With the introduction of the first smartphones in 2012 with in-cell touch displays, a battle is emerging between the traditional supply chain of ITO-based touch sensors that are externally combined with the display, and the display vendors themselves integrating the touch sensor layers inside their display stackup. Either party is in the pursuit of the thinnest and lowest-cost touch solution.
Cypress Semiconductor has been supplying capacitive touchscreen controller IC solutions since 2008 and has maintained a database with the sensor parameters and layer structures for each customer project that has entered mass production since then. Based on actual design project data from this database, we have derived some trends on the prevailing parameters of 'on-stack' sensors over the 2009-2012 timeframe. 'On-stack' refers here to structures where the touch sensor layers are supplied by standalone (ITO) sensor vendors and are combined with the display module after completing manufacturing of the display. The remainder of this article discusses the characteristics of the prevailing designs, their evolution over time, as well as some of the rationale for the observed trends.
We also highlight more recent structures with touch sensor layers that are integrated inside the display stack-up, so-called 'on-cell' or 'in-cell' touch. Promoted by the display companies, these display-integrated touch modules now compete against new 'on-stack' structures from discrete sensor vendors that offer a similar reduction in overall thickness.
Smartphone Screen Size
As might be expected, screen sizes have gradually increased from an average of 3.2-in in Q1 2009 to 4.3-in by Q3 2012 (the data in Q4 2009 is an anomaly). Figure 1 shows that especially during 2012 screen sizes have increased dramatically. The dataset used for this graph, and all subsequent ones, is only for small-screen projects up to 6-in and thus excludes any tablet or notebook touchscreen designs.
Figure 1. The evolution of average screen in under 6-in customer projects shows a dramatic increase in 2012.
Airgap, substrate and cover lens thickness
In an on-stack sensor, the sensor layers can be directly laminated onto the top glass of the display or there can be an airgap between both. While the presence of an airgap results in a thicker touch module, an airgap attenuates any noise from the display observed by the touch controller. A touch sensor vendor combines an on-stack sensor with a completed LCD display. As a result, the touch controller has typically no access to any display synchronization signals. This is because the display drivers that generate these signals are physically put on the display glass inside the LCD module, while the touch controller is on a separate flex tail attached by the sensor vendor. The touch controller scans throughout the active time of the display refresh period, and is thus susceptible to any noise generated by the TFT pixel drivers. An airgap mitigates this noise source.
Figure 2 shows that airgap thickness increased from Q1 2009 to Q4 2010, likely the result of a move to higher resolution displays which generate more noise. Afterwards, as on-chip touchscreen IC noise mitigation techniques improved, airgap thickness declined again to an average of ~0.24 mm by Q3 2012.
Figure 2. The evolution of cover lens, sensor substrate, and air gap thickness in under 6-in customer projects shows increased thickness starting in 2009.
Another clear trend is the reduction of the sensor substrate thickness — this is the material that holds the (ITO) touch sensor: from 0.5 mm in 2009 to 0.2 mm by Q3 2012. This is mainly the result of the introduction of film based (PET) substrates vs glass based substrates. PET can be made thinner and is less expensive, not only because of lower material cost but also since the capital investment, and thus manufacturing cost, for PET-based ITO patterning is significantly less than for ITO patterning on glass.
On the other hand, from our database of small-screen touch projects, we cannot see a clear trend for the thickness of the cover lens: it continues to hover around 0.7-0.8 mm on average over this time period. While reducing the cover glass thickness reduces module thickness, cover glass strength is key to a smartphone OEM to avoid glass breakage, especially for today's phone designs that have no raised bezel around the screen (and thus no protection against a drop).
Furthermore, while it would seem obvious that a thinner cover lens improves capacitive touch performance (i.e., less attenuation and thus a higher signal-to-noise ratio), there is actually a trade-off: a thinner cover lens complicates capacitive touch performance when the phone is ungrounded, such as when it is laying on a table without a charger connected. Thus, there are practical limits to the minimum cover lens thickness, both from a material strength perspective as well as due to restrictions imposed by touch controller signal processing.
You can also derive from this data that there was no similar trend towards thinner plastic-based cover lens materials as we mentioned earlier for sensor substrates: the vast majority of phones still use a glass-based cover lens, not a PMMA type lens.
Pattern Types
Over the years, there has also been continuous innovation in the sensor patterns used for capacitive touch. Initial designs used a 'diamond' type structure (Fig. 3). Capacitance was measured between each diamond node and a common ground level (self-capacitance). Later the same pattern, or a variant with some improvements, was used for detecting capacitance between neighboring diamond nodes (mutual capacitance) as also shown in Figure 3. Mutual capacitance sensing allowed the introduction of 'unlimited multi-touch' vs the initial self-capacitance controllers that were limited to single-finger touch or two-finger pinch/zoom gestures due to the ghosting effect associated with diamond-type self-capacitance sensors.
Figure 3. Diamond type pattern showing mutual capacitances detected for multi-touch was used initially for capacitive touch sensors.
