Eversmart vs drum scanners & Aztek plateau

[bglick]

> I have said this before. If the detail is not on the film you are not going to be able to pull it out, no matter the quality of the scanner or the resolution of the scan.


Yep, this is the point often overlooked and is the basis of all the confusion of how much information is in a 4x5 scan, (in terms of MP's). This also relates to the scanned film vs. digital capture comparisons. Consider this.....


A 4x5 Provia capture of a landscape image, at f22 = a max. recorded resolution of 32 lp/mm assuming 1500/f and 60 lp/mm for Provia at a given contrast that is practical. Lets assume at f22, the lens IS apt. diffraction limited. As we know, often you must stop down futher for DOF....of course, if you shoot an infinity shot, you can open up a stop or two, based on the lens MTF data... (most LF lenses only perform well at f16+, but some of the newer modern designs can get to f11, (such as the SSXL's) Anyway, so f22 is nice middle of the road f stop for 4x5. We also know that there is other issues, such as film alignment, depth of focus at film plane, etc. All slightly degrade image quality, but lets assume there is no losses here, since f22 provides decent leeway for a decent camera set up.


Now, assume you had a scanner that was 100% eff and could extract every bit of that 32 lp/mm from the film. 4x5 film is 90x120mm.... extract the math, at 64 dots per mm, you have a 44MP worth of clean, captured data on the film. Of course, to extract this data, you are forced to oversample by a factor of 3 - 6, you can end up with 150MP to 300MP file....but that is only because its second generation image, again, Nyquist (1/R) is the cause of these large file sizes. They key is, there is only 44MP of recorded data in the entire 4x5 film.


Now, lets look at a 60MP digital back capture.... lets use a hypothetical 40x50mm sensor size, which it would take a 7000 x 8750 pixels = 61 MP. This equates to 88 pp/mm (pixel pairs per mm) To get the same DOF in the smaller format at the same size final print, you can reduce the f stop by the enlargement factor (and also the lens fl, but thats not relevant here) , which is 2.3x. (2.3 x 40 = 90 mm)... f22/2.3= f9.6. Of course, this assumes same composure.

Now run the math....

9.6 f stop
157 lp/mm aerial rez of lens (1500/f stop)
56 lp/mm 1/R

2.3 format factor to match 4x5 final size
25 lp/mm at 4x5 size


This demonstrates how a typical 4x5 color image shot at f22, scanned with perfect efficiency will match a 60MP digital capture on a smaller digital format. So why do scanned files have 6x more data? It's due to second generation oversampling requirement, which is Nyquist, regardless of what Phil once said. It's can't be more obvious.


Now you can see why those touting a 4x5 film capture is equal to 150 - 300MP digital capture is simply NOT comprehending all the issues at play. Many have stated, they believe what they see. Well, this explains why Mike R's Luminous Landscape test, of scanned 4x5 film vs. 39MP capture were very close, with the film just barely edging out the digital capture. this example above, suggests he is right. And even if his eyesight is poor as suggested, it's the same set of eyes being laid on both images. Also, he had several other people come to the same conclusion.


But then you have variables...such as Velvia will record more resolution than Provia, so the 4x5 film capture will win by maybe 15 - 20%......if you go to Tmax, the 4x5 capture will destroy the digital capture, probably by 40%+. But, as a more fair comparison, to match the DR of the MF digital capture, one must go to color neg film, now the 60MP digital capture will exceed scanned 4x5 film probably by 15%.

Anyway, the numbers are close, except B&W films like TMAX. Hopefully this will shed some light on how many MP are actually in scanned 4x5 film......vs how massive the scanned files are.

[bglick]

> Then there are all the beginners and non-photographers who may not even see the difference we focus on. Who gets to say when "it doesn't matter", or that's "good enough?"


This discussion was based around, how to compare the two.... vs. at what size you will see the difference.... as this introduces printing technique, printer type, viewing distances, resolution of the persons eyes, etc.

