3K mode vs 5K anamorphic quick test

Started by Flynt, September 09, 2020, 02:56:07 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

Flynt

Last month I shot a few footages playing with different ML-modes. Some results in the video below.
Canon 5D mk III, Canon 35mm f 1.4L, Davinci resolve





Grognard

How did you manage with the crop factor? Which ens did you use for 5k anamorphic ?

IDA_ML

Flynt,

Although I appreciate your efforts to adjust the same field of view and film these comparative shots, they may create the wrong impression that the 5k anamorfic mode is much more inferior in terms of image quality than the 3k 1x1 mode.    The crop mode uses much smaller  sensor area, therefore the depth of field is much larger causing much more of the image to appear sharp.  The lack of sharpness in large areas of the anamorphic shots is attributed to the much smaller depth of field which creates an overall impression of a soft image.  Those areas are soft not because of lack of resolution or inferior image quality but because of defocus.  If you want to be correct with the comparisons, in your anamorphic shots, you will need to close the apperture to achieve equal depths of field in both modes.

In fact, image quality of the 5k anamorphic mode is so good that it is barely distinguishable from 3k.  This is evident from the pink flower shots where depth of field for both modes is large enough for the entire flower to appear in sharp focus.  I use 5k anamorphic for landscape videography all the time and it resolves fine detail perfectly.  Videos look gorgeous on large high-resolution screens.

paulheran

Quote from: IDA_ML on September 13, 2020, 12:16:39 AM
The crop mode uses much smaller  sensor area, therefore the depth of field is much larger causing much more of the image to appear sharp. 

How does it use a smaller sensor area, but the FOV is the same from the looks of it in this video.
5D3 113 | 16-35 2.8 | 24-70 2.8 | 70-200 2.8 IS

IDA_ML

He changed the subject distance, I guess, to achieve the same FOV.

whitelight

Quote from: IDA_ML on September 13, 2020, 12:16:39 AM
Flynt,

Although I appreciate your efforts to adjust the same field of view and film these comparative shots, they may create the wrong impression that the 5k anamorfic mode is much more inferior in terms of image quality than the 3k 1x1 mode.    The crop mode uses much smaller  sensor area, therefore the depth of field is much larger causing much more of the image to appear sharp.  The lack of sharpness in large areas of the anamorphic shots is attributed to the much smaller depth of field which creates an overall impression of a soft image.  Those areas are soft not because of lack of resolution or inferior image quality but because of defocus.  If you want to be correct with the comparisons, in your anamorphic shots, you will need to close the apperture to achieve equal depths of field in both modes.

In fact, image quality of the 5k anamorphic mode is so good that it is barely distinguishable from 3k.  This is evident from the pink flower shots where depth of field for both modes is large enough for the entire flower to appear in sharp focus.  I use 5k anamorphic for landscape videography all the time and it resolves fine detail perfectly.  Videos look gorgeous on large high-resolution screens.

IDA_ML, although I understand your point, there is in fact a great loss of detail in 5k, also in the pink flower shot (first rose I guess) which you consider sharp in both modes. The difference is immediately noticeable to my eyes.

IDA_ML

Quote from: whitelight on September 15, 2020, 08:19:42 AM
IDA_ML, although I understand your point, there is in fact a great loss of detail in 5k, also in the pink flower shot (first rose I guess) which you consider sharp in both modes. The difference is immediately noticeable to my eyes.

No, I mean the flower at about 30 sec.  The first rose seems to be out of focus.  You can try to film a wide angle landscape shot at F8 and compare both modes.  You will barely see a difference in detail unless you pixel peep at 300%.  If you watch the videos full size on a large high-quality monitor, even from a distance of 1m or less, the differences in image quality will be not perceptible.

Flynt

Quote from: IDA_ML on September 13, 2020, 12:16:39 AM
Flynt,

Although I appreciate your efforts to adjust the same field of view and film these comparative shots, they may create the wrong impression that the 5k anamorfic mode is much more inferior in terms of image quality than the 3k 1x1 mode.    The crop mode uses much smaller  sensor area, therefore the depth of field is much larger causing much more of the image to appear sharp.  The lack of sharpness in large areas of the anamorphic shots is attributed to the much smaller depth of field which creates an overall impression of a soft image.  Those areas are soft not because of lack of resolution or inferior image quality but because of defocus.  If you want to be correct with the comparisons, in your anamorphic shots, you will need to close the apperture to achieve equal depths of field in both modes.

In fact, image quality of the 5k anamorphic mode is so good that it is barely distinguishable from 3k.  This is evident from the pink flower shots where depth of field for both modes is large enough for the entire flower to appear in sharp focus.  I use 5k anamorphic for landscape videography all the time and it resolves fine detail perfectly.  Videos look gorgeous on large high-resolution screens.

