100mp Phase One vs Canon 5DSR and A7RII: Architectural photography

[jduncan]

Hi,

Here we come again. This time is different: When he tested the H6D he shoot the MF system himself and he is a dude that can't make the sigma 135mm art to work[1] . Not exactly the best to deal with MF quirks etc. He also admitted in a posterior video that he fas a feud with Hasselblad. In this location is different, he did give Phase One all the chances he could with a single exception:

1. All the pictures were processed using Capture One.
2. A Phase One expert shoot the P1 system.
3. This is the mistake: he compared the images at 100%, not at the same resolution.
4. He did not go into flash sync and the like, or the ability to use film etc. This is just for Architectural Photography.

The video: https://www.youtube.com/watch?v=vKhRvGqq7xQ&ab_channel=SonderCreative

I will say that his conclusions are interesting. MF has a lot to do. The days of using german sounding names and believing that was enough are gone. People are doing an actual comparison, I insist the "Häagen-Dazs" marketing trick is not enough. It's true that mirrorless is needed. Super high-resolution cameras demand almost perfect focus. Having the focusing system on the chip is far better than trying to align components. I do love the H series viewfinder. I remember the day, years ago, that I used for the first time the H4D and the 100mm f2.2, the details, the luminosity, a revelation. MF guys need to work a lot. Mainstream FF cameras are now including multishot, a technology that was the staple of MF a few years ago. And we don't have a H6D-400MS and I don't know if it will even make sense with the current Hasselblad glass.

The X1D and the Fuji are the path of the future[2]. I am eager to see P1 big sensor mirrorless. Not only is not all lost, but I believe that is a fantastic time for MF.



----------
[1] (https://www.youtube.com/watch?v=fNIDAaAURu4&ab_channel=SonderCreative)
[2] https://www.youtube.com/watch?v=wD3An09hmjU&ab_channel=KjellPost

 

[dchew]

It's true that mirrorless is needed. Super high-resolution cameras demand almost perfect focus. Having the focusing system on the chip is far better than trying to align components.

I keep hearing this but I don't understand it. All three systems mentioned in your title have live view. Shooting architecture is probably going to be on a tripod. What difference does mirrorless make in this case? I admit I don't use a DSLR any more, so maybe I am missing something.

Dave

 

[jduncan]

I keep hearing this but I don't understand it. All three systems mentioned in your title have live view. Shooting architecture is probably going to be on a tripod. What difference does mirrorless make in this case? I admit I don't use a DSLR any more, so maybe I am missing something.

Dave

Hi
You are totally correct. My comment was not about architecture, but in general. The context does not make it plain.

Best regards,

 

[Boinger]

Isn't this the same guy that said the hasselblad h6d was worse than an iphone...

 

[yaya]

It may sound a bit harsh but looking at the author's website it seems as if his work could benefit from using a camera with a greater dynamic range and possibly less distortion...or from practicing more careful capture and post techniques...

BR

Yair

 

[jerome_m]

Or we could save our time and not watch youtube videos with incompetent "photographers" who spend 10 minutes reading a pre-written 2000 words text in front of the video camera?

 

[drunkenspyder]

Or we could save our time and not watch youtube videos with incompetent "photographers" who spend 10 minutes reading a pre-written 2000 words text in front of the video camera?

Done. Or not, as the case may be.

 

[ErikKaffehr]

Hi,

The mistake he makes is quite significant, after all, he is comparing the two image for sharpness. The 100 MP back has something like twice the pixels, so looking at the images at 1:1 means the P1 images are around 45% magnified.

Doing a proper comparison is not easy. Downscaling to common size throws away resolution. Upscaling an image to 145% will mostly to cause it to fall apart in 1:1 view.

One way to do such comparisons is to assume a print size, say 30"x40" and resize to say 180PPI for that format.

Best regards
Erik

Hi,

Here we come again. This time is different: When he tested the H6D he shoot the MF system himself and he is a dude that can't make the sigma 135mm art to work[1] . Not exactly the best to deal with MF quirks etc. He also admitted in a posterior video that he fas a feud with Hasselblad. In this location is different, he did give Phase One all the chances he could with a single exception:

1. All the pictures were processed using Capture One.
2. A Phase One expert shoot the P1 system.
3. This is the mistake: he compared the images at 100%, not at the same resolution.
4. He did not go into flash sync and the like, or the ability to use film etc. This is just for Architectural Photography.

