Motion blur illustrated

Motion blur illustrated.

Recently on the forum, the issue of motion blur is stirring a lot of attention by folks like me for example who are very sensitive to this. Since my old CRT FP died on me early this year, I bought and sold 3 different PJs (Sanyo LCD/JVC RS2/Marants 11S2) and I have yet to find a display tech that handles motion as well as my old PJ. Some people are still skeptical and believe the topic of motion blur is mute. Hence I decided to take hold of my SLR camera and devised a little test to proof that motion blur DO exist and is more or less apparent depending on the display technology being used. Before I tackle the issue of motion resolution, it is important to differentiate between 2 major factors responsible for this blur: Slow Response and SAH(Sample and Hold) effect.

Slow Response is usually caused by pixel not being fast enough to change state for every frame. LCD/LCOS are especially vulnerable to this problem. You can notice it with fast camera pans, fast action sports, and it is especially obvious when group of pixels of close tonal range are moving around the screen. Motion blur will introduce softness to the image and can hide subtle tonal changes in people skins and textures.
In theory, it shouldn’t show with PWM (Pulse width modulation) display since they change pixel state(on/off) almost instantaneously(in micro seconds). For instance, DLP is totally immune to this type of blur (as we will see in the next images). Plasma might leave some phosphor trails and hence some models tend to exhibit this flaw to certain extent.

SAH (Sample and Hold) effect is due to how the eye-brain interprets the image perceived. Since light falling in the eye retina is persistent and the brain integrates light intensity over a short period of time, a certain amount of blanking time is required between frames for the brain to easily differentiate sequential images and interpret movements. Not doing so, the brain will integrate and join 2 subsequent frames and this will be perceived as fuzziness in the image. LCD/LCOS blanking time is almost nil since pixel state is always on for the duration of the whole frame. Single chip DLPs have some blanking time but definitely not as long as 3 Chip DLP due to its color wheel that turns several round per frame and display each color twice per round. 3 Chip DLP blanking time is directly proportional to pixel brightness intensity, and knowing the average pixel IRE hover in the 30s, add gamma to the equation and we’ll get on average very long blanking times with off state mirror time. CRTs have relatively long blanking times as well (about 14ms with 60fps material assuming 2ms phosphor decay time)


The Test:

My test is easy to reproduce by anyone who is feeding his PJ with an HTPC running Windows OS and is curious to assess his display and his motion sensitivity. There is a XP screen saver named “marquee” that scroll a text horizontally. I went to “marquee” settings and filled the text box with “ININI”. I then started playing with scrolling speed trying to determine the text legibility speed threshold above which I couldn’t read the letters anymore. I performed this test with LCD/Single Chip DLP/LCOS/CRT and took screen shots with my camera at different shutter speed. For reference, I decided to attach 2 screen shots for each display tech I tested. One taken with a 1/340s shutter speed that test pixel response ability to redraw the text at different location for every frame. I added another with 1/60s shutter speed that mimic what my eyes perceive (I tried about 20 different shutter speeds from 1/20s ->1/500s and found image produced by 1/60s best reproduce what MY eyes actually sees).

Results:

(See attached Pics)
LCD: The word lost its readability with scroll bar @13 notches below highest speed settings. The letters became unreadable.
DLP: The word became unreadable only 7 notches below highest speed.
CRT: I was able to read the word at highest speed.

Results analysis:

*LCD suffer from slow response as seen in the blurry 1/320s shot. It also suffer from SAH effect since the shot taken with 1/60s shutter is worse then the one with 1/320s.
*Single chip DLP have excellent response time but word “ININI” becomes fuzzy at high scrolling speed (See Pic. 1/60s) again due to SAH effect.
*CRT performed best since I was able to read the word scrolling at the highest speed and letters were legible. DLP followed next, letters remained sharp but ghosting took form at high speed (SAH effect). With LCOS/LCD, letters had very blurry edges and ghosting was also seen. From memory since I don’t own the RS2 anymore, the LCOS performed better then LCD and started to become blurry at higher speed)


PS: I know that films are not text and all….. But surprisingly enough those test results correlate very well with how I rated motion handling when screening real film/movie materials on my past PJs. The test is fast, easy to produce, and illustrate well the issue at hand.

It will be interesting to get other people inputs and results. And Since I never owned a 3 Chip DLP, I am very curious to know how such tech can coop with motion since in theory (Rapid pixel response and long blanking time between frames) should give it a performance edge.

If they insert 1 black frame with lcd/lcos, you ll address 1 problem and create 3. You ll somehow get less SAH since blanking time between frames is increased.

ON the other hand, you ll loose:

1)Half brightness
2)Half contrast (Since during black frame, lamp will remain on and thus u ll have half the available brightness for same amount of leackage....)
3)LCD response time is 2 slow for 60hz FPS, let alone 120Hz. Don't forget the manufacturers advertise on/off response time, a number much faster then grey/grey response. LCD will struggle with low changes in intensity since little voltage is applied on the panel.

A better solution with lamp model is to insert a "shutter wheel" with half disk transparent, and half black turning 60Fps. U still get half brightness but contrast won't suffer. Also, since DLP have perfect transition time (as seen in 1/320s Shutter speed), you ll have much better results then with LCD/LCOS

True, But since CRT can't display 24 without flicker, You need to run it at multiple this frame rate (pefereably 72Hz) and u ll get a 72fps scorlling bar . Now I have another program named (juddertest), you can google and the program, you get to choose the speed of the scroll(Nb of pixel per refresh rate) and number of Refresh Rate per frame. That should be enough to perform this comparision for 24fps material. I did the comparision andthe results somehow close to what you get with 60fps. ie: CRT>DLP>LCOS>LCD.