Samsung should add a 2:35 crop mode on the NX1 to fix rolling shutter

[Micah Mahaffey]

Someone should message Samsung and ask them to implement a 2:35:0 crop mode where it only records the image in that ratio. allowing the Camera to have better and more controlled rolling shutter results. Its the same with how the full 4k mode has less rolling shutter over UHD because its not scanning as much of the sensor do to a slight 1:85 crop to the top and bottom. Now image now much better it would be in 2:35?  Plus that aspect ratio works perfectly for a lot of filmmakers who'd be cropping their videos anyway.  I don't know how to submit suggestions to Samsung on future firmware features but I definitely think this would be a welcomed idea. 

[Guest Ebrahim Saadawi]

The sensor readout speed is a core specification of the sensor design, I don't thing cropping the top and bottom would allow then to make it go faster. The only way they seen to be able to go faster is by using the entire sensor and lineskipping/binning to 1080p, not reading the 7K sensor. It's a compromise wel'll have to take with a camera that reads a 7K sensor and downsamples it to 4K at a conaumer product. Perhaps a 1:1 crop of 2.5x could potentially reduce rolling shutter as it will only read a 4k area not a 7K one. But then again it would be a 2.5/2.6 crop, more like s16 and Pocket BM than large super 35 APS-C, whoch I'd take anyday over a small sensor. 

 

I suggest one uses the 4k mode but switch to the almost-global-shutter 1080p mode for fasting panning. You can also correvt rolling shutter to a degree in post. I've shot extensively with similarly bad RS cameras and never had a viewer notive (first Nilon D90 and a7s). 

 

In short, just get one it's awesome unless you spcifically need to shoot fast movenent at  4K, if it s the case' GH4 is nuch mode  For anything else it's a perfect little cmaera with one of my nost favourite aesthecs

[Micah Mahaffey]

Why is it that the full 4k  4096 x 2160 mode has less rolling shutter than the UHD mode? People keep saying its because its not scanning as much of the sensor do to the 1.85 crop. 

[MattH]

The readout time is less but the apparent skew is not.  You in effect create your proposed mode just by cropping in post.  You dont see the start of the scan and you dont see the end. So the image that is still visible took les time to scan.

So the readout time is less.  But you dont watch the wider image on a wider screen, so the magnification is the same as if you hadn't bothered cropping, and so the apparent skew is the same.

This is in contrast to the aps-c crop of the A7s which does result in a higher magnification and so the apparent skew is less.

[Geoff CB]

I completely disagree with MattH & Ebrahim Saadawi. 

If you have to read less lines of the sensor to produce a frame, the rolling shutter will be better. Period.

As for the feature, it would be a welcome one. The setting should be in industry standard DCI scope format: 4096×1716

 

EDIT: I am extremely mistaken in this assumption, apologies to MattH.

[Brian Luce]

Doesn't 1080p in the nx1 have very little rolling shutter? 

Moire is the one the bugs me the most...

 

 

[Geoff CB]

Doesn't 1080p in the nx1 have very little rolling shutter? 

Moire is the one the bugs me the most...

 

 

​There is moire on the NX1, but very little. Better than the D800, A7, A7r, A7 II, 7D, and 5D Mark 2. It's there but minimal, and there are no color artifacts (rainbow pattern) like the horrid moire from the A7 and D800.

[MattH]

I completely disagree with MattH & Ebrahim Saadawi. 

If you have to read less lines of the sensor to produce a frame, the rolling shutter will be better. Period.

As for the feature, it would be a welcome one. The setting should be in industry standard DCI scope format: 4096×1716

​

You are aware that adding the word ‘period’ on the end of a statement doesn’t make it true, right? 

Lets not take our communication style from fox news anchors. 

I notice you don’t offer any reasoning to support your statement.

Nevertheless I will bite and provide further reasoning for mine:

 

A sensor scans from top to bottom (if you discount lens flip) so vertical movements are not effected by rolling shutter effect but horizontal movements are.

From this we can understand that the vertical crop will have no effect on rolling shutter, but horizontal crop will.  Which I will support with this example:

 

Lets say that we have a sensor with a 40 millisecond readout time when in 16 by 9 mode.

We set the camera to 25 frames per second which gives a frame time of 40ms. So the read time is the same as the frame time.

 

If we then point the camera at a pole and pan the camera at a speed such that the pole travels 9/16ths of a frame width in 40ms, then the pole will be skewed at a 45 degree angle.

 

If we take a screen grab of that, then cut of the top quarter of the image and the bottom quarter of the image, giving a 16 by 4.5 ratio image, we will be left with an image that only took 20ms to scan.

But will the angle of the pole be any less?

