Luminous Landscape Forum

Equipment & Techniques => Medium Format / Film / Digital Backs – and Large Sensor Photography => Topic started by: ErikKaffehr on August 06, 2018, 03:32:36 am

Title: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: ErikKaffehr on August 06, 2018, 03:32:36 am
There was a discussion thread discussing observable IQ difference between X1D and D850.

Answering that kind of question is a bit hard, because any comparison is easily getting an apples to oranges comparison. To get a valid comparison we need:


DPReview published a good test of the Fuji GFX and include some comparisons with several 24x36 cameras.

Below is an image, developed in Lightroom with no sharpening or noise reduction and resized to 100 cm picture height at 180 PPI and identically sharpened in Focus Magic:

(http://echophoto.dnsalias.net/ekr/Articles/TMP/GFXvsA7rII.jpg)

At that size, the GFX definitively has some advantage.

Let's look a bit at the sources of that advantage.

Pixels is what carry the information

Let's compare the vertical number of pixels on Hasselblad X1D and Nikon D850

X1D629614% advantage for the GFX
Nikon D8505512

But, the quality of the pixels may also differ. One aspect of that is the fine detail contrast transferred, that is normally measured as MTF. Below is MTF data for the GFX and the Nikon D850:

(http://echophoto.dnsalias.net/ekr/Articles/TMP/MTFCOMP.PNG)

A measure LP/PH (Line Pairs at Picture Height) MTF50 (at 50% transferred contrast).

X1D191519% advantage to X1D
D8501614

A small note. MTF at Nyquist is 26% on the Nikon and 25% on the X1D. That means that both would do well with say 8000 vertical pixels, 85 MP on the X1D and 100 MP on the D850.

A usable measure of image sharpness is SQF, that has been proposed by Ed Granger of Kodak.

(http://echophoto.dnsalias.net/ekr/Articles/TMP/SQFCOMP.PNG)

If we look at SQF for 100 cm picture height we would have:
X1D89
D85086

It used to be said that it takes about 5 SQF points for a visible difference.

So, what I can see is that there is a material advantage to the GFX when comparing images between the GFX and Sony A7rII in DPReviews comparison.

Comparing MTF data between the GFX and the D850 there is a 19% advantage to the X1D.

Looking at the SQF data, there would not be a observable difference at 100 cm print size.

The image based comparison includes sharpening, while the MTF data have been calculated on unsharpened images. The MTF data gives a 19% advantage (in LP/PH) to the GFX over the D850.

Rules of thumb

Resolution and magnification are linear properties. A 44x33 mm sensor has 1.68 times the surface area of the 24x36 mm sensor, so that would yield a linear advantage 29%.

This discussion looks at image quality near centre. Looking at corners may be different.

Best regards
Erik


Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: bjanes on August 06, 2018, 07:14:24 am
There was a discussion thread discussing observable IQ difference between X1D and D850.

Answering that kind of question is a bit hard, because any comparison is easily getting an apples to oranges comparison. To get a valid comparison we need:

  • Shoot the same subject
  • In the same light
  • use comparable lenses
  • Develop identically
  • Shoot tripod
  • Be able to go back and repeat

DPReview published a good test of the Fuji GFX and include some comparisons with several 24x36 cameras.


Erik,

That is a nice and well done comparison, but one variable that is not well controlled is lens quality. Do you know what lenses were used on each camera? In the DpReview of the GFX they compared its IQ to that of several full frame dSLRs. Not surprisingly, the GFX generally came out on top but they noted the following in a footnote:

"** The performance of your lens may have more to do with ultimate resolution than body itself. "


Regards,

Bill
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: ErikKaffehr on August 06, 2018, 07:48:57 am
Hi Bill,

Yes, lens quality is a great part of the equation and it is a poorly controlled one. Good lenses tend to be quite good at centre and often achieve maximum performance at f/4 or f/5.6.

Just to mention:


The Nikon 85/1.8G was very good in Opticallimits's test.
(http://www.opticallimits.com/images/8Reviews/lenses/nikkor_afs_85_18_d3x/mtf.png)
Hasselblad has measured MTF data for their lenses, and similar data are available from Zeiss. Lensrentals/Olaf Testing does MTF tests that should be comparable. Most people don't use Zeiss Otus lenses on Nikons, but some do.

Jim Kasson has done a lot of good quality testing. It seems that his Fuji lenses have Otus class correction.

Lensrentals does not test MFD lenses, for reasons known by me. That means a bit that we don't have comparable data.

From what I have seen, the GFX and the X1D would be excellent performers. That said, I have not really seen any great raw images from the X1D.

Diglloyd had some excellent stuff on both, but he is quite restrictive about sharing/discussing his findings.

To sum up. I would think that X1D and GFX do deliver on promise, at least with good lens samples.

Just to mention, Lensrentals has some interesting write up on transportation damage of three Sigma lenses, and effort Sigma went to to analyse the issue:

https://www.lensrentals.com/blog/2018/07/mtf-tests-for-the-sigma-14-24mm-f2-8-art-series-lens/

See addendum.

