V600 capable of higher resolution than assumed?

[_T_]

I recently bought an Epson V600 with which to scan my 4x5 film. I looked up several available scanners on filmscanner.info and decided based on their tests that the V600 would be sufficient for such large negatives despite the low tested resolution, but once I actually got to scanning some 4x5s on it I quickly realized that filmscanner.info were sorely mistaken about the capabilities of the scanner and that the actual achievable resolutions are quite a bit higher than their tests achieved.

First I will present a 100% zoom of a 6,400 dpi scan of a piece of HP5. This is the highest possible resolution of the scanner.

As you can see we're resolving beyond the grain of the film. We can see the film grain and it is not only present, but it is definite beyond the mere absence or presence of each individual grain (i.e. beyond the pixel level.) I have intentionally left the bed of the scanner dusty in these tests so that you can see the resolution of the scanner as it relates to not only the finite resolution of the image in the negative, but also the infinite resolution of the physical dust.


Here now is a 400% zoom showing that the grain is not only present as the presence or absence of each grain, but that the scanner is resolving beyond the grain to show the contours of each grain, resolution beyond the information that the film contains.

Next I provide two images for comparison. The first scanned at 3,200 dpi, the resolution recommended by filmscanner.info, and another scanned at 6,400dpi. These scans were made one directly after another so that the film and dust in the scanner would not be disturbed. They show that there is resolution beyond what is achievable with a 3,200 dpi scan. I encourage you to open the images in multiple tabs or windows and switch between them quickly so that you can defeat change blindness.

3,200dpi at 3,200% zoom all matched

6,400dpi at 1600% zoom all matched

As you can see there is information contained in the 6,400dpi scan that does not exist in the 3,200 dpi scan. It is clear that the resolution has increased, but the question is does this increase in resolution correspond to an increase in actual information captured? For this I bring your attention to the two miniscule pieces of dust seen in the center of the red circle. We see at 3,200dpi that there is a small vaguely 8 shaped blob of lighter pixels, this could be anything and looks quite like grain in the film. But when we capture that same area at 6,400 dpi we can see that there are two diagonal lines of lighter color pixels much smaller than the surrounding film grain. This information was not contained in the lower resolution scan. They are not an artifact of the scanner which captures the image in horizontal lines. I believe this must be additional information captured. I do not believe it to be aberration or artifact but to reflect the actual capability of the scanner.

Please let me know what you think.

 

[brbo]

Seeing more (real) detail in 6.400dpi scan than in 3.200dpi scan (on that same scanner) is unfortunately no proof that scanner can resolve more than 3.200dpi. It's just proof that increased (real) resolution comes from oversampling (stepper motor having better resolution than lens). Very easy to check if you have a resolution target.

 

[Alan Edward Klein]

How do you scan 4x5 film with a V600? It's not made to do that.

 

[_T_]

Seeing more (real) detail in 6.400dpi scan than in 3.200dpi scan (on that same scanner) is unfortunately no proof that scanner can resolve more than 3.200dpi. It's just proof that increased (real) resolution comes from oversampling (stepper motor having better resolution than lens). Very easy to check if you have a resolution target.

Real resolution is real resolution. And the scanner is not designed to oversample at 6,400 at all 6,400 is the nominal optical resolution. I’m not claiming it achieves 6,400 dpi but rather that filmscanner.info performed their tests inadequately, leaving a significant amount of resolution on the table unreported.

How do you scan 4x5 film with a V600? It's not made to do that.

I made a tray and mask out of black card stock and scan in 2-3 pieces depending on the resolution before stitching in photoshop

 

[cramej]

A blob does not resolution make. Filmscanner.info's test methodology is solid because they're using a known resolution target to measure. Just because there are pixels smaller than something on a negative doesn't mean it can resolve it.

 

[_T_]

I agree but I’m saying their methodology, which is not listed, was flawed since it’s clear that the scanner is out resolving the film which would be impossible at 1350 dpi that’s nowhere near acceptable for scanning miniature negatives and yet the scanner performs admirably.

