Size of scanner CCD, in number of elements

[coisasdavida]

Some scanner's manufacturers publish the size of their scanners CCDs, so we know for sure how much resolution we can achieve with these units, for example, I have a PFU DL-2400p with 10500 elements in its CCD. Published are also these ones: the Cezanne has 8000 elements, the Pakon has 2000 elements and the Minolta 5400 has 5340 elements.

The most used linear CCD was for some time a 10200 elements one. Would anyone know if that is what's inside an Artixscan F1 or an Epson V700?

[Pere Casals]

Some scanner's manufacturers publish the size of their scanners CCDs, so we know for sure how much resolution we can achieve with these units, for example, I have a PFU DL-2400p with 10500 elements in its CCD. Published are also these ones: the Cezanne has 8000 elements, the Pakon has 2000 elements and the Minolta 5400 has 5340 elements.

The most used linear CCD was for some time a 10200 elements one. Would anyone know if that is what's inside an Artixscan F1 or an Epson V700?

V700 has 40800 pix (for each R,G,B colors) in the sensor.

But an interesting parameter is actual pixels per inch (dpi), and this depends on the lens than projects the film image on the sensor. The V700 has two different lenses (no zoom), the one that covers all the bed allows 4800 dpi, and the high res lens covering 5.9" width allows 4800 dpi.

The Cezane has only 8000 pix, but it has a zoom optics that allows to spread those 8000 pix in an small negative or in the entire bed.

The Hasselblad X1 and X5 also sports 8k pix only, but it also has a fine lens inside (Linos/Rodenstock) to zoom.

Then we also have the effective resolving power, the V700 allows worth 2300 effective dpi in one axis and some 2800 in the other when selecting (automatic) the lens that covers 5.9" width.

You can review this web site to get information about those intrinsics: https://www.filmscanner.info/en/EpsonPerfectionV800Photo.html

[coisasdavida]

V700 has 40800 pix

Sorry, but is nothing on the link to support this information. I used a search tool and nothing.
How did you get this info?

[Pere Casals]

Sorry, but is nothing on the link to support this information. I used a search tool and nothing.
How did you get this info?

see page 155 of the manual: https://files.support.epson.com/pdf/prv7ph/prv7phug.pdf

section: scanner specifications

the 56160pix is in the movement axis, so determined by the stepper motor, 40800pix is what the linear sensor takes with one shot in the transverse axis.

[coisasdavida]

40800pix is what the linear sensor takes with one shot

Sorry, those are not the specs of the sensor, if they were it would read way more clear than that.
That is a similar way that Microtek describes their F1/M1 scanner, relating to the choice of dpi in the interface.
This is not a positive information about the actual sensor.

I'm almost sure the V700 has a NEC sensor of 10,200 elements and the rest is only software.

[Pere Casals]

Sorry, those are not the specs of the sensor, if they were it would read way more clear than that.
That is a similar way that Microtek describes their F1/M1 scanner, relating to the choice of dpi in the interface.
This is not a positive information about the actual sensor.

I'm almost sure the V700 has a NEC sensor of 10,200 elements and the rest is only software.

That's wrong, if it was the case it would not be possible that the V700 would scan a 8x10 sheet with 4800dpi (hardware).

This scan was made with a V750, that has the same sensor than the V700:

https://www.flickr.com/photos/125592977@N05/32535835184/in/dateposted-public/

When the V700 scans at 9600dpi then it's interpolation. But at 5600dpi (5.9" scan width) or 4800dpi (full bed wide scan) there is absolutely no interpolation, it's the native hardware resolution, I'm completely sure, if you read well the manual you will find that.

Another thing it's the effective resolving power... that involves lens performance.

[coisasdavida]

The only mention of a 40800 sensor in the web is about a rgb sensor with 3x 13600 pixel lines.
https://www.google.com.br/search?q="40800"+pixel+linear+sensor

[Pere Casals]

The only mention of a 40800 sensor in the web is about a rgb sensor with 3x 13600 pixel lines.
https://www.google.com.br/search?q="40800"+pixel+linear+sensor

There are other sensors...

You also have, for example TCD2964BFG, Toshiba : 21360 elements × 6 line, this is 128,160 photosites, or 42720 R-G-B sites

https://www.glynshop.com/erp/owweb/Daten/Datenblaetter/Toshiba/TCD2964BFG.pdf

You may know that both in Area Scan sensors/cameras or linear sensors not always all pixels are used, so you have to search parts with higher than 40800 rgb pix to find matching possibilities.

There are several reasons for a device may not be using all available pixels, but a crop, we can discuss on that.

[Peter De Smidt]

When the V700 scans at 9600dpi then it's interpolation. But at 5600dpi (5.9" scan width) or 4800dpi (full bed wide scan) there is absolutely no interpolation, it's the native hardware resolution, I'm completely sure, if you read well the manual you will find that.



If so, then that should be easy to test with a high resolution target. I doubt very much that the system gives 4800 true dpi. The reply might be made that that's the sensor is that large, but the lens is poor. Why use a higher resolution sensor, which would no doubt be more expensive than a lower resolution part, when the system is limited to about half that because of the lens? That seems...unlikely. A more likely explanation is that it's marketing BS, just as Epson's dmax specs are.

[Pere Casals]

If so, then that should be easy to test with a high resolution target. I doubt very much that the system gives 4800 true dpi. The reply might be made that that's the sensor is that large, but the lens is poor. Why use a higher resolution sensor, which would no doubt be more expensive than a lower resolution part, when the system is limited to about half that because of the lens? That seems...unlikely. A more likely explanation is that it's marketing BS, just as Epson's dmax specs are.

The V700 delivers 2300-2400 effective dpi in one of the axis, and 2600 to 2800 in the other, from the 6400 hardware DPI (if scanning under 5.9" width). This is not a lot for roll film, but's really good for sheets. Owning a V850 and a Plustek 8xxx (or 120) covers most needs of a multiformat film shooter. An optimal tandem.

Of course, the V700 is a cheap scanner, a semi-pro one, not a pro device, so lenses could be better.

The lenses (there are two, only one is used at a time) are made of plastic, anyway it has a sound design, being a consumer device, and this allows great performance, in special for 8x10 it's hard to beat. You know, the Cezanne delivers 1000dpi hardware for 8x10, 800 effective if not scanning strips and stitching in PS... while the V700 delivers 2000dpi effective with the low res lens for 8x10.

For 4x5" a V700 outresolves a Hasselblad X5 that delivers 1800 effective DPI in that format, while the cheap V700 delivers at least 2400 effective (when scanning at 4800 with the high res lens). Of course for high velvia densities the X5 is way better, but when using multiexposure in the V700/...850 we have really good results with velvia if no extreme densities are important.

It is also true that with the Epson the image comes less digitally optimized than with a pro scanner, so there is some work to do in Ps.

[coisasdavida]

A more likely explanation is that it's marketing BS

Very likely.


Of course, the V700 is a cheap scanner, a semi-pro one, not a pro device, so lenses could be better.

Even plastic, they are probably very good and enough for the sensor and so are the mirrors. If they were bad, there would variation among the scanners, but all seems to suffer from the same limitations.

[Steven Ruttenberg]

I have been using a V850 for about a year ar and I can say with 100% clarity and hat at 2400 dpi settings the image sucks, period. At 3200 it is better and at 4200 it is excellent. One thing to note is that at 4200 dpi setting the resolution and is no better than my Canon 5DMKIII. Yes, files are large, but then so is 4x5 film.

Regardless of so called testing, so called effective dpi, etc. I know from a practical standpoint what the scanner can produce. Yes, I know many say I don't know what I am talked ng about, will absolutely disagree with me and post countless links to the so called testing, whatever. I am an end user and I know what this scanner does based on actual use. I also chose 4200 dpi not only since it matches my Canon 5DMKIII, it gives 1.79gb size color files and 600mb or so bw files it is a multiple of 300, and ie, 300, 600, 900, 1200, 1500, 1800, 2100, etc which Canon pringers like. Epzons I believe are multiples of 270.

Anyway, the V850 produces images extremely well excellent resolution and detail. If anything, its only shortcoming is with zone1-zone 0. I can see faint details i. These zones on light table I can make out some detIl, but scanner has hard time with it. Yes, I am sure a drum scanner will provide a far superior scan, but you know what? Who cares! As long as the image you scan privixes you with the results you wanted then it is the right scanner for you. If you got 25 grand juet burning a hole in your pocket, a Flextight or drum scnner is well beyond mosts dreams. And not really necessary.

[Pere Casals]

Even plastic, they are probably very good and enough for the sensor and so are the mirrors. If they were bad, there would variation among the scanners, but all seems to suffer from the same limitations.

for industrial applications, I've been using the Linos (Rodenstock) lens type that's inside the Hasselblad X1 (and X5), it's a good choice for linear cameras. That lens alone is more expensive than a full V700...

X1 moves up and down the lens to provide different magnifications, and it's a real challenge to make a lens that works very well in that range of magnifications.

The V700 (series) has two cheap lenses working each at a fixed magnification, so it's easier to make a lens optimized for that job. With the high pixel count, way outresolving the lens, they extract all what the lens can do avoiding variable magnification.

