The GetDPI Photography Forum

Great to see you here. Join our insightful photographic forum today and start tapping into a huge wealth of photographic knowledge. Completing our simple registration process will allow you to gain access to exclusive content, add your own topics and posts, share your work and connect with other members through your own private inbox! And don’t forget to say hi!

H3D39 as first step to MFDB/Technical (& vs DSLR)

timparkin

Member
You live in Britain so are more sensitive to green is a classic and because I don't agree with you I shouldn't continue with this conversation...... BRILLIANT! :ROTFL:
Actually I said if you can't see any difference in the colour in the pictures we've shown then there is little point in continuing. Some people are more sensitive to colour than others and if you're not that sensitive then the issue isn't bad enough to bother you.

However given that I showed you a set of scientific comparisons that demonstrated that metamerism was a scientific issue and that the P45 had been measured to be one of the worse performing sensors but you chose to selectively requote a small portion of my reply and ROTFL me - I think I can see how engaged with the actual topic you are. :)

I've put down as much evidence as I can and any more posts will be just finding new ways to state the same thing..
 

fotografz

Well-known member
No need to get hard-edged.

This article I think was quite good introduction at metameric error in digital photography for anyone interested:
http://dougkerr.net/pumpkin/articles/Metameric_Error.pdf

It is clear that cameras vary in their ability to capture colors so they render in the same way the eye see them. In a camera the issue is that many spectrums can get registered with the same R, G and B values and thus be registered as the same color, despite that the eye may see different colors.

We can argue about if this actually affects green in KAF-39000 sensor, if its easy to correct for and if it matters. I'm no expert in that area so I'm open for that it could be any answer to these, but I don't really trust the general "noone has complained therefore its perfect", in fact I've seen people "complain" indirectly by saying that Dalsa sensors have more accurate color.

After all this hard-edged discussion it doesn't seem to matter much though. The H3D-39 is okay, buy it if you find a good deal :)
I don't think anyone is saying any of these backs are perfect ... some excel in some areas and some in others. I've not seen general comments referencing Dalsa sensors as empirically producing better colors, as much as how they subjectively render skin tones to more people's liking.

The most accurate color rendering I've seen, and others also subscribe to, is from the Multi-Shot backs ... including the CF39 MultiShot that uses the same Kodak KAF-39000 sensor being discussed. While my H4D/60 could equal the CF39MS in file quality with one shot, it never matched the color accuracy of the MultiShot ... and the H4D/60 uses a Dalsa sensor.

I agree that the H3D/39 is okay, and can be a way into getting a MF look and feel ... or an entry into basic tech camera use which can be fun ... especially if the camera has full movements. For Landscape work specifically that is a subjective call IMO, and one I have no opinion on since I don't do landscapes.

-Marc
 

timparkin

Member
I don't think anyone is saying any of these backs are perfect ... some excel in some areas and some in others. I've not seen general comments referencing Dalsa sensors as empirically producing better colors, as much as how they subjectively render skin tones to more people's liking.

The most accurate color rendering I've seen, and others also subscribe to, is from the Multi-Shot backs ... including the CF39 MultiShot that uses the same Kodak KAF-39000 sensor being discussed. While my H4D/60 could equal the CF39MS in file quality with one shot, it never matched the color accuracy of the MultiShot ... and the H4D/60 uses a Dalsa sensor.

I agree that the H3D/39 is okay, and can be a way into getting a MF look and feel ... or an entry into basic tech camera use which can be fun ... especially if the camera has full movements. For Landscape work specifically that is a subjective call IMO, and one I have no opinion on since I don't do landscapes.

-Marc
I think accuracy of colour rendering depends on lighting conditions and the subject being photographed. And accuracy isn't the be all and end all of photography either. However, knowing about the potential colour 'issues' relating to any photographic device is useful and I would say if you photograph a lot of greens and were thinking about buying the KAF-39000 sensor then you should consider the comparisons we've shown. If you're happy with them then no problem. A lot of people are - some aren't..

The camera is without doubt capable of stunning results and from what I've seen is excellent at portraying skin, on which I have no opinion as I don't do portraits ;-)

Tim
 

ondebanks

Member
I am 100% with Tim on this colour debate.

Don't be silly! There is a huge range of choices for colour filter arrays and they come in varying densities too.
This is true.

On top of that, there is also manufacturing variation in the spectral responses of CFAs, certainly within a sensor (pixel to pixel), and probably also between sensor batches. What you see in the datasheet for a sensor is only indicative. Shashin, look up the KAF-40000 datasheet for your 645D's sensor. The title over the spectral response curve says "KAF-40000 Average QE", while the next plot is captioned "Typical GR - GB QE Difference" [my underlining].

Then look up the KAF-31600 datasheet (same sensor size and microlensing, just slightly bigger pixels) and you'll see a noticeably different set of curves.

You can't apply a profile that corrects one and not the other.
Exactly, and this is the nub of the matter. Shashin talked about getting Photoshop to correct the disputed metamerism. But a pixel's colour is just a set of 3 numbers for R,G,B. If a fern pixel has an R,G,B of 30,94,12 and a moss pixel has an R,G,B of 30,94,12 also, how can you ask Photoshop to treat them differently from each other?!