Since 2Q 2010, Cypress also uses a different pattern type, often called the 'Manhattan' pattern due to its perpendicular street grid layout of transmit and receive sensors (Fig. 4). This pattern has some benefits for use with mutual capacitance touch controllers.
Figure 4. Manhattan type sensor pattern has some benefits for use with mutual capacitance touch controllers.
As the smartphone market moved to multi-touch, continuous cost reduction efforts for the low-end smartphone market has led to the development of a low-cost pattern specifically for self-capacitance known as the LCS (low-cost sensor) pattern (Fig. 5).
Figure 5. Shares of different sensor patterns in under 6-in customer projects
Traditionally, two separate sensor (ITO) layers are required: one each for the transmit (TX) and receive (RX) layers in mutual capacitance sensors, or two sensor layers to avoid routing overlap between the horizontal and vertical buses that connect self-capacitance diamond-type sensors. For several reasons — including thinner modules, better transparency, and lower cost — there has been a drive towards sensors using only a single layer of ITO. This in turn has led to the use of diamond and Manhattan type patterns laid out on a single ITO layer with metal or ITO 'bridges' for the node cross points. The trade-off is the extra manufacturing cost to add these bridges, and their visual impact, vs the cost of an additional ITO layer and extra 'bonding' (OCA i.e. optically clear adhesive) layers for a 2-layer design.
More recently, touch screen controller vendors have started to supply 'true' single-layer touch sensor designs that still provide full multi-touch but do not require bridges. These patterns are proprietary for each touch controller vendor - Cypress calls its one-layer pattern SLIM (single-layer independent multitouch). The SLIM type pattern strives for a better cost / performance trade-off through better transparency (i.e., no loss of transparency from 2nd ITO layer), elimination of bridges that could impact visual uniformity, and no additional manufacturing steps required from 'bridges'. As always, there are challenges. For example, keeping full multi-touch and excellent touch accuracy may require a higher number of sensor nodes on the panel compared to a traditional sensor pattern. This leads to more routing channels with possibly vias and multi-layer routing in the screen bezel area and on the flex tail to route all signals to the touch controller IC. Furthermore, the larger sensor count in SLIM type designs makes it more challenging to have a thin bezel — a desired feature. The data shows that about one third of our designs in 3Q 2012 used this new pattern type. It can also be seen that LCS designs have faded over the past quarters: independent multi-touch is becoming a requirement even in the low-end of the market.
Stackup Types
In parallel to the evolution of sensor patterns, the stackup types that implement these sensor patterns have changed (Fig. 6). In Q1 2009, 100% of all designs used a glass sensor substrate with glass cover lens (G/G). Over time film-based sensors, first with two sensor substrates (GFF) and later with one layer either with glass cover lens (GF) or PMMA cover lens (FF), reduced cost and sensor module thickness. As ITO patterning on PET is less accurate than on glass, both sensor manufacturing methods had to improve and touch controllers had to become more tolerant to sensor pattern variances to realize this transition.
Figure 6. Shares of different sensor stackup types in under 6-in customer projects has varied over the years.
Sensor-on-lens designs have been done for quite some time. In this case, the cover lens (glass) is used also as the substrate for the sensor. While it seems a straightforward evolution to reduce the number of sensor layers, there is a trade-off between the quality of the sensor and the shatter resistance of the cover glass which is chemically strengthened. This may actually lead to a lower-rated cover glass breakage spec for SOL type designs compared to traditional designs with separate sensor substrate(s).
In addition, as the sensor is on the cover glass, there is a smaller distance between the finger touch location and sensor. As we discussed earlier, in the context of a thinner cover lens, this puts higher requirements on the touch controller to deal with certain touch scenarios (i.e., ungrounded case). Nonetheless, as we overcame these technical challenges by improved signal processing on the touch IC, the data shows a clear trend towards SOL type designs. The industry has come up with many acronyms for SOL, such as WIS (window-integrated sensor), TOL (touch-on-lens), and OGS (one-glass sensor).
LCD Drive Schemes
Active-matrix organic LED (AMOLED) type displays have emerged in recent years as a strong alternative to LCDs. Only taking into account customer projects with LCDs, there has been an interesting trend from LCDs with an AC VCOM drive scheme to DC VCOM type displays (Fig. 7).
Figure 7. Shares of AC VCOM vs DC VCOM drive schemes in under 6-in LCD-based customer projects highlight an interesting trend in LCDs.
