Ryan Patrick O'Hara
Veteran
I posted over in EX-1 about rolling shutter reduction. See what they say.
If they disagree with me, they are liars!! lol.
If they disagree with me, they are liars!! lol.
The end of 35mm adapers and hv20s....
- Thread starter Hunter Hampton
- Start date
Clip #2
Clip #2
http://www.vimeo.com/1643205
Please keep in mind, the color rendition of this clip does not represent D90's ability to reproduce color.
It may be interesting to see how the format handles people and objects moving around in various directions.
Clip #2
http://www.vimeo.com/1643205
Please keep in mind, the color rendition of this clip does not represent D90's ability to reproduce color.
It may be interesting to see how the format handles people and objects moving around in various directions.
Ryan Patrick O'Hara
Veteran
That is some foggy glass! You warned us! haha.
I was watching carefully at the end, when the camera quickly resets, and those telephone polls did not bend at all. A good sign at best. Don't abandon hope, rawfa.
PS: That was a bunch of people! Wow.
I was watching carefully at the end, when the camera quickly resets, and those telephone polls did not bend at all. A good sign at best. Don't abandon hope, rawfa.
PS: That was a bunch of people! Wow.
Ryan Patrick O'Hara
Veteran
He is shooting through a really dirty thick piece of glass.
Literally the window to the outside.
Literally the window to the outside.
Last edited:
Clip #2
Clip #2
Yes, I shot it through window and it was already dirty and dark color tinted.
Even though there was a rather clear window above this one, I had to fix the camera to avoid hand shake. Hand shaking was really bad. I can tell this from looking at LCD monitor. Also, I was told the area I was trying to shoot did not allow me to set tripod.
I was able to use tripod on previous shot because it was inside Nikon event booth.
So, instead of shooting shaky shot with clear view, I chose to shoot fixed shot with not so clear view.
I had to pick one and that was my decision. Now I think of it, I could have shot both but again, SLR camera was very intimidating.
Clip #2
Yes, I shot it through window and it was already dirty and dark color tinted.
Even though there was a rather clear window above this one, I had to fix the camera to avoid hand shake. Hand shaking was really bad. I can tell this from looking at LCD monitor. Also, I was told the area I was trying to shoot did not allow me to set tripod.
I was able to use tripod on previous shot because it was inside Nikon event booth.
So, instead of shooting shaky shot with clear view, I chose to shoot fixed shot with not so clear view.
I had to pick one and that was my decision. Now I think of it, I could have shot both but again, SLR camera was very intimidating.
I'll give it a try. It will be very technical, don't know if it will be understandable. Just ask then.
First I want to correct an earlier statement.
I should have put it like this instead:bearing said:
The scan rate is fixed, the only thing you can change is the exposure time. Exposure time is defined as the time difference between the resetting pointer and the reading pointer. Put in other terms you can say that shutter speed is fixed, only shutter angle is variable.
The sensor is read like a data matrix, starting from the top left corner, going to the right, pixel by pixel in a fixed rate, say 80 MHz.
When the first line is completely read some invisible pixels are read, then the reading continues on the next line. If the amount of horizontal pixels are 1280 then that would have taken 1280/80M seconds which is 16 microseconds. Since there are a number of invisible pixels, say 320, there are 1600 total pixels to be read per line, taking 20 us per line.
An amount of time before pixels are read their charge is reset. That is done in a similar way to how they are read. One at a time their charge is zeroed in the same rate as reading, that is 80 MHz in my example.
If the exposure time is to be 1/500 of a second (which is 2000 microseconds) then the reset of a pixel has to happen 2000 us before reading it. Since one line takes 20 us then a pixel is reset 100 lines before it is being read.
So, if the picture is 720 lines and a slowly moving "band" of 100 of them are exposed at the same time, then we can clearly see how when panning a vertical line becomes angled.
If the exposure time is longer, say 1/48 of a second, then the whole picture is exposed at the same time for the most part, but the start and the end of it is moving in the same slow speed, creating the same effect. Picture is also blurred to a greater extent.
Heres an example of a 16x6 pixel CMOS-sensor taking a picture of a grey wall, just to show how the resetting, reading and exposure is working.
Code:
Its dark, pixels are zero
0000000000000000
0000000000000000
0000000000000000
0000000000000000
0000000000000000
0000000000000000
Turning on lights, a gray wall.
Turning on camera.
0000000011111111 Camera is starting to reset pixels, 8 so far
1111111111111111 All pixels are taking charge from the light
1111111111111111
1111111111111111
1111111111111111
1111111111111111
1111111100000000 Continuing to reset pixels, 16 so far
2222222222222222 The eight first that were reset are
2222222222222222 starting to recharge
2222222222222222
2222222222222222
2222222222222222
3333333322222222 Continuing to reset pixels, 32 so far
1111111100000000
4444444444444444
4444444444444444
4444444444444444
4444444444444444
Starting to read, 32 pixels behind the resetting pointer.
Red data is in the right matrix.
7777777766666666 4444444444444444 Continuing to reset pixels, 64 so far
5555555544444444 4444444444444444 Reading pixels, 32 so far.
3333333322222222 0000000000000000
1111111100000000 0000000000000000
8888888888888888 0000000000000000
8888888888888888 0000000000000000
9999999999999999 4444444444444444 Continuing to reset pixels, 96 so far,
9999999988888888 4444444444444444 resetting is finished.