 

The answer is of course, no.  The pole will look exactly the same, we will just be seeing half of it.   So a movement of a specific frame width ratio will produce the same amount of rolling shutter skew regardless of the vertical crop.

 

We already agree that a horizontal crop and vertical crop combined (as in the crop mode of the a7s) will reduce rolling shutter skew, so I don’t need to explain how that works, but I will if anyone is interested.

[Micah Mahaffey]

That doesn't explain why the 4k DCI mode has better rolling shutter performance over UHD..

[Andrew Reid]

I think you need to reduce horizontal resolution to make the lines roll faster off the sensor.

Cropping the sensor vertically would have same effect on rolling shutter as doing the crop in post.

In the end, a line of 4K is 3840 or 4096.... that's a lot of pixels compared to 1920.

[MattH]

That doesn't explain why the 4k DCI mode has better rolling shutter performance over UHD..

​It doesn't have better rolling shutter performance, It has a marginally faster readout because it uses less of the sensor vertically, but that doesn't translate to a better performance for the reason I give above.  I was the one who worked out that it had a faster readout (in another forum) by the way .  Until I explained it people thought the DCI 4k mode would have the same or longer readout, mistakenly thinking that because 4096 is bigger than 3840 it must take longer.  But then I realised it doesn't make a difference anyway, unless you crop in horizontally as well and use a wider lens to give you an equivalent angle of view with a faster readout.

[MattH]

I think you need to reduce horizontal resolution to make the lines roll faster off the sensor.

Cropping the sensor vertically would have same effect on rolling shutter as doing the crop in post.

In the end, a line of 4K is 3840 or 4096.... that's a lot of pixels compared to 1920.

​If the camera does a full pixel read out, it's the native resolution of the active sensor that makes the difference in readout speed.  The resolution that it downscales to doesn't matter.

 

 

[hmcindie]

I completely disagree with MattH & Ebrahim Saadawi. 

If you have to read less lines of the sensor to produce a frame, the rolling shutter will be better. Period.

As for the feature, it would be a welcome one. The setting should be in industry standard DCI scope format: 4096×1716

​Well I agree with MattH, you seem to not understand how sensors work. The scanning speed is pretty much the same, no matter how much you do cropping from top and bottom. For example, the 5d in cropmode with magic lantern has the same scanning speed as in fullframe. The area is just much smaller so that same speed is enough for the smaller area.

Even if you start the scan from the middle of the screen, the speed of the actual scan line-by-line will be the exact same so in the end, cropping in post will do the same exact thing.

[Geoff CB]

Hmcindie & MattH & Andrew

You all are very much correct, apologies for the tech assumption on my part.

MattH,

Yes I did sound a bit to much like a Fox News correspondent there..yikes. :/

[Marco Tecno]

Perhaps a dumb question. Could rs be diminished via sw optimization? Or is it a pure hw issue, so not possible to improve via, say, fw update?

[Micah Mahaffey]

What about having a 2.5k filming mode? Not only would we have better rolling shutter but then we might even get up to 60fps in 2.5k resolution! I think this would be a good add on as well.

[Frozone]

I think you need to reduce horizontal resolution to make the lines roll faster off the sensor.

Cropping the sensor vertically would have same effect on rolling shutter as doing the crop in post.

In the end, a line of 4K is 3840 or 4096.... that's a lot of pixels compared to 1920.

​No.

3840 x 1920 = 7372800

3839 x 1920 = 7370880

3840 x 1919 = 7368960

If you remove 1 pixel from vertical you remove 3840 pixel, but when you remove 1 pixel from horizontal you only remove 1920 pixel.

[Micah Mahaffey]

What would be better is a 2.5k shooting mode! 

[MattH]

What would be better is a 2.5k shooting mode! 

​At this point I think it makes sense to either have hd or qhd output files, because 2.5k would necessitate that you either upscale or downscale no matter what.  If you wanted to end in 4k then It would obviously be better just to have 4k rather than upscale.

As for rolling shutter etc, I think it depends more on how the sensor is scanning the image and downscaling rather than the output resolution.  So if the sensor scanned every other line and then downscalled in camera to 1080.  I think that would be good.    Low compression 4:2:2 1080 at 60p with low aliasing and managable rolling shutter in h.264 sounds perfect to me.

Having said that, having a 2.5k mode wouldn't hurt.

[Nikkor]

​No.

3840 x 1920 = 7372800

3839 x 1920 = 7370880

3840 x 1919 = 7368960

If you remove 1 pixel from vertical you remove 3840 pixel, but when you remove 1 pixel from horizontal you only remove 1920 pixel.

​The sensor reads lines (not columns). The rollingshutter skew comes from the difference in readouts between lines, so the difference lies between consecutive lines, so it doesn't matter if you skip lines on the top and on the bottom, the difference between consecutive lines will still be the same.