And look at the test target scan they show. They have test patterns larger than the size of the eyelashes in my photo looking like a blurry mess.

There’s no way they didn’t get the scan out of focus somehow

 

[brbo]

Real resolution is real resolution. And the scanner is not designed to oversample at 6,400 at all 6,400 is the nominal optical resolution.

That's exactly how this desktop scanners are designed. Stepper motor resolution ("Y-resolution") is real, but unfortunately lens and sensor that would have that resolution ("X-resolution") are absolutely beyond the mechanical and economical envelope of those devices. Read this article carefully to understand how this class of scanners is designed. The resolution which you call "assumed" by many people is in fact measured and in line with the design. People testing this scanners may not go into trouble of finding the absolute best position of the film above the scanning bed so they in fact might be a bit out of focus and leaving some resolution on the table, but that's understandable. Their job is to test a scanner as supplied.

Get a resolution target and see for yourself that V600 can't get anywhere near 3.200dpi. It will probably be 1.500dpi out of the box, maybe close to 2.000dpi if you are really lucky or you play with the film holder position.

Epson 4990 @4.800dpi nominal resolution which is actually resolving about 2.000dpi:



(if you get considerably more from any 4990 or V600 you can have my drum scanner for free)

 

[Helge]

I agree but I’m saying their methodology, which is not listed, was flawed since it’s clear that the scanner is out resolving the film which would be impossible at 1350 dpi that’s nowhere near acceptable for scanning miniature negatives and yet the scanner performs admirably.

And look at the test target scan they show. They have test patterns larger than the size of the eyelashes in my photo looking like a blurry mess.

There’s no way they didn’t get the scan out of focus somehow

Not clear at all. Scanning is treacherous in making you think you “got everything”.
The way the matrix of the sensor and the grain (clumps) interact, creates this confusing illusion.

It might be out resolving the camera focus and the particular lens used for all we know, but not the film.

It’s also vexing how often HP5 pops up in scanner tests. It’s a film with many qualities, but it’s probably the worst film to test a scanner with of all. Grainy, with grain clumping, relatively thick emulsion and not particularly high resolution.

This scan has not “emptied out” the film.

 

[_T_]

That's exactly how this desktop scanners are designed. Stepper motor resolution ("Y-resolution") is real, but unfortunately lens and sensor that would have that resolution ("X-resolution") are absolutely beyond the mechanical and economical envelope of those devices. Read this article carefully to understand how this class of scanners is designed. The resolution which you call "assumed" by many people is in fact measured and in line with the design. People testing this scanners may not go into trouble of finding the absolute best position of the film above the scanning bed so they in fact might be a bit out of focus and leaving some resolution on the table, but that's understandable. Their job is to test a scanner as supplied.

Get a resolution target and see for yourself that V600 can't get anywhere near 3.200dpi. It will probably be 1.500dpi out of the box, maybe close to 2.000dpi if you are really lucky or you play with the film holder position.

Epson 4990 @4.800dpi nominal resolution which is actually resolving about 2.000dpi:



(if you get considerably more from any 4990 or V600 you can have my drum scanner for free)

I’m not saying you can get considerably more than any number you’ve given, just literally significantly more than measured by filmscanner.info. Countably more resolution than 1350dpi

 

[_T_]

And yes it is assumed you act like a website called filmscanner. God damned info is the same thing as a peer reviewed journal. This isn’t scientific in any way

 

[ant!]

And yes it is assumed you act like a website called filmscanner. God damned info is the same thing as a peer reviewed journal. This isn’t scientific in any way

Filmscanner.info is mostly a shop for film scanners, and they like Silverlight. I never used this software, but that this site says in most reviews how much better the results with this are is strange to me... Why should give using the manufacturer's SW or vuescan give other resolution values?
So, while there is a lot of useful information on this site, you need to interpret their reviews a bit...