Best regards
Erik




Erik,

That is a nice and well done comparison, but one variable that is not well controlled is lens quality. Do you know what lenses were used on each camera? In the DpReview of the GFX they compared its IQ to that of several full frame dSLRs. Not surprisingly, the GFX generally came out on top but they noted the following in a footnote:

"** The performance of your lens may have more to do with ultimate resolution than body itself. "


Regards,

Bill
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: eronald on August 06, 2018, 07:57:54 am
Erik,

 You chose a good picture, the Fuji shows a nice area of realistic vegetation, the 35mm shot shows digital mush.
 Look at it for a while, zoom it a bit, you will see what I mean, there's a notable difference in texture preservation.
 Judging by the image alone, the Fuji is one whole class camera above the 35mm and well worth the difference in price, at least to nature shooters.
 
 Re. the nyquist cutoff and after mtf, this will give a measure of the aliasing and false detail one can expect. As we all know false detail is an unavoidable consequence of sampling.

Edmund
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: ErikKaffehr on August 06, 2018, 11:33:11 am
Hi Edmund,

I have not exactly chosen that picture, it is one of the few available from DPreview that are usable for a comparison.

I would say that the way I have processed the images, it is a reasonable comparison. This is essentially what I would see in prints 100 cm high at reasonable viewing distances.

Just to say, I would not see those benefits with my P45+/Hasselblad V system. The P45+ is not a 50 MP back and the V-series lenses are not as good as the lenses for the GFX.

So, I think that GFX makes sense, if you print large enough. With the X1D, I don't feel I have enough data. The older systems like Hasselblad HC and Phase One, I would be more in doubt.

On the other hand, I was shooting for 48 years, and I still don't have a a really critical 100x150 cm print hanging on any wall...


Best regards
Erik


Erik,

 You chose a good picture, the Fuji shows a nice area of realistic vegetation, the 35mm shot shows digital mush.
 Look at it for a while, zoom it a bit, you will see what I mean, there's a notable difference in texture preservation.
 Judging by the image alone, the Fuji is one whole class camera above the 35mm and well worth the difference in price, at least to nature shooters.
 
 Re. the nyquist cutoff and after mtf, this will give a measure of the aliasing and false detail one can expect. As we all know false detail is an unavoidable consequence of sampling.

Edmund
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: Doug Peterson on August 06, 2018, 12:04:15 pm
On the other hand, I was shooting for 48 years, and I still don't have a a really critical 100x150 cm print hanging on any wall...

Over the past decade we have seen a huge uptick in interest among our customers for making prints of this size or larger. Factors include:
- The increased ease of printing large vs past decades
- The decreased price of printing large vs past decades
- Increased ability to capture images that have enough resolution* for printing large vs past decades
- Changing expectations of art buyers vs past decades
- New models for selling art vs past decades
- Desire to differentiate and stand out from an increasingly crowded field of photographers

Large prints don't just look impressive. They can tell a different story than a smaller print. I wrote a bit about this here (https://www.getdpi.com/wp/2016/01/what-can-you-do-with-100mp/).

Expectations of both photographers and clients (of varying types) change over time. In 1990 (http://articles.baltimoresun.com/1990-09-30/features/1990273133_1_horizontal-resolution-image-resolution-32-inch) a good 32" TV set cost 2 grand and none of my friends had something like that. Now, a 40+ inch (1 meter) TV at home is like having a microwave; not everyone has one, but it's stranger to find it absent then present. The prices came down, the ease of using them went up, technology and content arrived that better leveraged larger presentation sizes. It's not a perfect analogue of course, but it is similar.

When I started 10 years ago if a client said they wanted to "print big" they generally meant something like 30" x 20" (~75cm on long side). Now when a client says they want to "print big" they generally mean something like 40" x 60" (~150cm on long side) or even larger. I just placed an order for three pieces of my terra nudum artwork for a gallery show. They are 1 meter square and I would have gone considerably larger except this was the largest size I could get shipped via FedEx (with appropriate box/padding) and the specifics of the order made freight-shipping prohibitive. I can't wait to see them!

*Of course you can print a grainy 35mm neg to 4 meters if you want. Anything can print any size, especially if you restrict viewing distance (as with a billboard ad). Here I mean "and still look good when viewed close up".
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: Jim Kasson on August 06, 2018, 12:26:38 pm
Over the past decade we have seen a huge uptick in interest among our customers for making prints of this size or larger. Factors include:
- The increased ease of printing large vs past decades
- The decreased price of printing large vs past decades
- Increased ability to capture images that have enough resolution* for printing large vs past decades
- Changing expectations of art buyers vs past decades
- New models for selling art vs past decades
- Desire to differentiate and stand out from an increasingly crowded field of photographers

Large prints don't just look impressive. They can tell a different story than a smaller print. I wrote a bit about this here (https://www.getdpi.com/wp/2016/01/what-can-you-do-with-100mp/).

Expectations of both photographers and clients (of varying types) change over time. In 1990 (http://articles.baltimoresun.com/1990-09-30/features/1990273133_1_horizontal-resolution-image-resolution-32-inch) a good 32" TV set cost 2 grand and none of my friends had something like that. Now, a 40+ inch (1 meter) TV at home is like having a microwave; not everyone has one, but it's stranger to find it absent then present. The prices came down, the ease of using them went up, technology and content arrived that better leveraged larger presentation sizes. It's not a perfect analogue of course, but it is similar.

When I started 10 years ago if a client said they wanted to "print big" they generally meant something like 30" x 20" (~75cm on long side). Now when a client says they want to "print big" they generally mean something like 40" x 60" (~150cm on long side) or even larger. I just placed an order for three pieces of my terra nudum artwork for a gallery show. They are 1 meter square and I would have gone considerably larger except this was the largest size I could get shipped via FedEx (with appropriate box/padding) and the specifics of the order made freight-shipping prohibitive. I can't wait to see them!


