Is this Moire, Chroma, Diffraction or Aliasing noise? What causes it, non-APO lens or sensor (low 24Mpixel) resolution?

[lincoln_m]

Advertisement (gone after registration)

100% crop of a top lit lake with ripples causing some sort of Moire, Chroma, diffraction, or aliasing noise? M10 with Summilux 35 Asph FLE at F4 or F5.6. Not my best work but it does show the artefacts I'm curious about.

I initially thought that M10 with 35 Summilux didn't make this high-resolution chroma type noise because I had only noticed it (much worse) when using my Sony A7 with Leica 50mm Summicron F2 (I'll post an example of that next) but it seems that I can also get the colour noise particularly with top lit water ripples. Maybe the ripples act as a prism so each pixel gets a different colour?

I think the M10 doesn't have an anti-alias filter but the Sony A7 does. The Leica 35mm summilux is a highly corrected Asph lens where the Leica 50mm summicron is a 1979 non-asph design so presumably less corrected and more prone to chromatic aberration. The A7 even with M-mount adapter doesn't have the micro lenses designed for Leica lenses and maybe the Sony sensor design also makes this (high resolution) noise worse? I'm not sure why a softer lens and AA filter would make the chroma type noise worse unless it is the 50mm summicron's chromatic aberration? I'm thinking I shouldn't bother with using the Sony A7 anymore and perhaps get an SL or SL2-S instead.

But the real question is, if I sometimes get this noise in land/seascapes with the M10, do I need to upgrade to M10R (M11?) or SL2 and perhaps even APO 35mm , 50mm lenses? Quite expensive. Is the 35ASPH FLE too high resolution for the M10 sensor, so M10R or M11 would help or the M10 sensor is fine but the 35mmAPO would be better?

If I'd known at the time the M10 with 35mmASPH-FLE was giving this noise I could have done a test with the 50mm but I'd not seen it before and didn't spot it until I got home several days and many miles away.

I'm basically trying to work out if my lenses need updating to APOs (50mm APO first) or if the camera needs to be higher resolution M10R/M11 or SL2 to cleanly capture lake/sea water ripples?

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[lincoln_m]

Here is the Sony A7 with Leica 50mm F2 summicron 100 % crop of part of a glacial lake in Iceland showing much worse chroma/moire type noise?

But A7 has anti-alias filter and the 50summicron is non-ASPH (softer) lens, but maybe its the lens' chromatic aberration and the lens is high res even with Sony sensor? I don't know.

 

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[pedaes]

Very difficult at this resolution to see. What ISO as highly relevant.

The answer to you last paragraph is neither to both.

[Al Brown]

It looks like color moire.
Here is an explanation from the web: "Color moiré is an artificial color banding that can appear in images with repetitive patterns of high spatial frequencies, like fabrics or picket fences – or your computer screen. Color moiré is the result of aliasing in image sensors that employ Bayer color filter arrays. It is affected by lens sharpness, the sensor’s anti-aliasing (lowpass) filter (which softens the image), and demosaicing software. It tends to be worst with the sharpest lenses."

On the flip side, 35 FLE (at least my copy) has exhibited very strong CA, but only on the edges of areas where CA normally happens and mostly wide open. Your samples are not a typical CA scenario.

Edited July 15, 2022 by Al Brown

[lincoln_m]

Within 30mins of the A7 shot here is M10 35mm Summilux ASPH FLE shot of similar when there was less wind on the lake. So less chroma type noise but still visible now that I'm noticing it more with this pixel peeping!

So what next higher res sensor or APO-ASPH lenses? I suppose there will always be a limit no matter how much you spend on kit.

The peaks of the mountains are about 10km away according to Google Maps with the base being about 8km away. The ripples on the lake are therefore accidentally forming a high-res test chart pattern so perhaps its a combo of moire and chromatic aberrations?