With a high pixel count we have smaller pixels, so we may want a longer exposure time... Well, Vs have drawbacks, but also are exellent performers for most LF photographers, for a significative improvement in sheets we have to go to drums, or to a hassy if we consider velvia extreme densities.




One thing to note is that at 4200 dpi setting the resolution and is no better than my Canon 5DMKIII.

Steven, sorry, but I completely disagree. The MkIII sports hardware 21 Mpix, you are lucky if you obtain 18 MPix true "optical" after lens degradation and discretization effects.

I don't use a 5D but I know very well what a D610 and a D810 do... Even with a V700 a 4x5" blows miles away what a MkIII resolves.

An scan from 4x5" in a V700 has a resolving power worth of 4x5x2300x2300, this is 105MPix optical performance, so you enlarge x2 larger or x4 more print surface for the same quality.

If you obtain similar performance than with the MkIII then you should review your hybrid processing. The DSLRs have internal optimization algorithms to make images look sharp, if you want to compare "perceived" sharpness then you should apply similar processing to the scanned image in Ps or other dedicated size blowing algorithms. What's about resolving power there is no doubt, we may debate if it's 105Mpix, or 90 or 143, but we compare that to 18 or less.

And always, with a drum we can take from 200 to 400 MPix (optical) from a 4x5, if one needs that...

[coisasdavida]

Epson V700 CCD board rear view

https://farm2.staticflickr.com/1950/44737450445_91af9203de_o.jpg

[Steven Ruttenberg]

So the 5DMKIII has an effective pixel of 5760 x 3840. The sensor size is approximate 1.417 in x 0.944 in (36 mm x 24 mm) This then gives 4064 pixels per inch effective on the sensor both long and short side. I use 4200 due to Canon printing requirements preferring things in multiples of 300. So, If I were to scan the 4x5 with my 5DMKIII and then scan at my settings I have the same sensor resolution. I agree that a 4x5 will blow away the 5DMKIII. I was referring to what the sensor records digitally. Now the size of the pixel on a V850 vs 5DMKIII are different, pitch is different, etc however, I was simply referring to scanning at the same pixel density if you will of the 5DMKIII. And by doing this I get a large file that is quite nice in detail compared to scanning at lower settings. I started out at 2400dpi the recommended settings from every site on the planet and then compared to 3200 and 4200 and I can say that the 3200 and 4200 are way better than at 2400.

As for internal algorithms and sharpening that is if you save as jpg. I save the raw data only and there is no additional processing to that data as I have that turned off in the camera (I need to check this though again). When I bring the raw file into PS or LR, I do not use sharpening. I apply sharpening as last step with things like high pass, etc.

Science, pixel peeping, number crunching, etc aside, I am just saying that I get the results I want and would not have a problem making prints as large as I want. And I have made some large panoramas from my 4x5 scans and printing up to 32x40 presents no challenges and I could go larger still if I wanted to, but paper is way to expensive and I don't have a wall large enough. Something else to remember to is that just because we have large numbers of pixels does not mean we get super detailed super sharp images. There are examples out there where a sensor with much lower resolution yields a much sharper, more detailed image than a sensor with more pixels of the same image. It is all in how those pixels are used and the type of pixels used and there arrangement.

Not wanting another scanner debate, we have had enough of those and they get quite spirited. I just know what I get and it works. And from time to time, it is fun to have a brief discussion of the topic.

Thanks for your input Pere. I am going to check out my resolving power of my 5DMKIII here soon and will post those up alongside the results from my scanner. If I find something from my scanner tests that are contrary to what I am seeing, I will certainly point that out as well.

[Steven Ruttenberg]

https://farm2.staticflickr.com/1950/44737450445_91af9203de_o.jpg

This is?

[Steven Ruttenberg]

The size of the pixel pitch on the Canon 5DMKIII is 6.22um (2.44882E-4 in) or 4096 ppi. For the Epson, it is 1.5625E-4 in or 6400 ppi (3.96875um, approximately 4um) effective ppi. This by conventional wisdom would say that the Epson should be better in resolution than the Canon so long as the capability of the pixel and supporting hardware/software is capable to do so. If, I knew what the lens optics were, I could determine its optical resolution, capability and inherent error (but ray tracing is a pain in the butt). The Canon lenses are such that the sensors have finally caught up to the lens capability and in some instances surpassed and so Canon has now started coming out with newer MKII lenses and in some cases a MKIII (1.4 extender). As you pointed out, it is the total combination in the system that will yield a final true capability. But we should not be hamstrung on only using 1/2 or less of the mechanical capability of a system.

Sometimes testing by pixel peepers do not tell the whole story and are not stringent. In the end, if you get the image you want in detail, etc from a scan of your negative then that is all that counts.

[Pere Casals]

I am going to check out my resolving power of my 5DMKIII

Steven, it is very useful you do those tests on your own at least once in a lifetime because it is a pretty nice way to learn.

With a prime lens you will get some 16MPix with the 5D Mk3. This is DXO rating that in general I find pretty accurate:

183910


Then if do the same with a TMX 45 sheet shot with a good lens (Say Sironar N 150) at f/16 you will obtain around 450Mp seen with a microscope, and around 140 in the V700 scan, at least 100 I guess.

There are many ways to degradate a LF image in the hybrid processing. For example size reduction it's something critical...

In the path from negative to the print, as an exercise you may try to optimize every step to conserve the native IQ as much as possible.

[coisasdavida]

This is?

Epson V700 CCD board rear view, I was unable to identify anything as to give a clue of which sensor the V700 actually uses.

[Steven Ruttenberg]

Steven, it is very useful you do those tests on your own at least once in a lifetime because it is a pretty nice way to learn.

With a prime lens you will get some 16MPix with the 5D Mk3. This is DXO rating that in general I find pretty accurate:

183910


Then if do the same with a TMX 45 sheet shot with a good lens (Say Sironar N 150) at f/16 you will obtain around 450Mp seen with a microscope, and around 140 in the V700 scan, at least 100 I guess.

There are many ways to degradate a LF image in the hybrid processing. For example size reduction it's something critical...

In the path from negative to the print, as an exercise you may try to optimize every step to conserve the native IQ as much as possible.

I agree.

[Steven Ruttenberg]

Epson V700 CCD board rear view, I was unable to identify anything as to give a clue of which sensor the V700 actually uses.


Ahh, ok

[Ted Baker]

Epson V700 CCD board rear view, I was unable to identify anything as to give a clue of which sensor the V700 actually uses.

The sensor on the Epson are supposed to use pixel shift, with two rows of sensors ofset half a pixel wide.

[Steven Ruttenberg]

That is interesting. I had not heard that. Wonder what happens on the V850.

[Ted Baker]

That is interesting. I had not heard that. Wonder what happens on the V850.
I think they all have it, that why it's a 6 line ccd instead of 3. Also pixel shift does not quite work as well the marketing would have it. This seems to tally with performance.

[coisasdavida]

The sensor on the Epson are supposed to use pixel shift, with two rows of sensors ofset half a pixel wide.

A lot o people replicate this idea, but I have not found an actual source/confirmation of this info.
Back in 2008, a user of this forum even explained the math behind the 4800 dpi, 2 lines of 20400 pixels, offset half a pixel.

[coisasdavida]

This seems to tally with performance.

So it is a 2400/3200 dpi scanner, depending on the lens used.

[Pere Casals]

The sensor on the Epson are supposed to use pixel shift, with two rows of sensors ofset half a pixel wide.

The Epson V700-850 may include a sensor similar to the TCD2964BFG

In page 3 of the TCD2964BFG datasheet https://www.glynshop.com/erp/owweb/Daten/Datenblaetter/Toshiba/TCD2964BFG.pdf shows the pixel alignment

183922

It looks that the second row has shifted half a pixel, but I guess that the effective separation is close to a full pixel, because the sensitive area of the photosite is smaller than the theoric surface in the arrangement:

183923

In that image we see an area sensor (not linear) including microlenses on the photosites to use a larger effective surface.

Well, it's clear that in that kind of arrangement the information from a row 4um row it is completed after the sensor has advanced 64+64+6 = 134um.

So the a row of the film it is not scanned in one shot, it will take dozens of steps to allow to pass all sensor rows over the film row, so some dozens of film rows are being scanned in course at the same time.

[Pere Casals]

So it is a 2400/3200 dpi scanner, depending on the lens used.

No, it's a true 6400 dpi hardware scanner for a 5.9" wide scan, and true 4800 dpi hardware for a full bed scan. As manufacturer specifies.

There are 6 rows in the sensor: 2 green, 2 red and 2 blue, this is two RGB sets: a half of the rows are shifted to take the (inter-pixel) space that was not scanned by the "first" set of rows. ("first" because in fact both sets are interlaced, as the image shows in previous post)

Because of the opto-mechanical design (mirrors, lenses, vibrations) this delivers 2000 to 2800 dpi optical neat perfomance, depending on the scanning settings and the axis.

All scanners deliver less optical performance that hardware dpi... In the case of the Epson V7/8 there is a lower than usual yield because the targeted manufacturing cost. They distributed the manufacturing budget in the way they could obtain best performance at the intended cost. This is mass production mentality.

