Does any scanner have a 3 exposure HDR scan option?

[invisibleflash]

In that it would scan the film 3 times to get the highlights and shadows, then combine the scans? Or it could scan the film 3 times and spit out 3 different exposure scans that could be combined later?

Thanks

[Pere Casals]

In that it would scan the film 3 times to get the highlights and shadows, then combine the scans? Or it could scan the film 3 times and spit out 3 different exposure scans that could be combined later?

Many scanners support the multi-exposure feature (google: silverfast multi-exposure) that takes two images and delivers an optimal blend, you have to save 16 bits per channel in TIFF format (not jpg or bmp) to conserve the 16 bits.

Mostly it's not necessary to take 3 images because with 2 you reach the scanners physical limits and/or the film limits.

You may also set 3 manual exposures and then you can stitch/blend the resulting 3 images in PS, but this is unnecessary because silverfast does it perfectly.

___

The silverfast version that it's bundled in the epson V850 includes the multi-exposure feature, the V800 not, but you may purchase a software upgrade to have it, this is mostly important if wanting to recover very deep shadows in velvia slides.

[Peter De Smidt]

Most digital-camera-based-scanners could do this easily.

[interneg]

What Peter said. And it's probably really only likely to be a necessity with certain situations involving transparency films. There are scanners running HDR scans for cinema film up to 10K resolution, but it's largely needed only for dealing with print stock, not camera negative.

[Peter De Smidt]

By using a quality digital camera, one already has about a 1-stop advantage over an Epson scanner, as tested with a Stouffer step wedge.

[Pere Casals]

It has to be a flawed test with the stouffer... was the epson using Multi-Exposure ? was exposure well adjusted for the job?

If DSLR is compared to an V700 auto scan made with EpsonScan then prossibly the DSLR beats the Epson, but not if the Epson is properly operated.

This is the signal to noise ratio of the V700.

193187

DSLRs have around 14 bits, and the last bits have a lot of noise, the epson still has discernible signal over noise at the highest desnsitites, of course result won't be clean by 4.0D and we cannot say it makes 4.0D, it's good for 3.3D, but the same happens with a DSLR.

A DSLR is not better than the epson if the last is operated properly, the SNR ratings are the proof.

[Alan Klein]

Scan once. Use the shadow adjustment on your post processing program. It works faster and is better. You can;t get more out of a scan than the scan will give you on one scan. The scanner is using the maximum amount of light to penetrate the negative with sensors to match. That is what its dMax rating is all about. YOu can't get more out of it. Also, the range will not exceed the clipping point at either end. Just set them to extend a little past the range of the particular shot. Scanning twice will not give you "penetration" any more than shooting a camera twice at the same object with all the same settings will give you a brighter picture.

[invisibleflash]

OK, thanks for all the feedback!

[Peter De Smidt]

Some scanners do have a multiscan option. What you have to be careful with is whether the scanner scans the whole negative multiple times and then combines the data, or whether the scanner moves into position, takes multiple readins, moves to the next are, repeat. The first way can lead to a loss of sharpness in the scan if the multiple scans don't perfectly line up. The second way is preferable.

[Ted Baker]

Vuescan with epson V series and 4990 all have 2 exposure HDR options, it is just called something different :-)

It is only needed with transparency, its a simple and very effective improvement for almost nothing for kodachrome at least if you already have the software. (scan takes longer). Two passes is all that is needed three passes isn't really necessary. Not really required for negs if the exposure is set appropriately.

[Alan Klein]

Vuescan with epson V series and 4990 all have 2 exposure HDR options, it is just called something different :-)

It is only needed with transparency, its a simple and very effective improvement for almost nothing for kodachrome at least if you already have the software. (scan takes longer). Two passes is all that is needed three passes isn't really necessary. Not really required for negs if the exposure is set appropriately.



I have an Epson V600. I know of no way to change the scan to change the light output which is currently set on maximum anyways. So how would two scans work to improve seeing more data in the shadow areas?