I'm sorry, but I don't agree with you. There's no out of focus, softness, etc. Let me tell you something. Just watch other ML videos on my channel, and you'll understand that I know how to focusing, how to work with ML and so on. I'm working with Danne's firmware for about a year, and all this time I've been comparing it with 3K crop rec mode in different situations.
3K footages always look amazing - with any lens, at close- mid- long-range distances , F-number, in real size with 100% crop, in fit in view mode, etc. There are tons of detalization in 3k mode. The keyword is - detalization, not softness or something else.
Yes, at first sight 3K and Danne's mode look pretty close, maybe the same for someone. But Danne's mode has less detalisation, it's a fact, and everyone sees it here, even you. That's what I want to show in this comparison.
Do you really think that 1920x2340 anamorphic pixels upscaled to 5.5K have better detalization than real 3072x1728 pixels, especially with the same or less bitrate, especially in 100% crop comparison? No, even mathematically. And in practice, as i said before, even you see it. These fake anamorphic 5.5K pixels not even close to real 2x anamorphic lens with the same matrix resolution.
I know there are lots of Danne's fans here, and I don't want to say that Danne's mode is bad or something else. It's good when you need to get long-time rec, large FOV, pseudo-panoramic automatization in post, long-range distances. But when you need detalization 3K is better.
As for me, I would prefer to work with real 3K upscaled to 4K (like ARRI Alexa) than fake anamorphic 5.5K  downscaled to 4K, but my CF and SD cards don't support long-time rec in 3K mode. That's why I use Danne's firmware. As you can see on my channel 3K looks great in any situation, even with Youtube compression, in post-production you can crop it, or upscale, or whatever you want without details loss.
And also I think you should try 5.5K full matrix resolution mode in ML to understand what 5.5K really means) There is fantastic detalization even with ultra-wide lenses at long-range distance, full sensor area, etc.

markanini

A more interesting test IMO is: 5k Anamorphic mode using stock lens VS. 3k mode using a cheap ultra-wide like Canon EF-S 10-18mm. Or course make sure to match subjects and DOF or people will nitpick. I suspect the anamorphic mode will win in low light but I'm curious how much more detail you get using the crop mode and a cheap ultrawide lens.
Gear: Canon 600D & Magic Lantern Nightly.

Grognard

I also made  tests, i think than 5k anamorphic is really much better than full hd upscaled and very close to 1:1 mode. Not true, mathematically 5k anamorphic has more pixels than 3k at the same ratio (2:46) 1920x2340 =4 492  800 pixels, 3072x1248 = 3 833 856 pixels. It's true that 1:1 is sharper but it's not so easy to see a difference in detail even in 3,5k.

I agree with IDA_ML, to be more scientifique, close shoot is not the best choice because of the variation of depth of field.
A landscape should be a better choice.

IDA_ML

Flint,

From what you write, I am left with the impression that you felt offended by my words.  Please don't!  We all are here to share our thoughts, different opinions and experiences and discuss things.  This is what the forum is all about.  And yes, I would be glad to take a look at your YT channel but you should provide a link to it.

I never said that 5,5k anamorphic provides better detailisation than the 1x1 3k one.  All I said was that in terms of image quality, both modes are barely distinguishable from each other when watched full screen.  That's all.  When people watch movies they watch them full screen and don't pixel peep them at 100%, do they? 

In my experience and in many real-life filming situations the 5,5k anamorphic provides some quite significant advantages to the 3k mode:

1) No crop factor and therefore full frame field of view.  Perfect for close-up shots with a beautiful bokeh;

2) Wider field of view with available lenses;

3) Continuous or much longer recording times;

4) Lighter and smaller size files that are easier and faster to process and render in MLVApp and Resolve.

The only downside I see with 5,5k anamorphic is with super wide angle shots, (filmed at focal lengths from 12 to 24 mm), where you may end up with some aliasing on straight vertical lines and some mushiness in the areas of finest detail.  Even those are barely perceptible on full screen unless you know where to look for them.

Here is a short test that I shot on the cropped 100D camera in its 1128x1900 anamorphic mode 2 days ago (5 more days left for download):

https://we.tl/t-YHIgOJZCCA

It looks very good on my 30-inch screen and this is by far not the limit for the video quality.  I hope, one day we will have full-res 1736x2928 anamorphic on the 100D and when this happens, quality will be even better, with a wider field of view, accordingly.

With all that said, my final thought is that the 5,5k anamorphic mode has more pros than cons for most real-life filming situations and should be considered for serious work by those who want full-frame vision in their videos.

IDA_ML

Here is another test shot on the EOS-M at 1736x2928 anamorphic using the same EF 35/F2 IS lens:

https://we.tl/t-B5iS2H9fJa

The mushiness is mainly attributed to the quite foggy day that the clips were filmed on.  On the full-frame 5D3 video quality would have been much better.