 

[brbo]

I’m not saying you can get considerably more than any number you’ve given, just literally significantly more than measured by filmscanner.info. Countably more resolution than 1350dpi

They measured it at 1.560dpi. I'd bet nobody can get more than 2.000dpi out of that scanner. If 2.000dpi is significantly more than 1.600dpi, ok. But V600 does NOT outresolve film grain in ISO400 film.

Filmscanner.info is mostly a shop for film scanners, and they like Silverlight. I never used this software, but that this site says in most reviews how much better the results with this are is strange to me... Why should give using the manufacturer's SW or vuescan give other resolution values?
So, while there is a lot of useful information on this site, you need to interpret their reviews a bit...

It's called SilverFast and, yes, filmscanner.info does want you to believe that Silverfast can somehow get more resolution out of your scanner, which is not true. They also particularly like Reflecta and dislike Plustek scanners. But lets hope that demonstrating that their scanner tests are wrong will be backed with more than just an opinion...

 

[250swb]

Out of interest I wondered who these 'filmscanner.info' people were and lo and behold it turns out to be good old ScanDig, and I say 'good' with reservations. I think generally speaking you should always interpret their reviews and not treat them as Gospel. They are like the Ken Rockwell of the scanner world, so anything they say can be very useful, but any bits that particularly interest you should be cross referenced because they are biased and seemingly unable to put right inaccuracies in their articles that come to light. This has led to their shortcomings and out of date opinions becoming the go-to mythology of scanners and 'facts' provided by them get trotted out without checking.

 

[brbo]

Quite right. One should double-check not just ScanDig, but others too.

Having personally tested a number of scanners that were also tested by ScanDig I can tell you that ScanDig is probably much much closer in their assessment of what you can expect from an Epson V600 (out of the box) than anybody claiming the ability of this scanner resolving "beyond grain". They make money from selling this scanner, why would they knowingly lie about it's capability? To coax you into buying a more expensive scanner like V850? I doubt it, cause if you listen to some people here, they are grossly wrong about that scanner's performance too. So they are obviously not interested in selling V850 either or any other desktop scanner they are selling since the only scanners that perform really great in their tests are discontinued...

I have yet to see anyone post significantly better results from their scanners (using supplied holders!) than what is shown in ScanDig's reviews (apart from Plustek scanners, I mentioned that they either have extremely bad luck with them or straight don't like them (one model in particular)). I also find it quite odd that people bashing ScanDig reviews don't believe in using a resolution target for measuring scanner's resolution...


A nice "user" review of V600 that doesn't pretend to be what it's not and also has side-by-side comparison to Fuji Frontier 2.000dpi scans.

 

[koraks]

As you can see we're resolving beyond the grain of the film.

I don't see it.

the presence or absence of each grain,

I suspect your understanding of what grain is in relation to silver particles is not entirely sufficient to make the interpretation you think you're making.

Here now is a 400% zoom showing

mush. It's showing mush.

I've scanned quite a bit of HP5+, as well as other kinds of film, on a scanner that evidently outresolve yours as well as a flatbed that's in the same ballpark. What you've shown in your example is a relatively low optical resolution interpolated image to fit a certain dpi resolution. I can tell you that if you scan at a higher resolution, you'll notice the grain takes on a sharper and better defined look. The results you posted are pretty much on par with a scanner that resolves in the 1000dpi-1500dpi range.

Here's an example of HP5+ scanned at 3200 on a scanner with an optical resolution that's around 3000dpi (according to the people you don't trust):

This is scanned without further processing such as sharpening. Note that the grain (same film as yours!!) is quite well-defined and looks, well, sharp. If you were right in saying that your 6400dpi scan has already outresolved the film, then the above scan simply couldn't exist. Alright, my scan is 3200 dpi, yours is 6400, so you're skeptical. Let's do a little experiment.