[Peter De Smidt]

I don't know that scanner, but some work by having a longer exposure time for the second sample, but to repeat: multiple exposure is usually unnecessary. It's better to use a device that can capture the required data in one sample. If that's not an option, then it's better to have a system that makes multiple samples, either varying light intensity or exposure time, at each sample location before moving on. Multi-pass systems usually lead to a loss of detail, as the passes don't line up perfectly.

[Ted Baker]

multiple exposure is usually unnecessary.

Peter,

I found for 35mm kodachrome, with exposures all over the place (often 1 stop under or more) using the "2 exposure HDR scan option" was a no brainer, the S/N ratio in shadow areas was improved, I never bothered to check any drop in resolution but the Signal/Noise ratio improvement is very obvious. It just looked better for virtually no additional effort. Often the same thing could be achieved by a single pass if I manually set the exposure since the highlights were also dense, but this requires a preview scan and operator to do so.

This was for the 4990, it is a pity this feature is not properly implemented so that only a single pass is required. I never use this feature for negatives, but I found it was certainly useful for underexposed kodachrome. For properly exposed 4x5 perhaps less so, I don't normally shoot transparency.

[Peter De Smidt]

Fair enough, Ted. I've only scanned about 50 Kodachromes, and there were all taken by my father. They scanned without any problem using a single exposure on a Nikon Coolscan V. In a given system, minimizing noise/tonal concerns might very well trump max detail. As usual, one should try all the options with one's system to optimize the process for one's purposes.

[interneg]

I don't know that scanner, but some work by having a longer exposure time for the second sample, but to repeat: multiple exposure is usually unnecessary. It's better to use a device that can capture the required data in one sample. If that's not an option, then it's better to have a system that makes multiple samples, either varying light intensity or exposure time, at each sample location before moving on. Multi-pass systems usually lead to a loss of detail, as the passes don't line up perfectly.

I've seen no evidence that any of the aftermarket software is able to vary the light intensity or drive speed of consumer flatbeds - and given that most of them in regular operation seem to use 2 overlapping sensor lines per colour channel of about 1200ppi resolution & attempt to interpolate/ align the whole thing together not terribly well, adding a second pass is just going to make things worse. I spent/ wasted several hours last week testing the effects of combining & aligning (in the manner oft described under various pseudo pixel-shift workaround techniques) several native 1200ppi scans from a 3xCCD sensor to get to 2400ppi to see how they'd compare with a single 2400ppi native scan. The results were interesting - the combined scan was qualitatively worse in some aspects than either native scan, yet did seem somewhat higher resolving than the 1200ppi. Interestingly, the qualitative flaws were not dissimilar from the flaws seen in consumer flatbed scans. What I think it showed me is that for optimal results, the software needs to know exactly how far the sensor has been moved shot-to-shot in order to optimise alignment - and is pushing me in the direction of cameras with built-in pixel shift systems for any further exploration in this direction. I think pixel shift has potentially more to offer than multiple exposure does- and it would not be a million miles off how a Fuji Frontier effectively operates...

[Corran]

In my tests with both a Nikon LS-8000 and Epson 700 years ago, multi-pass exposures usually resulted in poor performance and little to no gain in the shadows, even on the much superior Nikon.

[interneg]

In my tests with both a Nikon LS-8000 and Epson 700 years ago, multi-pass exposures usually resulted in poor performance and little to no gain in the shadows, even on the much superior Nikon.

That tallies with 'multi-exposure' not really adjusting the exposure & instead adjusting the gain.

[Ted Baker]

These are two crops of 35mm underexposed Kodachrome 64, nothing at all special other then sentimental value from my Epson 4990-


https://i.ibb.co/thjj6nq/straigh-scan.png

https://i.ibb.co/qmyzmXH/2-pass-HDR.png

These are both straight scans with no operator adjustment, enlarged to 200% on 2100x3200 pixel scan so nothing special, I made no changes to the scan other a simple linear increase to bring these shadow areas up to a level slightly above what I think it should be SHOW the DIFFERENCE in NOISE.

