Ad
Ad
Ad
Pages: « 1 ... 7 8 [9] 10 11 ... 15 »   Bottom of Page
Print
Author Topic: Can you see a difference in small prints between an MFDB and a DSLR?  (Read 14347 times)
JV
Sr. Member
****
Offline Offline

Posts: 550


« Reply #160 on: March 05, 2014, 09:18:17 PM »
ReplyReply

It's absolutely fine for an artist to say "I like to work this way even if it won't deliver the best results". But for someone who is painstakingly testing technical quality, that argument does not fly.

+1.  At the end of day everybody should shoot what makes him happy and what inspires him and if that's a Sony then that is a Sony.

Why all these posts that don't prove anything?  Sell the damned thing, make a trip with the $15K, take pictures with your Sony and be happy!!!
« Last Edit: March 05, 2014, 10:57:22 PM by JV » Logged
ErikKaffehr
Sr. Member
****
Offline Offline

Posts: 7255


WWW
« Reply #161 on: March 06, 2014, 12:07:24 AM »
ReplyReply

Hi,

There are two bits to it. One is that we may have different preferences on sharpening. I felt that yours is a bit over the edge. I have tested both your sharpening and Paul's. I am quite happy with Paul's settings in LR, but sharpening image in FM may go over the edge. FM can choose different settings, of course.

It's good to have found out that the deficiency Eric, Synn and Jerome and some others have found seems to be lack of sharpening. I can look more into that, a good learning experience.

The other point is that any improved sharpening I can also apply to the DSLR image, so the difference will still be small. I actually tested using Capture One this time but didn't go to print, as colour from the two cameras was much different in C1. In LR both cameras gave essentially identical colour, with the same WB. I guess it is due to my DCP profile for the P45+.

I would also make a point on "MP" both Ken R and Jerome makes the point that about twice the resolution is needed for a significant difference in print. It happened in this case as the DSLR was cropped down to 20MP and the P45+ image was stitched from two images and was 40 MP after cropping. The subject didn't fit within the FOV of the 120 lens and I could not move back.

Jerome and Synn say that bad light doesn't work well with MF. This was heavy overcast with very lush greens. I don't really like the idea of a camera that doesn't work in the forrest under overcast conditions. That is stuff I like to shoot! Jerome and Synn perhaps suggest acceptable light is light overcast with some hazed sunlight?

As I also said several times the prints are very similar when viewed with the naked eye, but a 5.5x loupe shows a very significant advantage for the P45+. So the differences are carried over to the print.

Eyesight is involved of course. Young eyes may see a difference although elderly may not. I have tested with a somewhat younger friend experienced in the area. I guess that you need experienced observers, as the DoF in the MF image is quite thin, making most parts of the image being sharper.

A bit of a side note, someone mentioned that a technical camera with the best wide angles doesn't need sharpening. I have seen that in Doug Peterson's library images. Those images were very sharp even without sharpening. Combination of very sharp lens and no OLP filtering. Really Right Stuff.

I absolutely agree with JV-s point. Shooting opportunities like travel are probably a better spending than a MFD equipment if printed reasonably small.

This is what I found.

I guess it's time for me to end this thread. I will not lock it, comments are welcome. But I feel the issue has been explored to enough depth.

Best regards
Erik

The point of posting sharpened and non sharpened versions was to show that there's a lot more that can be done with erik's files than his conversions have shown so far. The point is also that when processed properly, prints from 39MP MFD files do look better than with a 24MP DSLR. Sharpness is only one part of the equation and this can be done to taste.

so is WB for that matter. Yellow grass or green, the points still stand.

Erik, all I hope to achieve with that demo is to make you open your eyes a bit and realize that MFD is capable of delivering much better results than what you have managed to get out of the files so far. Perhaps this will make you stop generalizing every possible scenario for comparing MFD and 35mm based on your suboptimal workflow. What I fail to understand is that you are interested in objective analysis, but when shown a better workflow, you stick to personal preferences that most certainly adds bias and user error to the equation. It's absolutely fine for an artist to say "I like to work this way even if it won't deliver the best results". But for someone who is painstakingly testing technical quality, that argument does not fly.

