1. Introduction

Extension of the digital TV market put recently a lot of pressure on the temporal behavior of flat panel displays. Indeed, HDTV requires almost 60 Hz working frequency and Liquid Crystal Displays are not intrinsically very rapid devices. There are two problems that affect the response time of LCDs. One is the nematic liquid crystal’s slow response to an external field; the other is the driving method. Numerous efforts have been made recently to improve the time response performances of LCDs. Nevertheless, the response time measurement itself is not straight forward especially when inter-gray levels with low differences are explored. In addition new driving strategies like overdriving leads to complex temporal behavior for the display emission and great care must be taken not only on the measurement itself but also on the measurement analysis.

When addressing motion artifacts on flat panel displays, the first thing typically considered is motion blur. Full understanding of the human visual system with regard to motion performance evaluation is complex. The new VESA FPDM section addressing this problem introduces moving edge-blur that can be measured by various instruments like pursuit cameras or fixed optical detectors[1].  This evaluation is generally tedious and expensive due to the cost of the instruments and their complexity. It results in moving picture response time behaviors versus gray levels generally measured  in very strict moving configurations.

In the present paper, our purpose is to relate standard gray to gray response time to motion artifacts using new Eldim Optiscope SA instrument. This system manages standard gray to gray response time measurements with a maximum of accuracy and is flexible in terms of optical configuration. One additional interest is that the experiment temporal behavior can be simulated using a regression approach and theoretical temporal behaviors. This approach is useful to extract accurately all the required parameters even for complex temporal behavior and offers the possibility to model the temporal behavior from any level to any level very rapidly. It is then also possible to calculated moving edge-blur to deduce moving picture response time or even moving of complex patterns in more realistic configurations.

Fig: 1 Photograph of the OPTIScope-SA system

 2. Description of the system

A photograph of the Optiscope SA instrument is reported in figure 1. The light coming from the screen is collected by an aperture limited objective. It is separated in two parts using a beam splitter. One part of the light is focused on a color CMOS sensor to get the video image of the object. The other part is directed to a photo-multiplier tube and the signal is sampled by a fully programmable acquisition card. A photo peak filter before the PM tube allows luminance measurements after a proper calibration. The dark noise is automatically corrected using a shutter. The angular aperture measured by the PM tube is fixed at ±1° following the recommendations of the FPDM VESA standards[2]. The imaging zone is about 3 times larger than the measurement zone on the surface of the display. All the optics, detection and electronics are integrated in the measurement head. A 4Mb buffer memory ensures fast and reliable acquisition with 14 bits sampling. Sensor setting and data transfer are realized via USB bus. The response time is then automatically computed by Windows friendly software. Low pass filter can be applied. The numerical procedure follows the FPDM VESA standards for this kind of measurement.

 3. Modeling of gray to gray response temporal behavior

For LCDs, the liquid crystal is always sandwiched between two polarizers and the optical response of each cell is due to the optical transmittance variation versus time. In simplified cases, it is possible to express the liquid crystal director reorientation with time. If φ is the phase shift due to the LC cell, a reduction of electric field leads to[3]:

Φ_m is the maximum phase shift and τ_d is the decay time characteristic of the cell which depends on the rotational viscosity of the LC and its dielectric anisotropy. In the same way, the φ dependence for an increase of electric field can be expressed as:

Φ₀ and Φ_∞ and the phase shifts before the electric field increase and after the reorientation. τ_r is the rising time characteristic of the cell. The time dependent normalized intensity can be calculated in each case by the following relationship:

One example of rising and falling shape is illustrated in the figure 2. In practice, the experimental rising or falling shapes are normalized and fitted adjusting τ_r or τ_d parameters and the time origin of the reorientation. In the case of overdriving, the same theoretical model is applied using a first rising shape characterized by τ₁ with a multiplicative parameter characteristic of the overdrive value followed, after a time interval ΔT, by a decay characterized by a τ₂ parameter. The same type of model is also applied to falling edges with under-driving.

Fig. 2. Response time measurement using computer controlled LED source

The response time measurement repeatability of the measurement system has been evaluated using a programmable LED source and applying a rising or falling signal of 4ms. The sampling rate is fixed at 100 kHz. The measurement has been repeated 60 times are the corresponding histograms are reported in figure 3. The repeatability is excellent (less than 0.1%) and should not depend on the response time value in the range 1-500ms adjusting the sampling frequency.

Fig. 3. Dispersion of response times on 60 different measurements. The standard deviation is ±0.009ms for the rising time and ±0.006ms for the falling time.

 4. Standard gray to gray response time measurement

Fig. 4. Standard grey to grey response time measurement on a LCD display: the measurement is made between 21 levels with iso-luminance variations

For testing purpose, we have realized a standard gray to gray response time measurement on a LCD display using the OPTISCOPE SA instrument and its ±1° angular aperture optics. The working distance was about 30cm and the mean measurement spot size of about Ø 10mm.   We have chosen to change the luminance level with 21 iso-luminance steps between black to white state. The advantage is that we really measured the temporal behavior versus the luminance levels which are directly related to the human eye sensation. The LCD show a simple temporal behavior without overdriving and the response time can be measured very accurately. All the results are summarized in figure 4.

 5. Local gray to gray response time measurement using moving edge

Then we have decided to measure the temporal behavior of the same display using gray level bars moving horizontally at a given speed. If the Optiscope SA is used with its standard optics, the measurement spot size is always quite large compared to the pixel size and the recorded temporal behavior is always and average than cannot be used easily to reconstruct the edge blur. On the contrary, if we adapt a microscope objective on the Optiscope SA with a medium magnification, the recorded temporal behavior address only about 1 pixel as shown in figure 5 and then the edge blur can be deduced easily. 

Fig. 5. Measurement zone of the Optiscope SA using a x20 microscope objective: the zone integrated by the photo multiplier is represented by the circle.

Fig. 6. Local grey to grey response time measurement on a LCD display using 120 pixel width moving bars at 8 pixels/frame: the measurement is made between the same 21 levels.

For comparison, we have made the experiment exactly with the same luminance levels as for the standard gray to gray response time but using now horizontal moving bars of on luminance level on a background of the other level. Temporal behaviors are analyzed exactly in the same way as for the standard measurement and the results are shown in figure 6 and 7. We see immediately that the results, even if there are noisier due to the lower efficiency of the optical collection, are exactly comparable to the previous ones. It shows that the temporal behavior of one pixel of the display is not dependant on its neighbors. So it is easy to predict that any moving artifact can be simulated using the conventional gray to gray response time measurement if the driving and moving conditions are clearly defined.

Fig. 7. Comparison of standard and local rising and falling time on the same display: the values are from dark level to all the other luminance levels.

5. Conclusions

In the proposed paper, we will present how to calculate easily and rapidly any moving artifacts using conventional gray to gray response time measurement. The method uses simple mathematical models for simulated all the possible temporal behavior of one pixel and simple assumptions to deduce brightness profiles of blurred edges. Results will be compared to conventional moving response time measurements performed on the same displays.