7777777766666666 4444444444444444 Still reading pixels, 64 so far.
5555555544444444 4444444444444444
3333333322222222 0000000000000000
1111111100000000 0000000000000000
Final picture will be:
4444444444444444
4444444444444444
4444444444444444
4444444444444444
4444444444444444
4444444444444444Ryan Patrick O'Hara
Veteran
So you are saying the rolling shutter is always rolling at it's fastest speed? Changing shutter angle is just exposing the frame for longer/shorter. Does that simply just add or subtract cycles?
May I then theorize that the shutter speed can never go faster then the constant speed of the rolling shutter?
I am kind of confused, but I think I know where you are coming from.
Thank you for taking the time to explain!
Focusing
Focusing
Hi Iceviking,
It was prime 80mm lens.
Focusing to me was not very easy.
I have some photographer friends and I have seen them checking if the pictures they took were focused or not by enlarging them. So, I was assuming I could not trust LCD screen much.
LCD screen is nice and beautiful but it was somehow different from Video camera LCD screen. And maybe that is only me. I don't know. It could have been peaking or something.
Focusing
Hi Iceviking,
It was prime 80mm lens.
Focusing to me was not very easy.
I have some photographer friends and I have seen them checking if the pictures they took were focused or not by enlarging them. So, I was assuming I could not trust LCD screen much.
LCD screen is nice and beautiful but it was somehow different from Video camera LCD screen. And maybe that is only me. I don't know. It could have been peaking or something.
Yes, always the fastest speed. And, yes, exposure time is the difference in cycles between the reset and the read events. Adding cycles is like widening the shutter angle.
That is correct, I think that such fast exposure times are not useful int practice though. If the difference between reset and write is only one line then the exposure time would be 1/50'000 of a second in my example.
Thanks, no deal, it was kind of fun.
Hey Bearing,
I'm thinking focal length has something to do with the RS effect? Will a longer lens (more magnification) magnify the problem. There was an old school basic rule of thumb that says if your shooting with a 200mm lens that has 4 times magnification.... It also has 4 times the vibration(movement) as a 50mm lens and you would set your shutter speed accordingly.
I'll have to look again at the bike footage.... Seems Sugoroku's lens was longer?
Thanks Sugoroku for the uploads!
frisco
I'm thinking focal length has something to do with the RS effect? Will a longer lens (more magnification) magnify the problem. There was an old school basic rule of thumb that says if your shooting with a 200mm lens that has 4 times magnification.... It also has 4 times the vibration(movement) as a 50mm lens and you would set your shutter speed accordingly.
I'll have to look again at the bike footage.... Seems Sugoroku's lens was longer?
Thanks Sugoroku for the uploads!
frisco
Yes, it has. That's because when you rotate a long lens a certain angle, objects will move a larger portion of the frame than if you would have used a short lens. But that is just comparable if objects are at the same distance from the lenses.
If you want to compare effects of panning in different clips then you should look at how many pixels objects are moved between two adjacent frames (given that clips have the same resolution). If objects are moved the same amount of pixels between adjacent frames in two different clips then rolling shutter effects should be the same no matter what focal length was used.
dfreerider
Well-known member
New movie footage from a production model D90 available for download at http://www.dpreview.com/news/0809/08090101nikond90samplegallery.asp.
Lee Wilson
Veteran
I am confused ! maybe my eyes are finally failing me and I need to get to the doctors pretty quick ! ;p
. . . . but these clips simply look no where near as good as any of the current crop of decent HDV cameras (Canon/Sony etc).
After some experimentation in photoshop I will change my charge of 'it looks like it has been run through a median filter' - to - 'it looks like it has been run through the dust and scratches filter - set to 1 pixel'. If you take a decent piece of focused footage you can get very close to the D90 look with this setting.
I am suprised at some of the claims being made for the D90s clarity/focus from these clips !
I am also seeing aliasing problems, particularly in the vertical plane (?)
Yeah, you're right. While I'm still pretty stoked about the D90, the PQ just isn't up to par yet. Hopefully it'll get there in the next couple of years with new cameras.
It's still a useful product though, with the 35mm DoF and low light skillz... I'm certain we'll see some cool shorts shot on it within long. But for 'serious' videography I don't think it's there yet. But hell, you get a kickass DSLR and the camcorder part is basically free.
EDIT: Forgot to add that with some post work it'll look even better (color grading, downsampling to SD resolution and perhaps a bit of grain etc.).
Lee Wilson
Veteran
I was thinking the same thing, this might produce a decent SD image after a little bit of post work.
I still hold out hope that with some on camera settings there is a good to decent picture to be had - so far examples have been shot in less than ideal circumstances - poorly (even incorrectly) compressed - with the exposure leaping around and so on. (I think this is partly due to the video mode being mainly tested by photographers who take only a passing interest in the video side?).
The stuff up on dpreview.com is very jerky, the focus is good (although the 'facet' problem is still there) and the exposure is doing it's epileptic thing - so again less than ideal to make a judgement from.
I won't give up hope until I see footage that is locked off, perfectly focused with a good lens and with the exposure button held in (if that is what it takes) and in the original M-Jpeg file.
I still think it is a great product, the next 6-12 months should be interesting.
Last edited:
Changing the shutter speed is going to change the look of your project - and no Gladiator wasn't cool... that effect is completely lame. I bet increasing the iso speed would have the same effect.