A winner mentality, because today they own a market niche.

[Ted Baker]

A lot o people replicate this idea, but I have not found an actual source/confirmation of this info.
Back in 2008, a user of this forum even explained the math behind the 4800 dpi, 2 lines of 20400 pixels, offset half a pixel.

It in the marketing documentation, they refer to it as ... Epson Double CCD technology

http://content.epson.it/maco/technology/scanners/ccd.htm

But there is no information on how it's implemented, when my scanner breaks I will pull it apart and find out for sure. I would also like to understand in more technical details on the scanner. For example what is the spectral sensitivity of the sensor etc.



So it is a 2400/3200 dpi scanner, depending on the lens used.

More or less but with added modest improvement that pixel shifting allows.


The Epson V700-850 may include a sensor similar to the TCD2964BFG

...

In that image we see an area sensor (not linear) including microlenses on the photosites to use a larger effective surface.

Well, it's clear that in that kind of arrangement the information from a row 4um row it is completed after the sensor has advanced 64+64+6 = 134um.

So the a row of the film it is not scanned in one shot, it will take dozens of steps to allow to pass all sensor rows over the film row, so some dozens of film rows are being scanned in course at the same time.

This is pixel shift, what not clear is how the algorithm in this particular scanner is working, for example how does it work when the stepper motor is driven at much greater steps. When you scan at say 1200dpi what happens? How is X pixel data merged with Y pixel data?

[Pere Casals]

More or less but with added modest improvement that pixel shifting allows.

Ted, I disagree. It is a 200% real improvement. Of course the lens has to match that performance to get the real improvement in the result.

If you review how linear sensors have evolutioned you will find that the pixel in the middle (of the second set) effectively takes just the area that was not covered by the first set.

It can be said that shift is half a pixel... but in fact is a full photosite shift, in terms of sensitive area, and this physically multiplies hardware dpi density by 2.

Look, the modesty of the improvement comes from optical limitations of the rest of the system. In fact a 6 rows arrangement has clear advantages over having the same pixel count in 3 rows, and is the preferred arrangement, single problem is that 6 rows instead 3 is a more expensive part, I think that this is well known.

[Ted Baker]

Ted, I disagree. It is a 200% real improvement.

I don't know for sure, I am just stating that is supposed to use pixel shift technologies, how it actually works in particular model is a mystery like many thinks about these devices... I am just attempting to reverse engineer, like everyone else... Certainly for pixel shift to work its best you also need to do it the Y axis at the same time...

We can look at modern DSLR like some of the Pentax and hasselblad models that also use this same pixel shift approach, also use a little "stepper motor" to move the sensor 4 times to take a shot. 2 in the X axis 2 in Y axis. My understanding is they don't get a 200% or 400% improvement...

However the pixel shift may be huge improvement, but the quality of the lens lets the system down, or indeed the pixel shift is not quite getting the improvement that you might imagine it is capable of. Without changing the lens its hard to say for sure.

Has anyone ever attempted to change the lens?

[Pere Casals]

This is pixel shift, what not clear is how the algorithm in this particular scanner is working, for example how does it work when the stepper motor is driven at much greater steps. When you scan at say 1200dpi what happens? How is X pixel data merged with Y pixel data?

This is easy.

> The controller board has in RAM a circular buffer of rows with matrix of pixels for all the image of the rows in course, say 40800(dpi) x 32(rows in course) x 4 (R+G+B+IR) x 2 (16bit ADC), this is 12Mb of RAM.

> After each stepper step the ADC receives all 6 rows, but it only converts the section that is scaned (it can be 6cm of the bed row), so the clocking can be faster for the not converted/stored pixels because we need not to wait for the ADC conversión happens for those pixels not wanted by user.

> The uC fill the RAM matrix with the photosites of the optical matric that have been scanning that time, for each step a row is completed and sent, binned (for lower resolution case), removed from the circular buffer sent to PC to next.

As with Area Sensors the capture is made al full dpi, and it is digitally binned later if the user wants less resolution, this makes the transmission to PC (and scanning process) faster.

I guess that the V700 series has only 2 advance possibilities, 6400dpi or 9600dpi, the rest comes from binning.

[Pere Casals]

I am just stating that is supposed to use pixel shift technologies, Certainly for pixel shift to work its best you also need to do it the Y axis at the same time...

This is not pixel shift, pixel shift technology is moving the sensor half a pixel (Hor/Vert) a couple of times to get several shots and combining that.

The Double CCD technology (6 row) is explained in the link you provided: http://content.epson.it/maco/technology/scanners/ccd.htm

In that case you remove the half of the pixels of the row (interlaced) of a color (green, red or blue) and you displace that a photosize distance back, so the photosite can be larger to effectively take all pixel area, because before that there was a blind zone in the inter-pixel spacing.

So we have 1) Higher sensitivity, 2) less electronic noise at same exposure, and 3) less discretization color noise.

See here how anyway a photosite has less sensitivity in its boundary, so there is little sensitivity-area overlap between both rows:

183924

The "Double CCD" is a technical improvement over placing all pixels in a single row, in two rows pixel count is the same but photosites cover all, because blind areas overlap sensitive areas.



However the pixel shift may be huge improvement, but the quality of the lens lets the system down, or indeed the pixel shift is not quite getting the improvement that you might imagine it is capable of. Without changing the lens its hard to say for sure.

The efficiency of "6 row" linear sensors are very well documented. Optically it's the same having all the pixels in a single row or in two, in a scanning system. But the 2 rows arrangement has electronic performance advantage. Top notch linear sensors are of that kind.

Why do you think they arrange the pixels in two rows instead in one? that's more expensive to do!

Sadly this technology is less suitable for industrial linear cameras, because the inspected object may have arbitrary translational speeds under camera.

[Ted Baker]

This is easy.

I might be easy to propose how it might be done, or even how it should be done... That's not the same as how it is done in this particular model/(s)

For example is each colour sensor (in the row) moved over the same spot or an approach similar to bayer matrix approach use to work out the colour ?...

and why use binning or a simple average when better techniques might be used? I think the OP like many (and myself) is trying to get real details on the innards and workings of these machines.


This is not pixel shift, pixel shift technology is moving the sensor half a pixel (Hor/Vert) a couple of times to get several shots and combining that.

The Double CCD technology (6 row) is explained in the link you provided: http://content.epson.it/maco/technology/scanners/ccd.htm

In that case you remove the half of the pixels of the row (interlaced) of a color (green, red or blue) and you displace that a photosize distance back, so the photosite can be larger to effectively take all pixel area, because before that there was a blind zone in the inter-pixel spacing.


The description of each each sounds exactly the same just implemented slightly differently... Of course manufactures of pixel shift cameras will want to claim their solution is unique, but there ain't much new under the sun.... So shall we call it Double CCD technology then? ;)

[Pere Casals]

So shall we call it Double CCD technology then? ;)

This is commercial jergon in the link you provided: http://content.epson.it/maco/technology/scanners/ccd.htm . To me it should be monolithic 6 line CCD color image sensor...


For example is each colour sensor (in the row) moved over the same spot or an approach similar to bayer matrix approach use to work out the colour ?...


Yes...

Area sensors (A7IIR) use the bayer arrangement for interpolation, each pixel has a single color, the other two colors are interpolated from neighbour pixels. The pixel shift feature allows not interpolating. But we need a tripod and a static subject.

In Scanners each spot on the film is exposed three times by different sensor pixels in different rows, so R, G and B are individually measured with no interpolation...

in fact the scanner performs a shift of the sensor with the stepper to place on each film spot R,G and B pixels, like the A7IIR with the pixelshift feature. Difference is that the A72R does it with an area sensor.



I think the OP like many (and myself) is trying to get real details on the innards and workings of these machines.

My view is that V700 etc have true 40800 RGB Pix. The 6 way arrangement is irrelevant for the resolving power, they modified the 3 way arrangement for the speed vs noise.

The 6 rows arrangement (vs 3 rows) is not intended to increase DPI, it's not about resolving power. The 6 rows way is exclussively intended to have larger photosites and lowering discretization noise for the same effective DPI, a great idea and a great Patent,

[Ted Baker]

the A7IIR with the pixelshift feature. Difference is that the A72R does it

Its my understanding that the Sony does not offset the pixels unlike the approach used by the Pentax and others. Perhaps that's the source of confusion. The Sony moves a full width but the Pentax, hasselblad and others move half

However I think you actually missed my point about what I was saying about how exactly does the Epson work.

Is each photosite visited by each respective sensor in the respective row and if so what evidence do you have to support that?

In summary there are some fine details about the sensor that we don't know and are not available through a software API or raw scanning form, where we might have some give us more insight or opportunity to control.

Otherwise it's just the scanner does what it does.

I am not particularly interested in any metric which like most statistics easy to misuse, just interested in a better understanding and perhaps a better method of use.

For example new software or a new lens.