Grognard


sm105


mlrocks

Quote from: Grognard on September 28, 2020, 08:27:48 PM
3,5 K(3584x1300) Vs anamorphic (1920*2340).
5D Mark III
https://imgur.com/a/1MdYyZn

Almost the same. 5.7k has a little bit less details when enlarged. But it is totally indiscernable in the full photo. Normal view distance will not show any difference.
Full frame mode has two stops better low light performance, and has much better shallow dof, tonality, gradation, color, dr, etc, due to 2.5 times larger sensor size.

mlrocks

When dslr revolution just began, about 2008, I watched Vincent Laforet's 5d2 demo video, and almost all of 5d2 videos available on vimeo, and when 7d came out, almost all of the 7d videos, I felt that 5d2 footage was much better looking than 7d's, although 7d was newer at the time, similar codec, line skipping, etc. Full frame footage just looks more organic, not just limited to shallower dof, although may not have clear scientific criteria.

Skinny

It is true, at least in my experience full frame video looks better even if it is compressed by youtube to such crappy quality as 240p or 360. I can notice it sometimes... I don't know exactly how, but details are more defined and all things are separated and clearly visible.
And if you look at the video from camera with very small sensor (phone for example) everything will just blur to some colorful mess.

The difference is small and sometimes not as visible, but I think it exists. It is just a matter of perception. And I don't want to argue about it, there is no point.. I can see it and for me it is enough.


And also with FF camera you have more options for vintage lenses. With crop sensor you almost always need wide-angle lenses which are less common.


I could be wrong of course, maybe I am just imagining stuff.. it is not really scientific or anything :)

Grognard

Quote from: mlrocks on March 11, 2023, 05:57:00 AM
Full frame mode has two stops better low light performance, and has much better shallow dof, tonality, gradation, color, dr, etc, due to 2.5 times larger sensor size.

What do you mean by two stops better low Light performance ?
Sensor is still the same. 3,5k is just a crop in the same sensor size so it should receive exactly the same amound of light.
3.5k with 1.6 crop factor is close to super 35 cinema so it's also a good choice.

Skinny

@Grognard it is because with full frame mode you are using pixel binning (for example 1:3 mode) where the information from three pixels averaged in one, so less noise.

In this mode you are actually using all the pixels, but some of them are averaged on hardware level. But you are still using the whole sensor area.
So we are collecting more light if you think about it.
I don't know if it's really two stops better though.

Grognard

Quote from: Skinny on March 12, 2023, 10:39:22 AM
@Grognard it is because with full frame mode you are using pixel binning (for example 1:3 mode) where the information from three pixels averaged in one, so less noise.


OK I understand what you mean. Thanks.
But I use to shot in fullhd pixel binning with my 5D mark iii and i've never noticed less noise in video (with pixel bining) than in photo (5.7k full sensor).
I will make some tests.

vastunghia

Maybe we should stop talking about noise and think in terms of SNR (Signal-to-Noise Ratio). If I take one billion noisy pixels and average them to one single pixel — ta-daa, no more noise! But zero SNR as well ;D
5D3 for video
70D for photo

names_are_hard

Quote from: Skinny on March 12, 2023, 10:39:22 AM
I don't know if it's really two stops better though.

It's not.  1 stop is twice the light, and mlrocks claims full sensor is 2.5x the area of the crop.  Assuming that's true, it's much closer to 1 stop.  4x would be two stops, 3x should be 1 1/2 stops, 2.5x is 1 1/4?  Can't be bothered looking it up, should be geometric mean based on the square, something like that?

If you're using the same lens for that full sensor capture vs the crop capture, there's no change in brightness, of course.  Same light per area, even though you capture more area (wider fov) full frame.

I think mlrocks is just being overly enthusiastic.  They also claimed colour would be improved, and I can't see any reason why that would be true.

Skinny

In audio if you sum two tracks with the same content (have 2x amount of data), you will get only 1,414 better signal to noise ratio.
Digital image is probably the same thing..


Colors could be different because of the chromatic aberrations.. Some vintage lenses have a lot of them even in the middle of the frame (with wide open aperture) and on crop-sensor it will be more noticeable. And it can slightly affect colors of bright objects. 
On the other hand, if you have clean center but a lot of aberrations closer to edges then crop sensor will provide cleaner picture and maybe slightly different color again.

And if you compare full-frame sensor and very tiny sensor like some phones or web cameras, then things like diffraction limited aperture can become visible. And usually DLA is considered limiting sharpness, but it also affects color even before the sharpness, at least I can see it usually...

mlrocks

Quote from: Grognard on March 11, 2023, 10:16:40 PM
What do you mean by two stops better low Light performance ?
Sensor is still the same. 3,5k is just a crop in the same sensor size so it should receive exactly the same amound of light.
3.5k with 1.6 crop factor is close to super 35 cinema so it's also a good choice.

I mean when I use 5d3 anamorphic mode, I can go iso 6400, in 3.5k 1x1 mode, I can go iso 1600.

mlrocks

A side topic, I have a kodak 645 pro back. It is 36x36mm. The image is so good in terms of color and gradation. I like it the best. I know it's ccd and 16 bit. My point is that larger sensor is better in many ways, sometimes not scientifically defined.