I do agree with Jerome though, the lighting in the scene is suboptimal. The solution is not to shoot a test chart, but just to shoot real life scenes in better light. It helps one develop as a photographer too.
« Last Edit: March 06, 2014, 12:39:37 AM by ErikKaffehr » Logged

synn
Sr. Member
****
Offline Offline

Posts: 398



WWW
« Reply #162 on: March 06, 2014, 12:27:41 AM »
ReplyReply


The other point is that any improved sharpening I can also apply to the DSLR image, so the difference will still be small.


You're still not reading what I am saying. The problem is not the sharpening. You can apply as much or as little as you want (Which is why I posted sharpened and unsharpened versions for your reference). Why are you talking as if the unsharpened version doesn't exist?

The problem is that your MFD workflow is flawed right from the start. You're using an inferior RAW processor, you're applying all the wrong settings in it and you're getting dud output. Which is not a problem by itself, but then you extrapolate that in every single thread saying that this is the best that can be achieved with MFD, ANY MFD by anyone; period. Which is just flat out wrong.


Jerome and Synn say that bad light doesn't work well with MF. This was heavy overcast with very lush greens. I don't really like the idea of a camera that doesn't work in the forrest under overcast conditions. That is stuff I like to shoot! Jerome and Synn perhaps suggest acceptable light is light overcast with some hazed sunlight?


This is also not what I said. I asked you to look for good light as a photographer to make pleasing images, which also happen to be good test images.

This isn't a forest, but it's a test scene I shot under overcast conditions and it happens to have quite a bit of greenery in it. I just processed it using a better workflow. (And please don't start the Kodak/ Dalsa thing again. Send me a Kodak sensor back and I will send back a  similarly processed file).



Cameras can be used under a variety of lighting conditions. MFD is no exception. It just so happens that no matter what camera you use, you get better results with better light. And understanding good light is an important part of being a photographer.


As I also said several times the prints are very similar when viewed with the naked eye

Again, prints made with your heavily flawed workflow. Why are you STILL not getting this? Doesn't matter how good the viewer's eyesight is; you're delivering a flawed product for him to view.


I guess it's time for me to end this thread. I will not lock it, comments are welcome. But I feel the issue has been explored to enough depth.

That's perfectly OK, but please don't start yet another thread comparing (for whatever reasons) images from two systems processed with flawed workflow that seriously negates the image quality advantage one has over the other. We can only point out these things to you so many times.
« Last Edit: March 06, 2014, 12:29:17 AM by synn » Logged

my portfolio: www.sandeepmurali.com
jerome_m
Sr. Member
****
Offline Offline

Posts: 522


« Reply #163 on: March 06, 2014, 12:57:03 AM »
ReplyReply

Just to make some points clear:
-I do not prefer images with more sharpening. I find the sharper image here unnatural. I even posted a picture of mine with what I consider more natural sharpness.
-the picture is actually nice. The light on the moss is nice. This is actually an image that you could print and hang on a wall. I am not saying that the light is bad, I am saying that the light is not adapted if one wants to judge sharpness at the pixel level
-we still do not know whether your camera, lenses and software work properly. I would still appreciate a simple picture of natural subjects (grass, nettles, stones) taken in flat daylight similar to the one I posted as example.
Logged
ErikKaffehr
Sr. Member
****
Offline Offline

Posts: 7255


WWW
« Reply #164 on: March 06, 2014, 01:09:29 AM »
ReplyReply

Hi,

Thanks, your points make sense.

Best regards
Erik

Just to make some points clear:
-I do not prefer images with more sharpening. I find the sharper image here unnatural. I even posted a picture of mine with what I consider more natural sharpness.
-the picture is actually nice. The light on the moss is nice. This is actually an image that you could print and hang on a wall. I am not saying that the light is bad, I am saying that the light is not adapted if one wants to judge sharpness at the pixel level
-we still do not know whether your camera, lenses and software work properly. I would still appreciate a simple picture of natural subjects (grass, nettles, stones) taken in flat daylight similar to the one I posted as example.
Logged

BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3461


« Reply #165 on: March 06, 2014, 02:21:45 AM »
ReplyReply

-we still do not know whether your camera, lenses and software work properly. I would still appreciate a simple picture of natural subjects (grass, nettles, stones) taken in flat daylight similar to the one I posted as example.