The 6 rows arrangement (vs 3 rows) is not intended to increase DPI, it's not about resolving power. The 6 rows way is exclussively intended to have larger photosites and lowering discretization noise for the same effective DPI, a great idea and a great Patent,

You missed your calling pere you should have been in marketing ;)

[Steven Ruttenberg]

In the end, unless we have an Epson Engineer willing to disclose the secrets of the scanner/s to the world, which I highly doubt will ever happen, we will never know exactly how these things are designed and set up. One thing we can be sure of, is they did a cost/benefit analysis on the design and made compromises to give something that would be a good alternative to a 25,000 dollar drum scanner. Most likely the sensor is no slouch and used in items from low grade consumer, to pro whatever. It is how it is used that makes the difference. I can have a Ferrari engine in a Ferrari and be happy, but I can have a Ferrari engine in a chevy vega and be underwhelmed. Both work, just one gets you there in much more style with more benefits.

From all the images I have seen and the back and forth on Epson V700/850, etc vs drum scanners, the Epsons more than hold their own. They may loose out in some areas such as sharpness or shadow detail. It is the implementation of the scanning technology that is the difference and the skill of the person doing the scanning.

Anyway, for a 1000 bucks for something that rivals (not surpasses) a drum scanner, I think I'm good and no one, but me or someone with access to my original negative and scan files will ever know the difference. :)

[Pere Casals]

Is each photosite visited by each respective sensor in the respective row and if so what evidence do you have to support that?


Ted, there is a single sensor, a monolithic one, sporting 6 rows instead 3. But the two paired rows (Green for example) are clocked like if it was a single row.

183926

Between Pix 1 and 3 there is a blind zone and a sorrounding very low sensitive zone. The pixel 2 just reads that area in the next step, so the firmware has to place the reading of the pixel 2 in a different row of the circular RAM buffer than the 1,3 that was read in the same exposure.



Otherwise it's just the scanner does what it does.

There is no room (my undertanding) for things to happen in a different way as described in the above point.



or a new lens.

I destroyed an Epson while trying to hack it by experiemting with lens replacement...

I'm considering to try it again, with less hurry, this time with a cheap V500 to have a less painful risk.

I'm considering using a Linos (like the hassies) spare lens I have laying around on a new frame with film placed at the bottom, so first step it would be disassembling the epson but in a way that it would still try to work diassembled.

IMHO a V700 sensor with a Linos can sign like Julia Lezhneva...

[Steven Ruttenberg]

Ted, there is a single sensor, a monolithic one, sporting 6 rows instead 3. But the two paired rows (Green for example) are clocked like if it was a single row.

183926

Between Pix 1 and 3 there is a blind zone and a sorrounding very low sensitive zone. The pixel 2 just reads that area in the next step, so the firmware has to place the reading of the pixel 2 in a different row of the circular RAM buffer than the 1,3 that was read in the same exposure.




There is no room (my undertanding) for things to happen in a different way as described in the above point.




I destroyed an Epson while trying to hack it by experiemting with lens replacement...

I'm considering to try it again, with less hurry, this time with a cheap V500 to have a less painful risk.

I'm considering using a Linos (like the hassies) spare lens I have laying around on a new frame with film placed at the bottom, so first step it would be disassembling the epson but in a way that it would still try to work diassembled.

IMHO a V700 sensor with a Linos can sign like Julia Lezhneva...

Sounds interesting if it can be done.

[Ted Baker]

Ted, there is a single sensor, a monolithic one, sporting 6 rows instead 3. But the two paired rows (Green for example) are clocked like if it was a single row.

183926

Between Pix 1 and 3 there is a blind zone and a sorrounding very low sensitive zone. The pixel 2 just reads that area in the next step, so the firmware has to place the reading of the pixel 2 in a different row of the circular RAM buffer than the 1,3 that was read in the same exposure.

With respect Pere I can read reasonably well... I did after all post the link ;) what is for sure it's not a detailed technical document, describing accurately the operation of the sensor, at the level we are now discussing...



There is no room (my undertanding) for things to happen in a different way as described in the above point.


I am not saying it doesn't happen that way, I am for real detail on how it works. In particular how does it work when the resolution other than 6400dpi is chosen, how it might vary when the stepper motor is driven at different speeds and most importantly how with evidence as opposed to how it could work...

Here are some ideas...

https://www.vision-systems.com/articles/print/volume-10/issue-5/technology-trends/ccd-sensors/pixel-shifting-increases-microscopy-resolution.html

Some other technical problems that I see with changing the lens is matching the stepper motor to the focal length (magnification) of original lenses. In the case of v750 it has two lens and two magnifications, how does the stepper motor cope with this fundamental problem... Perhaps this isn't a problem if the stepper and gearing match the ofset distance of the pixels on the sensor.. I am just suggesting there is a little bit more too this than meets the eye.

[Pere Casals]

With respect Pere I can read reasonably well... I did after all post the link ;) what is for sure it's not a detailed technical document, describing accurately the operation of the sensor, at the level we are now discussing...

Ted, I try to give you an interpretation about what's not explicit in the link you provided, from my technical background, let's say that I've designed several PCBs... do you really need I send you my Altium designs to give credit to those otherwise evident observations ?




I am not saying it doesn't happen that way, I am for real detail on how it works. In particular how does it work when the resolution other than 6400dpi is chosen, how it might vary when the stepper motor is driven at different speeds and most importantly how with evidence as opposed to how it could work...
[/url]

Some other technical problems that I see with changing the lens is matching the stepper motor to the focal length (magnification) of original lenses. In the case of v750 it has two lens and two magnifications, how does the stepper motor cope with this fundamental problem... Perhaps this isn't a problem if the stepper and gearing match the ofset distance of the pixels on the sensor.. I am just suggesting there is a little bit more too this than meets the eye.

Sorry, but the challenges you present are straight to solve.

Look, the V700s has an step resolution of 9600dpi, to always have square pixels with both lenses you need an stepper able to perform the "Greatest common divisor" of the inversions of 9600, 6400 and 4800, a pair of 9600 dpi steps makes a 4800 dpi step, but the native step has has to be finer than 9600 because you cannot do 6400 with 9600. Well, dividing by 9600/1200, 6400/1200 and 4800/1200 you have 8, 6, and 4, from that it resuts that native stepper resolution is exactly 19200 DPI or an entire fraction of it. Point the stepper model you want and I'll tell you the gearing.

Then... there is no critical speed control, there is an straight synchronization. The Stepper moves to next pixel, it fires the CCD trigger, waits integration time, and starts clocking the ccd to sequentially feed analog values (voltage) of the capacitors in each pixel to the ADC, wait for conversion (perhaps some in parallel) , and then... then the stepper move to next location....

[Ted Baker]

let's say that I've designed several PCBs...

Yes but did you do the one for Epson? Your still not getting it... I give up.

[Pere Casals]

Yes but did you do the one for Epson?

Absolutely no, but I think what I stated is correct.


Your still not getting it... I give up.

Ok, let's simplify, the V700 to V850 have 40800 hardware pixels in a row. RGB pixels in the sensor are arranged in 6 rows instead 3, but this is to improve performance over a 3 rows configuration.

[Steven Ruttenberg]

The bigger question is, does it provide the end product desired? If not, then we should improve upon it. But for 1000 bucks, it does a damned good job. I am sure the 25000 dollar drum scanners do better, but to quote John Travolta from Pulp Fiction, "how much is that shake? 5 dollars?!, pause, tastes shake, that's good, real good, but I don't know if its worth five dollars, but it is good..." My point is, is that little extra worth 25000 grand to do it yourself? Or worth up to 60 bucks or more for someone to do it for you? Or is it worth the headaches to purchase a used drum scanner or flextite scanner (still pricey) and you have to cobble together some archaic pos system just to run the scanner? And on top of that, they seem to be finicky and break often without parts available or knowledgeable techs to fix.

Anyway, spirited discussion as expected.

[Pere Casals]

The bigger question is, does it provide the end product desired? If not, then we should improve upon it. But for 1000 bucks, it does a damned good job. I am sure the 25000 dollar drum scanners do better, but to quote John Travolta from Pulp Fiction, "how much is that shake? 5 dollars?!, pause, tastes shake, that's good, real good, but I don't know if its worth five dollars, but it is good..." My point is, is that little extra worth 25000 grand to do it yourself? Or worth up to 60 bucks or more for someone to do it for you? Or is it worth the headaches to purchase a used drum scanner or flextite scanner (still pricey) and you have to cobble together some archaic pos system just to run the scanner? And on top of that, they seem to be finicky and break often without parts available or knowledgeable techs to fix.

Anyway, spirited discussion as expected.

Steven, if you had at home a V850 and a drum probably you would be making 90% of the LF scans with the V850, and you woudn't be the first one in that situation.

While a drum is a way superior machine it happens many times that the print size does not require more quality, having much more optical pixels in the image in that ppi in the print has no effect !!!

And also it can happen that the negative has no more information than the one that the V850 can retrieve, because diffraction, shake, focus... in a 3D scene not all objects are in the perfect plane of focus.

Then it happens that a V850 has more resolving power for 4x5 than a Hasselblad X5 !!!

There are situations where a drum makes a really big difference, but also for LF in many situations simply it's not worth to mount the negative for the drum, given what it's obtained with the V850.

[Ted Baker]

Absolutely no, but I think what I stated is correct.

