Minolta Elite 5400 - Exposure control

Discussion in 'Digital Darkroom' started by og, Oct 13, 2003.

  1. og


    I have just received my new scanner Minolta Elite 5400, and I am
    currently testing it with both Minolta Dimage 1.1.1 and Vuescan 7.6.65
    (note: I use mainly Color Negatives).

    The first thing I noticed is that Minolta Dimage does have an
    'Exposure' Tab where you can control exposure for each channel (Red,
    Green, Blue). Vuescan does not have this R,G,B equivalent on Exposure.

    1 - Does it really mean that I can control the analog part of the scan

    2 - Where is the control : Light (intensity of R,G,B sources) / Analog
    Gain on the CCD ... ?

    3 - In 16 bits mode, and for scanning color negatives, do you think
    that this Analog control will really allow me to get more useful
    information to work from ? (ie. otherwise I would need to remove the
    orange mask in an editing software, which would reduce the useful
    information from 16 to ?(maybe 14) bits in the Red channel)

    Any opinion ?

    Thank you, Olivier
  2. In my opinion the so called manual exposure control
    of Minolta scanner software is really just software manipulation
    of the image's histogram. I have an older Minolta scanner that runs
    the same software. It is very easy to verify that the exposure
    adjustment, regardless a prescan is initiated or not, only shifts
    the histogram graphs. I could achieve the same shifts by manual
    level adjustment in Photoshop. I have since discovered this never
    messed with the exposure adjustment any more.
  3. 1) Maybe

    2) Probably the exposure time.

    3) IF it changes the exposure times, it's probably already doing this in color neg mode.

    Not sure about the Minolta software, but Vuescan on my Polaroid 120 absolutely does alter exposure times--it does it automatically based on the prescan, or you can check "lock exposure" and a box appears where you can put a number in yourself. The problem is it only offers a single value, not separate RGB control. The only time the channels are different is in color neg mode where it automatically gives the green channel a fixed (relative to the red) 2.5x exposure and the blue channel 3.5x exposure.

    I find this immensely useful for B&W negs with a long range of tones--say TMY pushed to 6400. The red channel gets all the shadow detail, the blue channel saves the highlight detail. The sad part is that 3.5x isn't quite enough to get everything on the film, and the scan-to-scan registration on the Polaroid is so vile that combining separate scans isn't an option.

    What's sadder is that I have some negs that I could get a much better scan off of by using a channel difference of more than 3.5x, but Mr. Hamrick refuses to implement either separate RGB controls or at least a fixed 6x setting.

    I even offered to pay him to implement this change; I got back a flip remark about "I have a minimum fee of $10,000 for custom programming". It'd be worth at least a couple hundred bucks to me, maybe more, to get this kind of functionality, and I can't see how it would be rocket science to write. Maybe if some other customers with thick negs asked for the same feature he'd pay more attention...
  4. Olivier,
    I don't actually have Elite 5400, but I happen to have Minolta Multi Pro which has very similar hardware and the same software. You indeed control hardware gain in the exposure in each channel separately, which actually allows a lot of flexibility in the work flow (especially with negatives, since it is quite tricky to automatically produce good negative scan). Analog gain for all the channels corresponds to a Master gain on the exposure control tab. A lot of people use manual exposure settings to scan the negative as positive using corresponding gains in each (R,G,B) channel to compensate for orange mask and then invert in photoshop to get all possible information out of the negative. The actual workflow is described in this document: http://home.planet.nl/~goede190/scanhancer/downloads/scanhancer_manual.pdf
    You will need to use the same same workflow with the exception that you don't need scanhancer and you probably will have to keep the master control gain at 0;
    As to your third question you will definitely have more information pulled out the slide/transparency by manual workflow, plus you can compensate for the color cast (if you happen to have one).

    Enjoy your scanner
  5. og


    Thanks for the link, Anton.

    My findings with the 'Exposure' control : There is no modification of the color of the light, but exposure time is modified. Therefore, I believe that RGB exposure settings do control individual gain of the R, G, B channels. (1=Yes / 2=in Gain, linked with exposure time).

    I also checked the differences between 'Negative' and 'Positive' in Minolta Dimage:

    - 'Negative' settings clip shadows & highlights in the histograms, and doesn't provide control or feedback about it. I don't like that too much...

    - When I stop exposure time for a scanned Color Neg at 900dpi : as Negative = 40s / as Positive (set for dark exposure, without clipping) = 30s / as Positive (set for bright exposure, without clipping) = 1mn13s.

    A color negative has a Density Range of about 2.0 (1:100 from dark to white) whereas a color positive can have a Density Range of up to 4.0 (1:10000). The Elite 5400 has an 'official' Dynamic Range of 3.8, with a linear 16 bits A/D converter. It would means that it can see from 1:6310 and use 65536 values to store this. If I expose my Neg with a (very) dark exposure (let's say from 1 to 100), I take advantage of 1039 values. If I expose it with a (very) bright exposure (from 631 to 6310), I take advantage of 58982 values...

