From Youtube: Drum Scan vs DSLR vs Epson (Via Nick Carver)

138S · Feb 22, 2020

Lachlan Young said:
I understand it perfectly. And the implications it makes about your technical abilities, honesty and understanding. Do you understand this?

Once again, Did you perform that operation on the screenshot or the original files? Do you understand what the implications are of not trying the sharpening operation on the original files?

It's mostly the same because crops are 100%, compression do no provocates a loss because at 100% compressed file outresolves a lot original Image Quality. If you want download the original files from the Pali test (I did it) and check it on your own, you will find the same. If not, post the difference you find. Hey... do it...

In post #85 you say that the Epson is worth for 1x to 2x enlargement... totally false
(https://www.photrio.com/forum/threa...on-via-nick-carver.172770/page-4#post-2253293)

From MF even at 15x it matches the Hasselblad X5, if the Epson image properly edited.

Lachlan Young · Feb 22, 2020

138S said:
It's mostly the same because crops are 100%, compression do no provocates a loss because at 100% compressed file outresolves a lot original Image Quality. If you want download the original files from the Pali test (I did it) and check it on your own, you will find the same. If not, post the difference you find. Hey... do it...

In post #85 you say that the Epson is worth for 1x to 2x enlargement... totally false
(https://www.photrio.com/forum/threa...on-via-nick-carver.172770/page-4#post-2253293)

From MF even at 15x it matches the Hasselblad X5, if the Epson image properly edited.

I have tested it: download it & look at it, if you view it online it may be compressed for faster viewing. The inversions are very quick and approximate. That is the Epson with your claimed settings & everything downsampled to 2400 and screenshotted at 50%. You'd see a definite difference in prints - the Epson always looks softer.

The test negative is also problematic as is comprised of high contrast edges and little in the way of fine detail at lower contrasts & smooth tone scales which are a far tougher test of both scanners and sharpening.

And while you're at it, why don't you tell us why you are refusing to engage with the MTF tests done on the Epson scanner I linked earlier?

138S · Feb 22, 2020

Lachlan Young said:
I have tested it: download it & look at it, if you view it online it may be compressed for faster viewing. The inversions are very quick and approximate. That is the Epson with your claimed settings & everything downsampled to 2400 and screenshotted at 50%. You'd see a definite difference in prints - the Epson always looks blurrier.

You are a bit wrong with math

Images are at 3200 dpi, so perfectly binned 4 pix in one, at 3200 100% the image is still very soft, so nothing was lost.

that image: https://www.largeformatphotography....rum-Scanners&p=1479178&viewfull=1#post1479178

Is a 100% screenshot, and still soft because largely surpassing the original image quality. It is 1842x 1048 and not full HD because a crop of the screen was made.

If you don't belive me, then download the original full resolution images from the provided links that are close 1GB : https://www.largeformatphotography....rum-Scanners&p=1479169&viewfull=1#post1479169

Do it (like I did) and tell us what difference you find: You'll find none, but discover it on your own.

138S · Feb 22, 2020

I have more ammo...

This is the Collaborative Large Format Scanner Comparison https://www.largeformatphotography.info/scan-comparison/

This is the test 4x5" film image:

Here you see compared remarkable crops from Epson V750 and Howtek 4500 drum, and also an edited V750 image:

Well, the drum image is slight better that V750 edited image in a 20x, wich would be 2.5m wide print...

Also you may donload crops to check it on your own...

Helge · Feb 22, 2020

138S said:
Of course, but a lens resolving 300MPix effective in a 5.9" frame, in macro conditions, this is a fine lens working at optimal magnification, no doubt. I'm used to check performance of MF and LF lenses in macro conditions, belive me, a 300MPix in 5.9" projection is a really fine lens.

A lens of this quality would cost a lot if sold as retail photographic grear, try to buy two new lenses like those in a V700 and you will see that cost can be several times the price of a V850, but you know... industrial components mass produced have a 1/15 of the retail price or less.

Several false staments and lies have been thrown by Epson haters about the V700 glass, some were saying that lenses were bad plastic... At first I believed this lie, but when I saw that it optically resolved 2900pix over a 150mm field then I found that it was a coarse lie.

300 MP for what is essentially a large format macro lens is not very good.
It sure AF doesn't get anything close to everything out of a 4x5 or 8x10.

