DSLR Scanning - Signal / Noise

[interneg]

Pere, you don't need other people's negatives or anything of the sort. What you need to be is honest. Do you or do you not own an Epson V7xx or v8xx series scanner? A 'Yes' or 'No' is all that's needed.

If the answer is 'Yes' why do you refuse to re-scan your own HP5+ negative? A fresh scan of that extant negative following your claimed procedures is all it will take to show whether your claims have any foundation in fact or are repeatable on an Epson scanner. No need for histrionics or accusations or wild speculations. Either you can adequately image the granularity on the Epson at the resolutions you claim for it - or you can't. HP5+ in Xtol (or any other developer) is going to show visible granularity at the resolutions in question if the imaging chain has adequate MTF. Either show new evidence or accept that on the basis of your own extant evidence, your claims do not hold water.

[Pere Casals]

Pere, you don't need other people's negatives or anything of the sort. What you need to be is honest.

Lachlan, you say my Epson scans are faulty, isn't it ?

Is this faulty? https://www.flickr.com/photos/125592...5/32535835184/

You also say that this test is faulty:

https://www.largeformatphotography.i...-Drum-Scanners
https://www.largeformatphotography.i...=1#post1479178


You don't behave honestly, so we only have a possibility to know the truth: let's scan the same 4x5" negative with your X1 and with my V850, then we'll compare, this is the single fair way we have to see the reality, you have the privilege to select the negative. Why are you so scared to show the truth ?

You have my address, send me that negative and I'll demonstrate you that the Epson 4x5" scan is slightly superior than the X1 scan. If you also want to play fair this is the opportunity. Send me the negative.

[grat]

From what I've read, the Epson has been demonstrated to hit 2600 ppi on real-world resolution tests.

Replicating this in a single pass with a DSLR would require a resolution of 100+ megapixels-- made more difficult by the fact that most DSLR's use a 3:2 ratio instead of 5:4. So you're probably looking at a minimum of 4 stitched images, taken perfectly flat, with identical color balance. This isn't impossible-- you might even be able to adapt something like a FotoDiox 4x5 <-> DSLR adapter to a copy stand. If you're good with stepper motors and arduino, you could probably even build something to automate the process. Various software packages have gotten pretty good at stitching, although they still get a little confused now and again.

By using this process, increasing resolution far beyond the MTF of the lens that originally created the negative is certainly possible. It's only time and money.

But in terms of time and money, you're still way behind the guy mounting / dusting the 4x5 in a frame on the Epson, and telling Silverfast to scan, correct color cast, do a quick trim on the histogram and dump it into <insert editor here>.

[interneg]

Grat - the Epson's MTF response falls below 50% at just under 700ppi under the best possible test conditions. The last 70 years of imaging science have shown that the relevance of USAF-1951 imaging target is questionable at best, if the optical/ mechanical system does not have sufficient MTF in the first place. In the shortest terms: a nominally higher resolution imaging system with low MTF response across the board will produce visibly worse results than a nominally lower resolution system with very high MTF across the frequencies that matter in image reproduction. Most academic papers that look at the field of understanding the 'image content' potential of film or other sensors discuss this apparent resolution/ MTF relationship in detail, before inter-relating it to the noise of the sensor/ grain. The FUD about other scan systems is produced by those who have profoundly limited experience of Epson type scanners and almost nil of any other higher end scanning solution. The DASCH scanner used in the linked article has about the same resolution as the Epson supposedly has. The difference in results are very obvious - one can image the film granularity, one cannot.

[interneg]

Lachlan, you say my Epson scans are faulty, isn't it ?

Is this faulty? https://www.flickr.com/photos/125592...5/32535835184/

You also say that this test is faulty:

https://www.largeformatphotography.i...-Drum-Scanners
https://www.largeformatphotography.i...=1#post1479178


You don't behave honestly, so we only have a possibility to know the truth: let's scan the same 4x5" negative with your X1 and with my V850, then we'll compare, this is the single fair way we have to see the reality, you have the privilege to select the negative. Why are you so scared to show the truth ?

You have my address, send me that negative and I'll demonstrate you that the Epson 4x5" scan is slightly superior than the X1 scan. If you also want to play fair this is the opportunity. Send me the negative.

