LUMIX s1 - WIRD JIDDER in vertical lines (not moire)

[Lukas Kalinowski]

Hi Andrew hi Guys!

I got new lumix S1 and I was testing the vlog. Once I recorded some clips I start playing around and I discover something like jidder. in attached video Clip A is ok, Clip B is wird. Please look closely on the "windows frames" in the right corner in the video at the "B" clip (its not about the moire). Look at the frame of windows,  there is something like movement back and forth. Clips A&B are two different clips with the same settings(60p. 1/125, VLOG, 640 iso, f5.6, canon 70-200, at70mmwith stabiliser on) using sigma mc21 clips was recorded one after another,. The wird jidder shows up in specific areas on vertical lines but only in specific place not everywhere in the frame. Do You have any idea what is going on ? firmware in this unit is 1.2. 

Where I can report this to panasonic to let them know that this is fucked up ? and what more this is random thing once its there next time there is nothing like in clip "A".

Here are the raws clips from camera and short clips exported from resolve with proper exposure hat I attached below > https://we.tl/t-4jUgNVyQPa

after discovered this I recorded again 20 more clips and there is nothing besides moire, and I'm confused because not sure is my unit is failed or this is normal behaviour. Thanks for your times guys! 

 

[andjo]

It's there in both P1000119 and P1000120. It's less apparent in P1000119 because it's slighty out of focus but as soon as it locks (at 0:06 mark) it gets the same effect on the window frame.

Maybe someone got a better explanation but I would still call that moire. It has a tendency to make things "dance" when they are in a sweet spot (because of the pixel shift that takes place). Sadly this is an artifact that comes with the S1 in the aps-c mode.

In general thin lines and no anti-aliasing does make it difficult for the sensor to interpret as to which pixel to bind it to.

But you're saying the new clips does not have this issue?

[andjo]

Actually I take that back.

Now that I look at the trees they look very weird too, I didn't notice that at first. Like they are ghosting/smearing or something - almost like when there's data loss in a codec, like a frame only partially transfers over to the next.

[androidlad]

You left out a key piece of info - they were shot in 1080p60.

Non-crop 1080p on all DSLR/mirrorless cameras is derived from pixel-binning, or line-skipping or a combination of both.

The artifacts you are seeing is moire and aliasing as a result from pixel-binning and line-skipping, this is very common and completely normal, it's particularly severe on S1 due to the lack of an OLPF.

[Otago]

16 hours ago, andjo said:

It's there in both P1000119 and P1000120. It's less apparent in P1000119 because it's slighty out of focus but as soon as it locks (at 0:06 mark) it gets the same effect on the window frame.

Maybe someone got a better explanation but I would still call that moire. It has a tendency to make things "dance" when they are in a sweet spot (because of the pixel shift that takes place). Sadly this is an artifact that comes with the S1 in the aps-c mode.

In general thin lines and no anti-aliasing does make it difficult for the sensor to interpret as to which pixel to bind it to.

But you're saying the new clips does not have this issue?

I agree, it's present in both of them when it's in focus. 

I think the solution is to shoot 4K, which is downsampled from the 6K and won't show any of those artefacts ( if it's been done well and there seems to be no evidence it's not ) and then downsample again to 1080 when you are editing. You could also try a softening filter as a form of OLPF if 1080 in camera is necessary. 

15 hours ago, androidlad said:

You left out a key piece of info - they were shot in 1080p60.

Non-crop 1080p on all DSLR/mirrorless cameras is derived from pixel-binning, or line-skipping or a combination of both.

The artifacts you are seeing is moire and aliasing as a result from pixel-binning and line-skipping, this is very common and completely normal, it's particularly severe on S1 due to the lack of an OLPF.

Can you tell from it being inconsistent between the horizontal and vertical what they are doing with line skipping vs binning ? Do they have to use what the sensor provides or can customers make their own version by changing the microcode on the chip ? Do you know if the sensor control unit is addressable in anyway ? 

[androidlad]

40 minutes ago, Otago said:

I agree, it's present in both of them when it's in focus. 

I think the solution is to shoot 4K, which is downsampled from the 6K and won't show any of those artefacts ( if it's been done well and there seems to be no evidence it's not ) and then downsample again to 1080 when you are editing. You could also try a softening filter as a form of OLPF if 1080 in camera is necessary. 

Can you tell from it being inconsistent between the horizontal and vertical what they are doing with line skipping vs binning ? Do they have to use what the sensor provides or can customers make their own version by changing the microcode on the chip ? Do you know if the sensor control unit is addressable in anyway ? 

You can run a Fast Fourier Transform using Matlab with a high detail test scene to determine the actual sampled resolution.

On IMX410, it's usually horizontal binning and vertical line-skipping. IMX5XX added pixel-weighted binning which improved quality a little bit.

To save R&D effort and cost, they usually use preset readout drive modes.

