The GetDPI Photography Forum

Great to see you here. Join our insightful photographic forum today and start tapping into a huge wealth of photographic knowledge. Completing our simple registration process will allow you to gain access to exclusive content, add your own topics and posts, share your work and connect with other members through your own private inbox! And don’t forget to say hi!

Hi another comparison of DPReview samples

ErikKaffehr

Well-known member
Hi,

Here is another comparison:

Nikon D810Fuji GFX
Canon 5DsRPhase One IQ3-100MP

Central text: http://echophoto.dnsalias.net/ekr/Articles/DPReview/Comparison_1/CentralText.png

Corners: http://echophoto.dnsalias.net/ekr/Articles/DPReview/Comparison_1/Corner.png

Moiré: http://echophoto.dnsalias.net/ekr/Articles/DPReview/Comparison_1/Moire.png

Processing:

  • WB on colour checker
  • No sharpening or noise reduction in LR CC
  • Export to tiff 80x120cm at 180 PPI
  • Open in Photoshop and apply Focus Magick (default sharpening)
  • Compare at actual pixels

Best regards
Erik
 

TsurTriger

New member
Text: IQ, Fuji, Canon, Nikon
Corner: I will call it almost a draw between Canon, Fuji and IQ
Moise: IQ, Canon, Fuji, Nikon.

The winner is IQ3 for my opinion.
 

Quentin_Bargate

Well-known member
I have seen these resolution tests done between various cameras before and they rarely give any useful insight in to real world performance. I remember taking the late Michael Reichmann test in his Toronto gallery, in which he challenged visitors to say whether prints were from a MFDB or a Canon G10 compact. Like everyone before me, I did pretty poorly!

All this test, like the others before it, proves is that all good digital cameras are great tools and certainly at 36mp and more, the limiting factor becomes the photographer, not the equipment.
 

ErikKaffehr

Well-known member
Hi Quentin,

Yes I would agree. Most cameras are good enough for most purposes, indeed.

To that comes that optimal focus is only possible in a single plane. In architecture photography and landscape photography that may be at infinity.

It is more important that we have a system that works well for our purpose.

On the other hand, comparisons like this sort of give an indication on what you get for the expenditure.

One of the reasons I posted this was because Jerome's posting differed a bit from what I usually see.

Best regards
Erik


I have seen these resolution tests done between various cameras before and they rarely give any useful insight in to real world performance. I remember taking the late Michael Reichmann test in his Toronto gallery, in which he challenged visitors to say whether prints were from a MFDB or a Canon G10 compact. Like everyone before me, I did pretty poorly!

All this test, like the others before it, proves is that all good digital cameras are great tools and certainly at 36mp and more, the limiting factor becomes the photographer, not the equipment.
 

glenerrolrd

Workshop Member
Isn t the issue here that you can t compare ...straight out of the camera files ? First manufactures develop FIRMWARE that renders files even RAW files into a semi cooked status .(some require more sharpening in post than others ) . Second the CONVERSION SOFTWARE...applies a preset level of sharpening,noise reduction and a tone curve. Huge difference with a Leica M DNG thru LR verse C1 at standard presets .

To get a objective comparison of the cameras ..you would need to process each file with independent settings in the same software (generally Camera Raw or LR) ....to assess the SYSTEM you would have to mix in the “best in class” raw conversion software (where available Capture One ).

Of course LENSES MATTER ...and it appears that particularly the new Fuji lenses have built in camera processing into the formula ..correcting lens limitations BEFORE ...raw conversion .

I know Diglloyd has become unpopular on GETdpi ...BUT ..his tests look to be done the right way . Sean Reid also does a great job of comparing APPLES TO APPLES .....and has shown that in camera lens profiling might improve overall IQ at the expense of edge sharpness.
 

ErikKaffehr

Well-known member
Hi,

What I can see the IQ3-100MP delivers significantly cleaner images than the others, so why not hang onto it? But, those differences may not be obvious to you?

Also, keep in mind that this was basically a resolution comparison based on a 80x120 cm print.

It doesn't say a thing about DR or color rendition. But it shows quiet clearly what differences you would see between four different systems in an 80x120 cm print.