The video: https://www.youtube.com/watch?v=vKhRvGqq7xQ&ab_channel=SonderCreative

I will say that his conclusions are interesting. MF has a lot to do. The days of using german sounding names and believing that was enough are gone. People are doing an actual comparison, I insist the "Häagen-Dazs" marketing trick is not enough. It's true that mirrorless is needed. Super high-resolution cameras demand almost perfect focus. Having the focusing system on the chip is far better than trying to align components. I do love the H series viewfinder. I remember the day, years ago, that I used for the first time the H4D and the 100mm f2.2, the details, the luminosity, a revelation. MF guys need to work a lot. Mainstream FF cameras are now including multishot, a technology that was the staple of MF a few years ago. And we don't have a H6D-400MS and I don't know if it will even make sense with the current Hasselblad glass.

The X1D and the Fuji are the path of the future[2]. I am eager to see P1 big sensor mirrorless. Not only is not all lost, but I believe that is a fantastic time for MF.



----------
[1] (https://www.youtube.com/watch?v=fNIDAaAURu4&ab_channel=SonderCreative)
[2] https://www.youtube.com/watch?v=wD3An09hmjU&ab_channel=KjellPost

 

[DB5]

Ahhh youtubers. When naivety and confidence collide.

 

[dchew]

Ahhh youtubers. When naivety and confidence collide.

Classic Dunning-Kruger at "The Peak"...

 

[dave.gt]

Classic Dunning-Kruger at "The Peak"...

W. B. Yeates:

“The best lack all conviction, while the worst / Are full of passionate intensity."

Makes me wonder where I fit in... surely not at either extreme!

 

[MrSmith]

Ahhh youtubers. When naivety and confidence collide.

But what about those who post comments on these videos? (Along with those comments on the likes of DPreview/petalixel etc)
A new Low?
Is it possible to go beyond the nadir?

 

[TimG]

I spent a while looking at the files, and discovered that the ones from the 5D were very poor compared to the IQ in everywhere apart from the centre of the image (where both should be sharp, especially at F8 / F10);

From his files, on his video in the description; (and it's the same story all over the image except in the centre where both are sharp)

IQ100 / Rodenstock; (reduced in size to be the same as the 5D image)

5D - 24TSE; (native size)

The guy doesn't want to hear it though, he just has it in for medium format,

For comparison - there's a guy here who does what looks like a less biased comparision between a 5DS-R and an IQ140 with a rodenstock 40 on a RM3DI,.... The rodenstock/IQ easily wins as it should > https://www.youtube.com/watch?v=r_n2N0RQJVM https://www.youtube.com/watch?v=6Lc9_LztJL4

1. All the pictures were processed using Capture One.
2. A Phase One expert shoot the P1 system.
3. This is the mistake: he compared the images at 100%, not at the same resolution.
4. He did not go into flash sync and the like, or the ability to use film etc. This is just for Architectural Photography.

The video: https://www.youtube.com/watch?v=vKhRvGqq7xQ&ab_channel=SonderCreative

 

[ErikKaffehr]

Hi Tim,

I would suggest you mixed up something. I downloaded the TIFF images upsized the Canon 5DsR image to IQ3100MP height and this is what I see on the edge:




Why am I getting different results from yours? That is because you are looking at the wrong sign. The Canon 24TSE covers a wider angle than the Rodenstock lens on the Phase One. So the Phase One image does not show the sign that the Canon still includes.

So, what I think is that the images actually indicate that Mr Dawood may have been right. It is difficult to draw conclusions from just two images.

The Canon image I was looking at was at f/8. For the IQ3100 image I cannot tell the aperture as there is no EXIF.


These are 100% crops at 100 MP for the Phase One and over 100MP (upsized to match short dimension).

