Measuring rolling shutter: put a number on this issue!

That looks really impressive. Without doing any measurements, I'd guess 10ms or so. Eagerly awaiting a controlled test clip.
 
In prep for an upcoming VFX job I did a series of rolling shutter measurements today with my new GH5s... here are the results:

C4K 29.97 11.3msec
C4K 23.98 11.3msec

UHD 59.94 12.0msec
UHD 29.97 11.8msec
UHD 23.98 11.9msec

FHD 29.97 11.5msec

I use the rolling shutter analysis process in the motion tracking software, Syntheyes, which does a 66 pass iterative measurement method;

https://www.youtube.com/watch?v=LgWscUOse_c

I suspect all the UHD times are actually the same, but I haven't had the chance to run multiple video clips to take an average. Of course there are a few things I didn't get a chance to test; only the flavors that I might have to shoot.

Interestingly C4K is the fastest so far... perhaps because it requires the least amount of processing from the sensor? Anyway, an improvement over the GH5, and impressive for UHD/C4K!

Regards
 
I've seen it, and yes, I plan to use it to take the measurements. Hopefully it will work. I have to download it first, and my usual app for that is not working currently, YT changed something and now I have to wait for them to update the app.
 
Thanks Samuel. 24.3ms isn't too bad considering the full 6K downscale.

I do wish that test had included the "full frame" 30p mode (which is actually 1.2x crop, scanning 5K instead of 6K), as that's my "base" framerate for most things. Hopefully it's a good bit faster than 24/25p.

8.7ms in 1080p is great, I just hope the resolution is good in those modes. It's pretty awful in the a6300/6500.
 
I'm not really interested in the 4k 30fps mode, but I'd also like to see some 1080p tests, because softness there is what kept me away from the a6500: I do shoot slow motion often, and I don't mind cutting some 1080p shots with 4k footage but they have to be sharp 1080p shots.
 
With most cameras, it always gets better with more frames per second or knocking down your system to a lower resolution.

So 4K/30p should be better than 4K/24p, but may not be better than 1080/24p. Some cameras may be pretty even all across the board.
 
When that happens, it's because you're using less information in order to create the final image. It's either cropping, skipping lines, or switching to a lower-bit-depth ADC path. But depending on the shot it may be worth it.
 
Just added the Pentax K-1 to the list. It's not a camera I would consider, but somebody was interested enough to shot the test clip and send it to me, so, here you go:

Pentax k1 1080p24 - 15.2 ms (15.9-15.7-13.9)
 
Might want to add to the chart the result that DPReview saw when they tested rolling shutter on the A7III at 4k/30p: 16.7ms (!) with 1.2x crop. (Standard crop at that resolution.)
 
Yes, they also tested 4k/24p, where it matched your score of 25ms. Seems the slight crop pays big dividends in RS reduction.
 
I get lots of questions about this (well, not so many, but over the years they pile up) so I updated the opening post with a detailed description of how I calculate these numbers :)
 
You must log in or register to reply here.
Back
Top