The difference is noise is visually obvious. At normal size, most people won't notice, but it looks just cleaner. Like many things in photography maybe it add .5% to image, sometime those .5% all add up, but it is important see this all in context.

I kept these scans purely as example. It is a pity its not properly implemented in these scanners.

[Pere Casals]

In my tests with both a Nikon LS-8000 and Epson 700 years ago, multi-pass exposures usually resulted in poor performance and little to no gain in the shadows, even on the much superior Nikon.

V700 does not include Multi-Exposure in the bundled software, so obviously you could see no gain with multi-pass.




That tallies with 'multi-exposure' not really adjusting the exposure & instead adjusting the gain.

The V700 and V750 are true multi-exposure able (extrange that you are not aware), the V700 requires a Silverfast software upgrade to have it, it requires the SE Plus, instead the SE version.

I'd say that you are not aware of the great benefits of ME in the V700 with velvia very deep shadows, when I can I'll prepare a side by side: the difference is huge !!!

Single ME problem is that it is much slower.

Multi-Exposure in Silverfast is an exposure time change, not a gain change. Don't think that changing the exposure time is high technology, exposure time it's just a parameter you set in the firmware, a single value.


While an scanner can make the two exposures for a row before advancing the Epsons don't support that, so two passes are required, if film curls from illumination heat then the second pass may not match, but if one makes flat negatives then there is no problem, also the anr glass on the negatives prevents that, or ideally we can wet mount. IMHO wet mounting is always good/necessary with curled film, some find that wet mounting was much better because they pruduce curled negatives that are not in focus.

[Pere Casals]

the DIFFERENCE in NOISE.

Just the same I found.

[interneg]

Ted - that looks like a fairly major loss of edge definition - which is quite likely from the multiple scan registration issues discussed upthread. Colour noise is not difficult to clean up - however it looks like your multiple scans have been severely noise reduced by the methods of blending & alignment to the point that it's taken the edge off definition

Pere - unless you can point to exactly where Silverfast is altering the motor speeds in multiple exposure or increasing the output of the backlight, it's creating multiple exposures by manipulating the gain of the sensor - if it could boost the backlight or lengthen the actual exposure, the hardware dmax would be higher - and Silverfast explicitly does not make this claim.

[Ted Baker]

Ted - that looks like a fairly major loss of edge definition - which is quite likely from the multiple scan registration issues discussed upthread. Colour noise is not difficult to clean up - however it looks like your multiple scans have been severely noise reduced by the methods of blending & alignment to the point that it's taken the edge off definition

Pere - unless you can point to exactly where Silverfast is altering the motor speeds in multiple exposure or increasing the output of the backlight, it's creating multiple exposures by manipulating the gain of the sensor - if it could boost the backlight or lengthen the actual exposure, the hardware dmax would be higher - and Silverfast explicitly does not make this claim.

These are stepper motors that don't have a speed, like a normal motor, they step, when told to step. When the scan line is completed they step to the next position. In order to position the sensor correctly just for one scan line, the stepper motor needs to be fairly accurate, because just to do one line requires the stepper motor to move one than once. So for one scan line you need I think 3 passes or 6 I am not exactly sure. For two passes everything doubles.

I am just offering this an example, for 35mm kodachrome on my scanner (epson 4990) for my needs with that film I would always now use multiscan. For other films NO.

BTW light output has nothing to do with. If you reduced the light output by 1 stop, the scan would just take longer (two times), with very little change because noise does not change that much with those kind of exposure time changes. Easy to show if have an 1 stop ND handy. If you have a brighter light that does not improve you dmax at all because you cannot overexpose a CCD as once the wells are full they are full! You are limited by how clear the film is. A brighter light will allow you to go faster IF the rest of the hardware is up to it.