[dwhistance]

Bglick, many thanks for the well reasoned comments. Can I ask one question which bothers me - how does the Bayer interpolation (if I've got the right term!) affect this? Surely with a one shot back you need to reduce the effective number of pixels to take account of the distribution of the colour pixels? Also presumbaly the interpolation itself loses some resolution? If you do adjust for this how would the revised numbers look?

David Whistance

[bglick]

David, thanks for the kind words......

> Surely with a one shot back you need to reduce the effective number of pixels to take account of the distribution of the colour pixels? Also presumbaly the interpolation itself loses some resolution? If you do adjust for this how would the revised numbers look?


This is a great question....and it relates to what I described above as.... "color of subject", which is an issue with Bayer sensors, but NOT an issue with film (within reason, as even film has color recording characteristics). I assume the issue you raise relates to the RGGB Bayer color distribution.... The short answer is....


When shooting B&W targets, this issue is moot. White is compromised of RGB, so every pixel can detect white. Black is the absence of color, which all the pixels can sense this as well.


With Green targets, you have the benefit of 50% of the pixels being green. Since green is NOT a primary color, its a mix of Yellow and blue, the blue pixels sense the green subject. So Green targets hold about 75% of the total pixel count, or reduce the pp/mm by 25%. When run through 1/R, you will see Green targets resolve almost as well as B&W targets. I have been testing this for many years, and the test results always pan-out as expected.

But now it gets worse....

When shooting Red targets, obviously there is resolution degredation. The Blue and Green pixels can NOT sense red. You have effectively reduced your pixel count by 75%. The same applies to Blue. If you shoot TRUE RED or BLUE targets, you can easily see the resolution degradation. On avg, I have noticed about a 25 - 30% reduction in recorded resolution.

Of course, this makes the assumption, the subject is TRUE red, or TRUE Blue, often times this is not the case...for example, Blue often has a bit of green, making it recordable...

Of course, if you really want to "bring out the worse" in Bayer sensors, try shooting a Yellow-Orange.... Yellow Orange is the farthest from Red and Green on the color wheel...not fun.


Then, the next big issue is orientation of the targets. Obviously, the checkerboard pattern of digital sensors prefers the same orientation targets. Once the targets turn radial, again, another degredation of resolution.... to put this in perspective, working from memory, a blue radial target will produce half the resolution as B&W targets oriented the same as the sensor.

Of course, often these issues are only issues in testing, but sometimes they do surface in real world shooting. Of course, most of this loss of resolution is only evident when the image is enlarged to the point it becomes an issue....

[Michael Chmilar]

Green is a primary of additive (ie. transmissive) color:

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

Red, Yellow, Blue are primaries for subtractive (ie. reflective) color.

[dwhistance]

Bglick

Many thanks for the additional thoughts/explanation. I have followed the discussions on both this and the other thread with interest and enjoyed your original analysis which made sense to my non-mathematical mind (I gave up maths when I stopped being an engineer and became an accountant 25 years ago!). What slightly concerned me was that, when printing images for others, I have found that images made with one shot MF digital backs don't enlarge as well as your original analysis would suggest. Sadly, as I mostly print for landscape photographers, I haven't had the opportunity to print any images made in tethered multi-shot mode which would prove whether it is indeed the Bayer filtering that is causing this or whether it is something else - it could of course just be bad technique! However I have printed lots of images made with a Betterlight scanning back (I have one for my business) which hold up really well to enlargement so I suspect that the Bayer filtering is at least part of the problem.

David Whistance

[bglick]

Dave, how many MP of the BL back? How many MP of the MF back? Are you cropping down the MF? What colors are the subjects? What f stops are you shooting at? When MF lenses are you using? What size are you printing to? All these factors can significantly effect the recorded resolution, which in return will effect the final prints size, at a desired resolution.

[Donald Miller]

Quote Lenny Eiger "I think its superior, actually. I think its close in the way you are thinking, as I have a lot of faith in PMT's, as well as the scanners very good specs. However, that said, film has much more dynamic range that digital. I believe the next step in capture software will be the ability to both stitch and do HDR at the same time. I think that will be exciting, and offer many good possibilities, once the price of it comes down."