Hi Jerome,

I agree with the first part of your comment. However, while good natural subject rendering is the goal, I do not think that natural objects are good for isolating the issues. I think a few test shots of a Star target will give a much less ambiguous basis for analysis of the root causes(es).

I had a look at the Raw files that Erik posted, and had a very hard job achieving a good(!) looking sharpening that also holds up on enlargement. Horrible sharpening is easy, but natural sharpening was much harder. The results seem to lack real sharpness, but with aliasing added as a sort of compromise. Sharpening tends to 'enhance' the aliasing and stairstepping artifacts more than real sharpness. It's unclear to me whether the exposure time (and perhaps camera vibrations) is a cause, or the lens is not up to the challenge, or something else is spoiling a good party, but when I compare the almost razor sharp results from my 1Ds3 with AA-filter with what I see from Erik's camera+back, I'm disappointed in what the DB produces.

I also get much better demosaicing results from the same file with Capture One Pro, compared to ACR, allowing much more sharpening (e.g. with FocusMagic, or Topaz Labs Detail) before the image falls apart. Sharpening of Non-AA-filtered images is more difficult, but such images usually also require less sharpening, unlike Erik's images.

Color rendering and tonality are something that can be adjusted quite well in Capture One, but we know that Erik prefers the LR workflow.

I'd suggest a more methodical, objective, and quantifiable, analysis to isolate the Root causes before shooting all sorts of wind-blown variable lighting natural subjects, but that's just how I'd approach it.

Cheers,
Bart
Logged
Hans van Driest
Newbie
*
Offline Offline

Posts: 20


« Reply #166 on: March 06, 2014, 02:22:51 AM »
ReplyReply

I thought the original subject was whether one would see a difference between a MFDB and a DSLR, in a 'small' print. it is possible to argue about all sort of subtle differences, that might be important to someone, but not to somebody else. Erik suggested to, when considering going to digital medium format, download some files and print this at a size that one would normally print on.
sound advice.
I am an engineer and that perhaps makes one a bit funny. I like to break problems down in simple sub-problems and look at each of those individuality. for me it seems that resolution is a important part of the likely differences (I am aware there is lot more).
I already described before how I had, for myself, made a comparison between an 24Mp file and one with four time the linear resolution, by means of making a multi row stitch with a lens of four times the focal length. but this is unnecessarily complicated. for just examining whether resolution makes a big difference, one could make a picture of ones normal subject, a landscape, whatever, with say a 50mm lens. using a tripod. after that shot, just replace the lens, without moving the camera, say with a 105 or 135mm lens. the idea is just to roughly double the focal length. as a next step, crop the result of the shorter fl to the same field of view. The same crop as obtained with the longer lens . one now has two identical pictures (i hope), with one having over double the linear resolution (four times the pixels). print them both at half the linear size of a 'normal' print (so 10x12, would be the size if one normal prints 20x24). And now just look if it makes a significant difference. if not, resolution of a digital mf camera is not going to make a significant difference. other things might.
Logged
hjulenissen
Sr. Member
****
Offline Offline

Posts: 1666


« Reply #167 on: March 06, 2014, 04:03:20 AM »
ReplyReply

so happens that no matter what camera you use, you get better results with better light. And understanding good light is an important part of being a photographer.
I usually find it more pleasing to watch portraits of interesting/pretty people than mundane people.

I don't think that an evaluation of camera capabilities needs to include pretty people, though. I take it that people with an interest in photography are able to make that connection themselves.

-h
Logged
synn
Sr. Member
****
Offline Offline

Posts: 398



WWW
« Reply #168 on: March 06, 2014, 04:22:22 AM »
ReplyReply

I usually find it more pleasing to watch portraits of interesting/pretty people than mundane people.