    The 'Negative' settings are probably very good (minor clipping, adequate exposure to get enough information...), but I don't know exactly 'How' good... (maybe further tests, but those are longer).

    On Vuescan side: with Default settings for negatives I got an exposure of 35s. You can lock exposure and manually change it, but not fine tune each R, G, B channels. And if you want to keep the red information, the Green and Blue will be under-exposed (=less value to store them).

    So, if I want the best quality (exposure wise only) for a color negative Scan with the Elite 5400, I need to scan it with Minolta Dimage as a 'Positive' and expose it for a bright exposure WITHOUT clipping highlights of the negatives (ie shadows in the picture).
  6. Perfectly correct reasoning, Olivier.
    I don't actually see the reason why you would neet to control the color or the brightness of each channel. Existing R,G,B channels are in a sense orthogonal, so changing either the color of the channel will just interfere with the info from other channels. Doesn't do you any good. Brightness is not an issue either, since using different exposure values should give you the same effect (just some statistical reasoning).

    My experience confirms that I can produce much more superior negative scans using the described manual workflow (through positive scanning) comparing to automatically scanned negatives produced by Minolta Software. I really don't use Vuescan much with this scanner, since I personally find Dimage sw scans are better and ICE (at least on Multi Pro) is just superb.


  7. og


    In order to bring more 'visual effects', here is the 'exposure' tab in Minolta Dimage Scan for: Negative / Positive(Dark) / Positive(Bright) (from left to right - for the same color negative). For 'Negative', you can see that the histograms are cropped in both shadows and highlights. No manual modification (exposure/image correction). For 'Positive(Dark)', I modified sliders in order to get the lowest exposure possible without cropping histograms. For 'Positive(Bright)', I modified sliders in order to get the highest exposure without cropping histograms. From my theory, 'Positive(Dark)' and 'Negative' Scans should have less usefull data than 'Positive(Bright)'... I'll try to find a way to check this! By the way, why am I doing this? Well, I want to get the best Scans from my scanner, and find the proper way to get this (I use color negatives + black&white negatives - no positives). This may be usefull to others as well...
  8. og


    Anton, my point here is that I want to use the analog part of the scanner as much as possible in order to get the best source for the digital output. It's very close to what Michael Reichmann is doing here with his digital camera. I am modifying separately the Red (and G & B) channels mainly to compensate for the orange mask of the negative (=optimize exposition of the CCD for my final image).
  9. Hey guys, this is an excellent thread - I'm getting a lot of valuable ideas since most of my film consists of 35mm negs!

    I have Minolta's 'budget' scanner (Dimage Scan Dual III), so can I presume that most of this (if not all) is applicable?
  10. og


    OK... Maybe I found a way to test my theory about usefull data from 'Dark' vs 'Bright' exposure: I will count the number of colors in the Scans!

    I should get less values for Darker exposition. Moreover, posterization would come from the fact that I have less tonal levels in the 16 bits image (=less colors overall).

    Here are the results: Negative=1.019.792 / Positive(Bright)=1.028.697 / Positive(Dark)=1.014.899.

    Vuescan gives the same kind of results (Negative(no correction)=1.021.728 / Negative(auto-levels)=1.011.990 / Negative(long-pass + auto-levels)=1.015.322 / Positive=1.027.716).

    Hmm... It works, but there is only 1% difference between 'Negative' settings and 'Positive(Bright)', and it probably comes from the histograms cropping. I didn't notice any significant difference in noise either.

    Not exactly what I expected. So where do I stand now ?

    For my VERY important & difficult Scans I think I will apply the Minolta Dimage Scan "Positive + Bright exposure with Red/Green/Blue corrections" to avoid cropping. I am lucky: I don't have any important Scan : )

    For my usual hign-quality Scans, I will use the 'Negative' settings, either from Minolta Dimage Scan or Vuescan, because it is easier and less time consuming for equivalent results. Both programs clip histograms a bit (Minolta Dimage Scan = shadows & highlights / Vuescan = highlights but about 2x more).

    Your comments are very much welcome.

    Thank you - Olivier


    Phil: thanks for your comments. I think that those findings directly apply to the Scan Dual III as well (and probably to most CCD scanners).
  11. Olivier, I don't think that counting the colors tells you how good your scan is. The scan usually has a lot of noise (even if it's benign), so by counting the colors you end up measuring how much noise you have, not how much density you capture.
  12. og


    Anton, I have to agree that the results (# of colors) probably don't show anything...

    I still think that if you get a longer exposure time without clipping anything, you get a better scan. And that for a Color Negative, you get this longest exposure by tweaking RGB separately + you should try to get a neutral color to reduce digital color correction after that...

