Real-Time Compensation in Organic Light-Emitting Diode Television Displays Using Current Sensing Method With Charge Integrators

This paper proposes a novel method for sensing and compensating for the characteristics of thin-film transistors (TFTs) in organic light-emitting diode (OLED) television (TV) displays. The proposed method uses a charge integrator (CI) to sense the TFT currents and calculates the TFT threshold voltage (<inline-formula> <tex-math notation="LaTeX">$\text{V}_{\mathbf {TH}}$ </tex-math></inline-formula>) and transconductance parameter (K) using the two different sensed currents. The sensing of the two different currents can be performed within <inline-formula> <tex-math notation="LaTeX">$300 \mu \text{s}$ </tex-math></inline-formula>, which is the vertical blank time for an ultra-high-definition resolution with a refresh rate of 120 Hz, enabling real-time compensation. A new measurement method of all feedback capacitors (<inline-formula> <tex-math notation="LaTeX">$\text{C}_{\text {FB}}$ </tex-math></inline-formula> s) of CIs is proposed to overcome the variation of <inline-formula> <tex-math notation="LaTeX">$\text{C}_{\text {FB}}$ </tex-math></inline-formula>. The proposed sensing circuit uses differential sensing and auto-zero methods to provide high panel noise immunity and accurate sensing. The performance of the compensation was estimated using Smart-SPICE (Simulation Program with Integrated Circuit Emphasis) simulation under the conditions of ±0.5 V and ±10 % variations in <inline-formula> <tex-math notation="LaTeX">$\text{V}_{\text {TH}}$ </tex-math></inline-formula> and K, respectively. The simulation demonstrated that the current deviation decreased from a maximum of 98.8 % before compensation to 2.3 % after compensation. The compensation performance of the proposed method was verified using a 13.7-inch active-matrix OLED (AMOLED) panel with a resolution of 960 × 540. The results showed that the current deviation of the 13.7-inch AMOLED panel decreased from 40.4 % before compensation to 2.7 % after compensation. The proposed method offers a more accurate and faster compensation solution for high-resolution and large-sized OLED TV displays compared to conventional methods.


Real-Time Compensation in Organic Light-Emitting Diode Television Displays Using Current Sensing Method With Charge Integrators
Myung Gi Lim and Seung-Woo Lee , Senior Member, IEEE Abstract-This paper proposes a novel method for sensing and compensating for the characteristics of thin-film transistors (TFTs) in organic light-emitting diode (OLED) television (TV) displays.The proposed method uses a charge integrator (CI) to sense the TFT currents and calculates the TFT threshold voltage (V TH ) and transconductance parameter (K) using the two different sensed currents.The sensing of the two different currents can be performed within 300 µs, which is the vertical blank time for an ultra-high-definition resolution with a refresh rate of 120 Hz, enabling real-time compensation.A new measurement method of all feedback capacitors (C FB s) of CIs is proposed to overcome the variation of C FB .The proposed sensing circuit uses differential sensing and auto-zero methods to provide high panel noise immunity and accurate sensing.The performance of the compensation was estimated using Smart-SPICE (Simulation Program with Integrated Circuit Emphasis) simulation under the conditions of ±0.5 V and ±10 % variations in V TH and K, respectively.The simulation demonstrated that the current deviation decreased from a maximum of 98.8 % before compensation to 2.3 % after compensation.The compensation performance of the proposed method was verified using a 13.7-inch active-matrix OLED (AMOLED) panel with a resolution of 960 × 540.The results showed that the current deviation of the 13.7-inch AMOLED panel decreased from 40.4 % before compensation to 2.7 % after compensation.The proposed method offers a more accurate and faster compensation solution for high-resolution and large-sized OLED TV displays compared to conventional methods.Index Terms-Active-matrix organic light-emitting diode (AMOLED), charge integrator (CI), external compensation, OLED television (TV), real-time, thin-film transistors (TFTs), threshold voltage (V TH ).