I don't think that an evaluation of camera capabilities needs to include pretty people, though. I take it that people with an interest in photography are able to make that connection themselves.

-h

That does not negate my point, nor is it contrary.
Logged

my portfolio: www.sandeepmurali.com
paul ross jones
Newbie
*
Offline Offline

Posts: 17


« Reply #169 on: March 06, 2014, 04:25:21 AM »
ReplyReply

3 (the man showing the leather) and 12, 13, 14 (the middle age scenes)? But I am not sure about 23 (city) and 27 (car circuit) either.

 

Indeed. The shallowest depth of field can be had on 24x36 camera, because of the very fast lenses unique to this format.




hi jerome, you got the slot car circuit correct.
the shots with red frames are the p65/contax.

i have to say that the p65 is a very good file, heaps of res and headroom with the highlights (although, bad in the darks). but the "look" isn't that distinctive imo.

i think the lenses give more of a difference of look than just a format.

paul

Logged
BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3461


« Reply #170 on: March 06, 2014, 06:15:03 AM »
ReplyReply

I thought the original subject was whether one would see a difference between a MFDB and a DSLR, in a 'small' print.

Hi Hans,

That's correct, but it is relatively obvious that either human visual acuity, or inherent image resolution after upsampling (less likely for a 'small' print), is the deciding factor. Besides that, there are more subtle (MTF related) differences, although many of those can be bridged by proper postprocessing.

Quote
I like to break problems down in simple sub-problems and look at each of those individuality. for me it seems that resolution is a important part of the likely differences (I am aware there is lot more).

Indeed, resolution as in sampling density is an important deciding factor.

Quote
For just examining whether resolution makes a big difference, one could make a picture of ones normal subject, a landscape, whatever, with say a 50mm lens. using a tripod. after that shot, just replace the lens, without moving the camera, say with a 105 or 135mm lens. the idea is just to roughly double the focal length. as a next step, crop the result of the shorter fl to the same field of view. The same crop as obtained with the longer lens . one now has two identical pictures (i hope), with one having over double the linear resolution (four times the pixels). print them both at half the linear size of a 'normal' print (so 10x12, would be the size if one normal prints 20x24). And now just look if it makes a significant difference. if not, resolution of a digital mf camera is not going to make a significant difference. other things might.

In principle that should work, if not for the introduction of several variables at the same time. Different lenses have different performance, and different focal lengths have different MTF for a given level of detail (cycles/mm), and different Depth of Field. Resampling the smaller sized version for output also introduces a variable, but that is maybe the most stable one for a straight comparison between larger and smaller sensel array sizes.

If it is only the number of pixels available for printing that one wants to compare, it's perhaps easier to compare a (several) down-sampled version(s) of the same image and print those at the same output size. Of course a very good down-sampling method should be used. However, the conclusion will probably be the same as the one I mentioned first, although good up-sampling may narrow the gap with real pixel content to a degree.

With specific image resolution (sampling density) out of the way, the rest of the platform induced differences will be of an optical and image magnification nature, and color rendering design choices.

Cheers,
Bart
Logged
Hans van Driest
Newbie
*
Offline Offline

Posts: 20


« Reply #171 on: March 06, 2014, 06:52:53 AM »
ReplyReply

Hi Hans,

That's correct, but it is relatively obvious that either human visual acuity, or inherent image resolution after upsampling (less likely for a 'small' print), is the deciding factor. Besides that, there are more subtle (MTF related) differences, although many of those can be bridged by proper postprocessing.

Indeed, resolution as in sampling density is an important deciding factor.

In principle that should work, if not for the introduction of several variables at the same time. Different lenses have different performance, and different focal lengths have different MTF for a given level of detail (cycles/mm), and different Depth of Field. Resampling the smaller sized version for output also introduces a variable, but that is maybe the most stable one for a straight comparison between larger and smaller sensel array sizes.