I could upsample my scan to 6400dpi so you can compare it with your scan, to which end I took a randomly selected small sample of your scan so you can see the difference:

It's the same digital resolution (6400 dpi) which you can easily see is a pretty meaningless piece of information; you can upsample any low-res image to insane levels by whatever interpolation mechanism you prefer, or apply oversampling if you're fancy, but this doesn't restore image information that wasn't scanned in the first place. On the left, there was a little more information to begin with, so the upsampled image doesn't look as mushy as the one on the right, which evidently didn't have much image information to begin with.

You could approximate the optical resolution of your scanner by downsampling your scan to the point where things start to look like they're actually in focus. This starts to happen around the point where you reduce your scan in size to about 25% of the original. That translates to around 6400/4 = 1600dpi. Interestingly, that's pretty close to what some people who take the time to test this properly try to make you believe.

Alright, for shits & giggles, here's a flatbed scan from a scanner that's in the same league as yours. It's a little older, but in terms of optical resolution, it's somewhere in the same ballpark.

This is again HP5+, I think developed in Rodinal, and scanned at 3200dpi. Here's the same thing (well, a crop) upsampled to 6400dpi:

We're now looking at the same sort of mushy image as your original scan. You could brush it up a little by adjusting contrast and applying some sharpening; I did neither and just lazily cropped part of the jpg scan as I had it on my harddrive somewhere.
If I play with this a little, I can determine that the actual optical resolution of this flatbed scan would be somewhere around 1600dpi, pretty much like yours. That's again consistent with the German guys who rated the scanner I use at around 1800dpi if memory serves.

Now, you can always be overly optimistic and argue that the pixels you're seeing are all meaningful, but that doesn't necessarily have any bearing on the real world.

Sorry to be so critical, but if you're critical on someone else's work without showing the money, then you make yourself vulnerable for the same kind of argument.

PS: I have no interest in the filmscanner people commercially or otherwise. In the past, I did consult their work frequently and have always found it to be consistent with real-world experiences, either mine or other people's. Their testing methods I also find sufficiently transparent to be able to interpret the outcomes, so I'm not exactly sure what your problem is with them in this respect.

 

[Alan Johnson]

When I tested my V700 some time back I obtained a resolution close to that of scandig.com so I think their results reliable.

Resolution of the Epson V700

I made a 35mm Agfa Copex Rapid microfilm negative of a test target photographed from the correct distance with a Canon 50mm f1.8 at f5.6. Under the microscope 125 line-pairs/mm is visible on this negative. The attachment shows the result of scanning with the V700.Only a very small part of the...

www.photrio.com

However I noted that at low contrast more lppm may be seen, this may be what the OP is observing.

 

[Alan Edward Klein]

Quite right. One should double-check not just ScanDig, but others too.

Having personally tested a number of scanners that were also tested by ScanDig I can tell you that ScanDig is probably much much closer in their assessment of what you can expect from an Epson V600 (out of the box) than anybody claiming the ability of this scanner resolving "beyond grain". They make money from selling this scanner, why would they knowingly lie about it's capability? To coax you into buying a more expensive scanner like V850? I doubt it, cause if you listen to some people here, they are grossly wrong about that scanner's performance too. So they are obviously not interested in selling V850 either or any other desktop scanner they are selling since the only scanners that perform really great in their tests are discontinued...

I have yet to see anyone post significantly better results from their scanners (using supplied holders!) than what is shown in ScanDig's reviews (apart from Plustek scanners, I mentioned that they either have extremely bad luck with them or straight don't like them (one model in particular)). I also find it quite odd that people bashing ScanDig reviews don't believe in using a resolution target for measuring scanner's resolution...


A nice "user" review of V600 that doesn't pretend to be what it's not and also has side-by-side comparison to Fuji Frontier 2.000dpi scans.

For whatever its worth, I did a comparison on 4x5 Tmax 100 between V850 and Howtek 8000 drum scanner. The V850 compared favorably.