Your Epson scans are in line with essentially every Epson scan I have seen or made in terms of their fundamentally poor imaging of the film granularity. In and of themselves there is nothing wrong with them as average examples of the Epson scanner sensor. They are however deficient in the detailed grain rendering that an adequate MTF performance in the optical chain should deliver. It is therefore incumbent on you to show with new scans on your equipment with your negatives that you can produce imaging of the granularity on a par with high end scanners/ DSLR scans running at 1500ppi or similar, let alone 2000+ppi. It is striking that you never dare link to this post because it shows very, very clearly that the Epson lacks fundamental ability to handle fine detail in low contrast areas - indicative of poor MTF in the imaging chain.

It is also now clear from your unwillingness to give a simple yes/ no answer to questions over your ownership of an Epson V7xx or V8xx scanner that your actual ownership of one of these machines is doubtful and that your sum total experience of Epson scanners may extend solely to the handful of scans you repetitiously post. Either produce new evidence that demonstrates your claims or stop clogging up threads with stuff you have derived from the spec sheets, rather than practical experience.

If you can accept that an Epson can make an OK 1000-1200ppi scan, then that is closer to the reality of what the sensor array can deliver in useful (as opposed to theoretical) resolution. And for most people using/ playing around with LF, those sort of resolutions are fine for what they need to do.

[Pere Casals]

It is striking that you never dare link to this post because it shows very, very clearly that the Epson lacks fundamental ability to handle fine detail in low contrast areas - indicative of poor MTF in the imaging chain.

Yes.. In that post hes says: "I am sure the Epson file can be edited better but my main point here is that things just work with the ES and they look spectacular without any fuss."

Editing it better you have this difference:

Original:
https://www.largeformatphotography.i...=1#post1479176

Edited better:

https://www.largeformatphotography.i...=1#post1479178


Well, totally matching !!!!

When he made a serious side by side he found that results are equal. It is true that the Epson image requires a 2min careful edition to match the Creo super-scanner.

If you can accept that an Epson can make an OK 1000-1200ppi scan

LOL, Lachlan, the Epson is better than your X1 for 4x5", your X1 only resolves 1800 dpi effective for 4x5", total contrast extintion at 1800 dpi !!! The Epson resolves 2900 in the Horizontal axis so your expensive gear takes a beating from the Epson.

I say that the Epson makes a 4x5" negative slightly better than X1, let me reiterate: send me the negative you want, we'll compare the two scans side by side, this is the way you'll learn it. Send me that negative and you'll see.

[grat]

Whoa... Waitaminute.

You say:

The last 70 years of imaging science have shown that the relevance of USAF-1951 imaging target is questionable at best, if the optical/ mechanical system does not have sufficient MTF in the first place.

From the article you linked:

Testing was done by scanning a high quality chrome-on-glass USAF target that had a series of well defined and measured lines and spaces.

So first, that article "determines" the MTF of the Epson scanners by using the chart that you say can't be relied on. And then, rather than using the USAF target the way it was meant to be used, they used their own method for measuring MTF, and used THAT result to determine the effective resolution of the scanner. If that's not a classic case of confirmation bias, I don't know what is.

The whole point of the chart, is to test resolution-- If you can see this line, then your resolution is that. If you use a method of interpreting MTF based on your interpretation of the results of scanning that chart, and your resolution numbers wildly disagree with the chart's actual results, then perhaps you need to reconsider your methods.

It's entirely possible they measured the MTF of the USAF-1951 target. After all, it is a piece of glass, and as such, has a refraction index, and therefore, must affect the light passing through it.

Further, that article uses the word "assume" or "assumption" multiple times, along with phrases such as "I believe" and "probably", and just to be petty, either the author doesn't know the difference between the word "mute" and "moot", or it was never reviewed by an editor.

So your position is that based on a questionable test, the Epson series of scanners, which people have been buying for the past decade, perform far, far worse at scanning negatives than pretty much any other scanning device on the market, and no one has noticed but you.

Now, you'll note that I stated the Epson has been tested to perform at somewhere around 2300-2600 DPI, which is far worse than claimed by Epson, but at least it has a reasonable test to back it up (the aforementioned USAF resolution chart, which if the website had it in the wrong place, it would indeed, have a lower resolution than expected, because apparently the Epson does best in the middle of the scanner bed).