Best regards
Erik

Based on what I see I am selling my IQ3100 and getting the Canon 5D.:ROTFL:
 

jerome_m

Member
I see artefacts of the debayering process in some images. It is most visible on diagonal lines, for example in the corner samples. In one direction, the diagonal lines look like dotted lines, which is not normal. I'll include a sample with an enlargement to show what I mean.

The artefacts are visible on the source images on dpreview, so the error is to be found there I suppose. It is not all cameras. The artefacts are not visible on jpegs.


artefacts.jpg
 

Quentin_Bargate

Well-known member
I knew the Sigma Merrills were the best :D

I'm not sure I'm joking either.... Those low level artefacts causing colour aliasing, jaggies etc are not present in a true colour sensor.
 
Based on what I see I am selling my IQ3100 and getting the Canon 5D.:ROTFL:
Based on what I see your IQ3100 is still the best! The cross-format opponent for the IQ3100 is never the 50MP 5DSR, but products in the future! (Sony sensor engineer told me about 70MP in 35mm format, 114MP in 44x33 format and 170MP in 645 format, but who knows when these will be materialized?). Before that, you can safely enjoy the throne of resolution without getting harassment from smaller formats :thumbs:
 
I have seen these resolution tests done between various cameras before and they rarely give any useful insight in to real world performance. I remember taking the late Michael Reichmann test in his Toronto gallery, in which he challenged visitors to say whether prints were from a MFDB or a Canon G10 compact. Like everyone before me, I did pretty poorly!

All this test, like the others before it, proves is that all good digital cameras are great tools and certainly at 36mp and more, the limiting factor becomes the photographer, not the equipment.
Would you post some real-world comparisons to show that the lab tests are not correlated to real-world comparisons? (To be clear, comparison needs to be at pixel-peeping level, and done at the same time and place, with the same angle of view, by the same skillful photographer, so that photography skill is not a varying factor.)
 

ErikKaffehr

Well-known member
Hi Jerome,

It is known as the zipper artefact. It is a form of aliasing. Some demosaic algorithm can suppress zipper artefacts.

The reason it shows up here, especially much on the Nikon image, is that the Nikon image is upsized to the given print size. Sharpening is applied after the upsizing and it is pretty strong.

This thread may give some useful info: http://forum.luminous-landscape.com/index.php?topic=104708.0

Best regards
Erik


I see artefacts of the debayering process in some images. It is most visible on diagonal lines, for example in the corner samples. In one direction, the diagonal lines look like dotted lines, which is not normal. I'll include a sample with an enlargement to show what I mean.

The artefacts are visible on the source images on dpreview, so the error is to be found there I suppose. It is not all cameras. The artefacts are not visible on jpegs.


View attachment 125904
 

Quentin_Bargate

Well-known member
Would you post some real-world comparisons to show that the lab tests are not correlated to real-world comparisons? (To be clear, comparison needs to be at pixel-peeping level, and done at the same time and place, with the same angle of view, by the same skillful photographer, so that photography skill is not a varying factor.)
No, because the tests that matter are not the tests you ask for. There are too many variables apart from resolution, including but not limited to focus accuracy, ISO performance, shutter vibration, lens quality, image stabilisation, software used, sharpening, test conditions, sensor design, and more. The Reichmann test example I gave showed that. We live in a post-resolution world. Factors other than the headline figures have become so important that they render your suggested test largely irrelevant. Its all about subjectivity.
 
No, because the tests that matter are not the tests you ask for. There are too many variables apart from resolution, including but not limited to focus accuracy, ISO performance, shutter vibration, lens quality, image stabilisation, software used, sharpening, test conditions, sensor design, and more. The Reichmann test example I gave showed that. We live in a post-resolution world. Factors other than the headline figures have become so important that they render your suggested test largely irrelevant. Its all about subjectivity.
These can still be controlled. If you ask the same skillful photographer to carry out the tests, focus accuracy can be offset by focus bracketing (and cherry picking the best), ISO can be set to the lowest for each camera, shutter vibration can be minimalized by delayed release, the best available lens can be picked for each system (with sample variation taken into account), the best or the same software and sharpening technique can be chosen for each camera, sensor design is a fixed factor for each camera.

When you ask the same photographer to shoot a long series of pictures with two camera systems after he maximizes his proficiency with both systems, and take the average image quality of each camera system, it is likely that the comparison would be correlated to lab test results under controlled environment.
 

