Epson V800 2400ppi Only After Down Sampling From 6400ppi

[IanBarber]

I have read on numerous websites that the V800 in reality only produces an effective resolution of around 2400ppi. This I can accept but I also stumbled across another article which claims that in order to arrive at 2400ppi, you really need to be scanning at 6400 and then down-sampling in say Photoshop.

Has anyone done any testing to confirm this. personally, I havent seen much difference when I have tried it but I am curious to hear from others who may have.

[locutus]

I would not be surprised if this is true, it seems the limitation with many of these scanners is the optics. Quite some other scanners have the same issue.

For 35mm have a Plustek 8200 scanner and it behaves the same, i scan at full reso (7200dpi iirc) and then downscale the resulting image afterwards to 4330x2880 (largest 3:2 image on a 5k monitor). The resulting file is noticably sharper then having selecting a lower resolution directly in Silverfast

[IanBarber]

What method are you using to downscale

[Alan Klein]

Trust your own eyes.

[faberryman]

Has anyone done any testing to confirm this. personally, I havent seen much difference when I have tried it but I am curious to hear from others who may have.


Trust your own eyes.

filmscanner.nfo tested the V800 and found that it achieved its maximum resolution of 2300 when scanned at 4800. Scanning at 6400 increased the file size but not the actual resolution. I trust the testing rather than my own eyes because I do not have a USAF resolution chart to conduct my own testing. I don't downscale my files until printing. Storage is cheap.

[locutus]

What method are you using to downscale

Just a simple downscale in GIMP.

[Ted Baker]

I have read on numerous websites that the V800 in reality only produces an effective resolution of around 2400ppi. This I can accept but I also stumbled across another article which claims that in order to arrive at 2400ppi, you really need to be scanning at 6400 and then down-sampling in say Photoshop.

Has anyone done any testing to confirm this. personally, I haven’t seen much difference when I have tried it but I am curious to hear from others who may have.

AFAIK the test method for scanners simple takes the effective size between pixel, and divides it by 1 inch. In the case of V800 the sensor size is 1/3200 inch and then by overlapping two rows of sensors you get 1/6400 on an inch.

The 2400dpi or what ever is just some made up test using a target and your own eyes, not a bad idea really :-)

For my own tests with my 4990 scanning at the "highest resolution" 4800dpi results in the best scan, but 2400dpi is actually very close, and is much quicker.

When I do scan at 4800dpi I use the Lanczos resizing algorithm, to resize, there is definitely a small difference.

[faberryman]

AFAIK the test method for scanners simple takes the effective size between pixel, and divides it by 1 inch. In the case of V800 the sensor size is 1/3200 inch and then by overlapping two rows of sensors you get 1/6400 on an inch.
The manufacturer's resolution numbers are just the resolution of the CCD chip, and do not include the optical and mechanical systems. The actual test resolution is always lower, typically 50% lower.

[Ted Baker]

The manufacturer's resolution numbers are just the resolution of the CCD chip, and do not include the optical and mechanical systems. The actual test resolution is always lower, typically 50% lower.

I think that's kind of what I just said... ;)

However in order to get that 50% lower amount, it is necessary to use the system as it best. If you use it slightly less than its best you might end up with 55% lower amount...

[faberryman]

AFAIK the test method for scanners simple takes the effective size between pixel, and divides it by 1 inch. In the case of V800 the sensor size is 1/3200 inch and then by overlapping two rows of sensors you get 1/6400 on an inch.

The manufacturer's resolution numbers are just the resolution of the CCD chip, and do not include the optical and mechanical systems. The actual test resolution is always lower, typically 50% lower.

I think that's kind of what I just said... ;)
You said that it was a "test method". I was just pointing out the it is not a test method, but a recitation of CCD sensor specifications by the manufacturer. Some manufacturers add insult to injury by also adding software interpolation to the CCD sensor specification. It is how Reflecta (Pacific Image), for example, gets to a resolution of 10,000 from a scanner that is only capable of an actual resolution of 4100. (Still very good for 35mm, but not as advertised). That is why it is important to seek out actual test results when selecting a scanner, to determine 1) actual resolution, and 2) scan resolution to achieve actual resolution. I think we are in agreement of these points.