If it is only the number of pixels available for printing that one wants to compare, it's perhaps easier to compare a (several) down-sampled version(s) of the same image and print those at the same output size. Of course a very good down-sampling method should be used. However, the conclusion will probably be the same as the one I mentioned first, although good up-sampling may narrow the gap with real pixel content to a degree.

With specific image resolution (sampling density) out of the way, the rest of the platform induced differences will be of an optical and image magnification nature, and color rendering design choices.

Cheers,
Bart

I agree the lens would play a role, but when the difference is as extreme as a factor of two or more in linear resolution, I would expect (assuming two decent lenses), this to be secondary. Indeed down sampling an existing file is also an option, although the down sampling will increase the pixel quality a lot (assuming a Bayer mask sensor). but in the end the idea would be to see of how much influence a big increase in resolution would have on the quality of the printed output, for a certain size. either method would do fine.

In my own experiment, with a stitched file to get the same field of view with four times the linear resolution, I was surprised how 'little' of this extra quality, so visible at 100% on screen, was visible in a 20x30 inch print. Although I must admit that the longer I look at both prints, the more visible the differences become. This is at close distance. Looking from a 'normal' viewing distance, there is little or nothing to differentiate the two print. My eyes are above average.

kind regards,

Hans
Logged
jerome_m
Sr. Member
****
Offline Offline

Posts: 522


« Reply #172 on: March 06, 2014, 07:34:00 AM »
ReplyReply


hi jerome, you got the slot car circuit correct.
the shots with red frames are the p65/contax.

i have to say that the p65 is a very good file, heaps of res and headroom with the highlights (although, bad in the darks). but the "look" isn't that distinctive imo.

i think the lenses give more of a difference of look than just a format.

Certainly it is the lenses and what I used as a guide were lens aberrations. Interestingly, it failed me. What camera/lens combination and aperture did you use for the leather (3) and middle age shoots?
Logged
jerome_m
Sr. Member
****
Offline Offline

Posts: 522


« Reply #173 on: March 06, 2014, 07:44:58 AM »
ReplyReply

I agree with the first part of your comment. However, while good natural subject rendering is the goal, I do not think that natural objects are good for isolating the issues. I think a few test shots of a Star target will give a much less ambiguous basis for analysis of the root causes(es).

Except that this subject is designed to ignore the effects of what we want to test:
-it is a flat subject, so we ignore all lens aberrations which arise outside of the plane of focus and these are a very important part of the "MF look" and even the effect of sharpening (sharpening influences the rendering of infocus to outfocus transitions on tridimensional subjects)
-it is a high-contrast monochrome subject, so we are in an ideal case for demosaicing and noise reduction. In my experience, grass is one of the most difficult subjects for post processing.
Logged
ErikKaffehr
Sr. Member
****
Offline Offline

Posts: 7255


WWW
« Reply #174 on: March 06, 2014, 01:06:18 PM »
ReplyReply

Hi Hans,

You are right about the original subject, and I feel that your suggestions make a lot of sense.

Best regards
Erik

I thought the original subject was whether one would see a difference between a MFDB and a DSLR, in a 'small' print. it is possible to argue about all sort of subtle differences, that might be important to someone, but not to somebody else. Erik suggested to, when considering going to digital medium format, download some files and print this at a size that one would normally print on.
sound advice.
I am an engineer and that perhaps makes one a bit funny. I like to break problems down in simple sub-problems and look at each of those individuality. for me it seems that resolution is a important part of the likely differences (I am aware there is lot more).
I already described before how I had, for myself, made a comparison between an 24Mp file and one with four time the linear resolution, by means of making a multi row stitch with a lens of four times the focal length. but this is unnecessarily complicated. for just examining whether resolution makes a big difference, one could make a picture of ones normal subject, a landscape, whatever, with say a 50mm lens. using a tripod. after that shot, just replace the lens, without moving the camera, say with a 105 or 135mm lens. the idea is just to roughly double the focal length. as a next step, crop the result of the shorter fl to the same field of view. The same crop as obtained with the longer lens . one now has two identical pictures (i hope), with one having over double the linear resolution (four times the pixels). print them both at half the linear size of a 'normal' print (so 10x12, would be the size if one normal prints 20x24). And now just look if it makes a significant difference. if not, resolution of a digital mf camera is not going to make a significant difference. other things might.