I. INTRODUCTION
T ODAY, flat panel displays are widely used in various applications and the demand for high-performance displays with large sizes and high resolutions is increasing [1], [2], [3].In the past, liquid crystal displays (LCDs) dominated the flat panel display market, but organic light-emitting diode (OLED) displays are gaining market share in the mobile and television (TV) display market.OLED displays are attracting attention for their outstanding advantages, such as a wide viewing angle, high contrast ratio, and short response time [4], [5], [6].Additionally, OLED technology is seen as a key technology for future displays, such as transparent and flexible displays [7], [8].Because OLEDs are current-driven devices, a thin film transistor (TFT) is responsible for supplying the current to the OLED.The OLED luminance is proportional to the current supplied, so the quality of the display directly depends on the characteristics of the TFT.The drain-to-source current (I DS ) of the TFT is expressed by, where V TH is the threshold voltage, and K is the transconductance parameter that includes the mobility (µ), the ratio of width to length (W/L) of the TFT channel, and the gate capacitance per unit area (C ox ).The I DS of a TFT at the same gate-source voltage (V GS ) depends on the deviation of two key parameters, K and V TH .K and V TH of all TFTs are not identical because of process variations.Additionally, K and V TH may vary due to electrical and thermal stresses during operation [9], [10], [11].The non-uniform TFT characteristics and unreliable performance degradation result in non-uniform luminance across the display screen.To address this issue, it is necessary to compensate for the TFT characteristics so that the same current is supplied for the same image data.
There are two main methods to compensate for TFT characteristics: an internal compensation method and an external compensation method.The internal compensation method utilizes an in-pixel compensation circuit [12], [13], [14], whereas the external compensation method uses circuits outside the panel [15], [16], [17], [18].This results in simpler pixel circuits in the external compensation method.However, the internal compensation method reduces the aperture ratio as it requires a large number of TFTs and capacitors in a pixel [19].To increase the aperture ratio, external compensation is used in large-sized OLED TV panels [20].
In the case of external compensation, the TFT characteristics are sensed by an external sensing circuit.The deviation is compensated for by adding it to the data voltage (V DATA ) after the TFT characteristics have been calculated based on the sensed data [21], [22].As shown in Fig. 1(a), the conventional external compensation method has a pixel structure that consists of three TFTs and one capacitor (3T1C) [23].The driver IC includes an analog-to-digital converter (ADC) and switches that comprise the sensing circuit.V S continues to increase until the V GS of T1 reaches V TH , causing T1 to turn off.Thus, V DATA − (the saturated V S ) becomes the V TH of T1.During sampling, SW2 signal is set to a high level.The ADC then converts the saturated V S value to its corresponding digital value.The compensation data value can be calculated easily using the digital value and V DATA .
It takes more than tens of milliseconds to complete the three steps [24], [25].Thus, it takes more than tens of seconds to detect the exact V TH values of all the driving TFTs on a panel.Furthermore, larger or higher-resolution panels require more sensing time.This is because C LINE is charged by the I DS of T1 during V TH sensing, and larger C LINE requires longer sensing times [26].As a result, real-time compensation for V TH is not possible using the conventional method during operation, but it can be done just before power-off.On the other hand, mobile OLED displays that adopt the internal compensation method perform the V TH compensation in real-time, ensuring the image quality is maintained regardless of operation time.However, the image quality of OLED TVs can degrade over time as the operation continues without being powered off.
This paper presents a novel external compensation method that can be performed within a vertical blank time by sensing TFT characteristics in a significantly shorter time than the conventional method.Previous studies on real-time TFT compensation either require more time than the vertical blank time of the latest display products, or they do not consider variations in K [24], [25], [26], [27].
The proposed method maintains uniformity in TFT current by periodically sensing and compensating during the vertical blank time.To accomplish this, we propose a method for sensing TFT currents using a charge integrator (CI) and a method for calculating V TH and K values using two sensed current values.Although CIs are used as sensors in various applications, there has been no research on sensing and compensating for TFT characteristics in OLED TV displays.Additionally, the proposed method uses a measurement of all feedback capacitors (C FB s), differential sensing, and auto-zero methods to remove sensing noise and errors, respectively.The effectiveness of the proposed compensation method was verified on a 13.7-inch active-matrix OLED (AMOLED) panel.