That may well be the case, but it is not really a supported fact. It is a valued (perhaps expert) opinion from someone with a lot of hardware design experience is it not?



Ok, let's simplify, the V700 to V850 have 40800 hardware pixels in a row. RGB pixels in the sensor are arranged in 6 rows instead 3, but this is to improve performance over a 3 rows configuration.

I am sure the OP and anyone else who get this far IS interested in the details... Its reasonable to assume that if the the sensor area overlaps that some part of the overlap contributes to some form of image degradation. Some maybe even all of this degradation may be able to removed, with the correct technique.

The PDF you posted for the toshiba sensor was interesting, it showed spectral sensitivities which for example are the NOT ideal for colour negatives. It would be interesting to see one that was built for dedicated negative scanning purposes, or for that matter the real deal on the Epson chip...

[interneg]

The bigger question is, does it provide the end product desired? If not, then we should improve upon it. But for 1000 bucks, it does a damned good job. I am sure the 25000 dollar drum scanners do better, but to quote John Travolta from Pulp Fiction, "how much is that shake? 5 dollars?!, pause, tastes shake, that's good, real good, but I don't know if its worth five dollars, but it is good..." My point is, is that little extra worth 25000 grand to do it yourself? Or worth up to 60 bucks or more for someone to do it for you? Or is it worth the headaches to purchase a used drum scanner or flextite scanner (still pricey) and you have to cobble together some archaic pos system just to run the scanner? And on top of that, they seem to be finicky and break often without parts available or knowledgeable techs to fix.

Anyway, spirited discussion as expected.

Yes, the money, frustrating software & sometimes wonky hardware is worth it - the difference is jaw-dropping. And despite the obfuscation about the Epson supposedly having 'better' resolution for 4x5 than the Imacon/ Hasselblads, that is only from a partial test of extreme high contrast resolution, not an MTF test - if you saw the scans side by side, the Hasselblad scan blows the doors off the Epson. PMT Drum scanners can offer some further significant advantages, but are considerably less operator friendly. If you want to go beyond 4x5 (or 5x7 on a couple of models), might be better to look for one of the other high-end CCD flatbeds than the Hasselblad/ Imacon if you don't want to go down the PMT drum route.

[Pere Casals]

That may well be the case, but it not really a supported fact. It is a valued (perhaps expert) opinion from someone with a lot of hardware design experience is it not?


I've interfaced linear sensors, and designed PCBs for that, so I'm telling you what I find more than plausible, for example using a circular list in RAM to buffer all rows in course has no alternative a developer would consider. Even I give you details about the algorithms in the firmware...

Tell me any point and I will try to explain why things are done in that way.



Its reasonable to assume that if the the sensor area overlaps that some part of the overlap contributes to some form of image degradation. Some maybe even all of this degradation may be able to removed, with the correct technique.

A 6 rows linear sensor performs better than a 3 rows one in effective dpi, rather than an image degradation you have a quality enhacement, there is no significative overlap because we have blind areas in what it would be the overlaping. Microlenses on the pixels are tailored to optimize resolving power efficiency. EPSON, Toshiba, etc know very well what they do.



The PDF you posted for the toshiba sensor was interesting, it showed spectral sensitivities which for example are the NOT ideal for colour negatives. It would be interesting to see one that was built for dedicated negative scanning purposes, or for that matter the real deal on the Epson chip...

Well, that cannot be said, what counts is "how is" the illumination SPD multiplied by the 3 relative spectral sensitivities. These are the curves you have to consider to evaluate if the full combo suits your preferences, the curves on the datasheet does not determine the actual behaviour until you add the illumination SPD.

There is one thing where I could be wrong, those sensors have low resolution modes that can average the voltage of several pixels by interconecting the integrated charges from several pixels, I wouldn't use that but perhaps EPSON developers did it, it's the single point I'm not sure.

[Ted Baker]

There is one thing where I could be wrong

Surely not? :p

It's a complete waste of time if you have all the answers... I made a mistake you just don't get it.

[SergeyT]

>> if you had at home a V850 and a drum probably you would be making 90% of the LF scans with the V850, and you woudn't be the first one in that situation.
Or ditch the Epson and never look its way again :D The poor thing can't even focus on the film

[coisasdavida]

I am sure the OP and anyone else who get this far IS interested in the details...

Exactly. Specially when the manufacturer will try in every way to avoid spilling the truth.

Not sure there are CCDs made for scanning negatives. I believe today most things are done in software. Maybe the last CCDs made for negatives where on the Pakon and the HR-500, maybe those were software solutions as well.

[Steven Ruttenberg]

Steven, if you had at home a V850 and a drum probably you would be making 90% of the LF scans with the V850, and you woudn't be the first one in that situation.

While a drum is a way superior machine it happens many times that the print size does not require more quality, having much more optical pixels in the image in that ppi in the print has no effect !!!

And also it can happen that the negative has no more information than the one that the V850 can retrieve, because diffraction, shake, focus... in a 3D scene not all objects are in the perfect plane of focus.

Then it happens that a V850 has more resolving power for 4x5 than a Hasselblad X5 !!!

There are situations where a drum makes a really big difference, but also for LF in many situations simply it's not worth to mount the negative for the drum, given what it's obtained with the V850.

True, but I like to make one scan at the highest resolution I will ever need and since I will print up to 32x40 or larger, I want those extra peels. Plus I do get a benefit when I down size.

[Ted Baker]

Not sure there are CCDs made for scanning negatives. I believe today most things are done in software. Maybe the last CCDs made for negatives where on the Pakon and the HR-500, maybe those were software solutions as well.

The Arriscan XT is still manufactured and is built for negative scanning, not exactly cheap (absolutely no expense spared) and too small for most still formats. The recent Star Wars film afaik was scanned with it or the previous model. The FUJI and noritisu models are still made. The Arriscan even allows wet mounting automatically!

It is interesting to compare the various techniques, for example the Arriscan using separate Red, green, blue narrow spectrum LEDs instead of a broad spectrum illuminant.

Though I agree many use software techniques, as a 3D matrix should do a reasonable job, if done correctly. But the devil is in the detail if you interested in image processing. Much of clever stuff being held by Kodak, Fuji etc, neg/positive film is designed after all as a closed system.

It is interesting that the arriscan seems to use pixel shift as well instead they seem to refer to as "microscanning"

[Pere Casals]

>> if you had at home a V850 and a drum probably you would be making 90% of the LF scans with the V850, and you woudn't be the first one in that situation.
Or ditch the Epson and never look its way again :D The poor thing can't even focus on the film

You are not well informed, by 4x5" a V700 outresolves a Hasselblad x5 and Screen Cezanne. It is true that the Cezane can scan narrow strips of the 4x5" and stitch that in photoshop.

If you make curled 4x5 sheets then you can wet mount, when a v700 is not focusing the operator can be blamed.

184011

As you see in this graph (with the V700) you have a couple of mm to place your medium, well, if you are not able to place the medium there then you can blame the gear...

[Pere Casals]

Surely not? :p

It's a complete waste of time if you have all the answers... I made a mistake you just don't get it.

I've the answer that I have, if you have the level then prove that I'm wrong.

But let me explain something else about 6 rows technology. Tailoring the photodiode sensitive area in the pixel is by design, these are micrometer structures when IC manufacturing in is nanometers. In the same pixel we have the photodiode, a capacitor, resistors, transistors and circuitry. Even in de dual configuration the sensor manufacturer places microlenses on pixels to extend sensitive area to the optimal point por best resolving power. The blind areas generate two different kinds of noise. The size and shape of the microlens is the optimal.

Is it painful that the V700 sports 40800 RGB pix and a Hassy 8000 ???

Complain to Hasselblad, rather trying to discredit V700 strengths...

[Ted Baker]

Is it painful that the V700 sports 40800 RGB pix and a Hassy 8000 ???


Absolutely not they are two pieces of manufactured plastic, metal, and glass. Could not care less, just interested in how they work and how I might explore and improve upon there use.

Your arrogance is breathtaking.

What you don't understand is other readers can read, can learn, can reason, and have there own original thoughts, and maybe make there own contributions in image processing and general photography!


I've the answer that I have, if you have the level then prove that I'm wrong.

If you take the time to read what I said... There is no disputing your plausibility of your explanation, it may be the correct explanation, what you fail to understand is why someone might want to enquire further...

[Pere Casals]

Absolutely not they are two pieces of manufactured plastic, metal, and glass. Could not care less, just interested in how they work and how I might explore and improve upon there use.

Your arrogance is breathtaking.

What you don't understand is other readers can read, can learn, can reason, and have there own original thoughts, and maybe make there own contributions in image processing and general photography!

Here we are discussing about linear CCD sensor, I'm not arrogant: I've interfaced linear sensors, designed 3 PCBs for that, I'm attended two seminars about it and I try to explain to you that the 40800 RGB pixels of the V700 sensor are true 40800, with no degradation because of the dual enhacement.

[Pere Casals]

Well, let's see why the X5 desing priorities led them to a 8k sensor. Do you guess why ?

[Pere Casals]

The X5 is a pro machine that may run a business on their own. The priority is quality vs production optimization. The trade of speed vs noise led them to 8k sensors.