Howtek 8000 Drum vs. Epson V850 flatbed scanners

Here's a comparison between Howtek HiResolve 8000 drum scanner and Epson V850 flatbed. Negative is mine, Tmax 100, processed normal by North Coast Photo dip and dunk with Clayton F76+ developer (similar to Kodak D76). Taken on a Chamonix 45H-1, Schneider Symmar APO 150mm with B+W orange...

www.largeformatphotography.info

Also, here's a comparison on 4x5 Tmax 100 vs Tmax 400 on a V850. See the skies for grain.
Tmax 100 V850
Tmax 400 V850

Here's Tmax 100 in 120 format scanned with V600

 

[Alan Edward Klein]

A couple more for comparison.

35mm TriX scanned V600
35mm Tmax 400 V850

 

[_T_]

Now we’re getting what I’m here for. Evidence. If anyone else has any scans to compare I’d be ecstatic to see them.

I don't see it.



I suspect your understanding of what grain is in relation to silver particles is not entirely sufficient to make the interpretation you think you're making.



mush. It's showing mush.

I've scanned quite a bit of HP5+, as well as other kinds of film, on a scanner that evidently outresolve yours as well as a flatbed that's in the same ballpark. What you've shown in your example is a relatively low optical resolution interpolated image to fit a certain dpi resolution. I can tell you that if you scan at a higher resolution, you'll notice the grain takes on a sharper and better defined look. The results you posted are pretty much on par with a scanner that resolves in the 1000dpi-1500dpi range.

Here's an example of HP5+ scanned at 3200 on a scanner with an optical resolution that's around 3000dpi (according to the people you don't trust):
View attachment 315505
This is scanned without further processing such as sharpening. Note that the grain (same film as yours!!) is quite well-defined and looks, well, sharp. If you were right in saying that your 6400dpi scan has already outresolved the film, then the above scan simply couldn't exist. Alright, my scan is 3200 dpi, yours is 6400, so you're skeptical. Let's do a little experiment.

I could upsample my scan to 6400dpi so you can compare it with your scan, to which end I took a randomly selected small sample of your scan so you can see the difference:
View attachment 315507
It's the same digital resolution (6400 dpi) which you can easily see is a pretty meaningless piece of information; you can upsample any low-res image to insane levels by whatever interpolation mechanism you prefer, or apply oversampling if you're fancy, but this doesn't restore image information that wasn't scanned in the first place. On the left, there was a little more information to begin with, so the upsampled image doesn't look as mushy as the one on the right, which evidently didn't have much image information to begin with.

You could approximate the optical resolution of your scanner by downsampling your scan to the point where things start to look like they're actually in focus. This starts to happen around the point where you reduce your scan in size to about 25% of the original. That translates to around 6400/4 = 1600dpi. Interestingly, that's pretty close to what some people who take the time to test this properly try to make you believe.

Alright, for shits & giggles, here's a flatbed scan from a scanner that's in the same league as yours. It's a little older, but in terms of optical resolution, it's somewhere in the same ballpark.
View attachment 315508
This is again HP5+, I think developed in Rodinal, and scanned at 3200dpi. Here's the same thing (well, a crop) upsampled to 6400dpi:


View attachment 315510
We're now looking at the same sort of mushy image as your original scan. You could brush it up a little by adjusting contrast and applying some sharpening; I did neither and just lazily cropped part of the jpg scan as I had it on my harddrive somewhere.
If I play with this a little, I can determine that the actual optical resolution of this flatbed scan would be somewhere around 1600dpi, pretty much like yours. That's again consistent with the German guys who rated the scanner I use at around 1800dpi if memory serves.

Now, you can always be overly optimistic and argue that the pixels you're seeing are all meaningful, but that doesn't necessarily have any bearing on the real world.

Sorry to be so critical, but if you're critical on someone else's work without showing the money, then you make yourself vulnerable for the same kind of argument.

PS: I have no interest in the filmscanner people commercially or otherwise. In the past, I did consult their work frequently and have always found it to be consistent with real-world experiences, either mine or other people's. Their testing methods I also find sufficiently transparent to be able to interpret the outcomes, so I'm not exactly sure what your problem is with them in this respect.