[Alan Klein]

The proof is in the pudding. I've asked people to show their multi scan shots. (edited: None) have that showed any difference. I also checked a couple people posted with both the straight scan and the "improved" multi scan. WHen I used shadow adjustments on the normal scan, I was able to see the same results in the shadow areas as they got with so-called special scan software. Frankly, I think multi scan or other changes to the scan operation to improve scan results are a lot of hype to sell after market scan software. Basically, these changes are just post processing edits applied during the scan.

[Alan Klein]

V700 does not include Multi-Exposure in the bundled software, so obviously you could see no gain with multi-pass.





The V700 and V750 are true multi-exposure able (extrange that you are not aware), the V700 requires a Silverfast software upgrade to have it, it requires the SE Plus, instead the SE version.

I'd say that you are not aware of the great benefits of ME in the V700 with velvia very deep shadows, when I can I'll prepare a side by side: the difference is huge !!!

Single ME problem is that it is much slower.

Multi-Exposure in Silverfast is an exposure time change, not a gain change. Don't think that changing the exposure time is high technology, exposure time it's just a parameter you set in the firmware, a single value.


While an scanner can make the two exposures for a row before advancing the Epsons don't support that, so two passes are required, if film curls from illumination heat then the second pass may not match, but if one makes flat negatives then there is no problem, also the anr glass on the negatives prevents that, or ideally we can wet mount. IMHO wet mounting is always good/necessary with curled film, some find that wet mounting was much better because they pruduce curled negatives that are not in focus.
Let's assume you can get more shadow details slowing down one of two scans. How do you line up the two passes afterwards if the stepper doesn;t allow that?

[interneg]

Pere, if what you are saying is what Silverfast does, they'd be able to claim a DMax improvement. They don't, so the exposure change has to come from elsewhere in the system - boosting the sensor gain is much more readily done than anything else.

As for boosting the light intensity, if you have a really dense/ underexposed transparency, the wells are not going to fill as fully as they could if more light was pumped through the object, which is going to give cleaner results than cranking up the gain.

[Ted Baker]

Dmax has nothing to do with how bright the light or how sensitive your sensor is, you can have a very bright light and very very sensitive sensor but poor Dmax.

[Peter De Smidt]

"Nothing"? Don't you have to have enough light moving through the film to give a good reading on the sensor? Sure, you can prolong exposure to get more light, but don't sensors have reciprocity? So if one has dense film, doesn't having more light, up to a certain point, allow better scans through the dense areas? I expect that sensors have set dynamic ranges, and as long as one has enough light so that the darkest area fall within the sensors levels of good tonal differentiation, i.e. the straightish line area of it's response, then it's maximum dynamic range is set by the sensor characteristics. But if the film, when illuminated enough to get good highlight seperation, requires a bigger range than the sensor can handle, then isn't the only choices to either write off that are of the film, or take another scan, ideally without moving the system, and changing exposure to bring the trouble area withing the sensor's range, and then combining the two scans?

With digital camera scanners, at least, doing this leads to better separation throughout the tonal scale of very high dynamic range film, as tested with a Stouffer test scale.

Changing topics, but I'm in the process of finishing a new scanner optimized for 8x10s. If I get off the darned forum, I should be able to complete it today.

[Ted Baker]

"Nothing"? Don't you have to have enough light moving through the film to give a good reading on the sensor? Sure, you can prolong exposure to get more light, but don't sensors have reciprocity?

Sure you have to have an exposure time that allows for enough photons to be collected to create a measurable voltage, that will register on the first stage of the A/D convertor. The longer you wait to collect those photons, the more you are effected by electrical noise etc. Loss of charge while you wait is probably a factor. So your error rate just goes up.

But if that exposure time is reasonable then yes in this context then the answer is YES NOTHING! :)

[Pere Casals]

Pere, if what you are saying is what Silverfast does, they'd be able to claim a DMax improvement.

And they do !! from 3.1D to almost 3.4D

193224

https://www.silverfast.com/highlights/multi-exposure/es.html


The 3.4D (of the V700 with multiexposure) is way better than what the LS-9000 does without multiexposure...


and the LS-9000 with multiexposure reaches supreme levels !!!

[interneg]

And they do !! from 3.1D to almost 3.4D

193224

https://www.silverfast.com/highlights/multi-exposure/es.html


The 3.4D (of the V700 with multiexposure) is way better than what the LS-9000 does without multiexposure...


and the LS-9000 with multiexposure reaches supreme levels !!!

Effectively they're claiming a signal to noise improvement of about 1 stop - at the cost of apparently worse sharpness (you can even see this in their demonstration image - much like Ted's). Having thought about this over the last few hours, I'm still very strongly suspecting that there's no alterations to the drive mechanism or light, but two levels of image gain are blended on some specific basis & at least one is subjected to fairly vigorous de-noising or even outright blurring. In other words, something quite possibly not a million miles off the frequency separation techniques used in retouching. And it's all fairly pointless in an era when digital sensors can be used as scanners with potentially 13-14 clean stops of dynamic range in a single shot - and more to the point, how many badly underexposed transparencies are you trying to dig detail out of?

[Alan Klein]

These are Velvia 50's single scanned with V600 and adjusted in Elements or LR afterwards. The shadows are fine. On some of them I even let them stay darker. I liked the contrast. But, I could have lightened them up even more with the shadow sliders. These were exposed correctly. I don't "save" under or overexposed shots to try to correct afterwards. Regarding scan stops, a single pass picks up the full range usually between 0-200.

The other issue is that shadows usually contain less important info. Your eye goes to the lighter parts of the image and ignores shadow areas. Showing more detail there actually reduces the impact of most shots. Also, reducing contrast takes away the eye "candy".
https://www.flickr.com/search/?sort=date-taken-desc&safe_search=1&tags=velvia&user_id=55760757%40N05&view_all=1

[Pere Casals]

Effectively they're claiming a signal to noise improvement of about 1 stop - at the cost of apparently worse sharpness.

Interneg, I see no sharpness loss. ME is often used for very dark slides that are absorbing near all light and they can reheat, sometimes the film can move from thermal deformation specially (I guess) with curled film, if it is the case then we can wet mount or simply use the ANR glass holders.

just an opinion, I've not checked it well, but I saw no sharpness loss when I tested ME with an USAF 1951 glass slide.



at least one is subjected to fairly vigorous de-noising or even outright blurring.

Not IMHO, the V series are very "simple" animals, they have little image enhancing work inside, you have to always optimize (sharpen) image in Ps, delivering always a very raw image, while pro scanners optimize the image saving workforce needs, I prefer having a raw image and to optimize in Ps, but it requires an effort.






And it's all fairly pointless in an era when digital sensors can be used as scanners with potentially 13-14 clean stops of dynamic range in a single shot

No 13, 14 "clean" stops at all !!! :)

Those spots in the dark areas where first 10 bits are zero have only 4 bits to describe tonality !! if you rise those shadows you will find insane banding.




and more to the point, how many badly underexposed transparencies are you trying to dig detail out of?

The Epson V850 is not the right tool for that, of course a Pro scanner would do a way better job for that. The V700-850 does an excellent job with all BW and Color Negative film, you will never reach the scanner density limitations. Also it is a very good performer with challenging Velvia/Provia if Multiexposure is used...

...but in extreme deep velvia shadows you reach the V850 limits, anyway probably you won't make much from a badly underexposed slide, you may "save the shot" if it's an important scene, but probably it won't rise big money at Christie's/Sotheby's (irony)

________


One interesting thing (I guess) when scanning very dark velvia slides with a V850 (for 35mm, MF or sheets on glass) is masking the other holes in the film holder. The V850 takes 4 35mm strips or 2 MF strips, those holes take all the illumination power they can to produce flare, if you have an extremly dense slide then you may notice that flare.

The Cheap Nikon LS-5000 reaches 4.24D (Silverfast guys say) but the expensive LS-9000 it reaches 4.0D "only", it is interesting to explore why.

[Ted Baker]

I'm still very strongly suspecting that there's no alterations to the drive mechanism or light, but two levels of image gain are blended on some specific basis & at least one is subjected to fairly vigorous de-noising or even outright blurring.

Next we will be discussing if the moon landing happened... :) In the absence of the actual chip design, there is plenty of documentation that says it has this feature, it can be even found in the source code that epson released for a driver for these scanners some time ago. https://github.com/hean01/iscan/blob/iscan-2.29.1-5/backend/epkowa.c

Plus it simply works and behaves as expected...

Adding an adjustable amplifier is no more difficult than adjusting the sensor timing. This type of technology uses a digital clock, which at the beginning closes a circuit for each light sensitive area. This brings the voltage to zero and removes the charge that has accumulated in the sensor. After a certain number of ticks another circuit is created for each sensor area and the charge that has accumulated in response to the number of the photons, is then measured as a voltage. Waiting more ticks of the clock is not hard. The light and stepper motor work exactly the same. Adding all the circuitry to store and compare two values in the same pass definitely adds a lot more circuits.

It my understanding the flextights can also vary the timing, as can the coolscans. Interesting the flextight take a different approach to this problem an add active cooling to the sensor to reduce the noise. (i.e. better scan first time around)

Comparing one scanners (ancient) feature set against another is sort of meaningless IMHO, it only useful to helping you maximise the use of your existing scanner, or building your own.

FWIW the no expense spared arriscan uses two passes.

[Jim Andrada]

I think the IQsmart 3 and maybe Eversmart Supreme have a chip cooler.

I have to manage with a mere IQsmart 2 - boo hoo. And I don't try to scan anything more than once - it's my job to get the exposure right in the first place and I take my job seriously (well, most of the time:()

[shoshin]

And they do !! from 3.1D to almost 3.4D

193224

https://www.silverfast.com/highlights/multi-exposure/es.html



The 3.4D (of the V700 with multiexposure) is way better than what the LS-9000 does without multiexposure...


and the LS-9000 with multiexposure reaches supreme levels !!!



Usual Silverfastpropaganda. -.-

I did a few hundred testcans with the Epson V750 (Multi-Exposure) with the Coolscan 8000 (Multiexposure and Multipass) via Silverfast and Vuescan. To my mind the time required for multi-exposure and/or multipass isn't worth the increase in quality at all and there is no difference in quality between silverfast and vuescan.

[Pere Casals]

Usual Silverfastpropaganda. -.-

To my mind the time required for multi-exposure and/or multipass isn't worth the increase in quality at all and there is no difference in quality between silverfast and vuescan.

I don't agree at all, ME improvement it depends on the job... for regular BW and CN shots there is absolutely no difference, because those situations are no challenge for the V850 scanner, regarding film density.

For Velvia/Provia slides Multi-Exposure makes a huge difference in the deep shadows having high density, there is an amazing improvement, anyway even with ME the V850 won't be able to reach the most challenging velvia shadows.

Anyway those velvia ultra deep shadows may not have much pictorial interest most of the times.

And yes... ME is time consuming, but drum scanning is more time consuming...





I did a few hundred testcans with the Epson V750 (Multi-Exposure) with the Coolscan 8000 (Multiexposure and Multipass)

With ME you see a difference with the highest densities, but you have to mask the other the holes in the scanning area.

The V750 illumination covers four 35mm strips, if you have a 3.2D slide in one of the holder holes but you have nothing (or low density slides) in the other three holes of the row.. those clear holes are taking as much flare as they can.

The V750 lens is multicoated, but 3 open holes taking 5000 times more lights than the one than comes from the test slide is too much flare for any optical system.

I point that because tests are useful when they are well done. Did you mask the other holes when testing ME with high densities ?


To me, the information Silverfast gives about ME is completely fair, and in the conservative side.

[John Berry]

When scanning my mentors daughter setting in front of the TV with the moon landing on the screen, it wasn't easy for my 4990 to get in one scan. Here's a novel concept, scan like you would shoot HDR. I made one scan for the room and another just for the screen. Processed as you would any HDR. Worked for me