That arrangement is optimal for 35mm and MF, but is has shortcomings for sheets, in a way that even a V700 outresolves the X5 for 4x5. While the X5 targets the roll film market it sports perfect capability for a share of the LF jobs, but evidently not for all, beyond it cannot make 8x10.

While a V700 has shortcomings in small sizes (specially in 35mm) it shines the more as the format gets larger. A LF 5x7 / 8x10 BW photographer may fulfill his needs with a V700.

Probably one may want something better for rolls, but for LF a V700 is not a joke. So today, for a multiformat film shooter, an optimal combo is a V850 paired with a Plustek. That combo challenges expensive or pre-press discontinued gear in most situations.

IMHO those trying to discredit the V850 should first learn to use it in the optimal way.

[Ted Baker]

Well, let's see why the X5 desing priorities led them to a 8k sensor. Do you guess why ?

Perhaps some questions you might ask yourself are: Are the readers of this thread capable of making such a guess themselves? does it matter if they guess right or wrong? Could they evaluate the effectiveness of this supposed design priority? Could they evaluate the applicability of this design priority against current sensor technology?
Will it ever be possible to answer the Op original question with exact details on the chip, or will we need to rely on reasonable assertions?

[Pere Casals]

Perhaps some questions you might ask yourself are: Are the readers of this thread capable of making such a guess themselves? does it matter if they guess right or wrong? Could they evaluate the effectiveness of this supposed design priority? Could they evaluate the applicability of this design priority against current sensor technology?
Will it ever be possible to answer the Op original question with exact details on the chip, or will we need to rely on reasonable assertions?

Just I explain why a V700 outresolves an X5 for 4x5.

For those denigrating the cheap choice...

[Ted Baker]

Just I explain why a V700 outresolves an X5 for 4x5.

For those denigrating the cheap choice...

Dunno, who cares? Is the correct answer the combined optical/mechanical/sensor is greater in one than the other?

Perhaps I am mistaken, but the OP question relates to what the actual sensor in the unit in question, what it has and how it works. You have kindly provided your opinion based on your expertise.

Perhaps the sensor is flawless from a resolution perspective with no noise caused by light outside the assumed pixel area, the stepper motor is perfect and all that is needed is a better lens. I have toyed with idea if I ever get the time to implement it I will report the results.

Or how does this type of arrangement work with grain aliasising, is it better or worse?

Is there anything to be learned that might be applied in other circumstances, for example new hardware or software or modified hardware or software?

[Pere Casals]

Is there anything to learned that might be applied in other circumstances, for example new hardware or software or modified hardware or software?

The V700 series are suposed optimal in the electronics, as one may expect from EPSON. It can be hacked in the optics to get an enhacement.

One thing it can be done is Multi-Exposure in a single pass, but this implies making a new controller board because you won't want to reprogram the uC chips. EPSON is not to provide the source code, so this would require disassembling the firmware, reverse-engineering what the controller does, adding the feature and compiling again... it's faster to start from zero.

[Steven Ruttenberg]

Here is an example of a wetscanned image on V850 using the better scanning adjustable holder, optical mylar, kami fluid and ANR glass. The distance above the scanner glass is .125 inch or 3.175 mm exactly (I know because I used dial gauge to perform a cmm measurement of every tab on the holder and adjust so they were all the same height. This is my measurements from previous adjusting for optimum focus. I am going to go back and refocus once I get my scanner resolution target. This was scanned at 4200dpi and then sized to 16x20 at 300dpi and compressed 35% for final jpg. Additionally, ficker seems to be doing something to the images that over-saturates them. Others have noticed the same. Yes, this was properly converted from adobergb1998 to srgb. Film was Extar 100, Nikkor 75mm f/4.5 at f/32 for 30 seconds and developed in standard C-41 kit. The detail is great, sharpness is great (I probably should have used some swing, but I just used some rise/fall and fwd tilt. Before stopping down, the two points in focus after tilting was about half way up the scene in the distance and the closest part of the foreground. Stopping down to give me the height needed for the rest.

https://farm2.staticflickr.com/1925/45676253861_b11b977d6f_z.jpg (https://flic.kr/p/2cAfJWi)20181027_0129_Working_20181101 (https://flic.kr/p/2cAfJWi) by Steven Ruttenberg (https://www.flickr.com/photos/157376714@N08/), on Flickr

[Pere Casals]

Here is an example of a wetscanned image on V850 using the better scanning adjustable holder

Steven, to show what a scanner does you may post an small crop at 100% resolution, because you may not be able to post all information of the sheet.

Like that: https://www.flickr.com/photos/125592977@N05/32535835184/in/dateposted-public/

[Steven Ruttenberg]

Steven, to show what a scanner does you may post an small crop at 100% resolution, because you may not be able to post all information of the sheet.

Like that: https://www.flickr.com/photos/125592977@N05/32535835184/in/dateposted-public/

Okay, I will have to do that tonight from home where I can use the original image.

[Pere Casals]

Perhaps the sensor is flawless from a resolution perspective with no noise caused by light outside the assumed pixel area, the stepper motor is perfect and all that is needed is a better lens. I have toyed with idea if I ever get the time to implement it I will report the results.


In the V750 the horizontal resolution is better than the vertical one, average 2643 vs 2148, this report says: https://archivehistory.jeksite.org/chapters/appendixc.htm

My interpretation (guessing) is that there is a vibration in the vertical direction provocated by the driving (belt?), and this effectively limits resolving power in the vertical direction. To remove that possible vibration the scanning could be made slower or an improved driving would be necessary.

2643/2/25,4 are around 52 lp/mm (at extintion), so hacking the lens would require taking from film more than this.

I guess that replacing the original driving by an screew would increase the vertical resolution to the horizontal performance, but (non scanning) displacements would be very slower.

[Steven Ruttenberg]

Redoing image to remove a colorcast that I thought I liked, but now that it was pointed out, I am not liking it so much, so I am fixing and when done will post up the 100% crop vs the full image.

[Peter De Smidt]

Getting good color is a challenge, and, like most things, it gets easier with a lot of practice. There are two main ways. First, there's the Dan Margulis's method of going by the numbers. To do this, you have to know what kind of numbers things should be. That takes some time. The advantage is that you remove perceptual issues/monitor issues from the process. As such, a good editor can make a good edit on a poor monitor, in a brightly lit and colored room, when one is tired...... This was an especially good method when good monitors were very expensive, and one had to hit pre-press target densities. Jon Cone also recommends this in his Piezography manual. He thinks one should learn what gray in a print a specific Photoshop number value will give. All of this is similar to Ansel Adams's system of knowing what scene values will print at with standard exposure and processing. The subjective element comes back when one makes a proof print. The numbers can get you close, but prints allow you to fine tune the output.

The other method is the color-managed one, one with a very good monitor, which can be surprisingly inexpensive these days, hardware calibration and profiling, a dim-neutral colored room, and plenty of breaks/looking at something else, to give your perceptual system a reset. One tip is to always go overboard with adjustments, and then dial them back.

To check whether there's a color cast, using the curve eye dropper's can be helpful. Make a curves adjustment layer. Now use the light eyedropper on the nearest thing to white in the image, the middle eye dropper on the thing nearest to neutral gray, and the black eye dropper on the closest thing to black. Change the blending mode of the layer to 'color'. Turn it on and off. Does it make the image better or worse? If better, then you have color casts. You can go back to raw (with digital), use the layer as as, perhaps adjusting it's opacity, adjust the color in some other way.....But you won't know to do that if you don't realize that there's a cast.

A hue saturation layer check at the end can also be helpful. Make a hue saturation layer. Click to show each of the channels. Slowly move the saturation slider from one side to the other. Figure out where it looks best. Now go to the hue slider for a specific color, say, red. Move it slowly one way and then the other. Does it get better in one direction? Do the same for each of the colors.

Separating adjusts in luminosity, hue, and saturation can be helpful. I only work on one at a time. Blending modes makes this easy to do.

Let's also distinguish: accurate color, expected color, color casts, and color grading. Accurate color is when the output, whether print or screen, comes close to matching the colors in scene. You might need this if you're wanting to print color samples, where you want the final output to match the sample you have. Expected color is a range of colors that we, as humans, expect. For example, sky's are normally blue, grass is yellowish green, stop signs are red. As long as the colors fall in this range, people we feel that the colors are natural. A cast occurs when there's an unwanted overall hue shift from either what is in the scene (accurate color) or what is meant to look natural (expected color.) An image can look good to someone, especially someone who's been staring it an image for too long, but it might still be recognized as unnatural by viewers. Finally, there's color grading, which is shifting the color from what's natural for a specific effect. With art, the artist can pick whatever color he or she likes. The important point is to be able to control your color to get the effect you want. I suggest that being able to get natural color regularly is a stage on the way to being able to use color grading effectively. When color is unnatural, viewers will ask whether the effect supports the picture or undermines it. If it undermines it, then it's bad color (for them.). Normally, such as with upright veritcals, symmetry, sharpness......the image looks best either perfect, or decidedly, purposefully different from expected. If something is a little bit off, viewers are likely to chalk it up to ineptitude.

[Pere Casals]

Getting good color


I'm realizing that using 3D LUTs is a very powerful way to manage color and to get what one wants with precission and fast.

