Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Stuttering image while Panning

who can have a look at this one?

who can have a look at this one?

I stumbled into this thread after testing a dolly this afternoon. Should I allready be shooting in 50 fps here? You can see the shots on: http://www.vimeo.com/7551202, password is "shutter".
The camera was 4K 16x9, Redcode 28, 25fps and the shutter was 50.
I checked my timeline in FCP but can't find a solution or major change here.
 
The new camera line will have a faster read out time which should match film cameras....
How it will work on high contrast scenes we will see when they arrive....
 
Just a thought to the red engineers.

Given, that rolling shutter is caused by the fact that each of the pixels are scanned one at a time like a TV….

When I first learned about the Red Mysterium chip and its ability to record at high frame rates with 4K resolution,
I somehow got the impression that the Mysterium had a separate A/D converter for every row of pixels and that’s how
It was able to convert the image so fast. So instead of one A/D converter for the whole chip, you have 2000 A/D converters
working in Parallel in order to scan and create the image..

Well, no…It works like a ‘normal’ CCD, It’s just really big and runs really fast.

However, If the chip did work this way, or even better have 4000 A/D converters scanning the columns down, then
It would ‘see’ the image pretty much the exact same way a film gate does, and no more rolling shutter…..

………Just a thought, Maybe that’s one of the new tricks they have coming……..
 
Gavin Greenwalt said:
Just out of curiosity does anyone even have a theory as to what *hypothetically* would cause RED to stutter more than film?
Click to expand...

i see two possibilities, but have no means to "test" either.

1. because of the top down nature of the electronic scan (eg "shutter") in RED, if a camera is making a pan, the top of the image would be read before the bottom. This could be verified easily, do a still frame from a fast pan with a strong vertical object. If the vertical bends left or right in the still frame, this is correct, otherwise that is not the problem. edit: i recall seeing examples of this with 1/1000 focal plane shutters.

2. this is possibly an encoder or bandwidth issue. Assuming a fast enough computer for playout, try several encoders and encoder settings. One clue here is that the total detail of the scene affects it.

Of course the very high end monitors now have hardware de-judder engines, although if the issue is #1 above it will not help much.

Please anyone with the facilities to test both hypotheses are urged to do so. Please send me a message with the results.

EDIT: As with using cheap lenses, The idea of writing a convolution chart for barrel distortion and chromatic abberation extends to compensating for sensor decoding artifacts. I think i mention a "computer per pixel" in some early 70's lectures. At the chip design level, simply adding a vertical clockd gate layer will allow readout of a "snap" of all pixels down into a serialized output layer.
________

new subject, someone asked if the video look can be achieved using film, i point you to 60 frames per second shooting and projection. I recall seeing a piece at the LUXOR when it first opened, at first gasp i actually thought it was live actors.
 
Last edited:
Judder

Judder

Dollys, steadycam and cranes were primarily invented to keep the action stationary in frame!

In the old days operators had a panning speed rule of thumb, I prefer a metric that can be used for all frame rates and if the camera itself is moving.

My rule is;
The subject (the part of the frame you want to viewer to focus on ie not the background) should not move through the frame in less than 5-7 seconds.

This applies for productions destined for viewing in a cinema at 24fps (flashed twice to make 48 fps) or 50i or 59.97i TV


Some examples,
A tree, occupying 1/2 height of the screen in landscape shot will judder if it moves through frame faster than than 5 seconds.
This is a problem if you want the viewer to look at the tree.

But if a car enters frame and you pan with the car, the car will not judder.


So a poorly framed panning shot of a foreground moving subject where the viewers eyes are not properly directed will have the viewer noticing the background rather than the relatively stationary foreground subject.

How dominant in shot do we need to have the car so we don't notice the tree?

It is the craft of DP and operator to figure this out.

Next time you are at the cinema look at the background during a pan of a foreground subject moving across frame and the background will probably judder.

In drama high contrast verticals ie white door frames, are the kind of backgrounds that are quite noticeable if the subject doesn't hold the viewers attention.


With the advent of HD came 35mm like resolution but an increase in 2/3 depth of field. This threw the background into focus more and made those white door frames (think soap and cheap crime drama sets) more defined and so the judder became more noticeable.


In respect to shutter speeds;
Can a higher shutter speed create more judder?
Yes. But it is dependent upon the degree of depth of field and makeup of the background (namely are high contrast edges or verticals present) as to the degree of judder that is present on playback.


How does 1/24th sec shutter effect judder?
One of the revelations with using 1/24th a second when progressive capture came along, was that much faster tracking shots at wide angle could be achieved, for instance low angle aerials. Landscape close to the camera became blurred (didn't judder) whereas distant landscape remains sharp. Similarly shooting say a car commercial with foreground trees whipping through frame can be a problem, using 1/24th sec shutter can alleviate this as the high contrast vertical line becomes more blurred and so does not judder so much.

The viewing medium can slightly reduce the effects of judder, ie LCD panels blur all movement to a greater degree than CRT.


Whatever judder testing you are doing don't get fixated on the background because with a good operator and DP the viewer shouldn't be....

