Rolling shutter > global shutter experimental solution at DVinfo.net

Matteo Ricchetti · August 6th, 2009, 10:13 AM

Hello I am new here I bought a Canon 500D a week ago , this is my first CMOS sensor movie mode experience, as I cannot do anything about the horrible hard binning, I tried to find at least a way to bypass the rolling shutter effect.
Yesterday I made this very first video test, video is here rolling shutter - LCD shutter experiment on Vimeo

There are still 2 problems to solve, by now the LCD shutter period is settled by hand, as the vertical synch of the video out is not synched with the scanning (I guess to have 59.94NTSC, and 50 PAL very tight timing to not introduce colour decoding artefacts from composite video) . The second is better glass, less light loss (of course having to work with polarised light at least one stop is lost). The test here is made whith old LCD shutter I used almost 20 years ago to make a interlaced switched glasses to view 3d video (in the forgotten analog video world).
If someone is interested I'll add more details... Matteo Ricchetti

Jon Fairhurst · August 6th, 2009, 11:21 AM

Your results look very good.

I assume that the image on the right uses the LCD shutter, but it is out of sync.

Here is a good experiment to run on the 500D (without the LCD shutter): Film something with a shutter speed between 1/40 and 1/60 and manually trigger a very fast photo flash many times.

In some cases the flash will occur within one frame. Count the number of lines over which the flash occurs. (For this thought experiment, assume that you are shooting 1080p and the flash duration is 700 lines.)

Now find another area where the flash covers two frames. (Assume that each frame shows 170 lines of flash.)

From this information you can calculate the rolling shutter time and the shutter exposure duration.

Time between frames = 360 lines = 700 - (170 + 170).
Virtual lines per frame = 1440 lines = 1080 (lines in a frame) + 360 (lines between frames)
Rolling shutter factor = 75% = 1080 / 1440.
Rolling shutter time = 25 ms = 1/30 seconds * 75%.
Actual shutter time = 16.2 ms = 1/61.7 = (700 / 1440) * 1/30 seconds

In the case of the 5D Mark II, the Rolling shutter time is indeed 25ms, and the shutter durations are reasonably accurate when in Manual Mode. In Auto Mode, the displayed shutter duration is not accurate.

I would be curious to know the rolling shutter time of the 500D - as well as the D90 and the GH1. Of course, you must use 1/20 seconds as the frame time when calculating the results for the 500D.

Toenis Liivamaegi · August 6th, 2009, 01:50 PM

This smells like a must have 4x4 LCD-shutter-filter to be used in a mattebox for action scenes a least...

we need a 4x4 inch (or smaller) basic electro optical polarizer (some 3D glasses use that same flickering tech)
a controller of some sort
and a knob or microswitch(es) to tune the sync or just a 25/30 setting

...last time I checked they, behind the Great Wall, wanted $65 for a piece if 150 is ordered and that's for a $650 Electra's Beach Cruiser bicycle ;).

T

Marcus Marchesseault · August 6th, 2009, 02:49 PM

Impressive results and I'm hard to impress! I don't mind rolling shutter effects so much on the 5Dii but it is nice to know there may be a solution if I ever need one. Keep up the good work.

August 6th, 2009, 10:13 AM	#1
Matteo Ricchetti Regular Crew Join Date: Nov 2004 Location: Italy Posts: 39	Rolling shutter > global shutter experimental solution Hello I am new here I bought a Canon 500D a week ago , this is my first CMOS sensor movie mode experience, as I cannot do anything about the horrible hard binning, I tried to find at least a way to bypass the rolling shutter effect. Yesterday I made this very first video test, video is here rolling shutter - LCD shutter experiment on Vimeo There are still 2 problems to solve, by now the LCD shutter period is settled by hand, as the vertical synch of the video out is not synched with the scanning (I guess to have 59.94NTSC, and 50 PAL very tight timing to not introduce colour decoding artefacts from composite video) . The second is better glass, less light loss (of course having to work with polarised light at least one stop is lost). The test here is made whith old LCD shutter I used almost 20 years ago to make a interlaced switched glasses to view 3d video (in the forgotten analog video world). If someone is interested I'll add more details... Matteo Ricchetti

August 6th, 2009, 11:21 AM	#2
Jon Fairhurst Inner Circle Join Date: May 2006 Location: Camas, WA, USA Posts: 5,513	Your results look very good. I assume that the image on the right uses the LCD shutter, but it is out of sync. Here is a good experiment to run on the 500D (without the LCD shutter): Film something with a shutter speed between 1/40 and 1/60 and manually trigger a very fast photo flash many times. In some cases the flash will occur within one frame. Count the number of lines over which the flash occurs. (For this thought experiment, assume that you are shooting 1080p and the flash duration is 700 lines.) Now find another area where the flash covers two frames. (Assume that each frame shows 170 lines of flash.) From this information you can calculate the rolling shutter time and the shutter exposure duration. Time between frames = 360 lines = 700 - (170 + 170). Virtual lines per frame = 1440 lines = 1080 (lines in a frame) + 360 (lines between frames) Rolling shutter factor = 75% = 1080 / 1440. Rolling shutter time = 25 ms = 1/30 seconds * 75%. Actual shutter time = 16.2 ms = 1/61.7 = (700 / 1440) * 1/30 seconds In the case of the 5D Mark II, the Rolling shutter time is indeed 25ms, and the shutter durations are reasonably accurate when in Manual Mode. In Auto Mode, the displayed shutter duration is not accurate. I would be curious to know the rolling shutter time of the 500D - as well as the D90 and the GH1. Of course, you must use 1/20 seconds as the frame time when calculating the results for the 500D. __________________ Jon Fairhurst

August 6th, 2009, 01:50 PM	#3
Toenis Liivamaegi Major Player Join Date: Nov 2005 Location: Tartu, Estonia Posts: 579	This smells like a must have 4x4 LCD-shutter-filter to be used in a mattebox for action scenes a least... we need a 4x4 inch (or smaller) basic electro optical polarizer (some 3D glasses use that same flickering tech) a controller of some sort and a knob or microswitch(es) to tune the sync or just a 25/30 setting ...last time I checked they, behind the Great Wall, wanted $65 for a piece if 150 is ordered and that's for a $650 Electra's Beach Cruiser bicycle ;). T

August 6th, 2009, 02:49 PM	#4
Marcus Marchesseault Trustee Join Date: Nov 2005 Location: Honolulu, HI Posts: 1,961	Impressive results and I'm hard to impress! I don't mind rolling shutter effects so much on the 5Dii but it is nice to know there may be a solution if I ever need one. Keep up the good work.