[Pere Casals]

Hello Ian,

Here you have a table with an extensive test showing effective dpi results scanning at different nominal dpi. V750 and V850 are mostly the same, beyond illumination.

179150

As you can see is not necessary to scan 6400 to get most of the V850...

You may see that, usally as with any scanner, USAF 1951 based measurement depends on the pass, each time can be a bit different measurement because the smallest pattern you see may be more or less aligned with the pixels, having more or less aliasing.

Here it's reported that even 2900 dpi have been measured for vertical bars, for horizontal bars measurements are a bit lower.

When evaluating scanners should be considered the best value you obtain from several samples, because the lower measurements can be provocated by aliasing between the position bars position in respect to the pixels.

But well, this table reports from a trusted source that scanning 6400 dpi with vertical bars with have average 2643 dpi resolving power, and with horizontal bars we have average 2148 dpi.


If you scan 6400 and you want to downsample with Photoshop then go to "image size" dialog, at the bottom you can select the interpolation/binning algorithm, use "Bicubic ideal for reductions".


We have "3 kinds" of sharpening: Input, Creative and Output.

Normally we may sharpen the raw file from the scanner, while editting we can perform creative local sharpening like in the eyes in a portrait, but perhaps not in the cheek, and after resizing to match the output device (printer) we can perform another sharpening for the pixel level.

Those average 2400 effective dpi makes the V850 a very powerful scanner for 8x10, and a very good choice for 4x5. If extreme/important densities are there, then a drum is what works.

For rolls a Plustek 120 (or 8xxxi) is better if the shot is very sharp or if we want the grain structure.

[Ted Baker]

You said that it was a "test method".

Fair point, but the USAF test target is also not really a proper test method for scanners either.

There is actually a proper test method ISO 16067-2. Which uses a slanted edge and is much more appropriate.

[faberryman]

There is actually a proper test method ISO 16067-2. Which uses a slanted edge and is much more appropriate.
Do you have a link to any test of a scanner using ISO 16067-2? Do any manufacturers use it and rate there scanners accordingly? Sometimes we have to use what is available. Using a USAF test chart, even if less than optimum, gives us a common method whereby we can assign relative merit among scanners, even if we could obtain better precision though other methods.

[Ted Baker]

Do you have a link to any test of a scanner using ISO 16067-2? Certainly no manufacturer that I am aware of uses it. We have to take what we can get. Using a USAF test chart, even if less than optimum, gives us a common method whereby we can assign relative merit among scanners.

No, maybe they do maybe don't, but the point being is the USAF test chart was for the analogue world of optics and film. I agree it gives a measure of how good the scanner is. What it doesn't tell you is how to get the best from your scanner, which is believe at the heart of the question that Ian is asking.

[Pere Casals]

No, maybe they do maybe don't, but the point being is the USAF test chart was for the analogue world of optics and film. I agree it gives a measure of how good the scanner is. What it doesn't tell you is how to get the best from your scanner, which is believe at the heart of the question that Ian is asking.

well, we can also slant the 1951 target...

but also having hor & vert bars is a source of information, irrelevant for practical concerns, but having both measurements is technically relevant.

[faberryman]

No, maybe they do maybe don't, but the point being is the USAF test chart was for the analogue world of optics and film. I agree it gives a measure of how good the scanner is. What it doesn't tell you is how to get the best from your scanner, which is believe at the heart of the question that Ian is asking.

The OP's original question was what scanning resolution to use with a V850 to get maximum actual resolution. filmscanner.nfo answered that question with 4800 using the USAF chart. I doubt ISO 16067-2 would have given a different answer. Knowing maximum actual resolution can be achieved at 4800 rather than 6400 save time and disk space.

[Ted Baker]

The OP's original question was what scanning resolution to use with a V850 to get maximum actual resolution. filmscanner.nfo answered that question with 4800 using the USAF chart. I doubt ISO 16067-2 would have given a different answer. Knowing maximum actual resolution can be achieved at 4800 rather than 6400 save time and disk space.

