Dienstag, 26. Dezember 2017

Sigma "Merrill" micro-contrast

Micro-contrast exposed:

There are still many discussions about the pronounced micro-contrast of the Sigma "Merrill"cameras.
Some love it, others criticize it and prefer the softer look of the "Quattro" generation.
There are claims that, because of the harsh look, the "Merrills" are not good for portraits but better for landscapes etc.

The fact is: It's your decision if you want the image to be harsh or smooth. It's all "hidden"
in Sigma Photo Pro.

There are three main settings that affect the micro-contrast of the image:

  • SPP chroma noise reduction
  • SPP monochrom mode
  • exposure (before taking the image, not within SPP)

 1. Chroma noise reduction:

Less chroma noise reduction produces less micro-contrast
More chroma noise reduction produces more micro-contrast

2. Monochrom mode:

Using SPP monochrom mode produces less micro-contrast.
Yes you lose the color, but the solution is simple:
1. convert a color version of your image
2. convert a monochrome version of your image
3. Open both images in Photoshop as layers, put the monochrom image on top
and set the blending mode to "luminance"
Now you have the colors of the color image and the luminance information of the
smoother monochrome image.

3. Exposure:
More exposure while taking the shot results in less micro-contrast.
More exposure/light "fills" each pixel with more data/color information, thus,
Sigma Photo Pro doesn't need to apply its micro-contrast generating chroma noise reduction.

Lets have a look what this means in a real world sample:

Here is an "underexposed" and an "overexposed" photograph of the same scene.
(bracketing on a tripod)
Lets bring them to the same brightness in Sigma Photo Pro:
They look similar now. Now let's take a closer look at the ancient door:
Clearly visible that the lesser exposed image on the left has less color and stronger micro-contrast.

Chroma noise reduction
Let's apply more and less chroma noise reduction to the lesser exposed image.
Again quite clearly, more chroma noise reduction (+2) leads to more micro-contrast 

SPP monochrom mode
The same image (1/6s exposure) on the left with a SPP monochrom conversion as luminance layer.
Result: less micro contrast

And finally the two extreme samples side by side
Left: more exposure and monochrome
Right: less exposure plus maximum chroma noise reduction

These different methods can be used, tweaked and combined to get exactly the look you want.
Sigma gave us, presumably unintentional, a very powerful tool with the Merrills and Sigma Photo Pro. It's almost as if you have different cameras in one device.
Unfortunately, all the described options are gone with the Quattro generation.

Samstag, 14. Oktober 2017

Sigma Quattro’s 2-pixel noise artifact finally gets some attention at dpreview.

Quattro’s 2-pixel noise artifact finally gets some attention at dpreview.

I addressed this issue more than 3 years ago already in that forum.
I called it „ants“, „flecks“, „spots“ or „2-pixel noise“

Here from 2014:

In march 2016 the problem was not solved:

August 2016 (noise reduction can't remove it)

It seems that only forum member tagscuderia recognized this error also and mentioned it repeatedly.

Just like now again:

„As a dp0 owner let me say that the 2-pixel crawl artefact was evident from day one and has been touched upon on here... but seemingly only Maceo and myself could see it?! We did however both independently report it to SIGMA with examples“

Probably he won't be heard again.

Montag, 11. September 2017

How to avoid/minimize color casts using vintage lenses on the SD1 Merrill

How to avoid/minimize color casts using vintage lenses on the SD1 (Merrill)
(thanks to user 'motionride' from bilderforum.de)

The Sigma SD1 is known to have problems with vintage lenses. Depending on the lens and aperture setting, you get strong color shifts/vignetting.

Previous generations SD9/SD10/SD14/SD15 do not show this behaviour, or only very little.
Until now, the procedure has been as follows:

Set your camera to A (Aperture Priority) and choose the "Reference Aperture":
F5,6 for the SD9
F2.8 for the SD10
F1.0 for the SD14, SD15 and SD1*
And now you’re good to go..... *but not with the SD1

With the SD1 it has been found that color shifts can be minimized by setting the camera aperture to the same aperture as the lens. 

How does it work?

