Foveon Reds that aren't Red - is there a Fix?

[xpatUSA]

Hello Sigma fans or not, as the case may be . .

Firstly, this is not a Foveon vs. Bayer thread, nor is it a Sigma vs. the Rest thread! It's simply a question about hue.

I bought an SD9 for my watch photography mainly for the promise of Sharpness (the watch photographer's Grail). However, color rendition is also important when selling watches on-line. To that end, I've noticed that reds in my shots tend more to orange than red. Photoshop puts the SD9's hue as 20 degrees off compared to the same shot taken with a Nikon D50 in RAW. I've already discounted white balance as a cause and also LO res vs. HI for the SD9. The SD9 has the latest firmware in it.

Here's a comparison screenshot:



And here's a color wheel which shows just how far off the SD9's red is:



I find that messing with white balance, hue, color bias, saturation and even opening the file in Sigma's SP 5.1 doesn't really fix it - although upping the saturation helps slightly (although it shouldn't). I really don't want to go back to the D50 which tends to image softness, so any suggestions on a fix for the SD9 images would be appreciated.

Ted

[xpatUSA]

In desperation, I put the UV filter back on the lens, thinking that the fluorescent lamps' UV content might be swamping the sensor a bit (I know, I did say "desperation"). No change. I also went into PSE 6's color adjustment where you can increase or decrease reds, greens, blues individually. On upping the reds, the hue of the watches hour marker didn't alter much and, not surprisingly, other parts of the image took on a pink hue.

I was going to try different lighting but the D50 shot was with the same lighting, so the comparison itself stands as valid.

I'm not intending to mess with color filters on this camera, so really if this problem isn't solvable, the SD9 is going to have to go.

Do other Sigma owners have this problem, like with the SD10, SD14, SD15 for example?

[Colin Southern]

Hi Ted,

I'd simply spring for a Colour Passport and create a custom profile for the camera (to be used in ACR).

You can also adjust the camera colour "by eye" in ACR camera calibration tab.

Hope this helps.

PS: I doubt that a Foveon sensor will give you more sharpness per se - that's really more in the domain of accurate focusing and correct sharpening workflow.

[xpatUSA]

Hi Ted,

I'd simply spring for a Colour Passport and create a custom profile for the camera (to be used in ACR).

You can also adjust the camera colour "by eye" in ACR camera calibration tab.

Hope this helps.

PS: I doubt that a Foveon sensor will give you more sharpness per se - that's really more in the domain of accurate focusing and correct sharpening workflow.

Hello Colin,

I had to google to find out what a color passport is (blush). Not sure if it's compatible with my PSE 6's ACR 4.2 (which I'm grimly sticking with, until death do us part). And I haven't found a "camera calibration" in my version of ACR; it just has the one-size-fits-all tabs on the right of the screen (unless it actually remembers the camera or the file type, not sure now (another blush)). Maybe I should mess with ACR's default setup some more, I only played with it a bit.

The part of the sharpness grail I was seeking (acutance?) is supposed to come from the absence of a blur filter on the SD9 sensor, all other things being equal and notwithstanding dire warnings about moire. Of course I must agree on correct focus, sharpening workflow - neither of which got much attention in my comparison shot above.

Ted

[xpatUSA]

Maybe I should mess with ACR's default setup some more, I only played with it a bit.

Colin, I went back into ACR (4.2) played with the sliders some more - specifically the "vibrance" and "saturation" and ended up somewhere close to 100% on the saturation. The red bits on the watch were then quite close to the D50 shot (and the live subject) with H/S/B being Hue 3 degs, Saturation 100% and Brightness 71% - all close enough for government work:



Having said that, it does bother me having to crank a slider that much but perhaps that's price one pays for using a clunky old SD9. Just for a minute there I was seriously contemplating a D90 . . . . phew (sigh of relief).

best regards,

Ted

[Colin Southern]

Hi Ted,

Ah - you won't have that tab on the Photoshop Elements edition of ACR (you just said Photoshop).