One thing that hasn't changed is the wall space that folks have. Most folks who want one of my images want it on C-size or 24x30 inch paper. When someone wants a big print, I love it, but it doesn't happen much. I ship big prints unmatted/stretched and rolled.

Jim
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: Jim Kasson on August 06, 2018, 12:33:15 pm
There was a discussion thread discussing observable IQ difference between X1D and D850.

Answering that kind of question is a bit hard, because any comparison is easily getting an apples to oranges comparison. To get a valid comparison we need:

  • Shoot the same subject
  • In the same light
  • use comparable lenses
  • Develop identically
  • Shoot tripod
  • Be able to go back and repeat

DPReview published a good test of the Fuji GFX and include some comparisons with several 24x36 cameras.

Below is an image, developed in Lightroom with no sharpening or noise reduction and resized to 100 cm picture height at 180 PPI and identically sharpened in Focus Magic:

(http://echophoto.dnsalias.net/ekr/Articles/TMP/GFXvsA7rII.jpg)

A couple of things to consider. The way that DPR does these outside tests, atmospheric thermal turbulence is not controlled for.  Neither is focus.  When I do tests like these, I try to find nearly equivalent lenses, like the Otus 55 and Otus 85. Of all the issues, I think the air is probably the most significant.

I agree with Edmund about the grass.

When I tested those two lenses and cameras, I found a greater difference in the corners:

https://blog.kasson.com/gfx-50s/sony-90mm-macro-on-the-a7rii-fuji-120mm-macro-on-the-gfx-50s-corner/

Jim

Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: ErikKaffehr on August 06, 2018, 12:38:38 pm
Hi Doug,

4K TV corresponds to something like 10 MP. That was passed by APS-C in 2010.

We may see 8K coming in a few years, but to really see the 8K advantage, it may take an 85" screen, viewed pretty close. And 8K is still around 37 MP.

What percentage of you customers print larger than 40"x60" or larger on glossy paper?

Best regards
Erik


Over the past decade we have seen a huge uptick in interest among our customers for making prints of this size or larger. Factors include:
- The increased ease of printing large vs past decades
- The decreased price of printing large vs past decades
- Increased ability to capture images that have enough resolution* for printing large vs past decades
- Changing expectations of art buyers vs past decades
- New models for selling art vs past decades
- Desire to differentiate and stand out from an increasingly crowded field of photographers

Large prints don't just look impressive. They can tell a different story than a smaller print. I wrote a bit about this here (https://www.getdpi.com/wp/2016/01/what-can-you-do-with-100mp/).

Expectations of both photographers and clients (of varying types) change over time. In 1990 (http://articles.baltimoresun.com/1990-09-30/features/1990273133_1_horizontal-resolution-image-resolution-32-inch) a good 32" TV set cost 2 grand and none of my friends had something like that. Now, a 40+ inch (1 meter) TV at home is like having a microwave; not everyone has one, but it's stranger to find it absent then present. The prices came down, the ease of using them went up, technology and content arrived that better leveraged larger presentation sizes. It's not a perfect analogue of course, but it is similar.

When I started 10 years ago if a client said they wanted to "print big" they generally meant something like 30" x 20" (~75cm on long side). Now when a client says they want to "print big" they generally mean something like 40" x 60" (~150cm on long side) or even larger. I just placed an order for three pieces of my terra nudum artwork for a gallery show. They are 1 meter square and I would have gone considerably larger except this was the largest size I could get shipped via FedEx (with appropriate box/padding) and the specifics of the order made freight-shipping prohibitive. I can't wait to see them!

*Of course you can print a grainy 35mm neg to 4 meters if you want. Anything can print any size, especially if you restrict viewing distance (as with a billboard ad). Here I mean "and still look good when viewed close up".
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: Doug Peterson on August 06, 2018, 12:48:11 pm
One thing that hasn't changed is the wall space that folks have.

US House Sizes:
(https://amp.businessinsider.com/images/59e0ecdd1f848906058b5034-750-515.png)

I suspect that the chart for the top 1% and top 10% are even steeper.
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: ErikKaffehr on August 06, 2018, 12:53:26 pm
Hi Jim,

I would agree on all issues. I did not discuss thermal turbulence. Regarding focus, I would assume that DPReview does focus bracketing on the studio shots, but I would guess they don't do it on outdoor shots. Previously DPreview's lens tests were published on "SLRGear" and they stated they were using focus bracketing. Also, the amount of aliasing observable in the DPReview studio test images indicates pretty decent focus.

I did a comparison between MFD and smaller formats based on the DPReview studio scene a while ago. What I have seen was that:


So my major finding was that building a system around the 44x33 mm sensor made a lot of sense. But that sensor size needs a small pitch sensor.

The other finding was that a system designed for a 54x41 mm sensor may not yield optimum performance on a 44x33 mm sensor.

Best regards
Erik




A couple of things to consider. The way that DPR does these outside tests, atmospheric thermal turbulence is not controlled for.  Neither is focus.  When I do tests like these, I try to find nearly equivalent lenses, like the Otus 55 and Otus 85. Of all the issues, I think the air is probably the most significant.

I agree with Edmund about the grass.