Best regards
Erik

I spent a while looking at the files, and discovered that the ones from the 5D were very poor compared to the IQ in everywhere apart from the centre of the image (where both should be sharp, especially at F8 / F10);

From his files, on his video in the description; (and it's the same story all over the image except in the centre where both are sharp)

IQ100 / Rodenstock; (reduced in size to be the same as the 5D image)

5D - 24TSE; (native size)

The guy doesn't want to hear it though, he just has it in for medium format,

For comparison - there's a guy here who does what looks like a less biased comparision between a 5DS-R and an IQ140 with a rodenstock 40 on a RM3DI,.... The rodenstock/IQ easily wins as it should > https://www.youtube.com/watch?v=r_n2N0RQJVM https://www.youtube.com/watch?v=6Lc9_LztJL4

 

[dchew]

So, what I think is that the images actually indicate that Mr Dawood may have been right. It is difficult to draw conclusions from just two images.
...
Best regards
Erik

Erik,
All I know is, if I was doing that shoot, came back with those images using my 40hr/3100, opened them in C1 and saw that, I would kick myself and head back to reshoot the photo. Something isn’t right there. Either he screwed up focusing or he drop-kicked that lens at some point.

Dave

 

[TimG]

Definitely the right sign and definitely the right files. (they're both in the image, one under the female mannequin)

Pretty easy to tell the difference.

At a second pass, I think the Linhof techno is problematic, as my own 32 looks sharper on my cambo than this 40 to be honest.

Hi Tim,

I would suggest you mixed up something. I downloaded the TIFF images upsized the Canon 5DsR image to IQ3100MP height and this is what I see on the edge:

 

[Geoff]

At a second pass, I think the Linhof techno is problematic, as my own 32 looks sharper on my cambo than this 40 to be honest.

Sometimes the Techno can be set up with the parallelism of the front lens panel just a wee bit off. There are stops to help get this right, but it doesn't hurt to check left vs. right side. What f stop was used? It seems odd the 40 is so soft, the lens has a stellar reputation.

 

[ErikKaffehr]

Hi,

When I first compared the Canon image to the Phase One image it was the f/8 image from Canon to the f/16 on Phase One. So, the Phase One image was in all probability degraded by diffraction.

Since that I have checked out all the Phase One images, and it seems that the tested sample doesn't perform to expectations. I mostly looked at the left side as there is a lot of good detail.

An issue with this comparison is that we don't get the raw images, thus we cannot say how much sharpening and noise reduction was applied. Capture One applies a lot of both sharpening and noise reduction per default and that default varies system to system. This behavior is not specific to Capture One, Lightroom does it to, but Lightroom's defaults to very little sharpening.

What I can see is that the images certainly certify Mr Dawood's findings. On the other hand, I would guess that he has a good sample of the 24/3.5 TSE LII. Regarding the Rodenstock/IQ3100 combo, Mr. Dawood states that the images were made by a Phase One specialist. My guess that the lens used is a pretty bad sample.

Mr. Dawood did some other comparisons like comparing macro shots with his Hasselblad H6D100c with the HC macro 120/4 and the Canon 100/2.8L IS. Again, it is difficult to evaluate his comparison without access to the raw files. It may be that he is a bit unfortunate with lens choice to compare and bad samples on the MFD side.

I would say that the tests are quite honest. The findings are not what I would expect.

I can mention a parallell. When the GFX arrived, it was tested by both Jim Kasson and Lloyd Chambers. Lloyd Chambers generally found a lot of issues with the GFX while Jim Kasson was extremely pleased. Lloyd Chambers has found a lot of issues with focusing on GFX, issues that Jim Kasson has confirmed to some part. But, even with focusing issues, the GFX outperformed any other camera Jim Kasson had. It also seemed that Jim Kasson got all the good lenses and Lloyd Chambers got all the bad lenses.

Jim also found that the Hasselblad H lenses he had were not even close to the GF-lenses. Jim's testing resulted in selling all his Hasselblad V and H-system lenses. Leica M-lenses were also sold, but he kept the Leica R-lenses.

The Hasselblad V 250/5.6 superachromat was actually a good performer but it was not very convenient on the GFX.

Best regards
Erik

Sometimes the Techno can be set up with the parallelism of the front lens panel just a wee bit off. There are stops to help get this right, but it doesn't hurt to check left vs. right side. What f stop was used? It seems odd the 40 is so soft, the lens has a stellar reputation.

 

[dchew]

Hi Erik,
I agree, with a few caveats:

First, I am not sure what qualifies someone to be a "Phase One Specialist." Specialist is a vague, undefined term. Could be a 20+ year employee of Phase One, a dealer, someone who shoots or used to shoot Phase One or someone who demo'd a Phase One back 2 years ago. We have no idea what it means. My guess is it is someone who has some undefined experience using a Phase One product, which may have nothing to do with a technical camera and the 40hr.