[Lukas Kalinowski]

Thank You Guys for respond and some info, yeah I forgot do add that it was in 1080p and in FULL FRAME mode. I did some more tests and most wird thing what I noticed that this things happen in random. sometimes its there and sometimes is not. Tomorrow I will let ou know if its also apparent in 4K modes. Thanks! 

[Lukas Kalinowski]

Hi Again! I have been investigating this issue more and this happens IN EVERY mode 24,25,30,50,60 4K, FHD, FULLFRAME AND apsc , and this happens totally randomly, one time the footage is as it should no aliasing nothing, (moire is there but as I said before its not issue for me) and next time this wird aliasing is there.  Lines are just jumping around or pixelate like crazy along its vertex I'm 100% sure its happens in total random fashion. So it I more related firmware or internal error ? becouse its impassible that clips recorded with same lighting and same settings one after another has HUGE different and has very apparent Jitter/aliasing its looking like crazy 

(It happening in ACES and non ACES workflow also im not in half or quarter resolution mode)

[Deadcode]

Dont judge by the preview window, you can see rescaled footage there. It will be aliasing free in full screen mode on a 4K monitor

[Lukas Kalinowski]

23 minutes ago, Deadcode said:

Dont judge by the preview window, you can see rescaled footage there. It will be aliasing free in full screen mode on a 4K monitor

This is good point but, issues happens also in 1080p and 4K as well,im previewing this in FHD but You know 4K from Ursa G2 on FHD do not have it, pocket also , and GH5 so I suppose that something is wrong when in one clips this is there and on second one there is nothing like that ? 

[kye]

Just set the preview window to 100% and make sure the timeline resolution matches the source clip resolution and that should eliminate any preview window scaling issues.

It's worth checking this with known bad clips, and it's also worth checking the footage in another viewer.  For example, VLC can take snapshots of single frames and you can zoom into those to really see what's going on.

[androidlad]

You need to test it with pin-sharp and consistent focus and smaller aperture, some of the shots are clearly out of focus. View them in full screen and full quality

Use the DPReview video stills comparison tool: https://***URL removed***/reviews/panasonic-lumix-dc-s1-review/8

You can see moire and aliasing in every resolution, framerate and crop mode, with 4K FF <30P being the least noticeable, and 1080p modes being the most noticeable.

S1 doesn't have an OLPF so it's prone to this.

 

[Lukas Kalinowski]

2 hours ago, androidlad said:

You need to test it with pin-sharp and consistent focus and smaller aperture, some of the shots are clearly out of focus. View them in full screen and full quality

Use the DPReview video stills comparison tool: https://***URL removed***/reviews/panasonic-lumix-dc-s1-review/8

You can see moire and aliasing in every resolution, framerate and crop mode, with 4K FF <30P being the least noticeable, and 1080p modes being the most noticeable.

S1 doesn't have an OLPF so it's prone to this.

 

Hey Man! Thanks for interest. Yeah I understand You clearly. Those are just stills, I'm using peaking and focus are always finally where should be , but as You said I will make it on f8 or something like that. But something different bothers me, why this freaking huge aliasing pop up randomly ? I know there is aliasing but sometimes its HUGE like on this screen, (this is also very visible, in grabbed frame from resolve, also on via Blackmagic mini monitor declink  via HDMI on my preview monitor,  I see exactly the same . You know normal aliasing is not soo huge and visible as hell. This is wird 

[Mokara]

On 10/4/2019 at 2:35 PM, androidlad said:

You left out a key piece of info - they were shot in 1080p60.

Non-crop 1080p on all DSLR/mirrorless cameras is derived from pixel-binning, or line-skipping or a combination of both.

The artifacts you are seeing is moire and aliasing as a result from pixel-binning and line-skipping, this is very common and completely normal, it's particularly severe on S1 due to the lack of an OLPF.

The S1 is supposed to be oversampled. The problem at lower resolutions is the increased processing demands that places on the camera, so there are artifacts that result. Not necessarily pixel binning/line skipping as such, but rather crude approximations of the data which has a similar results.

[androidlad]

1 hour ago, Mokara said:

The S1 is supposed to be oversampled. The problem at lower resolutions is the increased processing demands that places on the camera, so there are artifacts that result. Not necessarily pixel binning/line skipping as such, but rather crude approximations of the data which has a similar results.

4K <30p is. 4K 60p is 1:1 readout. 1080p is not oversampled but pixel-binned and line-skipped. Period

[Mokara]

15 minutes ago, androidlad said:

4K <30p is. 4K 60p is 1:1 readout. 1080p is not oversampled but pixel-binned and line-skipped. Period

And you know this how?

I am confused why you think that 4K can be oversampled but all 2K is pixel binned/line skipped.

[subscatter]

I wonder if this has been fixed?? I had been testing S1 last month with the latest firmware + Mc-21 and Sigma art lens 20mm 1.4. And i saw this type of antialiasing issues and moire artifacts all over the footage... especially the repeated bricks on the buildings. And i also saw moire in still pictures as well. Truly frustrated. Can this be fixes in future firmwares?? Or we are just out of luck without OLPF!