IQ180 Vs. 8X10 film article

[clayh]

Just a quick note about my statement re 35mm film and a 39MP back. If you dig, you will find a shootout performed by Don Hutton on the LF forum where he went out with a friend who owned a P45 and shot some very controlled tests with a Leica MP, a leica ASPH lens and Adox CMS document film and then scanned with a drum scanner. He also got out a very high powered microscope to examine resolution charts shot with this camera/film/scanner combination. Bottom line was real resolution above 100lpm, which the 39MP back can't do.

Now all that said, this combo is pretty limiting. It is black and white. It is slow film. No way would a professional want to use this for any sort of volume work. But if you are doing photography for your own account and on your time, the point is that there are more options (and less expensive ones) to achieving super high print quality than running out and buying the latest digital back.

My point with the initial post was that the test was not done in a way that optimized and really showed what an 8x10 can do. The scan sample rate was at least 1/3 of what a typical 8x10 setup is capable of resolving with normal film. Come on, less than 1000dpi? Every test I have ever seen done on sheet film indicates that there is probably around 2000-3000 spi of real resolution on 8x10 film if it is shot at optimal apertures with good modern optics.

I am not saying that anyone should run out and buy an 8x10 with a brace of Sironar-S lenses and a Aztek Premier drum scanner. Shooting 8x10 is a pain in the ass, and it has some of its own inherent limitations. You are always fighting the battle between depth of field and diffraction-induced resolution loss because of the long focal length lenses needed (a 'normal' lens on 8x10 is 300mm). It is heavy, demands the use of a tripod and a huge bag of film holders. You fight dust, alignment issues, and all sorts of other challenges.

But there are more options for getting truly high quality digital files than spending the amount of money that would buy you a nice Beemer on brand new digital back.

[ErikKaffehr]

Hi,

A couple of issues:

1) You cannot compare Leica resolution to MF back, because the MF sensor is much larger. So 50 lp/mm on the sensor is close to 100 lp/mm on 135

2) Resolution is a poor parameter for sharpness. Edge contrast is the dominant factor for human vision. For instance a 50/1.4 lens at full aperture will have very good resolution but quite little edge contrast, because peripheral rays will dominate and those have higher aberration.

3) Large format is normally not limited by film but by lens and/or diffraction. LF lenses are normally used at small apertures, needed for both coverage and depth of field

Best regards
Erik

Just a quick note about my statement re 35mm film and a 39MP back. If you dig, you will find a shootout performed by Don Hutton on the LF forum where he went out with a friend who owned a P45 and shot some very controlled tests with a Leica MP, a leica ASPH lens and Adox CMS document film and then scanned with a drum scanner. He also got out a very high powered microscope to examine resolution charts shot with this camera/film/scanner combination. Bottom line was real resolution above 100lpm, which the 39MP back can't do.

Now all that said, this combo is pretty limiting. It is black and white. It is slow film. No way would a professional want to use this for any sort of volume work. But if you are doing photography for your own account and on your time, the point is that there are more options (and less expensive ones) to achieving super high print quality than running out and buying the latest digital back.

My point with the initial post was that the test was not done in a way that optimized and really showed what an 8x10 can do. The scan sample rate was at least 1/3 of what a typical 8x10 setup is capable of resolving with normal film. Come on, less than 1000dpi? Every test I have ever seen done on sheet film indicates that there is probably around 2000-3000 spi of real resolution on 8x10 film if it is shot at optimal apertures with good modern optics.

I am not saying that anyone should run out and buy an 8x10 with a brace of Sironar-S lenses and a Aztek Premier drum scanner. Shooting 8x10 is a pain in the ass, and it has some of its own inherent limitations. You are always fighting the battle between depth of field and diffraction-induced resolution loss because of the long focal length lenses needed (a 'normal' lens on 8x10 is 300mm). It is heavy, demands the use of a tripod and a huge bag of film holders. You fight dust, alignment issues, and all sorts of other challenges.

But there are more options for getting truly high quality digital files than spending the amount of money that would buy you a nice Beemer on brand new digital back.

[fotometria gr]

Hi,

A couple of issues:

1) You cannot compare Leica resolution to MF back, because the MF sensor is much larger. So 50 lp/mm on the sensor is close to 100 lp/mm on 135

2) Resolution is a poor parameter for sharpness. Edge contrast is the dominant factor for human vision. For instance a 50/1.4 lens at full aperture will have very good resolution but quite little edge contrast, because peripheral rays will dominate and those have higher aberration.

