Colour fringing with V750?

[walter23]

Before I go too far with this, I want to put this into some perspective. I've made beautiful prints from scans with this scanner, at pretty big sizes. I'm just nitpicking and trying to optimize things at this point. Now, on to the question:

Do you guys get noticable colour fringes with your V750? I find these if I scan B&W film in colour mode, etc, at high resolution (~4800 DPI).

This is from a 6x4.5 transparency scan, but I've seen it in detailed crops from 4x5 and 35mm, B&W and colour, both with silverfast and epson scan.

Resized:

http://ashphotography.ca/gunther/uploads/pages/unseen/AAAfullscan.jpg

Detail (6400 DPI), but also shows up in 2400 DPI scans:

http://ashphotography.ca/gunther/uploads/pages/unseen/AAAred.jpg

Note the red border along the buildings. This sometimes happens normally with lenses on cameras due to the different wavelengths of light refracting differently through the lens, but in a scanner is this normal or defect? Is there anything I can do to reduce its appearance? Is it an issue of focus? It's not the camera; it appears in scans of B&W negatives if I scan in colour (red fringes on one side of high contrast lines, blue or yellow (can't remember) on the other), and seems to contribute to softness in B&W scans (again, for some perspective, I'm talking at ridiculous enlargement. I've printed 8x10s from 6x4.5 scans that look great).

[Walter Calahan]

The best test is to scan this image on someone else's V750 to campare notes. It might be you've found the limit of this machine. I own a V750 and haven't notice, but then I'm not looking so extremely. Good luck and let us know what you find if you can test on another scanner.

[Doug Fisher]

FWIW, I have experienced this and have had others contact me with the same experience. The has happened scanning on the glass, using the holders, with and without using anti Newton glass, so it can be quite variable/random. Sometimes you can scan a second time without changing anything and not get the problem. Other times, you can just change the orientation of the film ever so slightly and the problem will go away. Others have minimized the problem by flipping the film over and scanning (then re-flipping in Photoshop). I have seen this issue on scans made from a 3170 all the way up to the V series. Some scanners brands may be better than others but this issue isn't limited just to Epson scanners.

Doug Fisher
---
www.BetterScanning.com

[walter23]

Okay, thanks Doug. I did have the impression that it was sort of variable (I was pretty sure that I rescanned the same B&W film twice and had it appear and then disappear). At least I won't worry that it's doing anything unusual.

[Kirk Gittings]

I think that much CA would only be visible in a 40x60 print. As far as the CA goes, how do you determine if it is in the film, i.e. CA from your camera lens, vs. the scanner lens?

For those scanning b&w in RGB then dumping all but the sharpest , least noisy channel, usually the green channel, the CA will not be an issue.

[walter23]

I think that much CA would only be visible in a 40x60 print. As far as the CA goes, how do you determine if it is in the film, i.e. CA from your camera lens, vs. the scanner lens?

That's a good point and one that I've considered, but as mentioned it shows up with variable intensity (with scans of the same slides), and more importantly, in B&W negatives scanned in colour. I've also compared a photograph taken with my digital SLR + macro lens to a scan of approximately the same film dimensions and the CA shows up in the film scan but not the macro photograph.

It's not such a big problem that I regret this scanner, but I do want to optimize my workflow to avoid it if that's possible (or exchange it if its a defect with my scanner, which appears not to be the case). I did wonder whether it was an issue resulting from lack of film flatness or bad focus, which I still haven't ruled out (though Doug's response suggests its not).

[Greg Lockrey]

For those scanning b&w in RGB then dumping all but the sharpest , least noisy channel, usually the green channel, the CA will not be an issue.

I've seen other folks talk about copying B&W in RGB and the "dumping" all but the green channel to get the sharpest image. At what step is this done? IE in the scan set up or later in Photoshop?

[Ted Harris]

Greg, it's done in PS.

[Michael Mutmansky]

CA is an issue all the time. Color Fringing will not be an issue with a B&W scan when done as Kirk describes.