You argument continues to be essentially "but it's Good Enough ™".
Well, it's not.
Even with a relatively modest print like Carvers in the OP you'll see a clear difference.
We are not talking nit picking and being anal retentive, obsessive compulsive, Asperger like, autistically obsessed with minute differences.
These are big and obvious differences, once you don't view the photos 1:1 on a screen.

A big part of of having a print on the wall is walking up close to it, to suck in all the detail and atmosphere in that detail.

I'm all for the low resolution aesthetic. I use halfframe all the time. Hell, I'd be using Minox if I could be arsed to get a cutter.
I get it.
But when you shoot larger formats and good, fine grain film, you want a high resolution result!

Sharpening is not the answer because:
A: You could of course do the exact same sharpening to the higher resolution scan, and get a better result.
More information is better. You can't invent something that is not there,
B. The Epson scans react poorly to sharpening. The false grain structure will get blown up, or even be mistaken for real objects in the frame by an "intelligent" sharpening tool.

All the amateur tests are fine. But we can't be sure of their conditions and stringency.
"Did they really get the focus right in the source?".
"How was their scanning technique and was the scanner calibrated?".
"What kind of PP did they actually use and what kind of PP where they not aware of being automatic?".
"Are they lying?".
"Do they actually have an agenda, even without perhaps being fully aware of it themselves?".
Etc.

138S · Feb 22, 2020

Helge said:
B. The Epson scans react poorly to sharpening.

It reacts poorly to sharpening made by people not having basic skills to make a good sharpening job. Me I'm not a master, but for MF and up I'm able to match what a V850 does with what does an X5, see this post:

https://www.photrio.com/forum/threa...on-via-nick-carver.172770/page-7#post-2255045

Helge · Feb 22, 2020

138S said:
It reacts poorly to sharpening made by people not having basic skills to make a good sharpening job. Me I'm not a master, but for MF and up I'm able to match what a V850 does with what does an X5, see this post:

https://www.photrio.com/forum/threa...on-via-nick-carver.172770/page-7#post-2255045

For a start please have a look at my addendum in the previous post.

That doesn't at all look like a 50 MP scan, simply for the fact that the grain is still so mushy. It looks upscaled.
That, or the focus is off somewhere in the optical chain, or a little all over.
What film was used? I don't think he is clear about that. Looks like HP5 or TriX. Edit: Oh, I see now, it was actually HP5! Hardly a good test of ultimate resolution and sharpness.?
In fact, if I play the devils avocado, they could be the very same image with a different grain filter put over.

Still if we go by our hunch that he is in fact honest. The high contrast clean areas are where sharpening has a very easy job.
Look at the shirt texture for example. Far bigger difference!

But overall a bad test.
Too many unknowns, and too many obvious problems.

138S · Feb 22, 2020

Helge said:
What film was used? I don't think he is clear about that. Looks like HP5 or TriX?

"The film was Ilford HP5+ rated at 250 ISO and developed in Ilford Perceptol 1:3."

On any doubt you may download the full size scans to check it on your own:

https://petapixel.com/2017/05/01/16000-photo-scanner-vs-500-scanner/

Tom Kershaw · Feb 22, 2020

Helge said:
Looks like HP5 or TriX. Edit: Oh, I see now, it was actually HP5! Hardly a good test of ultimate resolution and sharpness.?

Exactly. There is a massive difference between HP5 Plus and T-Max, Delta, Pan F Plus etc. I don't want to get into the Epson debate but my with my LS-9000 I wouldn't do scan quality tests with a fast traditional grain film.

138S · Feb 22, 2020

Lachlan Young said:
And while you're at it, why don't you tell us why you are refusing to engage with the MTF tests done on the Epson scanner I linked earlier?

For a 4x5" sheet the X5 has contrast extintion (MTF 0%) by 1800dpi effective, the Epson reaches 2900dpi. Of course for 35mm the X5 is way better.

Probably the Epson has some higher flare than X5, but this is easily compensated by adjusting levels in Ps.

138S · Feb 22, 2020

Tom Kershaw said:
Exactly. There is a massive difference between HP5 Plus and T-Max, Delta, Pan F Plus etc. I don't want to get into the Epson debate but my with my LS-9000 I wouldn't do scan quality tests with a fast traditional grain film.

Well... grain structure depiction can be critical for aesthetics... in that job in 35mm the x5 makes a superb job, the Nikon ED also works great, but the older nikon models had a too collimated illumination delivering a not nice Callier effect.