II. PROPOSED SENSING AND COMPENSATION METHOD
Fig. 2(a) shows the pixel structure and the novel sensing circuit proposed in this paper.The pixel structure is the 3T1C, which is the same as that of the conventional external compensation method.However, the sensing circuit includes an added component, a CI.
The CI measures the T1 current in a much shorter time, as the current flows through the smaller C FB included in the CI.This results in a quicker measurement of the current than the conventional method where the ADC measures the voltage caused by the current flowing through the larger C LINE .As a result, the proposed method provides far shorter sensing times than the conventional method.In addition, it is not affected by differences in C LINE with panel size and resolution.SW1, and SW2 signals are set to a high level.T2 and T3 are then turned on, and V DATA and V REF_CI are applied to the gate and source nodes of T1, respectively.During sensing, SW2 signal is set to a low level, allowing the current to flow to the CI while maintaining the V GS of T1.The input current decreases the output voltage (V SEN ) of the CI.During sampling, SW1 signal is set to a low level and SW3 signal is set to a high level, after which the ADC converts the V SEN value to its corresponding digital value.The I DS of T1 is expressed by, where C FB is the capacitance of the feedback capacitor of CI, t is the time taken for CI to integrate current, and V is the difference between the V REF_CI and the final V SEN .
V TH and K can be calculated from the corresponding currents in T1 for two different applied voltages (V GS1 and V GS2 ).The currents flowing through T1 can be expressed as shown in ( 4) and (5).4) Using ( 4) and ( 5), the values of V TH and K can be calculated as shown in ( 6) and (7).
The calculated V TH and K values are added to the V DATA value for compensation.The formula to calculate a new V DATA for compensation is given as follows: where V DATA_COMP is the new V DATA compensating for V TH and K variations, V DATA_IN is the V DATA determined by original input image data, V TH_CAL is the calculated V TH , and gain is a nonlinear parameter that compensates for the variation of K depending on V GS (i.e., V DATA_IN ).
The gain is given as follows: where K CAL is the calculated K that is assumed constant for V GS1 and V GS2 .As shown in ( 9), the gain is a function of K CAL and V DATA_IN .If K were not dependent on V GS , the gain would be (K CAL ) −0.5 .We measured TFTs to get the characteristics of K depending on V GS because K is a non-linear function of V GS .We implemented the nonlinear function expressed by ( 9) as a look-up table (LUT) based on the measurement results.The adjusted K, K ADJ , is obtained using the (9).Thus, the gain becomes (K ADJ ) −0.5 rather than (K CAL ) −0.5 .The calculated values of V TH and K are then added to the V DATA value to compensate for the TFT characteristics, which improves the luminance uniformity of the panel by removing the influence of V TH and K.After the initial variation is sensed and compensated for, the degradation of TFTs is sensed and compensated periodically in vertical blank time using the compensated V DATA .If TFTs degrade during operation, the degradation can be compensated for by calculating V TH and K in the same way as the first sensing and compensation for the initial variation of TFTs.In this way, our proposed architecture compensates not only for the initial characteristic variation of TFTs but also for their degradation.

III. SENSING PERFORMANCE IMPROVEMENT METHOD A. How to Compensate for Sensing Circuit Variation
C FB s are manufactured on a silicon wafer.As shown in (3), C FB variation directly affects sensing performance.Thus, it is necessary to overcome the C FB variation.To address this issue, we developed a new algorithm to measure all C FB s.
As shown in Fig. 3, a given current flows through a C FB .Then, we can easily calculate the value of the C FB by measuring V SEN at a given time.In this way, we can find out all C FB values for all CIs.Therefore, we can get all information about all C FB s for all driver ICs and remove the effects of C FB variation.

B. Differential Sensing Method
Accurate sensing of the currents is crucial for perfect compensation of V TH and K.The calculated values of V TH and K are based on the sensed currents.Any errors in the current sensing can result in inaccuracies in the calculation of V TH and K, leading to decreased compensation performance.Several factors can contribute to the decrease in the accuracy of current sensing during I DS sensing, including noise current (i NOISE ) from power sources and deviations in current caused by differences in V GS settings [29].
The power supply noise appears as i NOISE in the load capacitance of the sensing line, affecting the I DS during sensing.Additionally, v REF_CI , the noise in V REF_CI generates v SEN , the noise in V SEN as expressed in (10) [30].
As shown in (10), V REF_CI noise is amplified by the ratio of C LINE /C FB .
A power supply noise commonly affects the entire panel.The current noise at adjacent capacitive loads is usually similar.Thus, we propose a differential sensing method to remove the common i NOISE at the adjacent odd and even lines.As shown in Fig. 4, this method works by first applying V DATA to the TFTs of the odd line, which causes both I DS and i NOISE to flow.The TFTs on the even line are turned off, resulting in the presence of only i NOISE .As a result, the output voltage of the odd CI, V SEN_ODD , contains both I DS and i NOISE .However, the output voltage of the even CI, V SEN_EVEN , contains only i NOISE .As shown in Fig. 4, the difference amplifier removes the effect of i NOISE , allowing for the accurate, noise-free sensing of I DS by the proposed differential sensing method.
The differential sensing method not only removes the common i NOISE at the adjacent odd and even lines but also the amplified V REF_CI noise.This is because the amplified noises at the odd and even lines are almost identical.By switching the inputs of the difference amplifier, noise-free I DS can be accurately sensed at the even lines.The differential sensing method improves compensation performance by increasing immunity to power supply noises by only sensing TFTs on either odd or even line.