The Anti-Aliasing filter isn't a big deal - a simple capture sharpen @ 0.3px @ 300% fixes it.

[xpatUSA]

Hi Ted,

The Anti-Aliasing filter isn't a big deal - a simple capture sharpen @ 0.3px @ 300% fixes it.

I'm rather glad you said that, Colin (the reason why appears further down).

Todays SD9 adventures . .

Took the watch outside under a clear blue - and "kapow" nice red markers appeared. Took it back inside and shot it under two 60W "soft" incandescent light bulbs, the markers got even redder almost to excess according to the hue of 336 degs. Tried several different fluorescents and it was back to orange in all the shots which quite ticked me off, to be honest.

As an aside, I also received a 70mm EX DG macro today and have yet to find it any better than the 50mm EX macro that came with the camera :-(

So, based on the almost uncontrollable nature of my SD9's indoor reds, I have . . .

Ordered a refurbished Nikon D90 body from B&H. Yep.

'twas fun while it lasted but I'm not going to invest in special lighting, special software, special color checking stuff, etc, ad naus, just to get the SD9 to take shots as color-correct as the D50 that I already have. I do hope people will understand this point of view from a guy who, in spite of being somewhat pedantic, just wants to take a watch picture and post it on a Forum or eBay.

My Dad used to immerse bits of paper in sodium thiosomething and his pictures turned out pretty good, IIRC, without all this tech stuff. And yes, his camera had a bellows on it . . .

Ted

[Colin Southern]

Hi Ted,

Flourescents are probably the worst possible light for photography (alongside Sodium Vapour anyway) - they just have a very spiky output; our eyes can't see it, but the camera does.

[krispix]

Hi Ted,

Flourescents are probably the worst possible light for photography (alongside Sodium Vapour anyway) - they just have a very spiky output; our eyes can't see it, but the camera does.

You're right! Flouries are the bane of any photographer's life. Even setting the WB to Fluorescent or using a filter doesn't really do the trick. I think Ted would probably be better off using the incandescent 60W lamps and setting the WB accordingly, but the best bet would be a little studio flash setup, probably cheaper than a S/H D90 as well ...

[Colin Southern]

You're right! Flouries are the bane of any photographer's life. Even setting the WB to Fluorescent or using a filter doesn't really do the trick. I think Ted would probably be better off using the incandescent 60W lamps and setting the WB accordingly, but the best bet would be a little studio flash setup, probably cheaper than a S/H D90 as well ...

I agree.

Need to be careful with the light setup though ... watches are reflective little blighters.

[xpatUSA]

I agree, too, but . .

My specific problem with fluorescent lighting is the Foveon sensor's response to reds under such lighting. I've used those Sylvania mini-craft CFL's for watch photography for some 6 or 7 years with a Nikon D50 and, before that, a humble Kodak point-and-shoot. Never had that kind of trouble with reds. But, the first time I used the SD9 on a subject with reds in it, the subdued coloring was quite noticeable although not as bad as that of the watch markers:



You'll notice that the red stitching isn't quite red, confirmed by the hue in PSE being somewhat to the "left" of red at somewhere around 20 degs. And here perhaps is a facet of the "Foveon look" with it's more subdued colors, as opposed to the in-camera trickery of Bayer algorithms and (in my range of affordability) added vividness to please the punters.

On a Sigma forum elsewhere, the "red shift" is well known and the sad part is that advice offered there involves all kinds of trickery to improved the rendition at that end of the spectrum. Well, call it recalcitrance, but I'm not going to go away from fluorescent lighting any time soon - I'm not into flash and neither am I into halogen, et al. Some of my watch shots are illustrative quickies taken under the computer desk CFL lamp and that's not going to change either.

The curious part of this saga is that my spectroscope shows a really strong band at the red end for the Sylvanias, but I guess the reflectance spectrum of the watch markers missed that one :-(. No matter though, the overhead strip light showed a relatively smooth spectrum and they still came out brown. I do have a 250W heat lamp that has a red coating on it, maybe I should have tried that! Ouch.