[Steven Ruttenberg]

Hmm lots more to review. My dilemma is I will like the way image looks and yet know there is a cast and until I remove it and see difference won't consider the effect of the cast as not good. I have profiled Benq monitors so should be able to do better.

[Ted Baker]

This is very detailed and maybe very accurate document discussing both the Epson series and Nikon Coolscan 9000.

It seems to suggests that the Epson series v750 scanners do NOT really have 4800/6200 dpi pixel densities in both axes. Page 135. Only half in the Y axis.

http://adsabs.harvard.edu/full/2009ASPC..410..128S

I suspect it is not the full story, but it underscores that on occasion sometimes there is an alternative explanation. ;)

[Peter De Smidt]

I suspect it is not the full story, but it underscores that on occasion sometimes there is an alternative explanation. ;)

"A little learning is a dangerous thing; drink deep, or taste not the Pierian spring: there shallow draughts intoxicate the brain, and drinking largely sobers us again." Alexander Pope.

[Pere Casals]

This is very detailed and maybe very accurate document discussing both the Epson series and Nikon Coolscan 9000.

It seems to suggests that the Epson series v750 scanners do NOT really have 4800/6200 dpi pixel densities in both axes. Page 135. Only half in the Y axis.

http://adsabs.harvard.edu/full/2009ASPC..410..128S

I suspect it is not the full story, but it underscores that on occasion sometimes there is an alternative explanation. ;)

:) Sorry, Ted, but this doc is hilarating...

Page 140 says that V750 is not worth because critical focus cannot be achieved with glass plates, as there is no auto-focus feature... authors show themselves being pretty ignorant about scanners. If there is a medium that it's really easy to get in perfect focus with a V750 is glass plates :)

Really I don't understand why a proficient astronomer can't plot or at least understand how to use a graph like that to gauge the glass plates on the V750 bed with total precission :
184073

Also for 8x10 focus is just on the bed !

Well... This document is descredited by itself...

You can check what precissions do 3D printers (Prusa with Marlin) obtain in XY axis driven with belts and in the vertical Z that is made with an screew, to me this is what enlights what may happen with the Y of the V750.




Anyway, for 4x5, the V750 has more resolving power in the Y (2148 average) than a X5 that has (effective) 1800. Also most drum scanning services offer 2000dpi for 8x10...


More than 2000 dpi 8x10 should be really expensive from a drum !
https://cone-editions.com/drum-scanning/
184074

[interneg]

It seems to suggests that the Epson series v750 scanners do NOT really have 4800/6200 dpi pixel densities in both axes. Page 135. Only half in the Y axis.


The MTF performance for the Epson reported in that article would also go a long way to suggesting that the quality of the optical system in those machines is a major hindrance. I'm slightly shocked at how bad it actually is, if that article is correct.

[Pere Casals]

I'm slightly shocked at how bad it actually is, if that article is correct.

Well, I really shocked by the poor 8k sensor that the Hasselblad X5 and the Cezanne sport... not being able to go beyond 1800 dpi effective for 4x5. Compared to the 40k of the V750 !!!

And it's LOL that a V700 outresolves a $25000 Hasselblad X5 for 4x5". The X5 is way better for rolls, but this is a LF forum...


The HR lens of V700 is arround 3000dpi at extintion over 5.9" width, after descretization yields some 2600-2900 in the X axis, if it yields 2200 only in the Y it's because of the mechanics, I guess. But anyway the $500 V700 outresolves an X5 for LF, and makes 8x10 !

Look, it's about understanding what each machine does. The EPSON is a powerful choice for 8x10 and very good for 4x5. For most LF situations it holds very well compared to expensive gear that is optimized for rolls (X5), or to pre-press gear that was optimized for graphic arts.

[Ted Baker]

:) Sorry, Ted, but this doc is hilarating...



I think your missing the point, this is another plausible explanation about the operation of these devices if anyone so cares...

I found it BTW because I am researching spectral characteristics of scanners, and there methods of operation.

Anyway these guys with their little telescopes seems to have two further very comprehensive reports that cover the CCD of the epson V750.

Its also my understanding the MTF they quoted used proper slanted edge tests, to work out an MTF of 50% not just picking a number of the USAF test card. Its of course possible they didn't think to focus the image.

http://dasch.rc.fas.harvard.edu/papers/Scannerevaluation1.pdf
http://dasch.rc.fas.harvard.edu/papers/ScannersandScience.pdf

Probably about 35pages on the epson sensor... (a lot of duplicated)

[Pere Casals]

Probably about 35pages on the epson sensor... (a lot of duplicated)

Ted, please, analyze the wording: "If the (V700) pixels were of the 2.7um x 5.4 um variety that seems in general use at NEC, then the active line length would be 55.080 mm"

This is an academic exercise of an student... that disassembled a trashed V700:

"I managed to get a broken Epson v700 scanner on Ebay so that it could be taken apart to understand..."

Ted, please...

[Peter De Smidt]

Well, I really shocked by the poor 8k sensor that the Hasselblad X5 and the Cezanne sport... not being able to go beyond 1800 dpi effective for 4x5. Compared to the 40k of the V750 !!!



How often are you going to repeat this BS, especially after being corrected numerous times? I've made over 500 scans on a Cezanne. One is sitting behind me as we speak. I have an Edmund high resolution chrome-on-glass target, and so my resolution claims are not some marketing pipe dream. I scan my 4x5s at 4000spi. Yes, it takes two strips, but that's incredibly easy to do. Specify first scan area. Setup parameters. Scan. Now click on scanning area and move it over to cover the next area, which takes seconds. Hit scan. All the adjustments will be the same, and the file name will automatically be increased by 1. You can queue up a whole bed of scans.

[interneg]

How often are you going to repeat this BS, especially after being corrected numerous times? I've made over 500 scans on a Cezanne. One is sitting behind me as we speak. I have an Edmund high resolution chrome-on-glass target, and so my resolution claims are not some marketing pipe dream. I scan my 4x5s at 4000spi. Yes, it takes two strips, but that's incredibly easy to do. Specify first scan area. Setup parameters. Scan. Now click on scanning area and move it over to cover the next area, which takes seconds. Hit scan. All the adjustments will be the same, and the file name will automatically be increased by 1.

Because it suits some bizarre & egregious agenda which cannot admit that even at 2000spi, the Cezanne, Hasselblad etc walk all over the Epson in real-world optical performance. Thus the hiding behind high contrast targets & pretending that (unlike real world optics) MTF performance is irrelevant. The denial seems to come from an unwillingness to understand that the high end scanners are able to operate to the limits of their optical systems & that they can & will hit their specified resolutions at the lower resolutions because the demands on the precision of film placement & lens resolution are lower.

[Ted Baker]

Ted, please, analyze the wording: "If the (V700) pixels were of the 2.7um x 5.4 um variety that seems in general use at NEC, then the active line length would be 55.080 mm"

This is an academic exercise of an student... that disassembled a trashed V700:

"I managed to get a broken Epson v700 scanner on Ebay so that it could be taken apart to understand..."

Ted, please...

Seems very detailed for a student, must have been a good student, since he is now an associate prof at Harvard... ;)

BTW if you look at the MTF graphs, they actually tally pretty accurately with the results quoted in this thread from scandig etc, if you say use an MTF of around 10%.

I am not suggesting its 100% correct, but it worth a read if it is topic of interest.

[Pere Casals]

How often are you going to repeat this BS, especially after being corrected numerous times? I've made over 500 scans on a Cezanne. One is sitting behind me as we speak. I have an Edmund high resolution chrome-on-glass target, and so my resolution claims are not some marketing pipe dream. I scan my 4x5s at 4000spi. Yes, it takes two strips, but that's incredibly easy to do. Specify first scan area. Setup parameters. Scan. Now click on scanning area and move it over to cover the next area, which takes seconds. Hit scan. All the adjustments will be the same, and the file name will automatically be increased by 1.

Peter, see the contex: http://www.largeformatphotography.info/forum/showthread.php?148772-Size-of-scanner-CCD-in-number-of-elements&p=1467642&viewfull=1#post1467642

"It is true that the Cezane can scan narrow strips of the 4x5" and stitch that in photoshop.", I said that...

Well, for LF, if not stitching strips in photoshop, the V700 outresolves the Cezanne, do you like that way ?


but 160 lbs weight and that money to be stitching strips in photoshop ???

Anyway if stitching this allows an amazing capability for sheets... this also has to be said.

And also it's interesting that the EPSON is better if not stitching in the Cezanne, this shows the V700 value.

[Pere Casals]

Seems very detailed for a student, must have been a good student, since he is now an associate prof at Harvard... ;)

BTW if you look at the MTF graphs, they actually tally pretty accurately with the results quoted in this thread from scandig etc, if you say use an MTF of around 10%.

I am not suggesting its 100% correct, but it worth a read if it is topic of interest.

Probably a nice guy, I found interesting some reasonings... His Harvard department had to be poor, if purchasing trash at ebay, I belived that this only happened in my country's research departments.

But he was not understanding the 6 row configuration intrinsics at all, that's key to understand V700 design optimization, that's cost-performance oriented.

