Theoretical and experimental comparison on the 240 Hz LCD TVs based on pure ME/MC FRC and 120 Hz TV with Back Light Scanning Stanley H. Chan, and Truong Nguyen

Abstract

Most of the videos available in the market are 60 frame per second (fps). In order to display a 240Hz video stream, one has to increase the frame rate by 4 times, and display it using a 240Hz LCD. There are two major methods to increase frame rate. The first one is to use motion compensation. In this method, a 60 fps is increased to 240 fps by inserting 3 additional frame between every two consecutive frames. This method is costly, but temporal resolution is increased by 4 times. The second method is to first increase frame rate to 120 fps. Then we divide each frame into three horizontal bands and display them alternatingly. As shown in Fig. 1, we display sequentially the top and bottom band from frame 1, middle band from frame 1, top and bottom band in frame 2, middle band from frame 2, and so on. Clearly, the frame rate is doubled because one original frame is divided into two frames. In this project, we want to study and compare the blur associated with these two methods. Which one can reduce motion blur more? What are the blur widths?

Publications

• Stanley H. Chan and Truong Q. Nguyen, "Comparisons of 240Hz LCDs," submitted to IEEE International Conference on Acoustics, Speech and Signal Processing ICASSP 2010.

Results

1. LCD Impulse Response

   The LCD impulse response is a composition of scanning system and response of the luminance element. The type of the scanning system determines the period of the driving signal. For example, since the electron gun in CRT performs a raster scan at 60Hz, so each phosphor is illuminated once every 1/60 second. The illumination period is very short because as soon as the electron gun leaves the phosphor, the phosphor does not illuminate. Therefore the driving signal can be modeled as a short pulse. But for LCD (hold type in particular), the liquid crystal stays ON for the entire 1/60 seconds. Therefore, the driving signal for LCD has to be a long pulse with period 1/60 second. Recently, there are scanning types LCD so in this case the sample-hold time will be shorter than the hold-type ones.

   The response of the luminance element determines how fast can the device responds. By observing typical impulse response from experiments, an exponential decay characteristic is used to model the transient behavior of the liquid crystal when a signal falls from 1 to 0. In this model, the tunable parameter is the decay time constant $\alpha$. For better LCD panels, the liquid crystal responds faster and hence the decay time constant is larger. In the idea case where the device responds immediately (e.g, CRT), then the time constant is infinite.

   

2. Analysis

   To analyse the blur caused by the two systems, we consider an image with a white strip moving at a speed of v_x. The displayed signal is essentially the weighted average over a few previous frames, with the weights defined by the impulse response of the LCD. In the figures below, the input signals for the two different methods are shown. Note that in the equation below, $x$ is chosen to be a pixel at the far right.

   

   The input and output relationship of the LCD response is essentially the convolution. In the real computation, the convolution is both time and spatial domain. However we only discuss the time domain convolution and assume $x$ to be a pixel at the far right. Now with the input signal shown above, and the transfer function provided, we can compute the output signal as follows.

   For Method 1, the out signal is the convolution of input and the transfer function. Therefore, we have the following equations. The figures on the right illustrates the effect of different $alpha$ and different $v_x$. As see, the rise of the signal is stable, and the peak is unity.

   

   Similarly, Method 2 is also calculated. However, as shown in the figures on the right, the intensity is oscillating at different time instants. Also, the peak never reaches unity.

   

3. Results

   There are a few observations that one should pay attention to in the figures below. First, the plots below are intensity versus time. So the width of the transient period is NOT the blur width. However, we believe that the transient time has a close relation with the transient displacement. In other words, the longer the transient time it takes, the wider the blur width it should be. This will be justified in a coming paper. Second, there is oscillation in Method 2 where as there is no oscillation in Method 1. This oscillation is caused by the continuously black data insertion of the method. A consequence of oscillation is that the displayed image will have blinking, and it can become one of the factors to cause blur. Third, the oscillation also causes attenuation in intensity, and hence the observed image will look dimmer. As dimmer images are often perceived more blurry than brighter images, Method 2 will be perceived more blurry.

4. Discussion

   There are also a few LCD TV operating at 120Hz with back light scanning. In this case, the period of the driving signal is still 1/240 second because the black light switching removes half of the 1/120 second period. However, A 120Hz LCD TV should have liquid crystals with only 120Hz response time (for otherwise one should have driven it using 240Hz signal!). Thus the exponential decay function should fall slower than a 240Hz liquid crystal. In this sense a 120Hz LCD with back light scanning cannot be referred as 240Hz as the response time of the liquid crystal is still 120Hz.

Conclusion

1. Method 2 is behind Method 1 due to oscillations and intensity attenuation
2. 120 Hz LCD TV with Back Light Scanning should not be referred as 240Hz LCD TV to clarify the difference of the systems