And I am politely trying to point out that conclusion is wrong...

[faberryman]

And I am politely trying to point out that conclusion is wrong...
What is the correct answer?

Pere's chart shows 3000 for horizontal and 6400 for vertical, but the methodology and software is not specified (lthough the use of the word "target" in the footnotes. indicates they used the USAF test target). It is also well to note, that in Pere's chart, the testers did not test at a 4800 scanning rate. They jumped from 3000 to 6400. filmscanner.nfo tested at both 4800 and 6400, finding no increase in resolution at 6400. Perhaps if the testers in Pere's chart had tested at 4800, they would have come to the same conclusion.

[Ted Baker]

What is the correct answer? Pere's chart show 3000, but the methodology is not specified.

Actually pere charts shows 6400, if you read it carefully, and take both vertical and horizontal.

But there is more to it than just the resolution of USAF test chart, which is my point about the misleading conclusion from filmscanner.nfo

[faberryman]

Actually pere charts shows 6400, if you read it carefully, and take both vertical and horizontal.

But there is more to it than just the resolution of USAF test chart, which is my point about the misleading conclusion from filmscanner.nfo

Apparently there is a time delay between the time I revised my post and the time you responded. Please reread my post.

In what way is the filmscanner.nfo test misleading. It seems pretty consistent with Pere's chart, whichever "trusted source" it came from.

[Ted Baker]

Apparently there is a time delay between the time I revised my post and the time you responded. Please reread my post.

That filmscanner.nfo conclusion is still misleading, but perhaps now we are arguing for the sake of arguing...

Things to consider are:

1. USAF test target is not suitable to show how to extract the maximum capabilities of the unit. (not saying it is not a useful measure for comparison)
2. Does the scanner actually support 4800? It is not capable of the that resolution in the X dimension, the sensor size never changes, just the amount of overlap.
3. How does grain aliasing picture in this.
4. What resolution is most suitable for capture sharpening, or an actual debluring algorithm

[faberryman]

That filmscanner.nfo conclusion is still misleading, but perhaps now we are arguing for the sake of arguing...

Things to consider are:

1. USAF test target is not suitable to show how to extract the maximum capabilities of the unit. (not saying it is not a useful measure for comparison)
2. Does the scanner actually support 4800? It is not capable of the that resolution in the X dimension, the sensor size never changes, just the amount of overlap.
3. How does grain aliasing picture in this.
4. What resolution is most suitable for capture sharpening, or an actual debluring algorithm

Interesting questions. I still don't see what is misleading about the filmscanner.nfo review. Which of its conclusions is inaccurate? It doesn't answer every question and is not a how-to article, but that doesn't make it misleading. Do you have any contrary data?

[IanBarber]

We have "3 kinds" of sharpening: Input, Creative and Output.

Normally we may sharpen the raw file from the scanner, while editting we can perform creative local sharpening like in the eyes in a portrait, but perhaps not in the cheek, and after resizing to match the output device (printer) we can perform another sharpening for the pixel level.

For sharpening the Raw file, I usually use Un-Sharp Mask (low radius high amount). Are there any other methods which may do a better job.

[Ted Baker]

Interesting questions. I still don't see what is misleading about the filmscanner.nfo review. Which of its conclusions is inaccurate? It doesn't answer every question and is not a how-to article, but that doesn't make it misleading. Do you have any contrary data?

I have a 4990 and have run my own tests, and run testing continuously, admitted it is NOT a v800.

But you might find the following contrary data on filmscanner.nfo itself... at http://www.filmscanner.info/en/Aufloesung.html

The answer is clearly yes. During the scan of a film, there are interferences as aliasing, other interferences, etc. They are all related to the intermediate effects of the used resolution of the scanner and the actual image resolution. The higher the scan resolution, the lower are those interferences. Thus, it does make sense to scan an original with 4000 dpi that has a resolution of only 2500 dpi by itself.