Second, and this is the big one for me: it is exceedingly difficult to perform tests and make overarching conclusions when testing something one is not intimately familiar with, unless there is access to extensive testing equipment like DXO, Lensrentals, Jim Kasson, etc. As much as I value some of what Lloyd Chambers does, I would not put him in that category (and certainly not myself!). Over the years I have occasionally tried to do what I initially thought was thorough testing with something very familiar. Two relatively recent examples were the goofy red-shadows w/long exposure testing I got dragged into when the 3100 initially came out, and just last month that 3100 vs Trichromatic test I attempted. In both cases, halfway through I got that sick pit in my stomach thinking, "Oh crap, I'm in over my head yet again!" And these were systems I had years of experience with; just testing some new aspect or feature.

My point is, unless a person either has access to robust testing equipment, or has long-term direct experience with something, there are just too many variables that the novice (to that specific equipment) cannot understand because there is no context. There are far too may “unknown unknowns.” You mentioned several of them in your post:
Were the lenses within a standard deviation of expectations?
What sharpening?
Was one system focused “hyperfocally” to take better advantage of the scene?
Was the Linhof set up square?

My previous post, referencing the Dunning-Kruger effect may have been a bit tongue-in-cheek, but these “unknown unknowns” are one of the basic drivers of the effect. Roger @ Lensrentals often writes about how even their robust testing can so easily be taken out of context when applied to other conditions or situations.

My basic reason for being critical of this test has nothing to do with being a fanboy (or not - I have a bunch of Canon equipment too). It is based on my familiarity with the lens / back combo. Those results were not good, and as I said before, if it were me I would question my technique or send the lens back. And yes, I agree C1 is aggressive with sharpening. The default for my back is AMT 180. I changed that default as soon as I got the back, reducing it to the low 100’s. Made a similar change with my previous IQ180.

Dave

Hi,

When I first compared the Canon image to the Phase One image it was the f/8 image from Canon to the f/16 on Phase One. So, the Phase One image was in all probability degraded by diffraction.

Since that I have checked out all the Phase One images, and it seems that the tested sample doesn't perform to expectations. I mostly looked at the left side as there is a lot of good detail.

An issue with this comparison is that we don't get the raw images, thus we cannot say how much sharpening and noise reduction was applied. Capture One applies a lot of both sharpening and noise reduction per default and that default varies system to system. This behavior is not specific to Capture One, Lightroom does it to, but Lightroom's defaults to very little sharpening.

What I can see is that the images certainly certify Mr Dawood's findings. On the other hand, I would guess that he has a good sample of the 24/3.5 TSE LII. Regarding the Rodenstock/IQ3100 combo, Mr. Dawood states that the images were made by a Phase One specialist. My guess that the lens used is a pretty bad sample.

Mr. Dawood did some other comparisons like comparing macro shots with his Hasselblad H6D100c with the HC macro 120/4 and the Canon 100/2.8L IS. Again, it is difficult to evaluate his comparison without access to the raw files. It may be that he is a bit unfortunate with lens choice to compare and bad samples on the MFD side.

I would say that the tests are quite honest. The findings are not what I would expect.

I can mention a parallell. When the GFX arrived, it was tested by both Jim Kasson and Lloyd Chambers. Lloyd Chambers generally found a lot of issues with the GFX while Jim Kasson was extremely pleased. Lloyd Chambers has found a lot of issues with focusing on GFX, issues that Jim Kasson has confirmed to some part. But, even with focusing issues, the GFX outperformed any other camera Jim Kasson had. It also seemed that Jim Kasson got all the good lenses and Lloyd Chambers got all the bad lenses.

Jim also found that the Hasselblad H lenses he had were not even close to the GF-lenses. Jim's testing resulted in selling all his Hasselblad V and H-system lenses. Leica M-lenses were also sold, but he kept the Leica R-lenses.

The Hasselblad V 250/5.6 superachromat was actually a good performer but it was not very convenient on the GFX.

Best regards
Erik

 

[ErikKaffehr]

Hi Dave,

Thanks for responding. Regarding testing, I would agree with you. I have done a lot of comparisons, though.

Just to say, Jim Kasson does not have a lot of technical equipment for his testing. He has a Stackshot and so do I. The main asset that Jim has is inside in head, a well trained engineering mind.

Regarding the test by Usman Dawood (I hope I got it right), it seems to me that he takes an honest approach to it. I was a bit surprised at the results as the 40 mm Rodenstock is known as a very good lens.