C. Auto-Zero Method
We need to consider offset-induced errors, as the input offset voltages (V OS ) of the CIs can result in inaccuracies in current sensing.The deviations in current resulting from differences in the V GS settings can cause errors in the calculation of V TH and K.I DS is dependent on the V GS of T1.Differences in V GS settings can result in deviations in current and inaccuracies in the calculation of V TH and K, even if the TFT characteristics are identical.The offset of the CI results in a difference in the V S of T1, leading to different V GS settings during initialization.Thus, the offset must be removed during initialization.An auto-zero method can be used to achieve this [31].The auto-zero method is a technique that samples the offset of the amplifier and then removes it by applying the sampled offset value to the amplifier input.The offset voltage stored in C AZ is then added to V REF_CI .By applying V REF_CI + V OS to the non-inverting input of the CI, the offset voltage is effectively removed.

IV. SIMULATION RESULTS
We simulated the proposed circuit under ultra-high definition (UHD, 3840 × 2160) resolution and 120 Hz refresh rate conditions using the Smart Simulation Program with Integrated Circuit Emphasis (Smart-SPICE).We simulated using oxide TFTs and white OLEDs (WOLEDs).To compensate for TFT characteristics in real-time, sensing must be completed within the 300 µs vertical blank time.The simulation considered the constraint that the sensing operation could be completed within the time limit, with assumed V TH and K values of ±0.5 V and ±10 %, respectively.

A. Sensing and Compensation
The simulation used V DATA values of 7.2 V and 5.95 V, with V REF_CI of 4.5 V.The Scan and Sense signals had positive and negative voltages (V GH and V GL ) of 24 V and −6 V, respectively.The supply voltage of the driver ICs was 16 V.The total sensing time, including two initialization times of 2 × 50 µs, was 300 µs.We selected the V GS conditions considering the sensing time and compensation performance.Sensing must be completed within the vertical blank time for real-time compensation.Since we targeted UHD resolution and 120 Hz refresh rate, sensing must be completed within 300 µs.We examined the compensation performance for V GS values that allow sensing within 300 µs and selected the two best V GS values.We compensate by adding the calculated V TH to the V DATA value.Thus, for accurate compensation, the V TH compensation error should be the same as or less than one least significant bit (LSB) of the digital-to-analog converter (DAC) of driver ICs.We determined the V REF_CI voltage level considering the sensing range and resolution.We selected the condition where the sensing range is V TH = ±0.5 V or higher, and the resolution is such that V TH is less than one LSB of the DAC.Since the V TH of WOLED is higher than 4.5 V, it did not affect current sensing.
Fig. 6 shows the results of the proposed circuit simulation.As shown in Fig. 6, these results indicate that V SEN reached two different levels for two different current conditions within the 300 µs vertical blank time.The simulation was carried out for three different V TH conditions that resulted in different current values being sensed through the V TH difference.
The simulation was conducted under nine different transistor conditions (T1 to T9) with V TH = −0.5, 0, +0.5 V and K = −10, 0, +10 %.Table I summarizes the calculated V TH and K values based on the sampled V SEN values for each transistor condition.The initial V TH and K values are represented by V TH_I and K _I , while the calculated values are represented by V TH_C and K _C .The errors in the sensed V TH and K are shown as e TH and e K , respectively.The maximum V TH error was 9.1 mV, which is less than one LSB of a 10-bit DAC with a supply voltage of 16 V.The maximum K error was 0.2 %.Fig. 7 shows the deviation of each TFT's current before and after compensation at 1.45 V of V GS (V DATA = 5.95 V).The current deviation of the TFTs before compensation ranged from −98.8 % to 62.7 %, while after compensation, it ranged from −1.77 % to 2.30 %.After compensation, almost the same current flows through the TFTs, regardless of device variations.