When I tested those two lenses and cameras, I found a greater difference in the corners:

https://blog.kasson.com/gfx-50s/sony-90mm-macro-on-the-a7rii-fuji-120mm-macro-on-the-gfx-50s-corner/

Jim
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: Jim Kasson on August 06, 2018, 12:53:46 pm
US House Sizes:
(https://amp.businessinsider.com/images/59e0ecdd1f848906058b5034-750-515.png)

I suspect that the chart for the top 1% and top 10% are even steeper.

This shows a 12.5% increase in area over 20 years, which is 6% linear increase.

Jim
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: Doug Peterson on August 06, 2018, 02:23:22 pm
This shows a 12.5% increase in area over 20 years, which is 6% linear increase.

Point taken, though it's a 50% increase over 40 years (22% linear) which seems noteworthy. And that's median. As I said, I expect the trend is much larger among the demographic that could afford to buy the work of a photographer selling a 1.5 meter print.

Also I think it's not unreasonable to assume that as the total space increases the "discretionary" amount of space in which you might hang large prints increases at a faster rate. I don't expect a lot of people would hang large art prints in their laundry room, garage, or children's bedroom, but would hang large art prints in their dining room; it seems likely that a home that is 50% larger has a dining room more than 50% larger (or has a dining room vs not having a dining room). I could be wrong about that; I'm definitely not a housing expert; my wife and I's apartment is 800 square feet; a large improvement from the 240 square feet apartment when I was single. Still, I hung four 4 foot by 2 foot prints in that 240 square foot apartment, which, after accounting for the window and tall book shelf, was nearly 100% use of the walls!

My point isn't that all, or even most, prints are made 2 meters large; that is just not the case. My point is that we've seen a large uptick in interest in printing that size (or larger), and a slow redefinition of what "printing large" means. As best I can tell this interest skews more toward younger shooters (or older shooters who started photography later in life). I think people tend to sell what they show, so if someone hasn't previously sold large prints, then they won't show or promote big prints, so they are less likely to sell large prints in the future. Obviously I have no idea if that applies to you; I'm only speaking here about my experience working with US-based P1 clients over the last decade.
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: hubell on August 06, 2018, 02:29:11 pm
US House Sizes:
(https://amp.businessinsider.com/images/59e0ecdd1f848906058b5034-750-515.png)

I suspect that the chart for the top 1% and top 10% are even steeper.

No question. Go into certain affluent suburbs today in the US and you see houses with 3,000-5,000 sq. ft. of living space being torn down left and right and replaced with 10,000+ sq. ft. houses. They need lots of BIG ART to fill that wall space.
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: Jim Kasson on August 06, 2018, 03:08:32 pm
Point taken, though it's a 50% increase over 40 years (22% linear) which seems noteworthy. And that's median. As I said, I expect the trend is much larger among the demographic that could afford to buy the work of a photographer selling a 1.5 meter print.

Also I think it's not unreasonable to assume that as the total space increases the "discretionary" amount of space in which you might hang large prints increases at a faster rate. I don't expect a lot of people would hang large art prints in their laundry room, garage, or children's bedroom, but would hang large art prints in their dining room; it seems likely that a home that is 50% larger has a dining room more than 50% larger (or has a dining room vs not having a dining room). I could be wrong about that; I'm definitely not a housing expert; my wife and I's apartment is 800 square feet; a large improvement from the 240 square feet apartment when I was single. Still, I hung four 4 foot by 2 foot prints in that 240 square foot apartment, which, after accounting for the window and tall bookshelf, was nearly 100% use of the walls!

My point isn't that all, or even most, prints are made 2 meters large; that is just not the case. My point is that we've seen a large uptick in interest in printing that size (or larger), and a slow redefinition of what "printing large" means. As best I can tell this interest skews more toward younger shooters (or older shooters who started photography later in life). I think people tend to sell what they show, so if someone hasn't previously sold large prints, then they won't show or promote big prints, so they are less likely to sell large prints in the future. Obviously, I have no idea if that applies to you; I'm only speaking here about my experience working with US-based P1 clients over the last decade.

I know there's gotta be a market; look at the monster prints that find their way onto museum walls. It just hasn't found me, or I, it. I guess I don't know folks with deep enough pockets.

Jim
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: pschefz on August 06, 2018, 04:57:47 pm
this is a strange discussion.....most HUGE prints sold today are from 20-40mpix systems or film from all different sizes....so the idea that we need more mpix just isnt true.....i completely agree that a 33x44 100mpix sensor will be nice but i honestly feel that 40-50 pix is a very nice number with plenty of information to crop and still manageable enough.....i am much more interested in DR, resolution and color are pretty incredible with existing a7rIII, d850, gfx, x1d....

on a side note: a good friend just had a 50x60 gallery quality made from 8x10 negative, this is the first of an ongoing series from everything incl 6x7, 8x10 and digital (d850).....all formats look different but there is no problem with either in terms of resolution.....
also: these prints are not cheap.....i doubt there are a lot of people just popping those out for the fun of it.....
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: eronald on August 06, 2018, 05:31:36 pm
this is a strange discussion.....most HUGE prints sold today are from 20-40mpix systems or film from all different sizes....so the idea that we need more mpix just isnt true.....i completely agree that a 33x44 100mpix sensor will be nice but i honestly feel that 40-50 pix is a very nice number with plenty of information to crop and still manageable enough.....i am much more interested in DR, resolution and color are pretty incredible with existing a7rIII, d850, gfx, x1d....

on a side note: a good friend just had a 50x60 gallery quality made from 8x10 negative, this is the first of an ongoing series from everything incl 6x7, 8x10 and digital (d850).....all formats look different but there is no problem with either in terms of resolution.....
also: these prints are not cheap.....i doubt there are a lot of people just popping those out for the fun of it.....