The problem with CA is that it is in the source material, and so it will impact the scan, even if it is for a B&W image. Color Fringing (for want of a better term, that's what Ted and I have come to call it) is a result of (probably) parallax errors in the optical system. The are always predominately in one direction in the scanner and not in the other.

If you want to determine whether it's the scanner or the source, rotate the source and rescan. If the fringing appears to change it's preference (which side the magenta or green favors) then you have a problem with CF, not CA. If it remains the same, it is CA in the source. Most likely, you have a combination of both here, which will confound the issue a little.

This looks to me to be more of a CA problem, because the color fringe appears to be about the same on both axis of the building, and that is not how I see color fringing in the scanner normally happening.


---Michael

[Greg Lockrey]

Thanks Ted. I don't play with B&W that often but I thought that it would be a good thing to know.

[walter23]

Thanks Michael, I'll check that to see whether it's a taking lens CA issue or a scanning issue.

[Gene McCluney]

Fringing on a scanner can be because the transparency or negative is not being held completely rigid. It can also be due to multi-pass scanning which is normally done to get better shadow noise, but most scanners can't reposition the scanning head exactly for each scan, thus fringing.

[Michael Mutmansky]

Gene,

Color fringing is not caused by multipass scanning or film stability problems in my experience. It is related purely to the way the R, G, and B read the data on the original, and the fact that the three sensors cannot occupy the same physical location at the same time, and hence, they have a little bit of parallax error in the system.

I actually find it surprising that the typical scanner software doesn't do a mathematical shift to reduce this error. Maybe they do, but are unable to correct for the problem fully.

The problem you are describing in multipass scanning is a reduction in sharpness as multipass scanning occurs. This is very real, and is easily seen in a close comparison of single pass vs. multipass. However, it normally does not exhibit worse fringing when this happens. In fact, I think it could help obliterate the sharpness of the edges in the scan in a manner that could reduce the appearence of the color fringing.

We normally do a single- vs. multipass example at our workshops to help explain why we don't generally recommend multipass scanning unless there is a significant shadow noise problem in the image.


---Michael

[Chris Strobel]

I thought it was the blue channel, now I'm confused


I've seen other folks talk about copying B&W in RGB and the "dumping" all but the green channel to get the sharpest image. At what step is this done? IE in the scan set up or later in Photoshop?

[Kirk Gittings]

Chris it depends on the scanner actually. Pick the one that is the sharpest and the least noisy. I have seen it both on blue and green depending on the scanner. The red is always the worst.

[neil poulsen]

Others have minimized the problem by flipping the film over and scanning (then re-flipping in Photoshop). I have seen this issue on scans made from a 3170 all the way up to the V series.

Do you scan with the emulsion towards or away from the sensor? Are you suggesting that scanning with emulsion towards the scanner or away from the scanner will aleviate this problem?

[walter23]

Gene,

Color fringing is not caused by multipass scanning or film stability problems in my experience. It is related purely to the way the R, G, and B read the data on the original, and the fact that the three sensors cannot occupy the same physical location at the same time, and hence, they have a little bit of parallax error in the system.


I looked carefully at a B&W negative scan (scanned in colour transparency mode) and I would have to agree with this assessment. If you go to the channels list in photoshop and turn on single layers at a time you can see that the red channel is offset a bit from the green, and the green from the blue. It actually looked pretty uniform so I'd imagine if you cared enough about it you could write a script to perform an alignment in photoshop (e.g. with a slider to move the layers around by a couple of pixels, or just choose a set pixel value based on your measured distances). It looks like maybe a 2 to 4 pixel misalignment at 6400 DPI. It probably wouldn't be worth the effort for most printing efforts though.

Taking a single channel (any one) seems to give a bit more definition just by eliminating these other offset colour channels. But not much.