Sony recently announced a prototype 240fps 3D camera. I figure 120fps x 2 imagers.

It is likely that they are working on a 120fps camera-transmission-display system to eliminate judder.



Mike Brennan
 
I read through this whole thread and was not too sure if the problem is not display related also in some cases.

If you are filming in 24fps and then watch it at 50Hz or 60Hz, there will be some extra duplicate frames thrown in. The reason why 120Hz and 240Hz displays are popular is that 120 and 240 can be scaled up from either 30fps or 24fps material with no extra frames thrown in.

So on my setup, I would make sure that all my displays, if I am not using a CRT, have native refresh rates of either 24Hz, 120Hz, or 240Hz. If my LCD panel can only operate at 60Hz refresh rate, it will also contribute to the judder problem.
 
There is also an active thread on the RED Discussion forum talking over solutions for stuttering images in a fast action pans. The thread started initially as a question about jello effect, then morphed into causes of stutter in fast pans and how to avoid it:

Link: http://205.234.135.241/forum/showthread.php?t=37663
 
Cüneyt Kaya said:
hey dont do pans on high contrast scences.

...not to shoot high contrast
Click to expand...

I've seen this issue too on high contrast screens for a music video when the performers are moving about on the high contrast white screens.

I would suggest trying to shoot green or blue screen then changing the background colour in post.

It seems to be the issue with the sensor. The sensors on the RED cameras are CMOS, which use a rolling shutter. A global shutter would be best on the CMOS and I'm hoping RED will look at creating their sensors this way. Film is always going to be the best, but sensor technology is soon catching up, but not yet...
 
Need ideas on a post production solution to judder

Need ideas on a post production solution to judder

I have read through this entire thread... seems to be the best place to explore this issue. However, I am seeking a solution on the post side for improving footage that has significant judder.

Here is how it was shot... jib arm, 43fps, 3K, 35mm lens, 23.98 time base.

Jib arm passes over table top subject, judder comes and goes as jib floats at different rates.

Can't fix the issue in production now... 43fps is an odd frame rate IMHO. I have no idea what the shutter angle was. Looking for a solution in post at this stage. I am exploring ideas now... anyone have a technique for minimizing this they would like to share?

Per Steve Gibby's reco on getting glass smooth footage, over-cranking up to 48fps variable, windowing down to 3K at 23.98 resulted in glass smooth moves for him. This is not what we are seeing at all.

Any productive ideas are most welcome.
 
Disco,

If the camera is scanning from one corner to the opposite it means that each pixel is active for a 1/50/the number of pixels of a second, which is an infintisemally brief period of time. Thios would lend itself to the sharpness etc of the image which would highlight the artifact in high contrast scenes so schaubergers workaround is the only real way to obviate the problem; ie if each pixel is active for the set exposure time not the exposure time devided by the number of pixels.
 
Hey guys

I want to revisit this thread because i think it is still an issue with Digital Camera, been a while i know. I was watching Transformers 3 Bluray version and i cant help to notice stuttering images thru out the whole movie, some inter cutting is so obvious that it was shot on 2 different cameras eventho the color is the same. But the characteristic of the panning simply gives away. I think the movie was shot with Alexa and Arri 35mm film camera if im not mistaken.

I want to ask those who have their EPIC now. How's the panning characteristic in EPIC over RED ? Does it still suffer the same issue ? Or better ?

Thanks
Rivai
 
RivaiC said:
Hey guys

I want to revisit this thread because i think it is still an issue with Digital Camera, been a while i know. I was watching Transformers 3 Bluray version and i cant help to notice stuttering images thru out the whole movie, some inter cutting is so obvious that it was shot on 2 different cameras eventho the color is the same. But the characteristic of the panning simply gives away. I think the movie was shot with Alexa and Arri 35mm film camera if im not mistaken.

I want to ask those who have their EPIC now. How's the panning characteristic in EPIC over RED ? Does it still suffer the same issue ? Or better ?

Thanks
Rivai
Click to expand...

It's a 24P issue, not a film vs digital issue.
 
Hi Deanan

I understand 24P will strobe. But i dont know how to tell you unless i showed it to you personally. The feel is different between film and digital. Try to take a look at transformers 3 in big plasma or cinema and you will see the differences between the intercutting. It is obvious that the characteristic of the camera is different. While i also try to spot the same thing in Final Harry Potter but i couldnt see any stuttering the movie despite all of the fast action movement. Later i found out it was shot on film.

I dont know man...to me, there's something in there that i'm still searching why it is so. Why some movie stutter like crazy while some is none...all shot for cinema released. I still couldnt accept that this is purely 24P issue.

Rivai
 
The issue is definitely caused by the shutter.

There is a difference between strobing and flickering.
All 24/25p 180° image systems strobe, but flickering is an artifact of the shutter.

In a electronic shutter the transition between open and closed is exact, it goes straight from closed to open to closed.
Think of this at the pixel level, the scan of the sensor array is irrelevant eg. rolling shutter or global shutter.
The effect of this is that the motion blur characteristics, shows an very exact and hard edge at both the heads and tails of the exposure time.
This hard edge causes the stuttering or flickering to be exaggerated.