About $1K should get you a decent USED printer and the ability to make humongous digital prints for peanuts. Of course in a city center the real estate to put said printer will be worth $20K at least :)

Large prints are fun. Both my P45 and my D3x could already very usefully print large ten years ago, so I have no doubt that today's SLRs can do so too.

I am seeing a lot of adverts for turning photos into wallpaper these days :)

Edmund
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: ErikKaffehr on August 07, 2018, 05:08:05 am
Hi Paul,

To a certain extent, small pixels are always beneficial. The reasons are:


There are some disadvantages with smaller pixels, though. There is a small loss of DR when reducing pixel size and. Also, smaller pixels are more prone to vignetting and crosstalk.

My take is that the optimal pixel size decreases with newer technologies.

Here is a simple example: The crop below was shot on 6.8 micron pitch with a 100 mm lens.
(http://echophoto.dnsalias.net/ekr/Articles/Shoots/Aliasing/20150103-CF046029.jpg)

The same subject was shot from approximately the same point of view with a small sensor camera with 4 micron pitch, also with a 100 mm lens.
(http://echophoto.dnsalias.net/ekr/Articles/Shoots/Aliasing/20150103-_DSC6682.jpg)

The first image shows false color on the street sign and has moiré on the blinders. The second image is noisy, but has much cleaner detail.

(http://echophoto.dnsalias.net/ekr/Articles/Shoots/Aliasing/20150103-CF046029_vsmall.jpg)
(http://echophoto.dnsalias.net/ekr/Articles/Shoots/Aliasing/20150103-_DSC6682_vsmall.jpg)

Even with both images downsized to the same size, the 3.9 micron image (on the right) is smoother and with less artefacts. But, it can be argued that the large pixel image is sharper, quite possibly because the Zeiss 100/3.5 Planar used on 6.8 micron pitch sensor may be a sharper the Sony 70-400/4-5.6 G at 100 mm used with the 3.9 micron Sony A77.

If you check out the MTF data for both the Nikon D850 and the X1D, both sensors deliver about 25% MTF (contrast) at the Nyquist limit. For aliasing free rendition MTF at Nyquist would need to be around 10% or so at Nyquist. So the sensor doesn't make full use of the lens.
(http://echophoto.dnsalias.net/ekr/Articles/TMP/MTFCOMP.PNG)

You can also check the DPReview studio shot image:
(http://echophoto.dnsalias.net/ekr/Articles/TMP/AliasingSamples.PNG)

You can see that the although the images are scaled to 50MP, the 100MP Phase One IQ3100MP yields the cleanest image.

Now, it is a well known fact that Sony's next generation MFD sensor is 100MP in 44x33 mm and 150 MP in 54x41mm. It is quite probable that Sony will not design a new generation 44x33 mm sensor. So, next generation GFX and X1D will be 100 MP. I would suggest that resolution will make the lenses better justice.

But, I would agree that we can do with much less resolution for large prints.

Best regards
Erik


this is a strange discussion.....most HUGE prints sold today are from 20-40mpix systems or film from all different sizes....so the idea that we need more mpix just isnt true.....i completely agree that a 33x44 100mpix sensor will be nice but i honestly feel that 40-50 pix is a very nice number with plenty of information to crop and still manageable enough.....i am much more interested in DR, resolution and color are pretty incredible with existing a7rIII, d850, gfx, x1d....

on a side note: a good friend just had a 50x60 gallery quality made from 8x10 negative, this is the first of an ongoing series from everything incl 6x7, 8x10 and digital (d850).....all formats look different but there is no problem with either in terms of resolution.....
also: these prints are not cheap.....i doubt there are a lot of people just popping those out for the fun of it.....
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: JaapD on August 07, 2018, 06:25:09 am
Hi Erik, those artefacts (interference patterns) you’re showing only tell half of the story. One would be able to create exactly the opposite results with image content consisting of the same pixel pitch as with the 100Mpix sensor. Then the 100Mpix sensor creates all the artefacts while the 50 Mpix sensor of the X1D and GFX does not.
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: ErikKaffehr on August 07, 2018, 07:35:50 am
Hi Jaap,

Not exactly right. These artefacts arise when the lens 'outresolves the sensor', that is transfers significant modulation at the Nyquist limit. So an optimal lens/sensor combination would transfer very little modulation at Nyquist.

Jim Kasson has some discussion on the issue here: https://blog.kasson.com/the-last-word/whats-your-q/

One of his findings is that optimal pitch for a lens diffraction limited at f/11 would be 2.75 microns.

But, there are many other factors limiting lens performance than diffraction.

Aliasing is in no way specific to photography. It is a fundamental part of signal processing theory. What is a bit specific with photography that photographers and camera makers choose to ignore it...

Best regards
Erik




Hi Erik, those artefacts (interference patterns) you’re showing only tell half of the story. One would be able to create exactly the opposite results with image content consisting of the same pixel pitch as with the 100Mpix sensor. Then the 100Mpix sensor creates all the artefacts while the 50 Mpix sensor of the X1D and GFX does not.
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: DougDolde on August 07, 2018, 01:27:22 pm
The problem I have found with really big prints, say 36x48 is the cost of framing.  Unless you are doing frameless canvas wraps it becomes very expensive and challenging.  Yes I do frameless canvas wraps and it's a good solution.