Helge · Feb 22, 2020

Well, I tried to do some crops that are fairly equal without making it a job, just to show to the casual reader what difference in dataset we are actually talking about.
V700 one the left and Flextight on the right.

The Flextight obviously has a much better view of the grain, but still doesn't out resolve it. You can see that grain conforms to the pixel matrix in many cases.

There is something awry with both scans. For example, one thing that is immediately noticeable is that the shadow detail is poorer on the Flextight which shouldn't be technically possible.

Of course I had to scale them to upload them, but still I think it shows just how much PP went on and how much detail the sharpening had to infer.
Have a quick look at the original files if you are interested, it is actually fairly interesting.

Now some of the obvious differences is obviously down to contrast/backlight type. But I hope we a familiar enough with those pitfalls here to look through them, and see the real huge difference.

And again, this can't be stressed enough: He should have used TMAX, Delta or even FP4 or some other fine grain film, for this test to be really interesting.
Either he didn't know, which makes us question his expertise in general.
Or he had a bias.

villagephotog · Feb 22, 2020

Lachlan Young said:
And while you're at it, why don't you tell us why you are refusing to engage with the MTF tests done on the Epson scanner I linked earlier?

I'm reluctant to step in to this, and I have seen quite clearly with my own tests that both a Coolscan 9000 and a Flextight can produce better scans of a given patch of film than an Epson V700, in absolute terms. Camera scanning at high magnification can do even better.

But I'll just throw this thought in re: the paper you linked to, which I read. The author doesn't mention any efforts to optimize the focus of the Epson scanner that he tested (unless I missed it; if so, please correct me). It seems far-fetched to me that he would have neglected this, so I don't know either way, but, as I said, he doesn't seem to mention it, and that leaves it open to doubt.

As you know, this is critical with any test of a flatbed scanner, which, unlike the Coolscans and Flextights, will not have a focusable lens. The resolution of my V700 is quite poor using the stock Epson film holder at its default height. I use an adjustable BetterScanning holder, and I need to raise it 800 microns higher than the Epson default height to get best focus. I begin to see a resolution drop if I adjust the holder as little as 300 microns up or down.

warden · Feb 22, 2020

villagephotog said:
I'm reluctant to step in to this

Too late! haha

villagephotog said:
I use an adjustable BetterScanning holder, and I need to raise it 800 microns higher than the Epson default height to get best focus.

Me too, although I haven't measured anything I have found using the BetterScanning ANR glass holder mandatory with my scanner.

Helge · Feb 22, 2020

villagephotog said:
I'm reluctant to step in to this, and I have seen quite clearly with my own tests that both a Coolscan 9000 and a Flextight can produce better scans of a given patch of film than an Epson V700, in absolute terms. Camera scanning at high magnification can do even better.

But I'll just throw this thought in re: the paper you linked to, which I read. The author doesn't mention any efforts to optimize the focus of the Epson scanner that he tested (unless I missed it; if so, please correct me). It seems far-fetched to me that he would have neglected this, so I don't know either way, but, as I said, he doesn't seem to mention it, and that leaves it open to doubt.

As you know, this is critical with any test of a flatbed scanner, which, unlike the Coolscans and Flextights, will not have a focusable lens. The resolution of my V700 is quite poor using the stock Epson film holder at its default height. I use an adjustable BetterScanning holder, and I need to raise it 800 microns higher than the Epson default height to get best focus. I begin to see a resolution drop if I adjust the holder as little as 300 microns up or down.

Well you might have a point, or at least have grabbed onto an unknown.
The DASCH digitiser he mentions in the paper and that is the comparison point, only has a 16MP sensor, so the plate would have to be badly out of focus for it to have the reported advantage.

AFAICS the lens used to scan large format (that probably includes the plates of this size) is the same one used for scanning non transparent materials, so it should have a focus that lies at the outer surface of the glass platen.
The plates sizes are given as "15 mm (5/8 inch) wide by 180 mm (7 inches) long and ∼50 mm x 80 mm (2 x 3 inches)".

So if the above is right the test should still hold.

It would seem silly for the authors to go through so much trouble, and then miss such an obvious thing.
But of course anything is possible.

This is by the way the home page of the project:
http://dasch.rc.fas.harvard.edu/project.php

Edit: That they scanned the U.S. Air Force 1951 to that resolving level and especially Figure 7s pretty accurate confirmation of the lp/mm limit of the system, would indicate that they did in fact achieve correct focus.