1. Use the standard method: 
-Set your SD1 camera to A mode and set it to F1.0
-Set your preferred aperture on the lens (for example F5.6) and the camera will display the correct exposure. (for example 1/250)
The exposure will be correct now but unfortunately this causes color shifts.

2. Remember the suggested exposure time (1/250) and switch your camera to M (Manual Mode) 
Now set the time to 1/250 and set the aperture to the same aperture as on the lens (F5.6 in this case)

3. Shoot and be amazed how the color shift has gone.

The following screenshots from SPP show the difference between the two methods.

Even though the same exposure time and aperture was used for the images, the color shift is virtually eliminated:

SD1 with Cosinon 55mm F1.4 (M42-mount)
click for full res
SD1 with Carl Zeiss 135mm F3.5 (M42-mount)
click for full res

Donnerstag, 1. Juni 2017

High ISO comparison: Sigma sdQH (DNG) vs. dp1m @ ISO1600

Sigma sd Quattro H review (dpreview)

The DNG implementation seems to be poor and only usable at very low ISO:

Look at the difference between the sd Quattro H and the good old dp1 Merrill at ISO 1600:

Click for 100% 

Mittwoch, 24. Mai 2017

Quattro SFD mode still "Beta"

Quattros SFD mode still shows weird artifacts, even on static subjects.
Click for 100%

Forum comments are uncritically, as usual.

Dienstag, 21. März 2017

Ongoing debate on DPREVIEW after a "Photo comparison between DP2M and new SDQ H with 35mm Art"

Big discussion on dpreview...
...about the "shadow quality" and colors based on a DP2m photo that is underexposed by 2 stops.

Exposure: Manual - 1/1000 - F6.3 - ISO: 200
...doesn't make any sense.

Samstag, 28. Januar 2017

Sigma sd Quattro H vs. Sigma sd1 Merrill image comparison

Maro provided an image comparison between the sdQH and the SD1 with the 85mm 1.4 Art lens:


At first, the images look quiet similar. But the statements in the dpreview forum are pretty harsh:
  • "SD1 looks sharper!"
  • "Not only sharper. In Quatro I see a lot of color artefacts."
  • "If the SD1 had live view and an EVF, it would be the better camera.  Certainly it produces a better image."
  • "Merrill was/is clearly outstanding!"
  • "This just makes me think what sort of fabulous camera a mirrorless SD Merrill H would have been."
  • "Once again the comparison show that 1:1:1 layers structure of the Foveon Merrill of the SD1 wins over the poor and rubbish 1:0.25:0.25 structure of the Quattro"
  • "The SD1M will be my last Foveon camera unless Sigma resurrects the Merrill chip in an improved body."
While the Quattro in general seems to have an edge regarding luminance resolution, many people
seem to prefer the look of the Merrill generation.
So what is it?

Maybe these 200% crops help to understand:
(click images for full resolution)

Clearly visible how destructive the default luma-noise reduction setting works with Quattro files,
but then luma-NR off shows how noisy the Quattro is on pixel level. 
So, finding the balance between noise reduction and adding/lowering sharpness is the tricky part with Quattro, while the Merrill sensor seems to be more forgiving and stable.  

Sonntag, 1. Januar 2017

Image quality: SD Quattro low-res vs. DP2 vs. Fuji X-E1 (downsampled)

Tom Schum from the dpreview forum made a comparison between these cameras:
SD Quattro (low resolution 1:1:1 Foveon structure) vs. the original DP2 and the Fuji X-E1 (downsampled)


Some people claim, if you want true 1:1:1 Foveon quality with the sd/dp Quattros, you should use the low resolution mode. 

Lets have a closer look at 200%: (ignoring the Fuji)

    click image for full resolution:

The sd Quattro sample looks sharper. (probably different sharpening amount) 
But have a look at the red and yellow surfaces of the boats.
The dp2 looks natural, smooth and realistic while the sdQ is noisy and shows posterization.

Quattros shortcomings are evident even in low res (aka true 1:1:1 Foveon structure) which leads to the conclusion that the sensor itself is flawed. 

Additional information:

A new conversion from the sd Quattro image with luminance noise reduction set to minimum:

Posterization effect is reduced, noise/grid becomes more visible.