But hopefully I have opened your eyes to the possibility that there is more too it. I am not suggesting Ian scan everything at 6400dpi either! I gave the method I use earlier for his and anyones consideration.

[IanBarber]

Although I have scanned mainly at 2400 it dawned on me that I might not actually be getting 2400, hence the reason for asking.

I have now scanned about 6 or 8 4x5 negatives both at 2400 and 6400 (Linear) and I have to say, when inspected closer, unless my eyes are deceiving me, I would say the 6400 ones do look better. When I say better, they just appear have better edge definition and better detail in the deep shadows.

[Jac@stafford.net]

If your goal is to print your scan, you must consider how the printer re-sizes your image to make it possible. Think using a factor of 360 or 300 ppi. There is no way we can get better so far. It is all about print capability.

[faberryman]

But you might find the following contrary data on filmscanner.nfo itself... at http://www.filmscanner.info/en/Aufloesung.html

The answer is clearly yes. During the scan of a film, there are interferences as aliasing, other interferences, etc. They are all related to the intermediate effects of the used resolution of the scanner and the actual image resolution. The higher the scan resolution, the lower are those interferences. Thus, it does make sense to scan an original with 4000 dpi that has a resolution of only 2500 dpi by itself.
This is not contrary data, but consistent data. Filmscanner.nfo has tested, and advocates, scanning at a higher rate than the actual resolution of the scanner, at least with the two software packages - EsponScan and Silverfast. And amazingly, it is consistent with Pere's chart, and with your own testing. Hardly seems controversial, much less misleading.

[Pere Casals]

For sharpening the Raw file, I usually use Un-Sharp Mask (low radius high amount). Are there any other methods which may do a better job.

To me Ps unsharp masking works perfect.

Here there is a nice tutorial :

https://web.archive.org/web/20171207231000/https://www.cambridgeincolour.com/tutorials/image-sharpening.htm

https://www.cambridgeincolour.com/tutorials/image-sharpening.htm

[Ted Baker]

For sharpening the Raw file, I usually use Un-Sharp Mask (low radius high amount). Are there any other methods which may do a better job.

I sometimes use a variaton of that by I using some of the controls in rawtherapee to limit that capture sharping to edges only I set the threshold such that it only sharpens edge bigger than about 2pixels. You can do something similar if you know how in PS I don't unfortunately.

I actually prefer to use the RL deconvolution algorithm it seems to give a better result with out enhancing grain aliasing. I am still not sure what I like best. But I most certainly get a better result than I did when I scanned at lower resolution and did not understand scanning. I also don't like to waste time either!

[Alan Klein]

Anyone have the actual resolution for the Epson V600?

[Pere Casals]

Here there is another personal test with the epson v700

http://www.sebastian-schlueter.com/blog/2017/3/10/optimum-resolution-sharpening-settings-for-epson-scanners

He says:

scan resolution Vertical Horizontal
1800spi 1825 2048
2400spi 2299 2580
3200spi 2580 2896
4800spi 2299 2896
6400spi 2299 2896

[IanBarber]

Pier, the chart you posted for the V850 earlier on in the thread, would those figures differ much for the V800

[Pere Casals]

Pier, the chart you posted for the V850 earlier on in the thread, would those figures differ much for the V800

IIRC Epson said that the V8x0 series has optical improvements vs the older V7x0, but I'm skeptical that this could end in a practical difference.

IMHO it's the same, with the great improvement in convenience because the LED light source.

I'd say that the main difference is in the man that process the image. I try to learn from people that has magic in their fingers and make a sound edition with few clicks, being not much intrussive aganist the original (...or changing it a lot) but having an amazing aesthetical criterion to know what to touch.

[Pere Casals]

Anyone have the actual resolution for the Epson V600?

Here (http://www.filmscanner.info/en/EpsonPerfectionV600Photo.html) it says that it performs 1560 dpi effective if scanning at least at nominal 3200 dpi

[minh0204]

Anyone have the actual resolution for the Epson V600?

Just FYI the V600 does not officially support 4x5 or larger film sizes, you have to scan in strips and merge later.

