Difference between panning and Dollying with Ex1?

K

kubalsky

Active member
There's been a lot of talk lately of vertical skew with the Ex1 on fast pans and skew wobble with fast back and forth pans. Im just wondering then why, look through some of my fast dolly shots- well, handheld shots out a car window at about 60-100kph- dont I see any skewing of poles, buildings etc? Looks fine to me. Does skew increase when camera is pivoting on a fixed point (panning) as opposed to when camera moves past thin vertical objects on a track (dollying)?

If anyone could clarify the difference between dolly and panning in terms of the skew effect i'd greatly apreciate it. Excuse me if this is a dumb question. Im new to the world of live film making, coming from a stop-motion animation background where this wasn't an issue.

Attached are four screen grabs that take place within 3 seconds shot out a car window traveling at about 80kph. Vertical poles etc dont appear to skew at all to me. Why not?

Kubalsky
 

Attachments

  • noskew1.jpg
    noskew1.jpg
    99.9 KB · Views: 0
  • noskew2.jpg
    noskew2.jpg
    103.7 KB · Views: 0
  • noskew3.jpg
    noskew3.jpg
    106 KB · Views: 0
  • noskew4.jpg
    noskew4.jpg
    93.5 KB · Views: 0
Skew increases under two factors: the slower the frame rate, and the faster the relative motion of the object.

So, 24p = more skew than 60p or 60i. If you've been shooting at 60i or 60p, you won't see as much skewing.

Second, it depends on the relative speed of the object. If you want to see the wobbly skew, zoom in all the way and pan past some vertical objects, and then reverse your direction. Really waggle the camera back and forth and you'll see the rubbery effect. Once you've seen it, you'll have educated your eye: you can then get a feel for how much relative motion it takes to get noticeable skew in an image. Try zooming in on the UPC barcode on a product pacakge and waggle the camera back and forth across it, in 24p mode. That'll educate you as to how the skew works.

The wider your lens, the less skew you'll see -- because the relative motion of the objects in the frame will be that much slower. Example: if you're at full telephoto, moving the camera just a couple of inches will probably get you from one side of the barcode to the other, right? Now zoom out, and you'll have to move the camera a foot or so to get the barcode from one side to the other. The smaller the object in the frame the less relative motion it'll exhibit (i.e., motion relative to the sensor -- it doesn't matter if it's the camera moving or the object moving, what matters is if it moves according to what the sensor is seeing).
 
I see. Frame rate and relative motion are the key factors. So fixed point pivot panning or moving dolly at the same speed are equal in their effect to the skew equation.
Thanks for the clear explanation Barry. This forum is awesome.
Kubalsky
 
Last edited:
Barry_Green said:
Skew increases under two factors: the slower the frame rate, and the faster the relative motion of the object.

So, 24p = more skew than 60p or 60i. If you've been shooting at 60i or 60p, you won't see as much skewing.
Click to expand...

Barry,

This is not true on the EX1.

The skew will always be the same whatever your frame rate, as long as your playback frame rate is the same as your recording frame rate.

The EX1 scans at 1/60 second no matter what the frame rate, so skewing can be seen as a function of real world motion over time without regard to frame rate.

Kubalsky,

Think of skew as a 2D phenomenon which occurs on the sensor, without regard to how the image forms on the sensor. A vertical line moving across frame can be created through panning, dollying, even zooming, subject motion or some combination of those.

If the vertical line moves horizontally on the sensor at a given number of pixels per second (or pixels per frame), it will skew the same amount. With a wide lens you need to pan faster to make the vertical move across frame the same speed as a telephoto lens. And it is much easier to move objects across the frame quickly by panning rather than dollying.

That's why the effect seems to be related to panning and not dollying, and seems to be lens dependent. But it applies to all movement across the sensor equally.
 
Last edited:
Everything Barry said is correct - I was just offering another way of visualizing it - except for one thing: that skew is dependent on frame rate.

In fact, on the EX1, skew is the same at all frame rates.





"So, 24p = more skew than 60p or 60i. If you've been shooting at 60i or 60p, you won't see as much skewing."

That's the part that's wrong.
 
Last edited:
So skew is the same at all frame rates? That does seem backed up by some other tests on some other forums. The Ex scans at about 1/60 in all modes it seems. They say the Red scans at 1/72.
I had hoped that shooting 720 60p I might show less skew... but come to think of it I dont think Im ever going to do fast pans or anything that will show up this effect.
Its interesting to be learning all this stuff though, gotta know our cameras strengths and weaknesses.

Right now, somewhere in the world there's an Ex owner reading this forum, sitting in tears, his planned documentary on telephone polls ruined...just ruined.