In spite of an intense interest in all things optical and electronic, I have to say that if something looks red to my eyes both outside and inside under all forms of light, then it is red, not brown. So, from my perspective, Nikon et al have got it right and the SD9 will be back on eBay shortly - because the D90 has already been shipped and I am a little poorer today!

best regards,

Ted

[Manfred M]

Using fluorecent lighting is likely the issue here.

Fluroescent lights are deficient on the red end of the spectrum; i.e. they don't emit a lot of red light. The manufacturers do play with the phosphors, but in general the distribution is "wrong" versus daylight, and unless the "spike" you mention matches / is close to the red in the crest, you are going to get strange colours.

[xpatUSA]

Using fluorescent lighting is likely the issue here.

Fluroescent lights are deficient on the red end of the spectrum; i.e. they don't emit a lot of red light. The manufacturers do play with the phosphors, but in general the distribution is "wrong" versus daylight, and unless the "spike" you mention matches / is close to the red in the crest, you are going to get strange colours.

I've tried so hard to make it clear that my fluorescent lighting is not the issue. It does not explain why my D50 with it's 6MP CCD shows something as red while the SD9 shows the same item under the same lighting as brown, some 20 to 30 degrees of hue different. That's really my point - theoretically there should not be that great a difference, IMHO.

Yes, I've done quite a bit of research on CFL spectra and what you've said does agree with the literature. However, here's a pic I took of the said Sylvania CFL's spectrum about 3 months ago:



The peaks are seen quite clearly. By co-incidence, this pic was shot with the
SD9! There is a signicant peak somewhere in the red as you can see. I'll bet
the color of my watch marker was just to the right of the bright red band
(called Sod's law, where I'm from).

The grating for that shot was a 600 lines/mm piece of audio CD. I've since
changed it for a real 500 lines/mm and the spectrum hasn't changed. Maybe I
should try to get the same shot with my D50. Now that would be a test, eh?

By the way, the SD9 sold within minutes of putting it up on eBay. The brand
new Sigma 70mm macro is exciting less interest, to my great surprise.

Later,

Ted

[Colin Southern]

I've tried so hard to make it clear that my fluorescent lighting is not the issue.

Hi Ted,

Probably the best way that I can think to describe the problem is to say that as I see it, you have a problematic light source that other cameras seem to be doing a better job of compensating for.

Having said that, I can't see any reason why a custom colur profile wouldn't fix the problem - or even a manual tweak withthe right software.

[xpatUSA]

However, here's a pic I took of the said Sylvania CFL's spectrum about 3 months ago:



The peaks are seen quite clearly. By co-incidence, this pic was shot with the
SD9! There is a signicant peak somewhere in the red as you can see. I'll bet
the color of my watch marker was just to the right of the bright red band
(called Sod's law, where I'm from).

The grating for that shot was a 600 lines/mm piece of audio CD. I've since
changed it for a real 500 lines/mm and the spectrum hasn't changed. Maybe I
should try to get the same shot with my D50. Now that would be a test, eh?

Here we go, D50, kit lens, RAW (all sliders at 0) saved with minimum JPEG compression:



There is fair amount of red to seen.

best regards,

Ted

[xpatUSA]

Hi Ted,

Probably the best way that I can think to describe the problem is to say that as I see it, you have a problematic light source that other cameras seem to be doing a better job of compensating for.

Having said that, I can't see any reason why a custom colur profile wouldn't fix the problem - or even a manual tweak withthe right software.

Hello again Colin,

In response to people's concern with my CFL lighting, I dug the SD9 out of it's shipping box and shot the spectrum of an LED light source (28-LED flashlight, er, torch). Opened the image in SPP 5.1 and also ACR 4.2. Saved at max jpeg quality. Then did similar with the D50, opened in ACR and saved at max jpeg quality. The shots themselves are pretty poor but I just wanted to check the color rendition . .