3) Large format is normally not limited by film but by lens and/or diffraction. LF lenses are normally used at small apertures, needed for both coverage and depth of field

Best regards
Erik

So.... now we can all agree on: 1. The test was crap.   2. DBs are better in resolution and noise than film of equivalent area.   3. Film retains the edge on other aspects of photography.   4.LF film is dead because the resolution parameter has been diminished by DBs and the improvement of MF film.   5. View cameras are only needed for their movements and can be as useful they were but with DB and 120/220 film instead of sheet.   Regards Theodoros   www.fotometria.gr

[Graham Mitchell]

Bottom line was real resolution above 100lpm, which the 39MP back can't do.

100lpm? I assume you mean lines pairs per millimetre (lp/mm)? In that case you have a total of 7200x4800 lines (or approx 34 megapixels). That's high but it still doesn't "blow away" a 39MP digital back. I'd like to see these results for myself. I'm willing to bet the film is a lot more noisy.

Now all that said, this combo is pretty limiting. It is black and white. It is slow film.

Exactly, so why bring it up? We're comparing with a colour digital back here so let's look at a standard colour film (which I used to use):


(courtesy http://www.normankoren.com)

As you can see the green layer has a resolution of around 40 lp/mm (at 50% MTF), which translates to 2880x1920 lines for 35mm film, or 5.5 MP. Velvia is not much better:



Once you add lenses into the equation, it gets significantly worse.

My point with the initial post was that the test was not done in a way that optimized and really showed what an 8x10 can do. The scan sample rate was at least 1/3 of what a typical 8x10 setup is capable of resolving with normal film. Come on, less than 1000dpi? Every test I have ever seen done on sheet film indicates that there is probably around 2000-3000 spi of real resolution on 8x10 film if it is shot at optimal apertures with good modern optics.

All I can do is repeat what has already been said. The article stated that there was no more detail to scan, and the scans back up that claim. They are already lacking any single-pixel-level detail. You can go ahead and scan at 10,000dpi if you like but it won't reveal detail that's not there in the first place. 8x10 lenses are just not achieving the full potential of the film. If you really believe 8x10 is capable of much more, why do all these tests demonstrate otherwise? Even if the potential of 8x10 is higher, does it really matter if the full potential is so difficult to achieve in the real world? If you have an actual sample to show, in which a scan of 3000spi contains meaningful detail at a per-pixel level,  I'm sure we'd all love to see it.

[RobertJ]

The article stated that there was no more detail to scan, and the scans back up that claim. They are already lacking any single-pixel-level detail. You can go ahead and scan at 10,000dpi if you like but it won't reveal detail that's not there in the first place.

But the reason there won't be any more detail at a higher resolution scan is not because the IQ180 is higher resolution than 8x10, but because the 8x10 image itself was botched. 
Not adequate. 
Not properly done. 
Misfocused. 
However you want to put it.  There seems to be no area of focus, or there was camera shake, or whatever.... which might prove another point -- An image on 8x10 or film in general can be screwed up pretty easily.

But if you actually believe the IQ180 is better than 8x10, I can easily say you're wrong.  I've seen 4x5 scans that, at 100 percent on screen viewing, match or exceed the resolution of the IQ180. 

I'm talking about 100 percent crops from a scan and an image that's around 320 Megapixels (and incredibly large file size) that actually has detail, and lots of it.  NOT anything like what is shown in the crops from the 8x10 in this unfortunate test. 

How would a properly shot 8x10 chrome compare?  Would it be just a little better than properly done 4x5, or perhaps in a different universe?

[ErikKaffehr]

Hi,

I guess that the perfect 8x10 image may not be that easy to achieve.


Regarding 4x5", Charlie Cramer did some comparison on P45 vs. 4x5" and essentially found them to be on par, possibly with some advantage for the 4x5. The outcome may also depended on post processing. At that time there were a lot of discussion claiming that Charlie scanned at to low resolution. The IQ180 has about double resolution (in MP) compared to the P45, but sensor size is the same.

The lenses may have improved meanwhile. The new Rodenstock HR lenses are very impressive regarding MTF, they achieve around 50%+ MTF at 80 lp/mm, double of what normal MF lenses use be capable of.

8x10" has four time the surfacer of 4x5" so it should have 4 times the resolution all other factors being constant. In real life the other factors are not constant.

All scans I have seen from film were quite soft and noisy at actual pixels, including my own.