Easy to guess that chief designer of the V700 had limitations (mainly production cost) and also amazing corporate resources. Because of the corporate intrinsics they were able to assemble a $500 product outresolving a Cezanne (if not stitching) and an X5 with sheets.

It is very interesting how they were able to assemble a concept that seized a market niche...

[Pere Casals]

even at 2000spi, the Cezanne, Hasselblad etc walk all over the Epson in real-world optical performance. Thus the hiding behind high contrast targets & pretending that (unlike real world optics) MTF performance is irrelevant.

Interneg, are you saying that the (4x5) 1800 effective dpi of the X5 are better than the 2400 effective dpi of an V700 ?

You are very wrong...

Here we speak about resolving power, if you want to debate about DMax in the slides particular case... we can... It's clear that in some Velvia cases the X5 is better than the V700 with Multiexposure, I admit it.

What's about tonality/color management I see no difference, with 3D LUTs I'm able to do the same with any scanner.

[Steven Ruttenberg]

Why don't we each submit a full scan of same image using the device they have. I have v850 and wet scan. We can then do a real a idea by side. Since none of us actually designed any of the mentioned scanners, trying to explain their proprietary operation as fact of any sorts is like trying to prove God is real. We all know the concept behind it as nd there are infinite explanations, all plausible and all equally bs.

[Peter De Smidt]

I hate the idea of being run off a forum I've spent a long time participating in. I respect a lot of people here, and I've had great interactions with a whole bunch of knowledgeable and generous people here. But I'm really tired of dealing with deluges of misinformation, only to have it pop up over and over. It's no problem being wrong or making mistakes. That happens to all of us, but being corrected and not having the decency to move forward is disrespectful and tedious.

[Peter De Smidt]

Why don't we each submit a full scan of same image using the device they have. I have v850 and wet scan. We can then do a real a idea by side. Since none of us actually designed any of the mentioned scanners, trying to explain their proprietary operation as fact of any sorts is like trying to prove God is real. We all know the concept behind it as nd there are infinite explanations, all plausible and all equally bs.

No, at least for me. It's been done before. It's a waste of time. A client just ordered a large print. I'm going to spend my time working on that.

[interneg]

Interneg, are you saying that the (4x5) 1800 effective dpi of the X5 are better than the 2400 effective dpi of an V700 ?

You are very wrong...

Here we speak about resolving power, if you want to debate about DMax in the slides particular case... we can... It's clear that in some Velvia cases the X5 is better than the V700 with Multiexposure, I admit it.

What's about tonality/color management I see no difference, with 3D LUTs I'm able to do the same with any scanner.

No. Stop spreading self-evident nonsense. The Precision II delivers 1800ppi, every subsequent Imacon or Hasselblad scanner that does 4x5 delivers 2048ppi. No reputable test has ever produced any evidence to the contrary. The onus is on you to prove your claims with repeatable results. Instead you continue to spin tangled webs of deceitful nonsense that run contrary to current optical engineering which long ago abandoned the high contrast target in favour of the more systemic MTF approach. Unlike you I can do side-by-side testing between an A3 Epson and an X5 and can see the difference as can those I scan for. It is not a small or subtle difference. I'm not even going to go into your proclaimations about LUT's, but it's pretty clear that you are intent on misrepresenting them too.

[interneg]

No, at least for me. It's been done before. It's a waste of time. A client just ordered a large print. I'm going to spend my time working on that.

Definitely the best policy!

[Pere Casals]

Why don't we each submit a full scan of same image using the device they have. I have v850 and wet scan. We can then do a real a idea by side. Since none of us actually designed any of the mentioned scanners, trying to explain their proprietary operation as fact of any sorts is like trying to prove God is real. We all know the concept behind it as nd there are infinite explanations, all plausible and all equally bs.

Steven, here there is a collaborative scanner test made in this forum:

http://www.largeformatphotography.info/scan-comparison/

I fully agree about what was concluded there, with the exeption that some V750 crops from shadows would enhance the result is Multi-Exposure was used. The V700s have lower performance in the high densities if ME is not used, for the rest, anyone can judge from images.

[Pere Casals]

The Precision II delivers 1800ppi, every subsequent Imacon or Hasselblad scanner that does 4x5 delivers 2048ppi.

The subsequent Hassies deliver 4x5 2048ppi hardware, 1800 dpi effective like the PII. Being minimally smart one will soon realize that because of discretization a 2048 sensor never, never, never will deliver efective 2048 DPI. Thanks to the amazing Linos lens that the Hassies sport the loss is only 10% to the effective 1800 dpi.

If you don't belive me scan a 2048dpi target and then tell me what kind of aliasing effects you see. Weren't you aware?

[interneg]

The subsequent Hassies deliver 4x5 2048ppi hardware, 1800 dpi effective like the PII. If you are minimally smart you will soon realize that because of discretization a 2048 sensor never, never, never will deliver efective 2048 DPI. Thanks to the amazing Linos lens that the Hassies sport the loss is only 10% to the effective 1800 dpi.

If you don't belive me scan a 2048dpi target and then tell me what kind of aliasing effects you see. Weren't you aware?

Think about it for a second. It's the difference between resolving element 2 & element 3 in group 5 of your beloved 1951 target. If what you claimed was true, any competent test would have flagged this up because it would have been obvious. This is basic stuff. Stop hiding behind clouds of hand-waving excuses.

[Pere Casals]

I hate the idea of being run off a forum I've spent a long time participating in. I respect a lot of people here, and I've had great interactions with a whole bunch of knowledgeable and generous people here. But I'm really tired of dealing with deluges of misinformation, only to have it pop up over and over. It's no problem being wrong or making mistakes. That happens to all of us, but being corrected and not having the decency to move forward is disrespectful and tedious.

Peter, it's not mandatory to read my posts.

To me it's interesting to debate what performance has a cheap device compared to expensive equipment, I feel it's useful to the comunity that all opinions can be told.

If anybody gets angry because others say their technical opinions, it's up to him. You can say I'm wrong and I'm not to get angry. And if I'm really wrong it will not be difficult for you to discredit me.

I made my tests and I've perfectly the right to express what I think.

Any one can download samples from the collaborative scanner test http://www.largeformatphotography.info/scan-comparison/

and like I did, he can adjust in Photoshop the V750 samples to make it match the other samples, like (repeatable by anyone) this:

https://c1.staticflickr.com/9/8561/28420386682_d481942db8_o.jpg

It's true that the V sample admit more sharpening than the 4500 and that some crops of the shadows are worse in the V, but my guess in that ME was not used.


https://c1.staticflickr.com/9/8779/27910101114_b802ff6d33_o.jpg

[Pere Casals]

Think about it for a second. It's the difference between resolving element 2 & element 3 in group 5 of your beloved 1951 target. If what you claimed was true, any competent test would have flagged this up because it would have been obvious. This is basic stuff. Stop hiding behind clouds of hand-waving excuses.

interneg, please, how can you say that a 2048 sensor can deliver 2048 efective dpi? this is not possible. See Rayleigh criterion...

imagine you scan a 2048 lines per inch target (1024 black, 1024 white lines per inch) with a 2048 dpi sensor: you may take a perfect grey image if sensor pixels fall in the middle of the bars...

[Peter De Smidt]

Here's a quote from Daniel Haybron's Happiness: A Very Short Introduction:

184079

[Pere Casals]

Here's a quote from Daniel Haybron's Happiness: A Very Short Introduction:


If I'm that wrong, I'm only wrong, and it would be pretty easy to you to discredit me technically. Do it...

[coisasdavida]

I hate the idea of being run off a forum I've spent a long time participating in. I respect a lot of people here, and I've had great interactions with a whole bunch of knowledgeable and generous people here. But I'm really tired of dealing with deluges of misinformation, only to have it pop up over and over. It's no problem being wrong or making mistakes. That happens to all of us, but being corrected and not having the decency to move forward is disrespectful and tedious.

Couldn't agree more.
Man, how I regret posting this simple question...

[Steven Ruttenberg]

No, at least for me. It's been done before. It's a waste of time. A client just ordered a large print. I'm going to spend my time working on that.

That is much better endeavor. Don't let a few run you off. I prefer people with knowledge who are willing to help such as yourself especially when your willing to help a stranger out as witnessed by your pm. I am not fond of the infighting and know it alls. My post was simply meant to take away from the all to serious scanner debate that has yet once again cropped up. I find it a waste when it doesn't matter in the end. So I hope you don't disappear.

[Steven Ruttenberg]

If I'm that wrong, I'm only wrong, and it would be pretty easy to you to discredit me technically. Do it...

Stop being antagonistic. It does not help. And if you are right? So? Wouldn't it be better to use your knowledge to help someone make a better image whether a print or negative?

I think you have a wealth of knowledge you could share.

[Corran]

You are not well informed, by 4x5" a V700 outresolves a Hasselblad x5 and Screen Cezanne.

Still you spout this garbage?

As usual, everyone should take anything Pere says about the Epson with a grain of salt. Or the whole salt shaker. Oversharpened Epson scans do not equal resolution.

[Oren Grad]

Time for a time out.

Also a reminder that vBulletin has an "ignore" function.