I know the SD9 images could have been better but it's late and the camera ships out tomorrow, so this is it, folks. SPP 5.1 seems to translate the RAW to slightly better color rendition but I do find it a pain to use compared with ACR 4.2

It is seen that, even with the relatively smooth spectrum of the LED source, the SD9 image colors are still significantly different to those of the D50, the tendency to yellow not green and the lack of red is plainly obvious to the eye. Yes, I could tweaked this and that slider and so forth.

So . . . the Nikon D90 is inbound on Friday, along with a Sigma 50mm EX macro. Perhaps a whole new unprofessional photographic adventure awaits!

best regards,

Ted

[Manfred M]

Thinking about this a bit more last night, I still think that the fluorescents are likely causing the issue, but suspect the Foveon design might also be part of the issue, coupled of course with how our eyes see things.

Black body radiation (emulated by sunlight and tungsten light) produces a fairly even distribution of wavelengths across the spectrum that they are emitting light at (for incandescent lighting, this is biased toward the red / yellow wavelengths). Other forms of light that rely on photon emission, like fluorescents and LED lights will emit peaks at very specific wavelengths. Our eye / brain image processor nicely processes this data for us whereas the camera sensors records what it picks up.

This is where some of the issues with the Foveon design could come into play. With the more common Bayer array, designing the colour filter is fairly straight forward, as each filter over the photodiode can be designed to handle the full spectrum. The Foveon design works more like film, where the top layer absorbs / measures blue light, the second layer absorbs / measures green light and the bottom layer absorbs / measures red light. In a perfect world, no problem, but in reality any real filter has some bias and can accept or reject wavelengths in a non-linear manner. If there is a misalignment between the light source and the stack of filters / electronics, the colour that ends up being recorded could have some level of colour bias associated with it.

As I have not heard of this problem with this camera and daylight shots, the design point for the sensor / image processor seems to have been designed to work under these conditions. If the light sources has spikes at different wavelengths and these are not properly compensated for in the design, one would expect to see some issues with how the sensor captures and processes these wavelengths.

My guess is that is the most likely explanation of the results you are seeing. Of course, I am making the assumption that the image processor is not causing this issue…

[xpatUSA]

Thinking about this a bit more last night, I still think that the fluorescents are likely causing the issue, but suspect the Foveon design might also be part of the issue, coupled of course with how our eyes see things.

My guess is that is the most likely explanation of the results you are seeing. Of course, I am making the assumption that the image processor is not causing this issue…

Hi Manfred,

I tend to agree with your comments, especially regarding the Foveon sensor characteristics. On a Sigma forum elsewhere, comments about color bias, balance, etc, are legion. And as can be seen in my previous shots of an LED lamp spectrum (which is quite smooth at the red end) the sensor seems to struggle with both long and medium wavelengths. There is also a clear difference in what Adobe Camera Raw does with a .X3F file compared to Sigma Photo Pro (see the middle part of the top two spectra for the SD9).

Having said that I wouldn't stop using CFL's, I have lately been looking at LED lighting more seriously, especially as the phosphor used seems to have smoother, broader emission than tri-phosphor CFL's:



The only concern being that peak in the blue at about 430nm, not that my LED flashlight seen above seems too bad in that respect.

Ted

[Colin Southern]

Off memory Ted, Steaphaney is a Sigma user, and someone with a better understanding of the physics of these things we are not likely to find!

Might be a good idea to bring this thread to her attention and see if she can help shed some (spectrally neutral) light on the topic!

[Manfred M]

The fluorescent tube curve is along the lines of what I would have expected to see. Orange is light runs between 590-620nm and red is from 620-750nm, so the source is in fact more orange than red, and that is what we see in your image.

The LED light spectrum is not at all what I would have imagined. Downright weird. Any white light LED curves I've seen have more or less equal strength R-YG-B spikes, not the smooth curve that is heavily biased towards the shorter wavelength. The left side of the graph makes sense, but not the right. I had expected something like:



They must be using some interesting phosphors in the mix.