It is my understanding that there used to be significant variation between lenses, see Joseph Holmes's classic article:
https://www.josephholmes.com/news/2009/4/5/medium-format-digital-camera-optical-precision

I would hope that Linos has improved on their quality control since 2009 when the article above was published.

I would also assume that a Phase One Specialist would be able to recognize a bad sample of a lens. I would be pretty sure there was some issue with the lens.

There are some points in Usman Dawood's presentation I am very skeptical about:

One is comparing both cameras at actual pixels. In practice that means that the resolution advantage of the IQ3100 MP is thrown away, or you can say that the IQ3100 image is compared at a relatively higher magnification. He also compares the IQ3100 MP with the Sony A7rII sensor and finds that the Sony has cleaner shadows. In the real world I would be pretty sure that the IQ3100 MP is a bit biased to underexposure in order of protecting the highlights. You would need to check out the exposures in a program like RawDigger to see it. If exposure (in terms of ETTR) would be identical, I am pretty confident that the IQ3100MP would have the cleaner shadows.

Diffraction is only dependent on aperture, optical designs cannot change it. But, a sharper lens has higher microcontrast (MTF) to start with. Diffraction also has an MTF and the general assumption is that those MTF factors can be multiplied. So if we assume that a lens has an undiffracted MTF of 0.8 and MTF of diffraction is 0.8 the remaining MTF would be 0.8 * 0.8 -> 0.64. Would we have a lens having an undiffracted MTF of 0.6 and shot using the same aperture we would end up with an MTF at the sensor plane of 0.6 * 0.8 ->0.48.

Just to mention, there is much research in MTF and human vision, and it seems to be an agreement that MTF (at a given level of detail) needs to be larger than 0.35 for that detail being perceived as sharp. But we don't really want 0.35 MTF at the pixel level as that would create a lot of fake detail.

To sum it up, I am quiet a bit surprised at the results of that test. I would guess that the 40 mm lens used with the IQ3100 MP may not be a very good sample.

Best regards
Erik

Hi Erik,
I agree, with a few caveats:

First, I am not sure what qualifies someone to be a "Phase One Specialist." Specialist is a vague, undefined term. Could be a 20+ year employee of Phase One, a dealer, someone who shoots or used to shoot Phase One or someone who demo'd a Phase One back 2 years ago. We have no idea what it means. My guess is it is someone who has some undefined experience using a Phase One product, which may have nothing to do with a technical camera and the 40hr.

Second, and this is the big one for me: it is exceedingly difficult to perform tests and make overarching conclusions when testing something one is not intimately familiar with, unless there is access to extensive testing equipment like DXO, Lensrentals, Jim Kasson, etc. As much as I value some of what Lloyd Chambers does, I would not put him in that category (and certainly not myself!). Over the years I have occasionally tried to do what I initially thought was thorough testing with something very familiar. Two relatively recent examples were the goofy red-shadows w/long exposure testing I got dragged into when the 3100 initially came out, and just last month that 3100 vs Trichromatic test I attempted. In both cases, halfway through I got that sick pit in my stomach thinking, "Oh crap, I'm in over my head yet again!" And these were systems I had years of experience with; just testing some new aspect or feature.

My point is, unless a person either has access to robust testing equipment, or has long-term direct experience with something, there are just too many variables that the novice (to that specific equipment) cannot understand because there is no context. There are far too may “unknown unknowns.” You mentioned several of them in your post:
Were the lenses within a standard deviation of expectations?
What sharpening?
Was one system focused “hyperfocally” to take better advantage of the scene?
Was the Linhof set up square?

My previous post, referencing the Dunning-Kruger effect may have been a bit tongue-in-cheek, but these “unknown unknowns” are one of the basic drivers of the effect. Roger @ Lensrentals often writes about how even their robust testing can so easily be taken out of context when applied to other conditions or situations.

My basic reason for being critical of this test has nothing to do with being a fanboy (or not - I have a bunch of Canon equipment too). It is based on my familiarity with the lens / back combo. Those results were not good, and as I said before, if it were me I would question my technique or send the lens back. And yes, I agree C1 is aggressive with sharpening. The default for my back is AMT 180. I changed that default as soon as I got the back, reducing it to the low 100’s. Made a similar change with my previous IQ180.

Dave