Measuring rolling shutter: put a number on this issue!

This is something very personal, and even project-dependent, ans the amount of camera motion will determine what is enough and what is not. There must be something on the list that you've used and can relate to...

Personally I've had complaints with cameras that I rated at >25ms, and I'm currently very happy in this respect (and many others) with my 16ms p4k.

---------------

General note: now that I've moved to resolve and I'm not paying for Adobe CC, I don't have the tools I used for measuring this anymore (I have standalone licences for LR but not for PS), so, I've received a couple of clips more but haven't been able to measure them yet. I told those people that at some point I'd borrow a laptop with Adobe CC just to make these measurements, but I'm super busy and can't see that happening in the near term (and yes, if there was a camera I was interested in, I'd make up the time to do that, but currently I'm not longing for any new cameras so I'm paying a lot less attention to this part of the world).
 
WOW, very similar to the p4k but with a bigger sensor with more resolution. Impressive!

(I rounded to single-digit because, even if your numbers are super accurate, the rest are not, and I don't want people to get a misleading sense of accuracy in the comparisons)
 
I'm pleased and grateful that BlackMagic Design through CaptainHook is consistently releasing the rates for their cameras, and that fact alone may seal the deal in me picking up a Pocket PK 6k for fun and experimentation. It's something I wish all manufactures would do. It's great to be able to just drop those numbers in for operations like motion tracking without having to figure it out yourself.

Yes, .26ms doesn't seem like much, but let me put it in a real world VFX context; Imagine doing a constant speed 360 degree pan, one that takes a full minute to complete. I often do that to build an environment map on location (not that I've EVER forgot a still camera and my fish-eye and am forced to fake it with the video camera... ever :)). The bottom line is a tracking point at the start of the pan will no longer line up with the same image feature at the end of the 360 degree pan even with rolling shutter compensation unless you have the rolling shutter measurement amount just right. So being able to put in 16.26ms vs 16ms is a real world useful thing.

Think of the frame rate/sound sync issue of 30fps vs 29.97 or 24 vs 23.976. Sure, you visually can't see the difference of 30fps vs 29.97, but try and sync up sound and see how quick that "tiny" difference is noticeable...

Just saying there's a place for precision, and I appreciate it when it's given to me. :)
 
There you go: my answer would have been "nobody", but dvxusers never disappoint, that's the one person whou coul see a 0.26ms difference XD
 
Jim Arthurs said:
I'm pleased and grateful that BlackMagic Design through CaptainHook is consistently releasing the rates for their cameras, and that fact alone may seal the deal in me picking up a Pocket PK 6k for fun and experimentation. It's something I wish all manufactures would do.
Click to expand...

Samuel maintains a good resource here that i think is useful and since i have access to the numbers it's relatively easy for me to post them and try contribute in some way. Glad some find it helpful, thanks!
 
And I thank your for that... particularly since I moved away from Adobe CC and don't have the tools to measure these things accurately anymore, which means I'm only adding the BlackMagic numbers to the table, thanks to you. As long as Sony keeps on bringing 8-bit cameras to the <$3k space, that's all I need, actually :)
Can I say again how happy I am with my p4k? I'm VERY happy with my p4k :) :) :) :)
 
Added, thanks!

Can't trust the numbers too much but they'll be in the general ballpark. They measure it exactly like I do... but I use pixels instead of a ruler, and I take two inter-frame measures for every intra-frame measure, and I repeat everything at least two or three times and then average. Or I did, because now that I don't have Adobe CC anymore I can't run tests myself unless I borrow a laptop from a friend, which I said I'd do months ago for clips people have sent me but it never happened. The fact that I'm happy with my p4k and not searching for my next camera plays a big role here.

If anybody wants to take over and keep on taking measures, I can send you the excel and have a quick skype explaining what to measure and how. You need Photoshop.
 
You can't pull frames with any editor and then measure the pixel shift in gimp? I don't understand your process, so guessing here.
 
I can use Resolve and any photoshop clone that has a pixel-mesuring ruler tool. LR standalone (which is what I use for stills now) doesn't have that. Installing GIMP right now... :)

edit: ok, it works, thanks a lot! the UI is not high-dpi friendly and every time I paste a frame I have to remember to "finish the paste" by turning it into a new layer, which I forget half the time, but free definitely beats $240/year :cheesy:
 
Last edited:
Added to the table:

a7III 4k S35 25p -- 16.6 ms (16.6)
RX100 VI 4k ------- 17.4 ms (17.4)
GX9 4k 25p -------- 24.3 ms (24.0-24.7)
Z7 4k FF ----------- 29.5 ms (30.0-29.0)
X-H1 4k 24p ------- 30.9 ms (30.4-31.3)
 
Can we add the following:

ZCAM E2: 4K DCI 25fps ZLOG = 11.1ms (from Cinema5D)
ZCAM E2C: 4K DCI 25fps ZLOG = 36ms (from Cinema5D)
ZCAM E2C: 4K DCI 29.98fps ZLOG = 31.3ms (from Cinema5D)
Panasonic S1: UHD 4K 25p FF VLOG = 22ms (from Cinema5D)
Panasonic S1: UHD 4K 25p APS-C VLOG = 10.4ms (from Cinema5D)
Panasonic S1: 1080p 25p APS-C VLOG = 7.4ms (from Cinema5D)
Panasonic S1H: 6K 3:2 23.98fps FF = 29.7ms (from Cinema5D)
Panasonic S1H: 5.9K 16:9 23.98fps FF = 24.8ms (from Cinema5D)
Panasonic S1H: 5.9K 16:9 29.98fps FF = 21ms (from Cinema5D)
Panasonic S1H: 5.9K 16:9 25fps FF = 22ms (from Cinema5D)
Panasonic S1H: 4K DCI 23.98fps FF = 24.2ms (from Cinema5D)
Panasonic S1H: 4K DCI 23.98fps APS-C = 13.9ms (from Cinema5D)
Panasonic S1H: 4K DCI 59.94ps APS-C = 13.9ms (from Cinema5D)
Panasonic S1H: 4K DCI 25fps APS-C = 10.4ms (from Cinema5D)
Olympus OMD EM1 MArk II: 4K DCI 23.98fps = 10ms (from Cinema5D)
 
That Olympus is impressive.

And just a note, Affinity Photo is really nice and not very expensive. Been using it for a little while because it can process raw files from the Fuji cameras.
 
I'm not sure about the cinema5D numbers. Just one year ago their methodology was awful and their results very badly skewed: they said 9ms on the XT3 at 4k 25fps, where I measured >20ms. I consider my methodology to be good, not just because I put an adequate amount of thought into it, but also because when BM started posting official numbers here, they closely matched my measurements. More recently they may have switched to the flickering light methodology, which should be way more accurate than their old thing, and for example their p6k number closely matches the one BM posted here, so, I would be fine with those... but I need to see another proven success before I add here all the numbers that come from their new methodology.
 
Oh Sam, we were just getting into good RS numbers, but now everything is ff35 and Large format!
 
Hi,

any news on the Fuji XT4 in terms of rolling shutter?
Would be awesome to know!

Also might be good to add Arri Alexa Classic and Amira / Mini numbers. As a standard reference.
I believe Alexa is 6 ms and Amira is 2 ms?
But those were Cinema 5D's measurements, and I read a few posts above their numbers are quite optimistic...
 
You must log in or register to reply here.
Back
Top