In a classic rotating mirror shutter there is a transition time from closed to open to closed. This creates a fade in and fader out of the image in the heads and tails of the exposure time.
The effect of this is that the motion blur characteristics shows an soft edge at the heads and a soft edge at the tails.

You can se an approximation of the difference in these images from Tessive:

Bicycling%204k%20Still%20180sm.jpg


Bicycling%204k%20Still%20Tessivesm.jpg


Look closely at the ends of the bright trails in the background. the upper is a normal electronic shutter and the lower is with a ramped exposure from a tessive filter.

The solution to this is to change the sensitivity/exposure at the heads and tails of the exposure time. either adjusting the sensitivity of the sensor over time during the exposure, or introducing a filter in the optical path.

The Alexa studio and F65 has rotating shutter in the light path, and you can clearly observe that with the srotating shutter active you have the normal strobe, and with it disengaged you have the same strobe, but with the flickering.

Another solution is the Tessive filter, which is a polarizing attenuation filter synced with the electronic shutter of the camera. This creates an even better phase change than the mechanical shutter according to them.

Note that this issue is about the same on all current cameras with an electronic shutter: EPIC, Alexa EV, Sony F3.
DOF, Contrast/gammacurve and post sharpening will affect the phenomenon also.
 
Hi Anders

I think your explanation makes the most sense. I wont deny that film will strobe but in digital it seems more pronounced than it used to be. Sometimes during shooting, i dont need a big plasma to spot the stutter but 24" broadcast is good enough for me to tell if the stutter happens or not.

I notice that hard/soft edge from tessive demonstration earlier on, but i never think that it will contributes the stuttering effects even more. But i think i couldnt see it that much from its youtube resolution, im not sure if they provide higher resolution video so we can observe more carefully, i will check further on that.

Can you explain more about what u mean by :
The solution to this is to change the sensitivity/exposure at the heads and tails of the exposure time. either adjusting the sensitivity of the sensor over time during the exposure, or introducing a filter in the optical path.

How do u do this in the field ? It seems not a common practice tho...

tessive is interesting, good acceptance but nothing after that, i guess the price is still expensive

Rivai
 
RivaiC said:
Can you explain more about what u mean by :
The solution to this is to change the sensitivity/exposure at the heads and tails of the exposure time. either adjusting the sensitivity of the sensor over time during the exposure, or introducing a filter in the optical path.

How do u do this in the field ? It seems not a common practice tho...
Click to expand...

The solution is to eliminate the motion aliasing that the electronic shutter creates.
The current method is to start/stop integration by issuing a reset command, which creates a clean and exact start/stop, and has issues with delays because the chip is always integrating between photo site resets.
What "I propose" is that a future image sensor tech could change the sensitivity in a bell or sinus curve instead of the current square wave curve the current reset image integration method creates.

curves.jpg


Than means that the sensor would reset, change "analog sensitivity" from, for example, 0 iso to 800 iso and back to 0 iso until readout of that period in a perfect bell or sinus curve pattern.
This tech does not exist today....but would eliminate a lot of issues with temporal aliasing and readout/reset artifacts (rolling shutter)

Another way would be to approximate the curve by merging multiple intermediate exposures, but a way to controlling the "analog gain" would still be needed per intermediate exposure.
This would not be as good as a continuous exposure, but most likely good enough to eliminate most artifacts.
I don't think current sensors would make this possible as the image well will overflow at the same point regardless of analog gains.

A retrofit could be a gadget like the tessive filter which is an active filter to be mounted in front of the lens, creating a curve like variable exposure by using electronically controlled intensity filters (most likely polarizers) synced to the phase of the electronic shutter. This could of course also be integrated into a future camera...
 
Hi Anders

I view tessive samples in 1080P resolution and it is indeed smoother and better. The hard edge is less pronounced, smoother and more natural. Strobe is less noticeable (still there but much better).

I hope RED can solve this issue of its coming system, this problem (to me) exists even in holywood blokbuster movies. I am aware of 24P strobing but not this pronounced, it's very distracting at one point. It's quite easy to tell the differences...

I guess the solution for now is to depend on tessive. Good that they have rental system

Thanks for the share

Rivai
 
Thanks for all the kind words about our system. Our system definitely addresses strobing, which I'd define as aliasing from the 24Hz band. We still have some response at 24Hz, but it's very reduced compared with a normal square wave shutter. Please stop by our website and fill out the Contact Us form if you'd like a (free) Blu-ray sent to you with a better look at the footage.

The price for our system is $7900 (single units, quantity discounts apply). We have rental houses in LA and New York now. Also, feel free to contact us to arrange a demonstration, we love doing that.

Also, regarding the changing sensitivity of the sensor, that is one embodiment of our patent. Having the Tessive shutter built-in to a digital system is absolutely a goal of ours. At the moment, though, the Time Filter will work on existing cameras, which puts this in your hands immediately. Incidentally, the optimal waveform isn't a sine wave but looks a lot like one. It's a different mathematical function that's too boring to discuss here.

Tony Davis
CEO, Tessive
www.tessive.com
 
You must log in or register to reply here.
Back
Top