However I find smaller prints sell better because you can keep the price down.  Most people that buy in galleries are quite price sensitive.  And most galleries have size limits as well so you may be limited to showing only one really big piece versus two or three smaller ones.

I know for some people with BIG reputations this isn't such a big issue, but for me it is..
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: cgarnerhome on August 07, 2018, 03:45:08 pm
Real world for me is as follows. When I’m shooting in the field I usually leave some extra space around the subject to give me more flexibility in post processing.  If I’m shooting with a fixed focal length lens I may have to shoot even wider than I would like.  In addition, I might do a square crop or more cropping than I planned.  I also want the flexibility to print large.  Given all that, I find I almost never use my Nikon 850 and almost exclusively use my XF100.  In my case, 100Mpix is important. It’s not always about theory it’s about what works for you.
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: Doug Peterson on August 07, 2018, 04:00:59 pm
It’s not always about theory it’s about what works for you.

+1 except I would modify it to: "It's almost never about theory. It's about what works for you."
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: eronald on August 07, 2018, 07:28:09 pm
+1 except I would modify it to: "It's almost never about theory. It's about what works for you."

O learned Phase rep, pray tell us, did Phase also get some optimisations with the 50c chip?

Edmund
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: siddhaarta on August 07, 2018, 07:48:54 pm
O learned Phase rep, pray tell us, did Phase also get some optimisations with the 50c chip?

Edmund

According to this ... DR Fuji GFX vs. P1 Q250  (http://photonstophotos.net/Charts/PDR.htm#FujiFilm%20GFX%2050S,Phase%20One%20IQ250)

... at least they seem to have squeezed more DR out of the same sensor .. or is there a problem with real ISO vs ISO setting??
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: bjanes on August 07, 2018, 08:40:26 pm
According to this ... DR Fuji GFX vs. P1 Q250  (http://photonstophotos.net/Charts/PDR.htm#FujiFilm%20GFX%2050S,Phase%20One%20IQ250)

... at least they seem to have squeezed more DR out of the same sensor .. or is there a problem with real ISO vs ISO setting??

I don't think so. The Phase curves are shifted to the right, ? something to do with ISO amplification. The base ISO for the Fuji is 11.9 stops at ISO 100 and the Phase One is 11.92 stops at ISO 200. No significant difference in DR. However, there is a significant difference in price. A quick check on the web indicates that the P1 is US $35,000, whereas the Fuji is $5850. IMHO, the P1 is vastly overpriced.

Regards,

Bill
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: bjanes on August 07, 2018, 08:46:15 pm
*Of course you can print a grainy 35mm neg to 4 meters if you want. Anything can print any size, especially if you restrict viewing distance (as with a billboard ad). Here I mean "and still look good when viewed close up".

Why do you bring up printing from a grainy 35 mm negative? The thread is about comparing two CMOS sensors.

Bill
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: Steve Hendrix on August 07, 2018, 11:05:40 pm
I don't think so. The Phase curves are shifted to the right, ? something to do with ISO amplification. The base ISO for the Fuji is 11.9 stops at ISO 100 and the Phase One is 11.92 stops at ISO 200. No significant difference in DR. However, there is a significant difference in price. A quick check on the web indicates that the P1 is US $35,000, whereas the Fuji is $5850. IMHO, the P1 is vastly overpriced.

Regards,

Bill


Bill, I can't blame you for not being able to find an accurate price for 50mp Phase One options - Phase One pricing has often been a moving target:

https://captureintegration.com/phase-one-2018-summer-specials/
https://www.digitalback.com/product/phase-one-iq3-50mp/

$14,990 - IQ150/XF Kit w/80mm Schneider Lens
$26,990 - IQ350/XF Kit w/80mm Schneider Lens

Still quite a bit higher, but I think it is fair to the readers to inform them they don't have to spend $30,000 more than a Fuji GFX to get into a Phase One 50mp system, should they want to.


Steve Hendrix/CI
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: ErikKaffehr on August 08, 2018, 12:21:59 am
Hi Bill,

Bill Claff's data use nominal ISO.

On older backs that DxO-mark has measured Phase One always overrated ISO by about one stop, probably in order to protect highlights.

That is probably what we see in the histogram on the Phase One DR is constant over 100 and 200 ISO.

Best regards
Erik



I don't think so. The Phase curves are shifted to the right, ? something to do with ISO amplification. The base ISO for the Fuji is 11.9 stops at ISO 100 and the Phase One is 11.92 stops at ISO 200. No significant difference in DR. However, there is a significant difference in price. A quick check on the web indicates that the P1 is US $35,000, whereas the Fuji is $5850. IMHO, the P1 is vastly overpriced.

Regards,

Bill
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: ErikKaffehr on August 08, 2018, 12:32:49 am
Hi,

Note that DR is the same at ISO 100 and ISO 200. That essentially means that the both nominal ISO values are the same.

