times higher for the 46-inch touch screen.
4.3. Noise Cancellation
We have conducted experiments for a single TSC case with a 23-inch commercial touch screen with the system shown in Figure 8. We implemented the digital controller in an FPGA and the differential driving and sensing circuit in an AFE board. The differential integrator of the sensing circuit has been implemented in cascaded stages of subtraction and integration circuits.
Figure 10 compares experimental results of our TSC with differential sensing and the results of the single-line sensing method. For both cases, the output of the integrator is measured using concurrent TX driving with differential TX pulses for a 23-inch TSP. In Figures 10(a) and 10(b), the signal in blue color represents the difference between two RX lines in case of differential sensing. In Figures 10(c) and 10(d), the blue signal is the RX line signal for the single-line sensing method. The signal in green color represents one TX line input, and signals in sky blue and purple represent the integrator output.
(a)
(b)
(c)
(d)
The proposed CDDS scheme can cancel common mode signals, which can in effect remove most of ambient noise as shown in Figures 10(a) and 10(b). As described in Section 2, the integrator output of the proposed scheme becomes zero in case of no touch (Figure 10(a)) and nonzero in case of touch applied (Figure 10(b)). On the other hand, concurrent driving with single-line sensing (CDSS) scheme results in nonzero output in both untouched (Figure 10(c)) and touched case (Figure 10(d)). In this scheme, the integrator output gives larger values for touched case. Nonzero output of the untouched case is due to the accumulation of the signal difference between every two adjacent cells in the same RX line.
Therefore, the proposed CDDS method enables efficient noise reduction by cancelling ambient noise. This effect can be observed by inspecting that the peak-to-peak noise values in Figure 10(a) are much lower than that of Figure 10(c). Such lower noise values allow smaller deviation of the signal samples at the ADC from their average values and so lead to more accurate decision of touch detection. Therefore, the CDDS scheme gives a higher SNR, and then it can eliminate the need for additional noise filters. Figure 11(a) compares the SNR gain of four different driving and sensing schemes. First, we compare the two schemes of concurrent driving. While the SDDS scheme gives 69.13 dB, the proposed scheme CDDS provides an SNR of 67.82 dB showing negligible SNR degradation. Then we compare the 2 schemes of single-line driving. While the SDSS scheme gives 43.2 dB, the CDSS scheme gives 40.11 dB. These results show that the concurrent driving has little degradation in SNR compared to the single-line driving scheme. The differential sensing, however, provides a large SNR improvement, 27 dB, over the single-line sensing. Figure 11(b) shows that the proposed scheme CDDS provides 6.63 times higher scan rate than the conventional scheme SDSS for 5-finger touch case and 23.83 times higher for one-finger touch case. These experimental results, therefore, prove that the concurrent driving with differential sensing scheme provides substantial improvement in frame scan rate at little sacrifice in SNR.
(a)
(b)
5. Conclusions
We have presented a new architecture for touch screen controller which can be extended to a multichip structure for parallel processing. We have also developed a concurrent driving and high SNR sensing method for projected mutual capacitance touch screens. A two-step algorithm is proposed to reduce the time required for scanning large touch screens. With the concurrent driving and multichip architecture, our method can provide substantially faster frame scan rate than conventional methods, a crucial requirement for large touch screen panels (TSPs). Experimental results show that our proposed method can improve the frame scan rate by 6.6 times for a 5-finger touch and 23.8 times for a single-finger touch for a 23-inch TSP compared to a conventional method. They also show that its touch detection SNR gain is 67.82 dB, a 24.6 dB improvement compared with the conventional scheme.
Conflict of Interests
The authors declare that there is no conflict of interests regarding the publication of this paper.
Acknowledgment
This work was supported by the Center for Integrated Smart Sensors funded by the Ministry of Science, ICT & Future Planning as Global Frontier Project (CISS-2014011066).