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[lincoln_m]

Pedaes: The first shot Derwent Water, Cumbria, UK was shot at M10 ISO 200, while the Iceland shots were shot both at ISO 100 for M10 and A7.

For high-res landscape shots with the 35mm ASPH FLE I tend to shoot at ~F5.6 (M10 doesn't always guess correctly) because the lens spec MTF show it as best resolution.

Al: Colour moire does make sense as the water ripples form ever increasing spatial frequency patterns (on the 2D sensor) as they go into the far distance of the 3D real world scene.

So you're saying high-res lenses (even the 50summicron) make it worse, and so too does the AA filter (on A7). Would that mean lower res lenses on a high-res (no AA filter) sensor wouldn't show this colour moire? But a high-res sensor may show up other artefacts of the lens!

I have seem examples of M10R with the 35APO-Asph at 100% somewhere on the web (photos around Seattle) that are amazing even at 100% crop but they didn't have these lake ripples in their shots.

Granted it seems that only water ripples show this issue in my shots (landscapes without water are not an issue), so perhaps I just have to be aware in future. I'm not sure how best to reduce this noise given similar scenarios though.

 

[adan]

Advertisement (gone after registration)

It is mostly moiré/aliasing (they are essentially the same thing). I don't see any significant CA or noise in these pictures, in the areas that do not have the "ripple" patterns as well.

The lens is resolving repeated details or lines (the water ripples/wavelets), that are so small and delicate that they don't cover a full Bayer RGBG 4-pixel quad, and thus flip back and forth from red/magenta-biased to blue/cyan-biased, as either red or blue sensor pixels are "left out of the mix" at any given spot in the image.

A better spatial sampling rate (i.e. higher-resolution/Mpixel sensor) will reduce this.

A higher-resolution lens will make it worse!

We chant the mantra that "a better lens will make any sensor better, and vice-versa." But moiré/aliasing, especially in color, is the one place where that breaks down - it is a symptom of a lens actually "outresolving" a sensor by just the right amount.

(The de-mosaicing algorithms of your chosen raw coverter can also have an effect - Adobe vs C1 vs others.)

The Nyquist-Shannon sampling theorem says that the sampling rate (for our purposes, sensor pixels per mm) needs to be about twice the signal resolution (lens resolution: line pairs per mm (lp/mm)) to avoid aliasing artifacts.

So if your lens resolves 80 lp/mm (fairly common for "very good" 35mm-format lenses at their "best" aperture, with perfect technique), which is 80 black lines alternating with 80 white lines, for a total of 160 lines per mm, then you need a sensor that resolves at least 320** pixels per mm to avoid any aliasing. A 24-Mpixel full-frame sensor has about 166 pixels per mm, so it may have trouble with aliasing above about 40 lppm lens resolution.

Bayer color sensors do even worse, since only 1/2 of the pixels are red or blue (1/4 red, 1/4 blue) - thus the color/chroma sampling rate is even lower, and the odds of color aliasing/moiré even higher.

(Note - most of the time, 24 Mpixels in Barnack-format is still adequate even for M lenses: they are not necessarily resolving 80 lp/mm at all apertures or all subject distances or all parts of the frame, or if hand-held or even slightly misfocused. Nor are the finest details usually a repeating pattern/texture like wavelets or bricks or window screens )

___________

**320 pixels/mm amounts to around a 90 megapixel sensor in the Barnack format. But if you get a fancy new 50mm APO-Summiwhatsis that resolves 120 lp/mm, you'll need even more than that! Especially on a Bayer-color sensor.

[lincoln_m]

Adan,

Thanks for your detailed info. I work in Electronics for my normal day job so I'm familiar with Nyquist sampling but for some reason when it comes to pixels, light and lines per mm I somehow forget that is similar to frequency (in Hz as in audio or radio signals). The lens MTF specs only show up to 40lp/mm but you are correct with quality lenses such as 35FLE or 50 & 35 APO, 80lp/mm is probably quite possible in favourable conditions (good technique, best aperture and high spatial frequency ripples in the subject). I think you hit the main issue when you mentioned the Bayer 4 pixel filter mixing with the subject spatial frequency of the wavelets.