B. Differential Sensing Method
The performance of differential sensing method was evaluated by adding noise to V REF_CI and i NOISE .There are multiple power supplies used to drive an OLED TV panel, such as ELVDD, AVDD, V GH , and V GL .ELVDD supplies emission current to pixels.AVDD is the analog power supply for driver ICs.V GH and V GL are positive and negative power supplies for integrated scan drivers, respectively.For the i NOISE simulation, we superposed the power supply noises except for V REF_CI .The peak-to-peak voltage and frequency of the added noise to V REF_CI were set to 10 mV and 400 kHz, respectively.The superposed power supply noise had a peak-to-peak voltage of 20 mV and consists of multiple frequencies.V REF_CI noise and the superposed power supply noise were determined based on measurement data.
Fig. 8 shows differential sensing operation in the case of sensing odd-line TFTs.As shown in Fig. 8(a), V REF_CI noise amplified by the ratio of C LINE /C FB was sensed.This added about 250 mV peak-to-peak noise to V SEN_ODD .As shown in Fig. 8(b), i NOISE was added.This added about 190 mV of peak-to-peak noise to V SEN_ODD .As shown in Fig. 8, V REF_CI noise had a greater influence on sensing than i NOISE .The only noise was sensed in V SEN_EVEN .The noises generated on the odd and even lines were almost identical, so differential sensing method was able to cancel out the noise and only sense the I DS .This demonstrates that differential sensing improves immunity to power supply noise.

C. Auto-Zero Method
Auto-zero is a method that stores the offset voltage of the charge integrator in the C AZ .The stored voltage in C AZ is applied to the non-inverting input of the charge integrator to Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.remove the offset voltage.Thus, the size of C AZ must be large enough to accurately store the offset voltage while the integrator senses the current of the TFT.However, a larger C AZ will result in a larger chip size.We carefully considered the requirements of auto-zero performance and chip size.We concluded that a C AZ of 1 pF is sufficient to achieve satisfactory auto-zero performance while minimizing the chip size.
Figs. 9(a) and 9(b) show the simulated V SEN waveforms before and after using the auto-zero method, respectively.Fig. 9(c  that the use of auto-zero effectively reduces the sensing error caused by the CI offset by removing the offset.

V. MEASUREMENT RESULTS
The proposed architecture for TFT compensation was verified using a 13.7-inch AMOLED panel with a resolution of 960×540 which has the same pixel density measured in pixels per inch (ppi) as that of a 55-inch UHD display.The panel utilized oxide TFTs and WOLEDs.The driver ICs utilized 10-bit ADCs with a sensing range of 4 V, and the current of red sub-pixels was measured at 1.45 V of V GS (V DATA = 5.95 V).
Fig. 10(a) shows a photograph of the manufactured display system, which features a 13.7-inch OLED panel, four driver ICs, a source board, a timing controller board, and a communication board (ethernet PCB).
The driver IC was fabricated using a 0.18 µm CMOS process with 1.8 and 18 V CMOS devices.Fig. 10(b) shows a micrograph of a fabricated IC.The driver IC includes 960 driving channels and 240 sensing channels.

A. Error Reduction Performance
We measured I DS of the driving transistor (T1) in each pixel (a total of 960 × 540 currents), and collected corresponding digital data converted by the ADC.We proposed the use of differential sensing and auto-zero methods to reduce sensing errors caused by amplifier offset and power supply noise.To evaluate performance of our proposed method, we compared the sensed data before and after using differential sensing and auto-zero method, as shown in Fig. 11.
Fig. 11(a) shows a two-dimensional (2D) map displaying digital current data for all pixels without the application of differential sensing and auto-zero method.The x-and y-coordinates represent the measured pixel positions in the horizontal and vertical directions, respectively.This 2D map is referred to as the sensing map.The scale bar on the right side of the sensing map shows the digital values converted by the 10-bit ADC.We can observe horizontal and vertical lines in the sensing map as shown in Fig. 11(a).These horizontal lines were caused by common power supply noise as the TFT currents in a row were sensed simultaneously, while the vertical lines were caused by the CI offset due to the connection of the CIs in the vertical direction.However, after using differential sensing and auto-zero method, the sensing map in Fig. 11(b) shows that most of the horizontal and vertical lines were no longer present.We can expect that compensation performance will be improved as more accurate current sensing is achievable.Fig. 11(c) shows the difference in sensed values before and after the use of differential sensing and auto-zero method.The difference was approximately 80 LSB as an ADC value, which represents a sensing error of about 300 mV in terms of the V SEN value.