Best regards
Erik


According to this ... DR Fuji GFX vs. P1 Q250  (http://photonstophotos.net/Charts/PDR.htm#FujiFilm%20GFX%2050S,Phase%20One%20IQ250)

... at least they seem to have squeezed more DR out of the same sensor .. or is there a problem with real ISO vs ISO setting??
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: ErikKaffehr on August 08, 2018, 12:48:20 am
Hi Bill,

In Sweden we have the same dealer for Hassy, Phase and Fuji. Prices here in Sweden are:

Fuji GFX, body only: 69 995,00
Hasselblad X1D body only: 74 995,00
PHASE ONE XF IQ1 50MP: 258 750,00

The prices include Swedish sales tax and 1 SEK is 0.11 USD

Best regards
Erik

I don't think so. The Phase curves are shifted to the right, ? something to do with ISO amplification. The base ISO for the Fuji is 11.9 stops at ISO 100 and the Phase One is 11.92 stops at ISO 200. No significant difference in DR. However, there is a significant difference in price. A quick check on the web indicates that the P1 is US $35,000, whereas the Fuji is $5850. IMHO, the P1 is vastly overpriced.

Regards,

Bill
Title: IQ difference between 44x33 CMOS and 24x36 CMOS, some reflections
Post by: ErikKaffehr on August 08, 2018, 02:31:06 am
Hi,

Being the OP, I would say this thread was intended to discuss the differences between 44x33 mm CMOS and 24x36 mm CMOS.

Not really stated, but the starting point was to compare the present generation of mirrorless DMF to best of breed 24x36 mm. It was also the idea of focusing on things observable or measurable.

The original posting indicated that:


I would have expected lens quality to play a greater role than what the MTF calculations yielded. In part, that depended on that the lens used on the Nikon was quite good. The DPReview test chart has only slanted edges in the central part.

I would suggest that many posters under rate the importance of pixel size and pixel pitch. After all, it is the pixels that carry the information.

When 44x33 mm MFD goes 100MP, I would think we would see major improvement in image quality. It seems that both the X1D lenses and the GFX lenses are excellent designs. This assumption is based on the MTF data Hasselblad publishes for the lenses, Hasselblad's MTF data used to be reliable and on Jim Kasson's evaluation of Fuji lenses.

Best regards
Erik
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS, some reflections
Post by: eronald on August 08, 2018, 05:29:34 am
Hi,

Being the OP, I would say this thread was intended to discuss the differences between 44x33 mm CMOS and 24x36 mm CMOS.

Not really stated, but the starting point was to compare the present generation of mirrorless DMF to best of breed 24x36 mm. It was also the idea of focusing on things observable or measurable.

The original posting indicated that:

  • There was an observable difference between Sony A7rII and Fuji GFX in 40"x60", based on one of the few comparable images DP-review has published. This was expected.
  • Comparing the Nikon D850 with the X1D, it was found that the X1D had a measurable advantage, like 20% linear scale. That was also expected.
  • Some of the observed/measured advantage comes from the aspect ratio.

I would have expected lens quality to play a greater role than what the MTF calculations yielded. In part, that depended on that the lens used on the Nikon was quite good. The DPReview test chart has only slanted edges in the central part.

I would suggest that many posters under rate the importance of pixel size and pixel pitch. After all, it is the pixels that carry the information.

When 44x33 mm MFD goes 100MP, I would think we would see major improvement in image quality. It seems that both the X1D lenses and the GFX lenses are excellent designs. This assumption is based on the MTF data Hasselblad publishes for the lenses, Hasselblad's MTF data used to be reliable and on Jim Kasson's evaluation of Fuji lenses.

Best regards
Erik

I would say the Fuji has a 1 to 2 generation advantage in crispness over its 35mm rivals in the vegetation images published, this may be due to Fuji's manipulation of the fill factor as much as the sensor size. So I think the Fuji is more like an 80MP 35mm camera.

It's NOT obvious Hassy or Phase are the same. We would benefit from creating a serious thread/forum to discuss image quality assessment. I think enough people here are interested.

Edmund
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: bjanes on August 08, 2018, 07:12:08 am
Hi Bill,

Bill Claff's data use nominal ISO.

On older backs that DxO-mark has measured Phase One always overrated ISO by about one stop, probably in order to protect highlights.

That is probably what we see in the histogram on the Phase One DR is constant over 100 and 200 ISO.

Best regards
Erik

Erik,

Thanks for the comment on ISO. DXO uses Ssat for their ISO measurements as documented here. (https://www.dxomark.com/About/In-depth-measurements/Measurements/ISO-sensitivity) This corresponds to ISO 12232:1998 and is what I think should be used for serious work with raw files. The 2006 versions added REI (recommended exposure index) which allows the manufacturer to use any arbitrary value that they desire, but is supposed to be used only with sRGB rendered files. As you state, Phase One often assigns a higher ISO, presumably to protect the highlights.

As you stated, Bill Claff uses the  manufacturer's nominal values.

On his charts the open circles for ISO indicate values beyond the normal analog range (fake ISOs). For Example (http://www.photonstophotos.net/Charts/PDR.htm#Nikon%20D850,Phase%20One%20IQ250), the base ISO of the Nikon D850 is 64, but ISOs of 50, 40 and 31 are offered and are shown as open circles on Bill's graph. On the DXO plots the DR for these fake ISOs do not change as with Bill's plots. These are merely overexposure and do not affect the ISO amplification or dynamic range. The Phase One offers a fake ISO of 100 as indicated by the open circle. The plot is shifted to the right because of the ISO manipulation to preserve highlights, and this does not indicate better DR.

I hope this clarifies the situation for some readers.