Best regards
Erik

But the reason there won't be any more detail at a higher resolution scan is not because the IQ180 is higher resolution than 8x10, but because the 8x10 image itself was botched. 
Not adequate. 
Not properly done. 
Misfocused. 
However you want to put it.  There seems to be no area of focus, or there was camera shake, or whatever.... which might prove another point -- An image on 8x10 or film in general can be screwed up pretty easily.

But if you actually believe the IQ180 is better than 8x10, I can easily say you're wrong.  I've seen 4x5 scans that, at 100 percent on screen viewing, match or exceed the resolution of the IQ180. 

I'm talking about 100 percent crops from a scan and an image that's around 320 Megapixels (and incredibly large file size) that actually has detail, and lots of it.  NOT anything like what is shown in the crops from the 8x10 in this unfortunate test. 

How would a properly shot 8x10 chrome compare?  Would it be just a little better than properly done 4x5, or perhaps in a different universe?

[TH_Alpa]

I am on Graham's side here: we still wish to see such an example, being it 4x5" or 8x10.

But if you actually believe the IQ180 is better than 8x10, I can easily say you're wrong.  I've seen 4x5 scans that, at 100 percent on screen viewing, match or exceed the resolution of the IQ180. 

Not well. As stated by Graham, LF lenses are just not good enough with their extrem large ICs to be able to cope with 8x10" and use its full potential.

How would a properly shot 8x10 chrome compare?  Would it be just a little better than properly done 4x5, or perhaps in a different universe?

Thierry

[theguywitha645d]

8x10" has four time the surfacer of 4x5" so it should have 4 times the resolution all other factors being constant. In real life the other factors are not constant.

Resolution is a linear measurement, so an 8x10 should have twice the resolution.

But like you, I would like to see some results that disprove the results of this test. I agree with Graham that a higher resolution scan would not be pulling out more detail. Having worked in large format myself, motion blur would be obvious. You can see grain, so the scan is sharp. If the image is out of focus, where is it out of focus? The image shows it is covering the scene.

And T-1000 is right. We can say anything. That is why this test is good, it actually shows us something.

[fotometria gr]

Resolution is a linear measurement, so an 8x10 should have twice the resolution.

But like you, I would like to see some results that disprove the results of this test. I agree with Graham that a higher resolution scan would not be pulling out more detail. Having worked in large format myself, motion blur would be obvious. You can see grain, so the scan is sharp. If the image is out of focus, where is it out of focus? The image shows it is covering the scene.

And T-1000 is right. We can say anything. That is why this test is good, it actually shows us something.

Resolution may be linear for a scanner but is not for a lens, if a lens covers twice the diameter of another, it doesn't mean that it has twice the resolution on the image, actually its no where near than double the resolution. This is because the projection distance to the light recording surface increases and thus (for the same resolving power of glass) the "projected resolution" is decreased! The phenomenon is the same with digital sensors, where doubling the resolution of a sensor doesn't necessarily improve the resolution of the printed image but it does double the recorded analysis. That is why if you shoot "multishot" 16x with an old back, the resolution is (much) better than say an IQ 180 while image analysis is the same, with the 22mpx back the lens requires to cope with only the 22mpx for each of its shooting actions, while with the more condensed back the lens is required to cope with the increased analysis. The same of course applies to "multiscan" than "just scan", the scanners that can do multiscan perform much better in this mode and even more so the more multiscan they do, because they reject the minority of data and only keep the confirmed ones by the multiple scans. Regards, Theodoros, www.fotometria.gr 

[Fine_Art]

Resolution is a linear measurement, so an 8x10 should have twice the resolution.

But like you, I would like to see some results that disprove the results of this test. I agree with Graham that a higher resolution scan would not be pulling out more detail. Having worked in large format myself, motion blur would be obvious. You can see grain, so the scan is sharp. If the image is out of focus, where is it out of focus? The image shows it is covering the scene.

And T-1000 is right. We can say anything. That is why this test is good, it actually shows us something.

Clearly false. Resolution is based on area. The fundamental unit is the airy disk. If a system produces airy disks that overlap there is no resolution. If they are separate they are resolved. Test charts use converging lines because they are easy to distinguish. The resolution is then assumed to be the same over two dimensions.

An independent tester provided 4000, 6000dpi scans on page two. A scientist provided scans in the links on the original post. Some people continue to say it never happened.