No need to apologize. This is exactly what I’m here for. I’d rather be proven wrong and learn better than continue to be wrong without my knowledge.

My problem with filmscanner.info is one of methodology. They have sample sizes of one, they have no statistical analysis because of that, and they do not publish the precise steps they took to obtain their data.

It’s very difficult to draw conclusions based on anecdotal evidence like this. Hence my skepticism.

 

[brbo]

My problem with filmscanner.info is one of methodology. They have sample sizes of one, they have no statistical analysis because of that, and they do not publish the precise steps they took to obtain their data.

It’s very difficult to draw conclusions based on anecdotal evidence like this. Hence my skepticism.

What is Epson's, yours or anyone else's sample size, methodology? Where is all that published? Where does the confidence that ScanDig definitely has it all wrong come from?

 

[_T_]

There is no data anywhere. I do not have it, you do not have it, scandig doesn't have it. Nobody has studied this scientifically in any way; no significant sample size, no statistical analysis; yet scandig is making claims about the resolution of all these scanners they have basically never studied. Thus they are suspect.

 

[koraks]

Ah no, now I must point you at your own methodological flaw: assuming that what you are aware of is all there is

Now, I understand what you mean with the n=1 argument, but there are many counter arguments. One could rely on manufacturing tolerances, which are rather tight on this kind of equipment, especially today. Even an n=1 sample says quite a lot. Moreover, it depends on unit of analysis; if you shift that to the actual scanned image, you can achieve a very high N even with one scanner (at the risk of systematical errors; but see the remark on manufacturing technologies above).

And I could go on and on when it comes to methodology. (Believe me, I really can. And I can also virtually guarantee you I can continue a methodological debate with you until the point where you just give up in disgust and frustration. I'm not kidding, either, nor am I boasting or engaging in power play. In this sense, you kind of happened on the wrong guy.) The only conclusion you could ultimately draw is one of fundamental skepticism: yes, you could mistrust every source based on their inevitable flaws, leaving you with no verifiable knowledge but the knowledge you acquire yourself, until you realize that this is equally (or more so) fundamentally flawed.

This problem seems to be not one of scanner performance or measuring methodology, but one of epistemology. This one goes deeper. I'll leave you with it; my interest currently is in photography. Your challenge is an interesting one, but exclusively yours to face.

 

[brbo]

I linked to a pretty serious document that studied the performance of Epson V750. They came to a conclusion that this type of scanners are worth about 1200dpi (but they obviously have much higher demands about MTF contrast %).

There are so much desktop scanners around that we would surely see some evidence of much better performance than "assumed". Why we haven't see any?

I can post comparisons between my other scanners and Epson 4990 flatbed, but now I know I should better spend my time on something else, since I'm just a sample of one, so worthless...

 

[koraks]

Epson 4990 flatbed

What I find somewhat remarkable is that in terms of real resolution, flatbeds haven't really progressed beyond the 4990. At least not the affordable ones. I suppose the scan bar manufacturing technology kind of stuck where it already was back then. I bet it shares a lot with laser printer illuminator/led bar technology and that's also pretty much stuck at 1200dpi AFAIK. One place I worked at they appeared to have enough difficulty keeping that technology available, let alone even think about progressing. It's just no use pushing it any further; no way to recoup that investment.
Anyway, I digress. Interesting stuff in its own right, but all too easy to get lost in it.

 

[grat]

Read this article carefully to understand how this class of scanners is designed.

That article is laden with the words "assume", and "probably", and has very few facts at it's foundation when you dig into the details.

There was another article that was being bandied around that used a back-handed method for determining MTF that concluded the maximum resolution of the Epson V750 was 700 PPI.

Personally, having spent some time getting to know my V800, I can say there is a difference between a 2400 PPI scan and a 3200 PPI scan in terms of resolution-- something absolutely not possible if the maximum resolution is 1200 PPI.

If your math proves that a bumblebee can't fly, the bumblebee doesn't really care-- it just keeps flying.