Lachlan Young · Feb 22, 2020

138S said:
You are a bit wrong with math

Images are at 3200 dpi, so perfectly binned 4 pix in one, at 3200 100% the image is still very soft, so nothing was lost.

that image: https://www.largeformatphotography....rum-Scanners&p=1479178&viewfull=1#post1479178

Is a 100% screenshot, and still soft because largely surpassing the original image quality. It is 1842x 1048 and not full HD because a crop of the screen was made.

If you don't belive me, then download the original full resolution images from the provided links that are close 1GB : https://www.largeformatphotography....rum-Scanners&p=1479169&viewfull=1#post1479169

Do it (like I did) and tell us what difference you find: You'll find none, but discover it on your own.

Your condescending assumptions reveal that you haven't looked at what I posted, or ever done the work yourself to any worthwhile extent. If you had, you would know what resolution the Epson file was posted (as opposed to scanned at) at & that my downsample was coincidentally actually closer to the binning ratio you claim than yours would have been.

I based my analysis off using the actual full resolution un-inverted files. Not some load of concocted claims on a screenshot like you so blatantly did. I gave the Epson the best possible chance by downsampling everything else to within its supposed performance envelope and still it failed. If it had been a smooth neutral tone image with fine detail, it would have failed even faster.

That my results don't comply with your silly agendas aren't my problem. The point still stands that the Epson is not up to the job with films below the grain visibility threshold for its optical system and really not up to the job with films above that threshold.

I care about reasonably accurate, sharp representation of the grain and film sharpness characteristics (like in a good darkroom print) - and the Epson cannot deliver it. Period. It is likely because of the severity of the point spread function caused by the method by which it tries to blend the overlapping sensor lines. Because of the sensor shortcomings elsewhere, this cannot be easily corrected for. A pixel shifting CMOS sensor should comfortably outdo the Epson if competently used.

Samuel Dilworth · Feb 22, 2020

138S said:
Well... grain structure depiction can be critical for aesthetics... in that job in 35mm the x5 makes a superb job, the Nikon ED also works great, but the older nikon models had a too collimated illumination delivering a not nice Callier effect.

Does that mean that if a Nikon scanner has “ED” in the name, it’s fine?

The Coolscan IV ED (LS-40) sells for much less than the Coolscan V ED (LS-50) that I had in the past but am reluctant to pay for now, especially since the price has actually gone up in the last ten years.

The Coolscan IV and V seem to be mechanically similar, i.e. much better made than the Konica Minoltas, Canons, etc. That appeals to me, especially when buying a scanner old enough to vote.

Tom Kershaw · Feb 22, 2020

Some of these older scanner are still going for fairly serious money and with potentially limited repair options: https://www.ffordes.com/p/SOR-18-021312/scanners-accs/ls9000ed-scanner - my own example is in much better cosmetic condition.

138S · Feb 22, 2020

Helge said:
Well, I tried to do some crops that are fairly equal without making it a job, just to show to the casual reader what difference in dataset we are actually talking about.
V700 one the left and Flextight on the right.

The Flextight obviously has a much better view of the grain, but still doesn't out resolve it. You can see that grain conforms to the pixel matrix in many cases.

There is something awry with both scans. For example, one thing that is immediately noticeable is that the shadow detail is poorer on the Flextight which shouldn't be technically possible.

Of course I had to scale them to upload them, but still I think it shows just how much PP went on and how much detail the sharpening had to infer.
Have a quick look at the original files if you are interested, it is actually fairly interesting.

Now some of the obvious differences is obviously down to contrast/backlight type. But I hope we a familiar enough with those pitfalls here to look through them, and see the real huge difference.

And again, this can't be stressed enough: He should have used TMAX, Delta or even FP4 or some other fine grain film, for this test to be really interesting.
Either he didn't know, which makes us question his expertise in general.
Or he had a bias.

View attachment 240600 View attachment 240601 View attachment 240602

After processing V850 the scan in this way, a curve and a sharpening:

We get this:

Both images are equally sharp but grain depiction is very different at 20x enlargement.... The texture pointed by the red arrow is better in the Epson.

But at x10 we have exactly the same:

Helge said:
And again, this can't be stressed enough: He should have used TMAX, Delta or even FP4 or some other fine grain film, for this test to be really interesting.

And it also can be stressed less, with Portra, Ektar, Fuji 160Pro images would match totally, because negative color films have been re-engineered to be easier to scan.