Best regards,

Bill


Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS, some reflections
Post by: ErikKaffehr on August 08, 2018, 07:38:06 am
Hi Edmund,

I would agree with your description on the crispness of the grass. But, I would attribute it to two different factors:



The other factor is that it seems that most GFX lenses are extremely sharp, at least according to Jim Kasson's testing.


Any camera with the Sony 44x33 would merit from the extra pixels.

Lenses are a different thing. As far as I can recall, Jim Kasson used to have a Hasselblad H3D39 which probably had the same 49x37 mm sensor my P45+ has. So far I know he has sold it.

But, before selling the H-system, he also tested some of his HC-lenses on the GFX and they performed far below the GF-lenses.

I would assume that GF lenses are sharper on 44x33 mm than the HC lenses on 44x33 mm. They are designed for a larger sensor. My understanding is that optical performance scales with dimension. So if you scale down a lens by a factor of two it will perform twice at good in lp/mm terms.

So, I would not expect H6D50c perform at the same level as the X1D.

This was seen a bit in a comparison I made with DPReview test images. The Pentax 645Z was quite close to the Sony A7rII in that comparison, even if it was paired with one of Pentax latest generation lenses (90 macro using ED-glass). Diglloyd found similar. The Pentax 645Z was not superior to the Sony A7r, with the lenses that he tested.

Phase One claims that their German lens design are superior (*), that may be the case. But, it seems that they are pretty classic designs, while the smaller formats have moved on more advanced designs. I started a discussion on that at DPReview: https://www.dpreview.com/forums/post/61114409

This posting may be of particular interest: https://www.dpreview.com/forums/post/61122175

The way I see it, the GFX offers significant image quality potential at a reasonable cost. It is priced like "pro bodies" from Canon and Nikon and lens prices are a bit like Sony's G-Master series.

Furthermore, I think that the GFX has a lot of potential for the next generation Sony sensor at 100MP.

The X1D is a bit more expensive, especially on the lens side, but I guess it has the same potential as the GFX.

The final question is if all that image quality is needed and if photographers can take full advantage of it.

The way I see it, if a Panasonic G9 is good enough for my needs, I will see little advantage from my A7rII. Would Sony release a 60-70 MP body in the A# series, I am not sure I would buy.

It can be smarter to spend the money on lenses or travel...

Best regards
Erik

I would say the Fuji has a 1 to 2 generation advantage in crispness over its 35mm rivals in the vegetation images published, this may be due to Fuji's manipulation of the fill factor as much as the sensor size. So I think the Fuji is more like an 80MP 35mm camera.

It's NOT obvious Hassy or Phase are the same. We would benefit from creating a serious thread/forum to discuss image quality assessment. I think enough people here are interested.

Edmund
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS, some reflections
Post by: eronald on August 08, 2018, 11:06:27 am
Hi Edmund,

I would agree with your description on the crispness of the grass. But, I would attribute it to two different factors:


Best regards
Erik

It's not just the grass, it's the vegetation of that test on Jim's links.
I think it's partly the reduced fill - ie less low-pass smearing, described on the Sony link you kindly provided.

We should open a thread on technical issues and image quality to discuss this.
As it stands, if my conjecture on the meaning of this Fuji announcement is correct the Fuji may be substantially sharper than the Hassy, with the same lens.

Edmund
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: vjbelle on August 08, 2018, 04:14:39 pm
Not having used or tested the Hassy lenses for the X1D I can attest that the GFX lenses that I own are the sharpest I have seen using any of my layman tests.  This includes any Otus lens that I have owned and all of my large format lenses ( Schneider Digitars ).  The differences between my Digitar lenses and the GFX lenses is visible at 100% pixels but very subtle.  So at 100% pixels some difference can be seen at best apertures (f5.6 for GFX, f8-10 for Digitars).  Again, I want to stress these differences are visually very small.  I print 40 inches long side all of the time..... would I see this difference?  Never......  Hope this puts things into some perspective. 

Victor
Title: Re: IQ difference between 44x33 CMOS and 24x36 CMOS
Post by: ErikKaffehr on August 08, 2018, 04:35:39 pm
Hi,

Thanks for sharing. Just a few comments...

If a lens reaches maximum performance at f/5.6 it is a pretty strong indicator that is is sharper than a lens reaching maximum performance at f/11. The reason is that diffraction is the factor limiting a lens when stopping down and diffraction is just a function of aperture. So if a lens reaches optimum at f/5.6, it is by definition sharper than a lens reaching maximum performance at f/11.

That said, the lens reaching optimum at f/11 can still outperform a lens that is optimal at f/5.6, if it is combined with a better sensor.

The comparison I made was 100 cm on the short side, BTW. There were two reasons for that. One reason was that aspect ratio is different between 44x33 and 24x36. It is often seen that the 4:3 aspect ratio is more practical than 3:2. The other factor was that I wanted to go a bit beyond what I usually print. I have been pretty happy with my 30"x40" prints, from any format I have used. 

Best regards
Erik

Not having used or tested the Hassy lenses for the X1D I can attest that the GFX lenses that I own are the sharpest I have seen using any of my layman tests.  This includes any Otus lens that I have owned and all of my large format lenses ( Schneider Digitars ).  The differences between my Digitar lenses and the GFX lenses is visible at 100% pixels but very subtle.  So at 100% pixels some difference can be seen at best apertures (f5.6 for GFX, f8-10 for Digitars).  Again, I want to stress these differences are visually very small.  I print 40 inches long side all of the time..... would I see this difference?  Never......  Hope this puts things into some perspective. 

Victor