Fuji use an XTrans sensor filter on their APS-C but still have the Bayer filter on the GFX Medium Format cameras so there is no getting away from it unless on goes to the M10M where these is no colour.

I find it interesting that the answer to avoid these colour-moire artefacts is a higher-res sensor (M10R, M11, SL2) but the same or lower res lenses. Or put another way if one goes for the high res APO lenses then you also need a high-res sensor or the artefacts will be worse.

I should probably stop viewing/zooming at 200% in Lightroom on my iMac Retina 5K display, I mean how often do I actually print that big (>24x16inches) anyway.

Maybe an M10R is worth considering next. I had initially thought if I needed/wanted more pixel just upscale in Lightroom for a bigger file but for some subjects (water ripples, distant buildings, fences) that might not work, while others (natural subjects with random non-repeating paterns, rocks, trees) may be OK to upscale.

I'm getting a better understanding of what's happening here thanks to these discussions.

Lincoln

[pedaes]

6 hours ago, lincoln_m said:

ISO 100 for M10

The native low ISO of the M10 is said to be 160, so 100 is a "pull" ISO which will not help the quality of you images. 200 is probably the sweet spot.

[pedaes]

1 hour ago, lincoln_m said:

I should probably stop viewing/zooming at 200% in Lightroom on my iMac Retina 5K display

I regularly zoom to 300% on same display for local adjustments or to check sharpness and do not see these artifacts from M10 with 50mm or 90mm APO- Summicrons, or 35mm FLE Summilux. FWIW

To try and be helpful, I will risk posting this with similar water pattern - M10 / 50mm APO-Summicron 

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

Edited July 15, 2022 by pedaes

[lincoln_m]

Pedaes, can you make a 100% crop of the water in the middle distance? I think there is a little (slight) colour moire. The low res full frame forum posts are effectively doing an anti-alias filtering. 
Lincoln

 

[Jon Warwick]

I used to see such moire/aliasing on a very frequent basis with my M240, which I typically paired with a M 50 APO. Eg, I’d see it in ripples of water like the OP’s example, and very often when photographing a distant mountainside that had rocky scree.

The ultimate solution (for me) has been the M10 Monochrom, where the image capture is so “pure” without the RGB color filter array that I tend never to see moire/aliasing at all …..or my GFX100S due to the higher resolution sensor that can much more accurately capture things like rocky scree than a 24mp camera ever could.

For “static” subjects like a mountainside with scree, a multishot mode on a camera like the SL2 has will also help achieve a moire free image much better than single-frame capture.

Edited July 16, 2022 by Jon Warwick

[pedaes]

9 hours ago, lincoln_m said:

Pedaes, can you make a 100% crop of the water in the middle distance? I think there is a little (slight) colour moire. The low res full frame forum posts are effectively doing an anti-alias filtering. 
Lincoln

 

I have sent you a PM with the file for you to 'play' with. I have pushed it to 1,2 and 300% and can't see anything, but perhaps that is me!

Edited July 16, 2022 by pedaes

[lincoln_m]

Pedaes,

Thanks for your PM but the forum has again compressed the file so its too small (729KB)to be able to see any details.

Anyway I understand what's happening with certain images on occasion and how the distant water ripples can act as a high res high spatial frequency test chart and cause potential moire.

Higher resolution helps M10R, M11, SL2 and no Bayer filters helps, such as M10M or an XTrans filter (Fujifilm) but there isn't a >40Mpixel XTrans Full Frame sensor AFAIK.

Maybe next time when I have a similar scene I'll save both DNG + JPG so I can compare the M10 camera de-moire v Lightroom's de-moire. I'll hunt around in LR to see if there are any de-moire tools to help.

Lincoln