[walter23]

Here's an animation from a maximum resolution 6400 DPI scan. First frame is an RGB scan of the negative (highly saturated for effect, but only for the sake of this composite frame - the individual channels are from the original non-saturated scan), the next three are the R, G, and B channels. It's not a big deal at 2400 DPI or so I'd guess. I think the red is in fact the softest, the other two (blue, green) are about equal. They are in fact a bit offset though; interesting to know. Clever software should be able to correct this if it matters (which it probably doesn't).

http://ashphotography.ca/gunther/uploads/pages/unseen/threechannels.gif

This was scanned with silverfast in 48 bit colour mode without any sharpening applied. I think based on this that the best B&W scanning method would be to scan in colour with no sharpening applied, grab a channel, and then sharpen that.

For some context, the scan was from this (ugly test) negative (taken from a fraction, maybe 1/5th the area, of the little red frame):
http://ashphotography.ca/gunther/uploads/pages/photoaday/lawnchairbrooding1IMG_0735.jpg

[Ted Harris]

To follow on with Michael's post. When he and I did our first series of tests on a bunch of different scanners (some two years ago) we tested color fringing and it is a fact-of-life aross the board with all consumer scanners, some are a bit better than others but it is always there. In fact even the high end scanners exhibit a small amount of color fringing, but very very small compared to the consumer scanners.

[Kirk Gittings]

Nice demo Walter. I would also apply a steep curve (makes noise stand out) to the green and blue channel examples and see which has the least noise. You won't need to retest this scanner for awhile as they don't change much in this regard as they age.

[walter23]

Hey, good idea. I'll do that.

[Ben R]

Can I understand that it is better to scan as RBG eventhough you will have misalignment of the channels, because you can then use only one channel which will offset any of the misalignment problems. Would it not be better to scan as B&W in the first place or is the RGB channel method using 1 of the channels still better? Do you not loose any of the tonality by doing that?

[Michael Mutmansky]

Ben,

We're talking B&W images here, so there is no useful color information in the three separate channels, and in most cases, the three channels are mostly duplicates of each other.

That said, due to the color fringing problem, scanning in B&W mode will almost always be less sharp than scanning in RGB and then dropping the R and B channels in PS. The scanner software essentially takes the R, G, and B channels and makes a gray value equal to the average of the three. That will lead to the color fringing problem reducing the sharpness a little. Scanning in RGB does not do this averaging, so the individual channels will be better than the composite. This is all easily tested and verified on your personal scanner.

One caveat; when scanning in RGB mode, it is important that you make sure that the channel you are going to be keeping has not been clipped by the software. That can happen, and is especially a concern with pyro-style negatives where the three channels will have considerably different intensities.


---Michael

[sog1927]

Just a dumb question: If you have doubts about the transparency itself, why not check for fringing by looking directly at it with a good loupe? I've seen 10x and 12x Schneiders on *bay for a song lately (and bought them, by the way). I would think this amount of CA would be readily visible under that level of magnification, and wouldn't take more than a couple of seconds to evaluate.

Steve


The epson is a cheap scanner (in comparison to Creo flatbeds, drums scans, etc). Remember, the 4990 and 700 really were only designed for 4-6x enlargments (although some have argued up to 28x). Point is you have cropped way pass the 16x20 point, so of course your going to see some problems. Now the question is where is the problem? It is less likely the film, and more likely the scanner(cheap lens,guts) or your camera lens. Get a drum scan done, and if your drum scan improves, you know it is your scanner, if it doesn't improve you know it is the camera lens, and less likely the film.

[snaggs]

Guys.. this is simple. the lens in the V700 is not a APO lens.. it can't focus the RG & B wavelengths at the same distance.. There is a reason Nikon Coolscans have ED glass.

Daniel.

[Tim Lookingbill]

Yeah,

Flip through each RGB channel in PS and note the slight off alignment shift from one another in each gray channel. I get this on my 35mm scans off my refurbed Epson 4870.

snaggs' explanation makes since as to the cause of this.

[walter23]

That might be a cause, but the other possibility is an inability to perfectly align the information coming into the different R, G, and B scanner sites - I assume the design is similar to a digital camera, but with a linear array of pixel sites instead of a rectangular sensor area. Imperfect treatment of a parallax problem caused by this could lead to some offset in the colour channels.

Pure speculation though.