[theguywitha645d]

Resolution may be linear for a scanner but is not for a lens, if a lens covers twice the diameter of another, it doesn't mean that it has twice the resolution on the image, actually its no where near than double the resolution. This is because the projection distance to the light recording surface increases and thus (for the same resolving power of glass) the "projected resolution" is decreased! The phenomenon is the same with digital sensors, where doubling the resolution of a sensor doesn't necessarily improve the resolution of the printed image but it does double the recorded analysis. That is why if you shoot "multishot" 16x with an old back, the resolution is (much) better than say an IQ 180 while image analysis is the same, with the 22mpx back the lens requires to cope with only the 22mpx for each of its shooting actions, while with the more condensed back the lens is required to cope with the increased analysis. The same of course applies to "multiscan" than "just scan", the scanners that can do multiscan perform much better in this mode and even more so the more multiscan they do, because they reject the minority of data and only keep the confirmed ones by the multiple scans. Regards, Theodoros, www.fotometria.gr 

Actually, my comment was much simpler. Resolving power is given in l/mm which is a linear measurement and so area is not a factor.

Projection distance (i assume you mean focal length or image distance) has nothing to do with the resolving power at the image plane. Given equal quality, the f-number/aperture (or the angular size of the aperture from the image plane) will determine the resolving power. This has nothing to do with the size of the image circle either, beyond the constraints of the optical design.

Actually, resolving power (or MTF) is additive. And so it is a systemic problem. You can get benefit from over sampling--an image sensor/film with a resolving power of 100 l/mm can benefit from a lens working at 200 l/mm, matching the resolving power of two components halves the the systemic resolution. However, the system is always limited to the weakest link. So a better scan might be able to pull out some more detail, but it will ever add any more.

Multi-sampling in a scanner is simply for noise reduction. Multishot backs that shift the sensor are adding resolution. These are really not the same technology.

[theguywitha645d]

Clearly false. Resolution is based on area. The fundamental unit is the airy disk. If a system produces airy disks that overlap there is no resolution. If they are separate they are resolved. Test charts use converging lines because they are easy to distinguish. The resolution is then assumed to be the same over two dimensions.

An independent tester provided 4000, 6000dpi scans on page two. A scientist provided scans in the links on the original post. Some people continue to say it never happened.

Really. I thought resolving power was measured in lines per mm, which is clearly a linear measurement.

You are speaking of resovling point sources/targets, which is not resolving power. It is actually harder to resolve points than lines and so the distinction is made. To be accurate, resolution still is easily achievable when the Airy disks start to overlap.

[TH_Alpa]

Tangential and sagittal rays, propagating in the tangential and in the sagittal planes: both planes are orthogonal to each other.

Source:

http://en.wikipedia.org/wiki/Astigmatism

Thierry

The resolution is then assumed to be the same over two dimensions.

[theguywitha645d]

Don't tell someone he is wrong if he likes what he does and is successful with it.

But no one is doing that.

[TH_Alpa]

Johannes,

Just a correction. I have been with Sinar for more than 20 years. Sinar did never make any research in the resolution field of films (or sensors), nor did they publish anything under their name. Sinar did possibly use general data available from the film or lens makers, but nothing more than that.

Luckily enough, all the (Sinar) experts are still alive as per today, retired (forced or legally) but alive.

 

Best regards
Thierry

The resolution of the various films (trans, neg, b&w) in different formats (35, mf, lf) is well known and was researched over many years by companies like Kodak, Agfa, Zeiss, Schneider, Rodenstock, Leica, Nikon, Sinar, Hasselblad and so on. It was compared in technical regimes (very high contrast) and for general photographic applications. Also typical errors in the field were analyzed.

You can find all that in technical publications of the companies, advanced photography textbooks and scientific publications in an university library. Sadly, most of the experts are retired or dead now. If one of these experts could see the behind the scenes photo that shows the 8x10 on that tiny tripod with the ball head he would probably laugh, I guess.

Johannes

[ErikKaffehr]

Hi,

My impression is that we get diminishing returns on high PPI scans. The samples you referred to look quite soft at actual pixels. I downloaded the 6000 PPI sample and did the following experiment:

1) Sharpened it to taste
2) Duplicated the image
3) Resized the duplicated image to 50% using bicubic
4) upsized the duplicated image to 200% using bicubic (so I downrezzed and than uprezzed by the same amount)
5) Pasted the downsized/upsized image above the original image
6) Added a layer mask over about 1/3 of the image

It is very hard to see any difference between the two images. I added a screen dump with the top layer in subtract mode. The mask is clearly visible, and the black area is the difference between the two images.

My conclusion is that there is very little excess information in the 6000 PPI file compared to the same file downscaled to 3000 PPI.