I don't know how much lower the price needs to be but I will post a couple of links that you may find interesting.

http://www.autopano.net/

http://www.paris-20-gigapixels.com/en/

Hope this provides information that is helpful.

[dwhistance]

Hi, the BL back is a 6000-HS - 6000x8000 pixels - 48 MP although I think Betterlight class it as 144 MP. I typically use it with Super Symmars (80XL, 110XL, 120HM, 150HM) and Apo Symmars (210, 240) generally at f11 or f16.

The MF back belongs to a client/friend. It is a Phase One P45. He uses it on a Linhof 679 with predominantly Scheider digital lenses. I am not sure what aperature he typically uses which could of course have a big impact on image quality.

The images I was printing that gave rise to my comment were at 32"x40" and were essentially uncropped (clearly they did need cropping slightly to fit this format but no other cropping was done). They were all autumnal Scottish landscapes if that helps describe the colours. I also scanned (on a Howtek 4000) and printed a few 4x5's for the same photographer at the same size (for the same exhibition, hence the format of the digital images). Of the two I felt the 4x5 images were sharper and showed more detail at this print size. When I printed them I did wonder about his technique with the digital images as, having photographed with him, I know he finds using movements much harder than he did with 4x5. He was also relatively new to the camera/back which may explain why the digital images didn't hold up so well.

I only use the Betterlight indoors but would have no hesitation in printing images of this size or even larger with it. I have never tried to compare it with 4x5 film on the same image (life is too short) but have always felt they were fairly evenly matched, its just that, for my purposes, the workflow with the Betterlight is much quicker and easier. Before I bought it I did, briefly, consider a MF digital back as an alternative but felt that for my purposes the Betterlight was better as well as cheaper.

I hope this is vaguely helpful, albeit hardly conclusive or scientific.

David Whistance

[bglick]

> I am not sure what aperature he typically uses which could of course have a big impact on image quality.


not a "huge" impact, but rather a HUGE impact :-) AT f4.5 on Digitar lens, a P45 can resolve up to 66 lp/mm... if the image is at infinity, this is close to reality. The same lens stopped down to f22 would record 37 lp/mm...almost half. See the significance here? That means at 3 lp/mm final print size, at f4.5 you can print 35 x 43" (using 40mm height of sensor). But at f22, at the same final print rez, you can only print to 19x24".


But more importantly, these are "best case numbers"..... add in some camera shake from wind, subject movement, long exposure, lens tilt which would greatly degrade the f4.5 shot, orange / red colors will really degrade the above values, etc. Again, details matter.... they REALLY matter....



> Of the two I felt the 4x5 images were sharper and showed more detail at this print size.


As you suggest, technique could be part of the problem, or the other variables I mentioned previously.... But, regardless, herein lies the beauty of 4x5 film....I will assume the photo used Velvia on the landscape, and assume f22, with good technique... This will record about 37 lp/mm to film (using 1/R), throughout most of the entire image, even if the scene had lots of DOF or at infinity. As at these higher f stops for LF (unlike MF), the resolution at the point of exact focus becomes nearer to the defocus positions (near / far). so at 37 lp/mm, your Howtek should extract most all this data, lets assume 90%, or 33 lp/mm, this will yield a 44x55" print at 3 lp/mm resolution on the print.


The punchline is here, the devil is always in the details..... MF digital capture is impressive, however, when a shot requires a lot of DOF, and you must stop down, its recording capacity is greatly minimized. In addition, if you view the MTF charts for MF digital lenses, their designs are optimized in the f5.6 - f8 range.... smaller apt. further degrade image quality, as apt. diffraction is not the only aberrations at play. OTOH, LF lenses are optimized in the f16 -32 range...some of the longer fl's in the f22 - f45 range.... this is why LF capture is a perfect match for Landscape photography....


I concur on your position on the BL back...if the subject is always static, you get more bang for your buck with a scanning back. The problem most people have is, they still need a one-shot capture back for non static subjects, so often the scanning back becomes a bit redundant. The added RGB per pixel site certainly adds a bit to resolution and tonality.....but certainly NOT the 144MP the maker claims.