[IanBarber]

Here there is another personal test with the epson v700

http://www.sebastian-schlueter.com/blog/2017/3/10/optimum-resolution-sharpening-settings-for-epson-scanners


Interesting article which got me thinking. Slightly off topic but it refers to the height the film is above the scanner glass to obtain optical focus on the V series.

I never use the supplied holders, I use the wet mount holder but in a dry environment for MF 35mm and 4x5

179197

I place the film onto of the wet holder and then place a piece of etched glass onto of the film which really keeps it flat. For ease I just made a hinge to make it easier to lift the top piece of glass.

Sometimes I do place a black mask on top to stop light scattering but the times I haven't I have not seen any scattering.

Thinking Behind This Approach
The wet holder is of a fixed height, with no available adjustments so I figured that it must have been designed and manufactured with the optical distance in mind.

Whats your thoughts on my thoughts... ?

[Pere Casals]

height

Time ago I saved a these graphs of a gmikol user at photrio:

179199

179200

This shows optical performance depending on holder height.

The V850 holders can adjust height and have ANR glass.

Well... IMHO if film is flat and not curled the thing it's easy, if not then a way to flatten the film is needed

[Peter De Smidt]

As was touched on earlier, getting the best performance from your scanner is important, but so minimizing quality losses at later stages. Re-sizing, sharpening, and compression types matter, as does avoiding having to rotate the photo in Photoshop.

[Pere Casals]

this is... IMHO

[Ted Baker]

Interesting article which got me thinking. Slightly off topic but it refers to the height the film is above the scanner glass to obtain optical focus on the V series.

I never use the supplied holders, I use the wet mount holder but in a dry environment for MF 35mm and 4x5

179197

I place the film onto of the wet holder and then place a piece of etched glass onto of the film which really keeps it flat. For ease I just made a hinge to make it easier to lift the top piece of glass.

Sometimes I do place a black mask on top to stop light scattering but the times I haven't I have not seen any scattering.

Thinking Behind This Approach
The wet holder is of a fixed height, with no available adjustments so I figured that it must have been designed and manufactured with the optical distance in mind.

Whats your thoughts on my thoughts... ?

It was my understanding the setting for the whole scanning area is supposed to be used?, this switches to a similar lens that is used in the epson 4990. It is a wider angle lens that is focused on the glass or nearly so and has a greater depth of field.

[IanBarber]

It was my understanding the setting for the whole scanning area is supposed to used, this switches to a similar lens that is used to in the epson 4990. It is a wider angle lens that is focused on the glass or nearly so and has a greater depth of field.

In the scanning software, I choose Film with Holder in order for the high resolution lens to be used and place the film in the middle of the glass.
The wet mount holder suspends the film 3.3mm above the scanner glass by default.

I might be barking up the wrong tree here doing it this way of course.

[Steven Ruttenberg]

All I know is that when I use my V850 at 2400 the scans don't look all that great, when I use it at 3900 they look great.

I am guessing, but it sounds like the lenses of these scanners are not as capable as the imaging sensor. Which is why beyond a certain number of pixels, no more benefit is gained since the lens can provide no more detail to be taken advantage of.

Hence, why I am looking at building a dslr rig to scan my negatives. I will most like use a Canon 5DS and a 100mm or 185mm macro lens at 1:1 to scan in.

[Bernice Loui]

With all the known difficulty of scanning-digitizing sheet film then putting the data into software for post processing then the printing process, why not keep the entire post developed film process as optical printing in a wet darkroom?

DSLR scanner is quality limited to the ability of the DSLR and it's optical system. Add to this the same digital post print process system.

If one were to go after the very best digital image file, why not use a medium format digital camera system. This skips the scanning process and in theory reduce the image digitizing process.

Have yet to see a digital B&W print that appeals to me.


Bernice

[Pere Casals]

IMHO traditional wet printing is amazing, also it adds the value of autenticity. Personally I'm in that way.

...but if one uses film, hybrid process has also the strong advantage of easy image manipulation and internet sharing.