Logged

tho_mas
Sr. Member
****
Offline Offline

Posts: 1696


« Reply #175 on: March 06, 2014, 04:23:14 PM »
ReplyReply

Synn - great image!
Is it a single image or part of a series? If so I'd love to see it.
Many thanks!
Logged
synn
Sr. Member
****
Offline Offline

Posts: 398



WWW
« Reply #176 on: March 06, 2014, 07:26:47 PM »
ReplyReply

Synn - great image!
Is it a single image or part of a series? If so I'd love to see it.
Many thanks!


Hi Thomas,

Thanks, that's just my regular "Test shot scene" out of my window. Smiley

I use this frame as my test scene every time I get some new piece of camera gear. This was taken to test the Mamiya 35mm lens. Shot was made with a 3 stop Lee grad ND on the lens.
Logged

my portfolio: www.sandeepmurali.com
Fine_Art
Sr. Member
****
Offline Offline

Posts: 1063


« Reply #177 on: March 07, 2014, 01:30:07 AM »
ReplyReply

I mentioned months ago that there seemed to be a problem with the lens. If all the lenses are the same than the body must be badly adjusted.

I will donate a raw in overcast light to the discussion. It is part of a pano with not much meaning by itself (although quite pretty), so it can be out on the net. This is the type of mountain landscape with snow that Erik likes.
http://www.sendspace.com/file/ke2pce

The .nef is in a blue box. you should see the .nef extension when you hover your mouse. Ignore their ads.

In overcast there should be abundant D/R.
It should be sharp at 100% view. If your trees look much worse check your equipment.
Logged
tho_mas
Sr. Member
****
Offline Offline

Posts: 1696


« Reply #178 on: March 07, 2014, 03:23:05 AM »
ReplyReply

that's just my regular "Test shot scene" out of my window. Smiley
Grin however, it's interessting. The scene looks a bit artificial (the subject, not the photo) and this is why I like it.
Logged
BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3461


« Reply #179 on: March 07, 2014, 04:21:10 AM »
ReplyReply




I agree with the first part of your comment. However, while good natural subject rendering is the goal, I do not think that natural objects are good for isolating the issues. I think a few test shots of a Star target will give a much less ambiguous basis for analysis of the root causes(es).

Except that this subject is designed to ignore the effects of what we want to test:
-it is a flat subject, so we ignore all lens aberrations which arise outside of the plane of focus and these are a very important part of the "MF look" and even the effect of sharpening (sharpening influences the rendering of infocus to outfocus transitions on tridimensional subjects)
-it is a high-contrast monochrome subject, so we are in an ideal case for demosaicing and noise reduction. In my experience, grass is one of the most difficult subjects for post processing.

Hi Jerome,

I disagree. To ferret out potential lens issues, one should use a methodical test that eliminates as many variables as possible. So, a flat subject is preferred, it will show (when focused well) the maximum in achievable resolution. When that resolution is lacking, the rest becomes moot. When camera shake is detected, technique can be improved. When all corners, individually focused for the best resolution to eliminate field flatness and other lens aberration issues, show the same kind of resolution, then the lens is well centered, otherwise there is a lens decentering issue (not uncommon, especially with second hand lenses).

For all that, it is best to reduce the influence of the Raw converter, so a monochrome grayscale target is preferred. It also produces a nice test object to compare Raw converters with the same image.

When we have established that the lens is adequate, then we can grill the Raw converter and Capture sharpening workflow.

When all that is done, we can test the lens rendering quality in front or in the back of the focus plane, at various apertures. At least we have now established a solid correctly focused basis to compare with.

So we agree that real image performance is the ultimate goal, and that may lead to purchasing different lenses for their specific image qualities. But for a methodical test benchmark, problem solving, it's best to reduce the number of variables to those things that can be individually quantified and perhaps addressed.

Cheers
Bart
Logged
Pages: « 1 ... 7 8 [9] 10 11 ... 15 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad