DPR Forum

Welcome to the Friendly Aisles!
DPRF is a photography forum with people from all over the world freely sharing their knowledge and love of photography. Everybody is welcome, from beginners to the experienced professional. Whether it is Medium Format, fullframe, APS-C, MFT or smaller formats. Digital or film. DPRF is a forum for everybody and for every format.
Enjoy this modern, easy to use software. Look also at our Reviews & Gallery!

Three new Nikon film scanners

>Posted by Thomas Tuszinski (Ttu) on Tuesday, November 11, 2003 - 4:48 pm: > >David, >[Deleted some text here, DAW] > >Other thing is, if you have another scanner attached, the TWAIN >interfaces may screw each other occasionally and at apparently random >intervals. > >regards Thomas

Well, I can't resist. How many folks know what "TWAIN" stands for? Don't answer if you already know.

DAW
 
Hi Mike,

It's perfectly possible to have a light source that looks white (or red, green or blue) to the human eye, but has certain spikes or notches in its spectral distribution. There is no way to fill in the resulting gaps using filters, because they can only absorb not add.

Similarly, the dyes in film may absorb unevenly across the relevant part of the spectrum, but the colors still look correct under spectrally even white light because the distribution of the peaks and dips still balances out to the right colour as far as the human observer is concerned. However, as soon as you combine the two irregular spectra by shining the light through the film, unless the spikes in the light source's spectrum line up with the film emulsion's absorbtion peaks you may end up colour values that don't follow what your eye observes under a broad-spectrum illuminant.

The best illuminants are actually incandescent lights, because they use the heating of a metal element to generate light, and this gives a very continuous spectrum (the element is what physicists describe as a "black body radiator"). If you consider that Kodachrome was formulated in the days when the only likely use for them would be projection, you can see why the using dyes with even spectral characteristics was not a top priority.

The same sort of issues arise when you are viewing prints (particularly inkjet prints generated using older pigment inks). Viewing lights are specified not only in terms of their colour temperature, but also the Colour Rendering Index (CRI) which is an indicator of spectral continuity. It's not easy to obtain a high CRI using a discharge tube, because these devices tend to emit strong spectral peaks corresponding to the resonant peaks of the gas in the tube. An extreme case is the orange sodium street light, which has a spectrum consisting of about three spikes, with nothing in between.

Make sense?

Regarding the current round of scanner releases, I think you're dead right, I think we will see a new product from Canon in the next six months at the outside. It's well worth waiting a bit to see what happens, if you have a unit that you are reasonably happy with in the meanwhile. I only took the plunge with the 5400 as I was really at the end of my tether with the LS-30's severe lack of bit depth.

-= mike =-
 
I'd say at least those who used photoshop or similar and clicked "select source" - "aquire"... etc.

Thanks David, I'll take the cue to stop posting. TT
 
Mike Nunan,

O, okay, I understand what you meant now. I got confused with color temperature (thought that this is what you were trying to say when you mentioned spectrum variation. My bad).

But then it's not really applicable in case with LEDs. LED is not a discharge l&, there is no gas burning inside it.

Also, technically, CRI is of no use here. It would have been applicable if we were talking about "white LEDs". But since Nikon uses three LEDs in red, green and blue (LEDs are anyway normally monochromatic, as far as their color is concerned), CRI of the entire system depends only on the combination of all three LEDs. So, any speculation about actual value of CRI would be a bad idea without knowing the characteristics of LEDs used. Also, since we don't actually know the color temperature of these, color-correcting filters of some sort will probably be required anyway, unless the whole system is designed to perceive different color-temperature as "daylight" (it's possible that it's done on software level as well, but I wouldn't know for sure).

Unfortunately, either way, I can only speculate on what's going on inside that scanner, since I didn't design it, nor do I know what Nikon's competitors do (e.g. Minolta or Canon). But the point is still the same - CRI of three LEDs as a whole is hard to determine without knowing the actual specs.

Let's say, we take Lumileds's new "white" LEDs. Their color temperature is in 3200K range, their CRI is 85. BUT, these are "white" LEDs!

I just searched on the Internet and found some papers with documented experiments in creating "white" LEDs by combining red, green and blue. CRIs achieved were in the range of 20 to 70! Now, that's a huge range, so it gives us little info to go about.

I also found one really interesting document.

Apparently there is some issue with CRI itself. Very often incandescent lights are thought to have CRI of 98-100. The problem though is that the whole Color Rendering Index system is basically built around this (as far as I understood the paper) [ here is the link - http://www.lrc.rpi.edu/programs/solidstate/pdf/CRIForLED.pdf ]

Which means, they take some "standard source" and compare to it in order to determine CRI. Now, if you read the "discussions" section in that paper, they make a very interesting and valid point - "CRI of a light source is a measure of a degree to which perceived object colors illuminated by a test light source conform to those of the same objects illuminated by a standard source such as an incandescent light source".

There are two implications here. First one is that technically there is no reason why we should care about CRI of light sources inside of a digital imaging system (and scanner is such a system), because there is no human eye involved! All that matters is the final result presented to our human eyes. So, as long as the scanner records all the colors in a way that can be later processed and produced (e.g. printed on paper) with colors matching these we see on the slide (and negative, although in this case it's harder to know what is it that we actually see on negative) - life is good.

Second implication is that considering that CRI is a comparative index, it means that it's actually fairly subjective.

For instance, since I'm considering buying a new light table, CRI is important, but less so than the color temperature in my opinion (I always preferred the 5400K bulbs over 5000K bulbs anyway. Just looks better, subjectively speaking). And when comparing two options - 5000K/CRI94 and 5400K/CRI91, I would probably choose the latter.

Now, I suspect that CRI is more important when it comes down to film exposure, but that's already a different issue alltogether and not really relevant in the scope of this thread.

And yeah, let's wait and see if Canon will release a new film scanner as well.

I'm repeating myself again, but you didn't say how well does that manual focus dial works on Elite 5400?

Mike.
 
Claus,

Yup, you're absolutely right, Nikons indeed always used LEDs in their film scanners, I simply never looked at their specs before, since wasn't really interested in getting my own film scanner.

Mike.
 
Hi Mike,

Sorry for the delay in getting back to this, I've just had a couple of busy days. We're certainly giving this one the full treatment! No harm in that, it's an interesting interlude and touches on some important issues. Taking things in order...

LEDs are not discharge tubes, but being semiconductors there will be electrons jumping around between atoms of the various different elements in the substrate, and presumably these interactions release the photons. As an electron moves from one atom to another, it will shift energy levels with light being released when the electron has moved to a level of lower potential, with the frequency of the light being related to the energy difference. Because there are only a handful of energy levels to choose from (being determine by the particular elements that have been combined into the substrate of the device) the spectrum of the emitted light will be discontinuous. (It's over 15 years since I studied any of this stuff, and even then it wasn't a core subject for me, so I'm sure a physicist could come along and correct the details and terminology, but I'm fairly sure I'm on the right track.)

I don't want to get hung up on CRI, I only brought it up to help with the explaination. I'm sure it's not the last word in light source metrics, and is specific to white light as you point out. Interestingly, I took a look at that paper and while they certainly make a good argument against CRI, they also comment that CRI is a measure of how closely a given light source correlates to a standard incandescent source. Given that we are talking about scanning film here, and that film has been developed to be projected using incandescent bulbs in projectors and enlarger heads, CRI may actually be more relevant to scanner light sources than it is to viewing lights.