Best regards
Erik

Clearly false. Resolution is based on area. The fundamental unit is the airy disk. If a system produces airy disks that overlap there is no resolution. If they are separate they are resolved. Test charts use converging lines because they are easy to distinguish. The resolution is then assumed to be the same over two dimensions.

An independent tester provided 4000, 6000dpi scans on page two. A scientist provided scans in the links on the original post. Some people continue to say it never happened.

[fotometria gr]

I tried to explain how I understand the discussion that revolves around: "You are wrong - I'm right."

Best,
Johannes

Johannes, "you are wrong - I'm right" starts with the OP and is inevitable, otherwise we would have to accept all the crap anybody publishes or quotes as true and no criticism should be allowed. There can be two different approaches that can both be correct because of different values or priorities (like the guy with the Holga you suggested or digital vs. film for that matter) that people approach things with, but stating simplified inaccuracies all the time, like some (very few) are doing  constantly, is quite different. Regards, Theodoros. www.fotometria.gr
 P.S. I'm sorry that I took advantage of your statement to "picture" others, I wanted to avoid getting drugged to a pointless argument with that same crap as I state above.

[ErikKaffehr]

Hi,

I guess I meant information. Thanks for putting things right.

Best regards
Erik

Resolution is a linear measurement, so an 8x10 should have twice the resolution.

[ErikKaffehr]

Hi,

Sometimes around 1985, when Sony presented the original Mavica I arrived at the 6MP figure. If you look at Modern Photography resolution tests from that time anything above 50 lp/mm was excellent. On the other hand they have found out that with the best lenses, slowest film, best technique and focus bracketing they could achieve a bit over 100 lp/mm. But, resolution means barely resolvable, very little fine detail contrast.

In addition at that time, focusing accuracy yielded perhaps 30-40 lp/mm both AF and manual focus in tests also made by Modern Photography.

So I calculated that you need 100 pixels to resolve 50 lp/mm. A 135 image would thus be represented by 3600x2400 -> 8.64 MP. The 6MP figure I got came from using 3x2 cm as film size for easy calculation.

Best regards
Erik

You are right.

I have always thought that a 6mp file was close to typical 35mm film quality.  I say typical, not Tech Pan scanned at 8000ppi on a drum scanner.

Scale that up to 4x5, and we are talking about roughly 60mp, which I happen to think is about right (years ago I think I posted somewhere on this forum that you would need about that to match 4x5).  With four times the film area, that would mean 240mp is needed for 8x10 film equivalence (OK, knock a bit off for bigger issues with film flatness, but it is still more than double 80mp).

Of course everything I have just typed above is pure nonesense, because words like "typical", "equivalence" "quality" etc are so subjective, dependent upon use, and are only one small part of the entirte imaging story, meaning (of course) this entire thread is a hopeless waste of time  

By the way, I am more than happy with my 50mp Hassy H4D-50 which has so many actual advantages over film of any size I really don't want to go there....  If someone thinks 80mp matches 8x10 film, or 4x5 film, then they are "right" and "wrong" at the same time.

[fotometria gr]

Actually, my comment was much simpler. Resolving power is given in l/mm which is a linear measurement and so area is not a factor.

Projection distance (i assume you mean focal length or image distance) has nothing to do with the resolving power at the image plane. Given equal quality, the f-number/aperture (or the angular size of the aperture from the image plane) will determine the resolving power. This has nothing to do with the size of the image circle either, beyond the constraints of the optical design.

Actually, resolving power (or MTF) is additive. And so it is a systemic problem. You can get benefit from over sampling--an image sensor/film with a resolving power of 100 l/mm can benefit from a lens working at 200 l/mm, matching the resolving power of two components halves the the systemic resolution. However, the system is always limited to the weakest link. So a better scan might be able to pull out some more detail, but it will ever add any more.

Multi-sampling in a scanner is simply for noise reduction. Multishot backs that shift the sensor are adding resolution. These are really not the same technology.

You are clearly wrong on all of the above, l/mm of a lens are measured on a certain FP distance, if FP distance is increased (and thus image circle magnifies) resolution drops. This simply means that if two lenses with the same glass resolving power are used to record the same AOV at different formats, the lens that records on the smaller image area will appear to have more resolution. Nobody said anything about same technology or scanners adding resolution, I think you (and others) are confusing resolution with analysis! Please re-read the post and make sure you understand it, don't quote in irrelevant matters.  Cheers, Theodoros. www.fotometria.gr