[Alan Klein]

I was in Venice Florida a couple of years ago visiting Clyde Butcher's gallery. https://clydebutcher.com/galleries/

Clyde's long history of LF wet printing four and five foot wide BW prints of the Florida everglades and elsewhere has been well established. The prints are georgious. However, he does have digital versions of the same photos at about 40% cheaper. I'm sure he's using the best scanners and best printers.

Due to health, he switched to taking digital photos (A7III or A&Riii)a few years ago and those are available in digital prints as well. I've got to tell you, he's got it down pretty good. The digitals are superlative and I would guess that 98% of the viewers couldn't tell the difference.

However, I am not suggestion that anyone should do something they'd rather not do. After all, I still shoot film (MF if not LF) (but don't wet print it), and people look at me like I'm crazy for using film at all after I convince them that film is even available any longer.

[Steven Ruttenberg]

With all the known difficulty of scanning-digitizing sheet film then putting the data into software for post processing then the printing process, why not keep the entire post developed film process as optical printing in a wet darkroom?

DSLR scanner is quality limited to the ability of the DSLR and it's optical system. Add to this the same digital post print process system.

If one were to go after the very best digital image file, why not use a medium format digital camera system. This skips the scanning process and in theory reduce the image digitizing process.

Have yet to see a digital B&W print that appeals to me.


Bernice

If I had 30,000 dollars or in some cases upwards of 100,000 dollars for a medium format digital camera and lenses, I might just give it a whirl.

[Ted Baker]

In answering the OP original question it is worth considering how these devices actually work.

In the case of the V700/750/800/850 scanners there are two rows of sensors. Each row of sensors has an equivalent spacing of 1/3200 of an inch, the alternate row of sensor is 1/2 pixel offset. This allows the resolution to be increased in the X dimension.

So the "resolution" in the X dimension never really changes, it is never less than 3200dpi. I have not been able to figure out when the alternate row of sensor is activated. i.e. is it used all the time and the resolution is 6400dpi in the X dimension or only when a resolution higher than 3200dpi is chosen. If this device was being designed today, you would assume that the offset row was always being used. In any case it is never less than 3200dpi.

In the Y dimension it is a bit different, the sample size it is still equivalent to 1/3200 of an inch, except the stepper motor controls the spacing between each sample.

As an extreme example if you scan at 300 dpi, you will have one sample 1/3200 of inch then a space of nearly 10/3200 and then another sample 1/3200 and so on.

When you scan at 6400 dpi then the stepper motor makes the sensors overlap.

Now most of this is just reverse engineering and there are probably a few inaccuracies, but it may help to understand why scanning at the highest resolution and re sampling gives the best result. Conversely driving the stepper motor so that the spacing is greater also gives good results.

[faberryman]

If I had 30,000 dollars or in some cases upwards of 100,000 dollars for a medium format digital camera and lenses, I might just give it a whirl.
The Fujifilm GFX and Hasselblad X1D are both substantially less expensive.

[Bernice Loui]

There are Leaf, Phase One and other MF digital backs and cameras on the used market, no short supply of used MF digital gear. The cost is not excessive by any means.

The root question is what is the image maker's goal?

Film or Digital, they are mere tools and means to achieve an intended goal. Nothing more, nothing less. IMO, both have good and bad, neither is ideal for all image making needs.


Bernice



The Fujifilm GFX and Hasselblad X1D are both substantially less expensive.

[Pere Casals]

The digitals are superlative and I would guess that 98% of the viewers couldn't tell the difference.

Well, I would not tell a difference between the Mona Lisa and a medium quality falsification of it, even if using a loupe, nor the 99.9% (at least) of the viewers.

I agree that digital prints are superlative, but a collector usually doesn't buy that, so that market is addressed mostly to ornamental applications.

A sound print made with pure optical means has a value that's beyond technical quality. The artist's hand can be seen in each copy, and there is the authenticity of the light's footprint.

Of course (IMHO) there is not more or less art in an optical print than in a digital print, but a sound traditional optical print is a unique hancrafted object that required uncommon skills.