Anyway TMX is very easy to scan with the Epson.

138S · Feb 22, 2020

Lachlan Young said:
Your condescending assumptions reveal that you haven't looked at what I posted, or ever done the work yourself to any worthwhile extent. If you had, you would know what resolution the Epson file was posted (as opposed to scanned at) at & that my downsample was coincidentally actually closer to the binning ratio you claim than yours would have been.

I based my analysis off using the actual full resolution un-inverted files. Not some load of concocted claims on a screenshot like you so blatantly did. I gave the Epson the best possible chance by downsampling everything else to within its supposed performance envelope and still it failed. If it had been a smooth neutral tone image with fine detail, it would have failed even faster.

That my results don't comply with your silly agendas aren't my problem. The point still stands that the Epson is not up to the job with films below the grain visibility threshold for its optical system and really not up to the job with films above that threshold.

I care about reasonably accurate, sharp representation of the grain and film sharpness characteristics (like in a good darkroom print) - and the Epson cannot deliver it. Period. It is likely because of the severity of the point spread function caused by the method by which it tries to blend the overlapping sensor lines. Because of the sensor shortcomings elsewhere, this cannot be easily corrected for. A pixel shifting CMOS sensor should comfortably outdo the Epson if competently used.

Lachlan, Fact: at 10x from MF the Epson and the X5 shows the same, at 20x the grain looks different but both images are equally sharp.

Lachlan Young · Feb 22, 2020

Tom Kershaw said:
Some of these older scanner are still going for fairly serious money and with potentially limited repair options: https://www.ffordes.com/p/SOR-18-021312/scanners-accs/ls9000ed-scanner - my own example is in much better cosmetic condition.

Probably worth bookmarking this page on repairing the more recent Coolscans.

Lachlan Young · Feb 22, 2020

138S said:
Lachlan, Fact: at 10x from MF the Epson and the X5 shows the same, at 20x the grain looks different but both images are equally sharp.

I know what that level of sharpening looks like on a print and it's UGLY. Disgustingly, obviously so. And the X5 scan is still clearly actually resolving better throughout, especially of fine detail. The results of which will be painfully obvious in a print. Quite apart from the X5 scan pretty clearly having been done with the USM still switched on which doesn't help matters.

138S · Feb 22, 2020

Samuel Dilworth said:
Does that mean that if a Nikon scanner has “ED” in the name, it’s fine?

The Coolscan IV ED (LS-40) sells for much less than the Coolscan V ED (LS-50) that I had in the past but am reluctant to pay for now, especially since the price has actually gone up in the last ten years.

The Coolscan IV and V seem to be mechanically similar, i.e. much better made than the Konica Minoltas, Canons, etc. That appeals to me, especially when buying a scanner old enough to vote.

True, but in the Minolta light collimation can be adjusted by using the "Grain Dissolver" feature. IIRC in the Nikon can use an "Scanhancer diffuser". I don't remember exactly what Nikon models were too collimated and what of those can be diffused with an "Scanhancer diffuser"... just saying that coliimation can be a concern in some Coolscans models.

http://www.scanhancer.com/index.php?art=15&men=15

138S · Feb 22, 2020

Lachlan Young said:
I know what that level of sharpening looks like on a print and it's UGLY.

Well, now that we can exactly match the crops in the monitor you argue that the print would be different...

I never had a poblem to print nicely those images that look nice in the monitor, you should review your workflow.

OK, I concede that there are nuances, but the Epson makes a close match to the X5 for MF, no doubt.

The funny thing happens in 4x5", while the Epson keeps exactly the same performance than for MF in the X5 case 4x5" performance is the half than in MF. Weren't you aware ?

I guess that after paying a lot for an X1 realizing that a "cheapo" Epson does the same... it has to be has to be quite irritating

Helge · Feb 22, 2020

Lachlan Young said:
I know what that level of sharpening looks like on a print and it's UGLY. Disgustingly, obviously so. And the X5 scan is still clearly actually resolving better throughout, especially of fine detail. The results of which will be painfully obvious in a print. Quite apart from the X5 scan pretty clearly having been done with the USM still switched on which doesn't help matters.

The Flextight scan is just way too contrasty. Even though you can see the that there is extra detail, it's very hard to bring forward because there simply isn't the tonality.
Did the author bork it on purpose?

From Youtube: Drum Scan vs DSLR vs Epson (Via Nick Carver)

Privacy & Transparency

Privacy & Transparency