B. Compensation Performance
Fig. 12(a) shows the sensing map before compensation.The center of the panel appears the darkest, indicating the lowest current levels.On the other hand, the lower-left area was the brightest, indicating the highest current levels.Fig. 12(b) shows the sensing map after compensation without differential sensing and auto-zero method.There are horizontal and vertical lines caused by power supply noise and amplifiers' offsets.Fig. 12(c) shows the sensing map after compensation using differential sensing and auto-zero method, demonstrating that almost the same current values were sensed at all pixels after compensation.Fig. 12(d) shows the histogram of the measured current levels before and after compensation.The obtained V TH and K values in the manufactured 13.7-inch panel were −0.28 V to 0.24 V and −3 % to +2 %, respectively.The current deviation before compensation ranged from −40.44 % to +29.69 %.After compensation without differential sensing and auto-zero method, the current deviation ranged from −6.83 % to +7.01 %.Although the mura of the panel has been significantly reduced, there was still a current deviation due to noise, as shown in Fig. 12(b).After compensation using differential sensing and auto-zero method, the current deviation ranged from −2.65 % to +2.16 %.
The reported threshold luminance value for a mura with an irregular shape was approximately 7.4 % [32].Thus, it can be confirmed that the compensation has been effective.Fig. 13 shows the measured waveforms of the control signals for sensing and V SEN .The SW1 to SW3 outputs of the timing controller were 1.8 V signals, while the other signals were the same as the simulation conditions.As shown in Fig. 13, the V SEN completed the sensing operation for different current conditions within 300 µs.This confirms that TFT characteristics can be compensated in real-time under the UHD resolution and 120Hz refresh rate conditions.
In a display with a 120Hz refresh rate, it is possible to sense even or odd pixels of one horizontal line every 1/120 second.Thus, in a WRGB pixel structure where four sub-pixels share one sensing line, it is possible to compensate for TFT characteristics in real-time every 144 seconds (1/120 × 2160 × 4 × 2) for OLED products with UHD resolution.Here, a refresh time of 1/120 s indicates the frame time at which the image is updated, while 2160 represents the number of row lines in the display.The number 4 refers to the number of sub-pixels used to create each pixel (WRGB), and 2 indicates the differential sensing (even or odd pixel).
Fig. 14(a) shows a photograph of the panel before compensation, and as previously discussed, the center of the panel had a dark mura.It is apparent that the sensing map shown in Fig. 12(a) and the panel photograph in Fig. 14( 14(c) shows a photograph of the panel after compensation using differential sensing and auto-zero method, which resulted in a significant reduction of the noise.The proposed method successfully compensated for TFT characteristics, as demonstrated by the luminance uniformity.We have demonstrated that the TFT characteristics in a 13.7-inch AMOLED display panel can be successfully compensated using external compensation circuits incorporating CI, along with the utilization of differential sensing and auto-zero methods.The proposed compensation approach effectively reduced sensing errors caused by amplifier offset and power supply noise, resulting in improved luminance uniformity and mura reduction, as shown in Fig. 14(c).
Table II summarizes the performance of the proposed external compensation compared to that proposed in previous studies on real-time TFT compensation.Previous studies took more than 300 µs for sensing time and only compensated for V TH without considering the change in K.In contrast, the proposed external compensation method can compensate for both V TH and K within 300 µs.Thus, the proposed method is superior to the previous studies in terms of both sensing time and compensation performance.

VI. CONCLUSION
In this paper, we have proposed a novel real-time external compensation method that uses CI to reduce sensing times and compensate for TFT characteristics.By utilizing two different sensed currents, V TH and K can be accurately calculated.A new technology to measure all feedback capacitors of all CIs has been proposed to improve sensing performance.By using differential sensing and auto-zero, we successfully reduce the sensing errors and improve the compensation performance.The performance of the proposed sensing and compensation approaches was verified using a 13.7-inch AMOLED panel.As the sensing can be completed within 300 µs, which is the vertical blank time for UHD resolution and 120 Hz refresh rate, the proposed method can compensate for TFT characteristics in real-time.Therefore, the proposed sensing and compensation method is an effective solution for high-resolution and large-sized OLED display applications.

Fig. 1 .
Fig. 1.(a) 3T1C OLED pixel and simplified block diagram of the conventional external sensing circuit.(b) Timing diagrams of the voltage waveforms for V TH sensing.

