Samsung NX1: Here is the rolling shutter measurement

M

Mattoid

Active member
I conducted this measurement several months ago with the help of users on dpreview. I'm posting it here now to make it official. Feel free to use the measurement in your chart of readout speeds Samuel H.

User Usee filmed the test. The shot was taken on a tripod and involved smooth consistent speed panning from right to left, with vertical lines extending from the top to the bottom of the image. In other words: perfect for an accurate measurement.

User hootsmon transcoded the video and uploaded it here: https://drive.google.com/file/d/0BzSCb-RC8txOZTQtVlUyTGw1b0E/view?pli=1

The first image shows a single frame in which the skew is measured.
The second image shows the same frame but with a cutout from the following frame overlayed in the bottom right hand corner. This shows the location of the slanted line one frame later. As you can see, in the cutout the line has the same slant showing that the panning speed is consistent.

NX1 rolling shutter frame 1 skew measurement.jpg

NX1 rolling shutter measurement.jpg

The green line shows the amount of skew of the vertical line across the entire height of the frame. The orange line shows the distance traveled by the line in one frame. The actual measurement in mm or whatever doesn't matter, It is the ratio between these two lengths that we are looking for. So I zoomed as closely as possible to get the greatest accuracy. I found that the ratio is 0.75 or 75%

Since we know that the frame rate is 24fps, we know that the time in milliseconds represented by the orange line is 1000/24 which is 41.666. We know that the green line is 75% of the orange line so the time represented by the green line is 41.666 * 0.75 which is 31.25. So this is the approximate readout time in milliseconds for DCI 4k mode.

Its worth noting because both DCI 4k and UHD modes use the entire sensor width, that UHD actually uses more of the sensor vertically. 1.0666 times as much. The readout time in UHD mode will therefore be 1.066 times as much. Which is 33.33 milliseconds.

So
DCI 4k = 31.25ms readout
UHD = 33.33ms readout.

Usee did provide a 1080p clip but I was away when he posted it and the link is no longer valid.
However, this video compares 1080p rolling shutter to the A7s and also NX1 4k. 1080p is good. 4k is bad: https://www.youtube.com/watch?v=fxWPHkxDE5k
 
I'll have a look at this tonight. From my experience running these numbers, no matter how good the setting seems to be, it is important to take at least two measures, from a left-to-right and a right-to-left pan. I say this because I would expect RS in UHD to be equal or better to RS in 4K mode.
 
Like this?

nx1 second rolling shutter measurement comp.jpg
(I uploaded this at 1920 wide but DVXuser downscaled it to 1000)

With this one I measured a ratio of 0.74 which would give a measurement of 30.83ms for 4k (32.88 for UHD)

So there are slight variations, but its definitely in this ball park.


As for UHD having equal or better RS, I think that's an understandable misconception.

After all, both 4k and UHD have a height of 2160, but 4k has a width of 4096 vs 3840 for USD.
So DCI 4k should be wider right? I mean, It's wider in the GH4.
But the GH4 uses a native crop of the sensor.

The NX1 doesn't use a crop. With UHD it uses the entire sensor width. So how can 4k be wider?
Without the sensor doing an impression of a vehicle in an 80s cartoon, it cant. It can only get wider by getting shorter, therfore using less sensor height which obviously takes less time to read out.

However, thinking about it, people are very unlikely to be watching 4k on a true DCI 4k monitor, It will just be letterboxed. So viewing 4k in most cases will have the same magnification as viewing UHD, so the shorter readout time wont translate into a perceptual decrease in rolling shutter skew.

So having said that, I think the easiest thing is just to say the NX1 has a readout of 33ms in 4k. 4k meaning both 4k and UHD.
 
Yes, like that. I'll download the files tonight and run the numbers. But you're absolutely correct: if UHD uses the total width of the sensor, and so does 4K, then UHD is a taller crop than 4K, so more lines of photosites have to be read (6480x3417 instead of 6480x3645), and UHD should be slower (in fact, by exactly the amount that you measured: it should go from 33.33ms to 31.2469ms).

I'd be shooting 4K all he time, and then cropping in post to slightly reframe. I see no downside in doing that.

My list of *IF* conditions is still in place, but I'm more and more into this little wonder toy.
 
Last edited:
So, I downloaded the 4K video and ran my usual measurements: 30.6, 31.6 and 31.4ms, for an average of 31.2ms in 4K mode. Which is impressively close to what you got :)

That puts the NX1 at the bottom of the table, basically tied with my a7S in FF mode.
 
geffyb sent me all the tests: 4K, UHD, and 1080p. Thanks, mate!!

These roughly confirmed the measures we already had for 4K and UHD (30.9ms and 32.6ms), and also that the 1080p mode is a lot faster. Like A LOT faster. It's actually the fastest camera in the chart, and by a very wide margin: 7.9ms. Amazing.


I think right now, if you can't get an FS7, this is the camera to get for green screen, and also for fast action shots. If they add a gamma mode that matches the DR of my a7S+slog2, it's the all-round winner.
 
Yeah, thanks geffyb for corroborating the results, and providing the measurement for 1080p. It was clear that it was fast but a 7.9ms reading really puts it into perspective.
 
eddavid said:
what is the disadvantage of shooting 1080p vs 4k for the NX1 - I need to test this asap.
Click to expand...

- Worse low light, I wouldn't use 1080p past ISO 800 unless for documentary work. It doesn't get noisy, it gets BLOTCHY

- Detail is not the best (not as clean as say a C100), but very usable, on par with a Sony A6000.
 
slashcam.de to the rescue:
http://www.slashcam.de/artikel/Test/Samsung-NX1---Codec--Dynamik-und-Farben-in-der-Praxis-.html
http://www.slashcam.de/artikel/Test...-4K-Konkurrent---Aus-dem-Messlabor.html#Aus_d

This is NX1 in UHD then downsampled to 1080p:
1303-Samsung_4K-Samsung_NX1_ISO340_4K.jpg


This is NX1 in 1080p:
1303-SamsungFUllHD2450-FullHD_100p.jpg


This is a6000:
sony_a6000.jpg


This is 7D2:
1282-ISO340-Canon7DMk2-ISO340_HDX.jpg



BTW the slashcam.de tests point to UHD being quite a bit sharper than the already-very-very-sharp 4K.
 
geffyb said:
- Worse low light, I wouldn't use 1080p past ISO 800 unless for documentary work. It doesn't get noisy, it gets BLOTCHY

- Detail is not the best (not as clean as say a C100), but very usable, on par with a Sony A6000.
Click to expand...

Shot in 800 asa. Don't really notice worse low light.
Don't notice the blotchiness. Can you send me examples?

I think in my real world test 1080p is not that bad - may take the digital edge off of the camera. And I rather have better rolling shutter than super super high res images.
 
ita149 said:
New firmware 1.21 available.
Someone on DPforum says the rolling shutter has been improved in 4k.
Click to expand...
Darn Samsung!
It could be all these tests are rendered useless!
Since the 1080p RS is so good, I have my fingers crossed they improved 4K.
 
I just tested with firmware v 1.21 - uhd still looks very bad - rolling shutter badness/jello galore
 
Thands to geffyb, I ran the numbers with the new firmware, and found no significant improvement over the old one. One could want to see one, but a statistician would complain. Details here.
 
You must log in or register to reply here.
Back
Top