Whatever, the underlying point remains the same: any mismatch between the spectrum of the source, the absorbtion characteristics of the film dyes and indeed the response of the sensor will potentially result in pixel values being recorded that do not correctly relate to the observations of a human being viewing (for ex&le) a slide via projection or against a true daylight backlight. You say that LEDs are monochromatic, but if you're using the physicist's definition of that word then you mean to say that they kick out just ONE frequency, so the combined spectrum would consist of three spikes and nothing else whatsoever. Used in more general terms, you can have a light source that is apparently pure green, but has a spectrum shaped like a mound with the peak centred somewhere near the frequency for true monochromatic green. IIRC, the CIE defines the human visual system's response to colour in terms of three overlapping curves of this type (this is distinct from their definition of an actual colour in terms of tristimulus values, which ARE based upon true monochromatic components).

The ideal situation for a scanner would be to have an illuminant that exactly matched this combined spectrum, with sensor elements having an individual response matching that of the three component spectra. (Well, it would be ideal subject to the accuracy of the match between the CIE's three spectra and your own eyes' response, but it seems ambitious to think that we could do a better job of deciding what those spectra should be and hopefully you get the point -- in this situation the scanner "sees" exactly what you do.) Back in the real world, there is no chance of achieving this. Incandescent sources generate too much heat to be used in a scanner, so we rely upon discharge tubes or LEDs. Then the sensors add their own response to the mix. Profiling can at least compensate for the resulting value shifts, but it's easy to see why accurate colour is hard to obtain unless you profile for each emulsion type. Also, for direct digital capture, different camera profiles are required for different illuminants.


Manual focus works ok on the 5400, but I must admit I don't often find occasion to use it. If I'm being critical, I tend to set the AF point to the most important part of the subject and this seems to produce nice sharp scans. There is not much difference between setting manual focus via the software or with the control knob, although if you're after a certain amount of defocus (which can help alleviate the grain problems to a certain extent) it's marginally more convenient to use the knob. Either way, you will definitely want to have the focus feedback bar visible, since otherwise you'll have no idea what's going on at the sensor. For sharpest focus, you just have to twiddle the knob to get the longest indicator bar you can manage, which is a function that the AF can achieve more quickly and just as accurately. The bar is really the most useful part of the whole thing, because if you can't get a high reading then you know you've picked a part of the frame with insufficient contrast to allow accurate focusing, meaning you should try moving the focus point around a bit until you've found somewhere better.

HTH

-= mike =-
 
Mike Nunan,

I think I understand what you're saying about LEDs. Now it would be actually interesting to hear how things have improved in last decade in someone closely involved in this industry.

Technically I think we should just wait and see when Nikon releases them and few people test them. It would be definitely a better measure of its color-scanning accuracy and overall performance qualities.

Mike.
 
Yeah, it would be nice if a techie from Nikon could drop by the forum to comment, but I guess the chances of that are slim! It won't stop me from trying out the 9000 ED, and maybe even replacing my if the 9000 can do a good enough job of the 35mm K-chromes. I don't underestimate Nikon's ability to produce good scanners, they have been at the head of the field when it comes to sharpness in the past, and have supplied important features such as the ability to scan a whole uncut roll (with the appropriate adapter for the 4000 ED).
 
Yepp, as far as their engineering approach is concerned they're certainly not bad. Canon is more "pushy" with their marketing, whereis Nikon usually tries to please by offering something that you might find useful. Not that I will switch to Nikon cameras, but they're nicely engineered.

Mike.
 
Posted by Mike L (Mikel) on Wednesday, November 12, 2003 - 2:41 am:

Mike Nunan,

Well these "diffusers", or whatever they're, can technically correct a problem with spectrum variation in a way similar to color-correcting filters. Unless I'm missing something, I think that should work.

--- snip ---

I feel strongly that there is no diffuser which can smooth our any "spikey" spectrum produced by any LED. On the other hand, if the individual LED's are precisely matched to the RGB sensitivities of the film emulsion, maybe this sort of spectrum isn't a problem. Some solid state LED's do produce a pretty narrow light spectrum, centered around their primary output color.

DAW
 
Back
Top