But that's my real problem with these kinds of tests. People either deliberately (or accidentally), handicap the product they're testing, and then provide only partial information on their method and results-- don't tell me you used the USAF chart, and then show a partial crop of the result, show the full result-- including the rest of the scan bed, because that might be relevant. These days, any amateur who can put three sentences together can produce an article or a blog and claim to be an expert without any understanding of basic scientific method.

If that article's all you've got to "prove" the horrible MTF of the Epson scanners, versus all the people who frequently upload high resolution images scanned with the Epson scanners, then good luck-- you're going to need it.

Oh, and a drum scan beats your DSLR scanning, according to at least one professional photographer I'm aware of, who scanned the same negative with a DSLR, an Epson v850, and a professional lab with a Noritsu drum scanner.

[interneg]

Grat: a Noritsu is, no matter how much you want to believe it to be, not a drum scanner. It's clear that you haven't done basic due diligence here. The USAF target (indeed, nearly any sort of bar chart target) can be used to determine MTF if used appropriately. High contrast resolution alone (which is what most who buy USAF-1951's use them for) is a bad determinant of useful imaging capacity (MTF, noise, latitude all matter much more in determining the useful resolution and information storage capacity). If you want more detail, there are innumerable academic journal articles that cover the maths in considerable detail. The upshot of all of them is that total 'resolution' matters much less than a high MTF response in the visually important part of the image - including resolution of some degree of image granularity, which the Epson fails to do well, because of its poor interpolation of 2x 1200ppi sensor lines.

[grat]

Grat: a Noritsu is, no matter how much you want to believe it to be, not a drum scanner. It's clear that you haven't done basic due diligence here.

I've done exactly as much diligence as you, in that I'm quoting other people on the internet. I have no personal experience with commercial scanning solutions. Doing some quick searches, it appears you are correct, the Noritsu is not a drum scanner. Unfortunately, I can't be bothered to go back and check to see if I misremembered what the professional photographer said (possible) or he misspoke (also possible). I can't afford one, and even if I could, there are other things I'd rather purchase. Also, I have no need to prove my resolution is bigger than yours.

The USAF target (indeed, nearly any sort of bar chart target) can be used to determine MTF if used appropriately. High contrast resolution alone (which is what most who buy USAF-1951's use them for) is a bad determinant of useful imaging capacity (MTF, noise, latitude all matter much more in determining the useful resolution and information storage capacity).

And the article you linked, did it badly. Something you still haven't addressed.

If you want more detail, there are innumerable academic journal articles that cover the maths in considerable detail. The upshot of all of them is that total 'resolution' matters much less than a high MTF response in the visually important part of the image - including resolution of some degree of image granularity, which the Epson fails to do well, because of its poor interpolation of 2x 1200ppi sensor lines.

And the fundamental problem here is that you're using math and MTF to "prove" that a scanner that thousands of people are happily using to scan negatives, performs at only 15% of it's published capability, and no one but you is aware of it. Just as the article you linked to used MTF to produce numbers at odds with the actual test equipment they used, you are using MTF to disprove real-world experience.

This is like the old theory that people wouldn't be able to breathe if their cars went over 30 miles per hour, or that bumblebees don't have enough wing-to-mass ratio to allow them to fly. If the evidence doesn't support your math, then there may be a problem with your math.

Fundamentally, you have yet to disprove my statement that while using a DSLR to "scan" negatives can surpass the resolution limits of an Epson perfection v700 or greater, the time and effort to do so is non-trivial. You'll get bonus points if you can do it without personal attacks.

[interneg]

Grat - anyone who has actually made decently sized prints from an Epson scan and looked at them alongside the same image off any decently high end scanner will be able to see that there are significant defects in the Epson scan.

With moderate competence, DSLR scans can do a much better job than the Epson in terms of sharpness and granularity resolution. I have seen the results and worked on the files to my satisfaction - there are issues that need refined, but they are merely ones of colour correction. A reasonably current 24mp APS-C sensor can do a better job than the Epson.

If you like the Epson, fine, but it's not in any way comparable to what either high end scanners or intelligently used CMOS sensors can do.