Fig. 1 (
b) shows the voltage change of the sensing line (V SEN ) and the timing diagram of the control signals for V TH sensing.The V TH sensing procedure consists of three steps: initialization, sensing, and sampling.During initialization, Scan, Sense, and SW1 signals are set to a high level.Then, T2 and T3 are turned on, causing the V GS of T1 to become V DATA − V INIT .During sensing, the Scan and Sense signals remain high, and SW1 signal is set to a low level.The gate node of T1 is fixed to V DATA , and its source node becomes floating.As the I DS of T1 charges the capacitive load of the sensing line (C LINE ), the source node voltage (V S ) increases.
Fig.2(a)shows the pixel structure and the novel sensing circuit proposed in this paper.The pixel structure is the 3T1C, which is the same as that of the conventional external compensation method.However, the sensing circuit includes an added component, a CI.The CI measures the T1 current in a much shorter time, as the current flows through the smaller C FB included in the CI.This results in a quicker measurement of the current than the conventional method where the ADC measures the voltage caused by the current flowing through the larger C LINE .As a result, the proposed method provides far shorter sensing times than the conventional method.In addition, it is not affected by differences in C LINE with panel size and resolution.Fig. 2(b) shows a circuit diagram of a CI amplifier.Fig. 2(c) shows the waveforms of the control signals and the voltage of the sensing line.The operation of the proposed method consists of three steps, similar to the conventional method: initialization, sensing, and sampling.During initialization, Scan, Sense,

Fig. 2 .
Fig. 2. (a) 3T1C OLED pixel and simplified block diagram of the proposed sensing circuit.(b) Circuit diagram of a CI amplifier.(c) Timing diagram of the voltage waveforms for the proposed method.

Fig. 3 .
Fig. 3. Schematic diagram of an algorithm to detect the value of C FB by applying a given current through the C FB .

Fig. 5 (
Fig. 5(a) shows the auto-zero circuit, which uses three switches and a capacitor (C AZ ) to store the offset.While SW2 is on, auto-zero operates to remove the offset and initialize V S of T1.Fig. 5(b) shows the waveforms of the switching signals during the auto-zero process, which is performed in two steps: offset sampling and offset zeroing.During offset sampling, SW2 and SW4 signals are set high to store the CI offset voltage in C AZ while SW5 is low.During offset zeroing, only SW4 and SW5 signals are changed to low and high levels, respectively.The offset voltage stored in C AZ is then added to V REF_CI .By applying V REF_CI + V OS to the non-inverting input of the CI, the offset voltage is effectively removed.

Fig. 6 .
Fig. 6.Simulation results of V SEN depending on three V TH conditions.

Fig. 7 .
Fig. 7. Current deviations depending on device variations before and after compensation.

Fig. 8 .
Fig. 8. Sensing error between before and after differential sensing method due to (a) V REF_CI noise and (b) i NOISE .

Fig. 9 .
Fig. 9. Simulated V SEN with V = ±20 mV (a) before and (b) after using auto-zero.(c) Timing diagram of control signals for the auto-zero.
) shows the control signals timing for the autozero operation.The offset voltage was set to ±20 mV to accommodate variations in the IC manufacturing process.The offset sampling step was set to 25 µs to ensure accurate sampling of the CI offset, as shown in Fig. 9(c).As shown in Fig. 9(a), the V SEN values varied depending on the CI offset with a difference of approximately 140 mV.Fig. 9(b) shows

Fig. 10 .
Fig. 10.(a) Photograph of the manufactured display system.(b) Micrograph of a fabricated IC.

Fig. 11 .
Fig. 11.(a) Sensing map before using differential sensing and auto-zero method.(b) Sensing map after using differential sensing and auto-zero method.Difference between (a) and (b).

Fig. 12 .
Fig. 12.(a) Sensing map before compensation.(b) Sensing map after compensation without differential sensing and auto-zero method.(c) Sensing map after compensation using differential sensing and auto-zero method.(d) Histogram of current levels for all TFTs in (a), (b), and (c).

Fig. 13 .
Fig. 13.The measured waveform of control signals for sensing and V SEN .

Fig. 14 .
Fig. 14.Photographs of OLED panel adopting (a) no compensation (b) compensation without differential sensing and auto-zero method (c) compensation using differential sensing and auto-zero method.

TABLE I CALCULATED
V TH AND K DEPENDING ON TRANSISTOR CONDITIONS