    Anton, in the Scanhancer manual, Erik put the histograms rather to the right (bright). Where you doing the same with your own Scans ?
  13. og


    More feedback would help, but here is what I have got:

    - Several persons reported that by using 'Positive' and adjusting Exposition you could get better Scans of Negatives, especially in the Minolta Multi-Pro threads... (cf worklow proposed by Erik de Goederen).

    - If you adjust Exposition, you will have a tendency to center the histogram = increase exposure, when compared to the default 'Negative' setting.

    Absence of clipping + longer exposition + less post-processing + whatever(personal feelings/magic... ;) seems to improve quality of scanning. I guess that in 16 bits, the first two are the most important ones.

    This is what I am trying to do by scanning Color Negatives as 'Positives' and maximizing exposition without clipping highlights.

  14. Oliver, just a simple statistical reasoning that by increasing the exposure time (pushing the histogram as much to the right as possible) you will reduce the variance in the number of photons captured by CCD cell. In common words, this would mean that you will significantly smooth the colors and reduce the amount of noise captured. Perhaps the very same effect we observe on slow films (ISO 100 and less) - more exposure proves to significantly reduce the noise and produce smooth color rendering. The conclusion is - yes, I would push the histogram as much to the right as possible (certainly avoiding its clipping) to improve the scan quality. Perhaps a specialist in CCD technology can tell us if there is a significant non-linear effect when shifting the histogram to the left/right from the center that should be avoided.
  15. og


    Last but not least: I was thinking that you should 'negative'(?) your image (color negative scanned as 'Positive') before applying any gamma correction on it (you would loose data in the shadows).

    Thus it is better to output your scan as 16 bits linear, without Color Profile, 'negative' it in your Editing software and assign there the Color Profile with the right gamma correction.

  16. og


    Here are some results [I corrected the colors to get the same visual aspect + Gamma x0.45 (2.2=>1.0=linear) for Negs (in 16 bits)].

    Please note that for '1350dpi' Minolta's software does a 2700dpi scan and then reduce by 2 in both directions (ie: averaging 4 values). Check here : http://www.photo.net/bboard/q-and-a-fetch-msg?msg_id=006Heo

    as NEGATIVE (4 on the left)

    Single-Pass Multi-Scanning(4x) doesn't seem to improve the scans (I tried Vuescan & Minolta).

    Vuescan at 2700dpi and reducing in both directions by 2 ("2700dpi/2") does a good job to smoothen the noise, and to average data as well.

    Minolta at 1350dpi (=2700 real dpi /2) has a similar result [check noise in the shadow of the blue window, on the right], but deep shadows get clipped when you scan as 'Negative' [in the tree]. Too bad : (

    As POSITIVE (last)

    Minolta at 1350dpi (=2700dpi/2) scanned as a 'Positive' with max exposure (16 bits linear) - then manually negatived(?): there is no clipping, and the noise seem even slightly reduced when compared to Vuescan 2700dpi/2.


    Well, it looks like scanning a color negative with a max Exposure (without clipping), as 'Positive' in Minolta's software, gets you a slightly better scan.

    I remind you that this part of the picture should be a lot darker, and that you would only see 'black' for the tree (almost no noise). So depending on your pictures and how you work, you may sometime take advantage of this method in dark areas + extreme editing.

    There are things more important here: Multi-scanning is useless and 'Up-sampling' Scans is a must (scanning at higher resolution and reducing size in both directions - obviously you can't do more if you want 5400dpi :). The clipping of shadows made by the Minolta software (as 'Negative') may become an issue when doing heavy editing (not so often, and result is not so bad... => should be ok for me for 90%. If not, I will use the Max 'Positive' method). Vuescan can get excellent results here.


    PS: Thanks for all comments so far : )
  17. Maybe I`m a bit off topic, but I wanted to let you know what I get when batch scanning (I try to get the scanning process automated with all boxes unchecked but auto levels, saturation in the auto box in exposure controls.
    Next response will be with an other setting. Would be very interested what you think.
  18. Same setting: Image 2
  19. This time I unchecked all the boxes and just ran a batch scan.
    I have to add that the source is a Kodak VS 100 slide scannes at 2700 dpi ICE enabled.
    What do you think? It looks advantageous to use the auto setting!? Do you agree or did I miss something? Regards Franz
  20. Image 2!
  21. og


    I prefer the images with the Auto-Settings (levels & saturation) than those without any adjustment. However, I found the saturation a bit too high for me; maybe you should try auto-levels only ?<p>

    Some slide Users prefer to go another way : get exactly on the Scan what they see on the slide. To do that, they will profile their scanner and their favorite films (using Vuescan and buying some IT-8 targets). Just to let you know... (this is not for me either: I prefer the larger Dynamic Range of color negatives + work on the Scan to interpret it the very way I want...)<p>


Share This Page