ErikKaffehr

Well-known member
Why this comparison was done and what I see…

Hi,

What I wanted to see was what differences would be visible in a decent size print. The print size is around A0. So, if we crop the images to 50% (linearly) and print in A2 this is about what we would see at 180PPI.

It seems that 180 PPI is needed for what most viewers regard to be a very good print.

Incidentally, A2-size at 180 PPI is pretty close to 50MP.

Four images were chosen. The rationale was to see the effect of increase in resolution, the effect of increased sensor size. The cameras were:

  • Nikon D810 - lowest resolution
  • Canon 5DsR - highest resolution 24x36 mm
  • Fuji GFX - 50 MP in the 44x33 mm MFD format - same resolution as Canon
  • Phase One IQ3-100MP. Highest resolution MFD camera at full frame 645

So, what did I find?

  • The difference between the Nikon and the 50 MP cameras was quite significant.
  • The Canon and the Fuji GFX were amazingly close. With good lenses, pixel quantity may be more important than sensor size.
  • The Canon did really well with it's cheap 85/1.8 lens, at least at optimum aperture.
  • As may be expected, the Fuji GFX was most moiré prone, but the only camera without significant moiré was the Phase One IQ3-100 MP.

This comparison doesn't cover a lot of important areas, like:
  • Autofocus, flexibility and accuracy
  • Lens programme
  • Frames per second
  • Dynamic range
  • Weight
  • Price

I would guess that any of the cameras would make great images at A2 size.

Would cost, weight and flexibility not matter, I would buy the Phase One IQ3 100 MP. No question about it. But cost, weight and flexibility do matter.

Best regards
Erik
 

Quentin_Bargate

Well-known member
Pity, Erik, the Pentax KI was not included, in pixel shift mode. Pixel shift completely avoids moire and is visibly superior to unshifted mode by some margin - maybe 60mp? One test I do plan on is the KI versus the GFX.
 

jerome_m

Member
Hi Jerome,

It is known as the zipper artefact. It is a form of aliasing. Some demosaic algorithm can suppress zipper artefacts.

The reason it shows up here, especially much on the Nikon image, is that the Nikon image is upsized to the given print size. Sharpening is applied after the upsizing and it is pretty strong.
I am not sure I understand correctly, but if continuous lines are resolved visually differently according to their orientation, what are we examining? Are we examining the capacity of the camera or are we fooling ourselves and examining the capacity of the demosaicing, upscaling and sharpening algorithm?
 

ErikKaffehr

Well-known member
Hi,

Take a bit of checkered paper and draw a few lines across it with a pen of half the thickness of the checkered pattern.

You will see that the area of the boxes covered by the line varies, the line will sometimes pass trough a single square sometimes across two. This is one of the reasons for spatial aliasing.

Once you are at the pixel level, there is no way a line can be correctly rendered if it is not correctly aligned with a pixel row or column. But, even two adjacent pixels will have different filters.

Correct rendition is not possible. So OLP filtration is needed to spread small detail information over more pixels. But, that obviously reduces resolution.

Check what a resolution trumpet looks like in a Foveon image, developed in Iridients raw developer:
Screen Shot 2017-03-28 at 23.07.57.jpg

Even if the sensor lacks Bayer filtering, correct reproduction of the resolution trumpet is not possible.

Just to say, that is a great reason to have many pixels…

Best regards
Erik





I am not sure I understand correctly, but if continuous lines are resolved visually differently according to their orientation, what are we examining? Are we examining the capacity of the camera or are we fooling ourselves and examining the capacity of the demosaicing, upscaling and sharpening algorithm?
 

jerome_m

Member
You are not explaining why two crossing lines at 45° respective to the sensor array give different results...
 

algrove

Well-known member
I was only joking as nothing can pry me away from the IQ3100.:grin:

Hi,

What I can see the IQ3-100MP delivers significantly cleaner images than the others, so why not hang onto it? But, those differences may not be obvious to you?

Also, keep in mind that this was basically a resolution comparison based on a 80x120 cm print.

It doesn't say a thing about DR or color rendition. But it shows quiet clearly what differences you would see between four different systems in an 80x120 cm print.

Best regards
Erik
 
Top