What does this mean? In an LCD it is important to not have a long-term DC voltage bias across the liquid crystal material. This would cause image burn-in and lead to long-term reliability issues. To avoid this, the polarity of the electric field across the LCD is reversed periodically. Think of this as a differential drive scheme to eliminate any common mode noise in electronic circuits. There are two ways to realize this polarity reversal:
- Both the LCD reference (i.e. VCOM) and LCD pixel drive voltages switch between a low and high voltage potential, causing a -V or +V differential drive across the LC material (AC VCOM)
- The VCOM voltage is kept constant at mid-level of the display driver output range, and only the source drive voltage is switched, creating a -V/2 or +V/2 drive (DC VCOM)
In the first case, twice the differential voltage across the LCD is obtained for the same LCD output driver voltage supply range. Early LCD panels required a larger voltage drive across the LC material, so AC VCOM drive was required. As LCD technology has evolved over the past couple of years, a lower voltage differential has become acceptable, and thus DC VCOM has become sufficient.
DC VCOM generally leads to better image quality since the periodic VCOM reference voltage switch cannot be observed as image flicker. Also, as screen sizes increase, an AC VCOM drive scheme requires higher power consumption. While AC VCOM does create more noise for the touch controller, it is actually a predictable noise component since the switch frequency is fixed and related to the pixel frequency. Thus it can be accurately estimated and compensated for by the touch controller. As a result, other noise sources in LCD displays are dominant for the touch controller, and the trend towards direct lamination (i.e. no air gap) between the LCD top glass plate and the touch sensor only makes it worse.
The Industry's Current Battle: Display-integrated Touch vs. Sensor-on-lens
The data so far has only been for 'on-stack' designs. For these designs, the touch controller IC is sold to a discrete sensor supplier, who manufactures the sensor and bonds a flex tail with the touch controller IC to it. At this point, either the sensor vendor, display company, or a third-party module vendor combines the touch sensor with the display. Finally the cover lens is added, another separate process since the cover lens can depend on variations of the phone model (e.g. white vs black decorative film, different logo prints, etc). This supply chain is inherently complex.
Integration of the touch layers into the display module aims to simplify the supply chain, and in the process allows the LCD or OLED vendors to capture more value. That's the theory. In reality, only a few display suppliers are already shipping in-cell in mass production, and news stories are easy to find about yield and supply challenges with in-cell touch. In their pursuit to minimize the number of extra layers — and thus manufacturing steps — when display vendors want to add touch to a display, they want to re-use certain layers between the touch and display function. This now makes the sensor also dependent on the type of display (OLED vs LCD) as well, in the case of an LCD, the pixel drive method; e.g., in-plane switching (IPS) vs vertical alignment (VA). As an intermediate step to in-cell, the touch sensor can be inside the display module but remain on dedicated layers above the top color filter glass — layers that are not shared with any display functions. This is called 'on-cell' touch.
While a detailed discussion of the various touch integrated display structures is beyond the scope of this article, in essence in-cell or on-cell touch display modules re-use substrates that are already present for the display function for the touch substrate as well: either the TFT or color filter glass for LCDs, or the encapsulation glass for OLEDs. The display industry touts the advantage of a thinner module vs. 'on-stack' touch since discrete sensor substrates are eliminated. While this is true compared to traditional sensor types we discussed earlier, we also showed this is not really true for sensor-on-lens (SOL) type of sensors.
This duality in the industry between the discrete sensor vendors pushing SOL and the display vendors pushing in-cell or on-cell touch is a key battle. As the existence of the complete industry of standalone sensor vendors is on the line — their role is eliminated in the case of display integrated touch — this is also a hard-fought battle. We do not yet see a clear winner and believe both will co-exist. This means that touch controllers ICs will need to support both configurations. The support for both on-stack and display-integrated touch puts additional requirements on the product definition of new touch controller ICs. It also changes a silicon vendor's go-to market plans since these products now also need to be promoted directly to display vendors.
Conclusion
The capacitive touchscreen industry has come a long way over the past 5 years. Not only has readily observable performance evolved from single and dual-touch to 10-touch, there have been many changes under the hood, including sensor stackups, sensor patterns, stackup types, and display types. These changes will continue following the electronic industry's ongoing pursuit for smaller module thickness, higher performance, and lower cost. Supplying capacitive touchscreen controller ICs is a business that requires an understanding of sensor materials technology, signal processing algorithms, mixed-signal and digital IC circuit design, and display systems. Perhaps this is why only a handful of IC companies are shipping in volume to today's hyper-competitive supply chain to smartphone OEMs, although many IC vendors have had aspirations to do.
There is no indication that any of the evolutions outlined here are behind us. Rather, the complexities are moving to larger screens over 10-in screen size. These larger touchscreen applications were once the playground of niche technologies such as multi-touch resistive or optical touch technologies. Now projected capacitive touch can be found in applications such as "all-in-one" PCs with screens beyond 20-in (although, admittedly, these are not yet at the right price point).
We are only at the start of seeing similar evolutions play out for these larger sized displays. Additional changes specifically addressing the large sensor material cost for these large-are sensors are coming as well. Alternatives to ITO sensor materials cause a new set of manufacturing, supply chain, and signal processing challenges for the touch controller IC, ones that are definitely not for the faint of heart.
The author would like to thank Jagadish Kumaran for the data mining on the Cypress touch project database which provided the source data for all plots shown in this article.