Noel,
Im still in Melbourne. That footage was from November. Ill be in Tokyo Feb 2nd.
See you then.

Kubalsky
 
matthew77 said:
....

"So, 24p = more skew than 60p or 60i. If you've been shooting at 60i or 60p, you won't see as much skewing."

That's the part that's wrong.
Click to expand...


Hmmm... for some reason that just doesn't sound right since for a constant pan speed each frame of the faster frame rate is capturing less horizontal movement.

Can you please elaborate a bit more on your position?
 
ESTEBEVERDE,

It's not my "position" but a fact that one can easily derive from testing the camera, as I have.

The sensor scans the image for a duration of 1/60 second. This is the same at all frame rates. For slower frame rates it waits longer between the scans. If you pan, say, 2° in 1/60 second, then no matter what frame rate you shoot at, the skew will be the same.

The overall position change per frame will be different at different frame rates, as expected, but not the skew.
 
Last edited:
Kubalsky,

The skew is there, but it will take a really fast pan (or dolly) and a short shutter speed (1/2000 of a second) to see it. A long shutter speed, like 1/120 of a second or longer causes a fuzzy edge, which makes it harder to spot the skew.

Assuming my calculations are correct, if an object takes 3 seconds to move from one edge of the screen to the other, the skew would be 0.57 degrees. It doesn't matter if it's a pan or dolly, as long as the object makes it from one edge (left or right) to the other in 3 seconds.

Increase the time to 6 seconds and the skew would be 0.28 degrees. So in that case, it's so small, it's going to be hard to see.

If you want to see the skew, you must set the shutter to 1/2000 of a second and do a VERY fast pan, like 1/2 a second for an object to move from one edge of the screen to the other. If you look at a single frame you will see a vertical object leans by around 3.4 degrees.


There's still more research going into this, but this is what I know so far: The rolling shutter appears to take 1/60 of a second (really 1/59.95) to move from top to bottom on the CMOS chip. It appears that for all modes the time to go from top to bottom is still the same 1/60 of a second.

A 1/2000 of a second shutter is 32 lines high and moves down from top to bottom at 1/60 of a second.


Bob Diaz
 
Bob,

Right. There really is no shutter, but a virtual shutter created by a time gap between reseting and reading the chip.

The chip resets and clears all of the data from the pixels and then reads the data, 32 lines later.

Each of these functions takes 1/60 second to get from top to bottom of the chip (bottom to top, really, as the image is upside down) and they are initiated 1/2000 second apart.

And this cycle is repeated at the frame rate that the camera is set on.
 
Last edited:
Bob,
Thanks for joining the thread! One question as a person new to digital, Is the mechanism that drives this 1/60 second scan something that is built into the camera as hardware or something that could be adjusted down the track with a software update?
 
If the CMOS sensor could of been sampled faster, Sony would of implemented it.
Same thing with the RED ONE camera's CMOS. The reset time is close to the EX1.
The RED maybe a bit faster, but not by much.
 
Not really ;)
I'm praying that somehow there could be an upgrade, but i'm 99.99% sure it can not be done.
 
matthew77 said:
Bob,

Right. There really is no shutter, but a virtual shutter created by a time gap between reseting and reading the chip.

The chip resets and clears all of the data from the pixels and then reads the data, 32 lines later.

Each of these functions takes 1/60 second to get from top to bottom of the chip (bottom to top, really, as the image is upside down) and they are initiated 1/2000 second apart.

And this cycle is repeated at the frame rate that the camera is set on.
Click to expand...

So if this is true -- reading the chip can not take longer than 1/60th of a second -- how can you get a 1/48th shutter speed which is a common shutter speed for 24p shooting? Other info I've read has suggested a 1/48th shutter speed can be set. What is happening in that case?
 
how can you get a 1/48th shutter speed which is a common shutter speed for 24p shooting?
Click to expand...

The reset takes place and then, 1/48 second later, the read takes place.

In the mean time, between the read and the reset, photons are accumulating ("integrating") on the sensor (for a total of 1/48 second on any given line). There's a brief period when neither resetting nor reading is happening anywhere on the chip - where the whole chip is open to the accumulation of photons*. That is to say that the "window" between resetting and reading the chip, that scans downward, is taller than the chip.

That's how the theory goes.


---
*this is the moment when it is safe to fire a flash in front of the camera, as the photons will fall on the entire chip at once, after it has been reset, and all of those photons will be read by the read operation which follows.

Flashing during read during read or reset operations will cause the dreaded "partial exposure."
 
Last edited:
You must log in or register to reply here.
Back
Top