Once the image capture has taken place, you can play with and rescale the ratios between the 3 R/G/B numbers in post processing, and this gives the illusion of infinite colour flexibility, or the illusion of perfect colour accuracy if you've taken a lot of trouble to profile things. But no playing or profiling can drive equal valued pixels away from each other. You can make them dance all over the gamut diagram, but they are always moving as a unit.

If you want a simple test whether cameras sensors can cause metameric shifts, try taking a photograph of a GATF/RHEM light indicator under a light source where the colours all look the same for your eye. I guarantee the colours will look different to the camera sensor - that is metameric failure.
Or do this simple test: shoot a street or garden in daylight, and again at night under low pressure sodium street lighting. All the varieties of object and foliage colour in the day shot collapse to essentially the same, almost monochromatic colour in the night shot. Ah, you say, but that's a metamerism problem due to the light sources, not the camera sensor. But wait, there's more! Do the test twice - use both a Kodak sensored MFD unit and a Dalsa sensored one, or if you can't manage that, use both a Kodak sensored MFD unit and a DSLR. Equalise their colours to each other as best you can in the daylight shots - and then apply those settings/profiles to the night shots. And be prepared for the shock of vastly different colours! Kodak sensors, at least the older ones like the KAF-18802 and KAF-2002CE that I have used in my MFDB and DSLR, record the yellow-orange (589 nm) emission of traditional low pressure sodium lights as yellow-green! [See my examples below]. That is an undeniable example of getting the colour "wrong". It can't be profiled away, nor fixed with a white balance correction; if you attempt to do so then you throw off the colours of anything in the photo which was not merely reflecting the sodium lights...such as car headlights and the objects they illuminated, or the Moon, or the remnants of a twilight sky.


Moon, Venus, and a few stars. 200/2.8 APO and Kodak DCS645M on a Mamiya 645AFD, 4 second exposure. I took this from just outside my house. The foreground bushes, which had bare brownish-gray branches in daylight, were very close to a sodium streetlight (it was just over my right shoulder). They should therefore look yellow-orange, but instead they are a stark yellowish green. The Moon, slightly warmed up by the hazy twilight, shows that there was no major white balance problem elsewhere in the scene. If I were to click white balance off the moon to make it more neutral grey, it would "cool" the colours of everything else in the scene, which would only drive the bushes further from the warm orange they should be.

So I have seen clear cases where Kodak sensors deliver weird colour because of where the CFA response curves rise, fall and cross each other in wavelength space. The green filter is either not falling quickly enough at 589nm, or the red filter is too slow in its rise into the orange region, or both.

If this is a pronounced issue in monochromatic light, it will be a more subtle - but no less real - issue in daylight.

Now, while this greenish colour is clearly "wrong", I sometimes prefer it to the "right" orangey colour, because green trees and bushes are more natural than orange ones! The yellow-green is not so lurid when the objects are not as close to the steetlight(s). See the next example.


Second example - Orion, Jupiter and the crescent Moon. A pano made with the same camera and back, 45/2.8 S lens, 30 second exposures. Where a high-pressure sodium security floodlight caught the treetops, they are the correct orange-yellow colour (because high-pressure lights are broadband, not monochromatic). But look under the treetops at the rest of the trees, especially over on the right - there's that yellow-green again! Those parts of the trees are dimly illuminated by low-pressure sodium streetlights. It's the monochromatic nature of this light that is so unforgiving: if your CFA filtration bandpass is slightly off, then there's no fixing the colour.

To the eye, both types of light were yellowish, but to the camera, reflecting off the same material, they record as completely different colours. [I guess this is like the inverse of metamerism - is there a name for that?]

Ray
 

MartinE

New member
So Tim
The bottom line is that this sensor is not able to differentiate small differences in greens. But it may be better in other colors. But I still see a big difference in the end result due to the size of the sensor. I shoot with 35mm also and its always different. Ok MartinE
 

timparkin

Member
So Tim
The bottom line is that this sensor is not able to differentiate small differences in greens. But it may be better in other colors. But I still see a big difference in the end result due to the size of the sensor. I shoot with 35mm also and its always different. Ok MartinE
It's not just greens but more specifically a certain substance of green. In this case chlorophyll. So the sensor can probably differentiate all sorts of greens but you get a hue shift when you try and photograph chlorophyll greens.

This hue shift just happens to be in the 'wrong' direction, making a lot of chlorophyll greens more yellow which is similar - Velvia also has a shift but it's 'good' (in my eyes anyway) as the shift seperates colour rather than merges them.

Tim
 

MartinE

New member
Thanks for the explanation. Still recommend this sensor as a great starter in medium format. I used 35mm for many years and the change to medium format was major for me on many different levels. I was thinking of moving to H4d40 but it does not add value. H5D50 is the next logical step I believe.
 

Ken_R

New member
Hi, looking at the image from one of the links you posted: http://static.timparkin.co.uk/static/landscapegb/56/camera-test/dpi3.html
its really hard to make any solid conclusions, specially in regards to resolution since I see quite a bit of motion blur on the h3d39 image. The D800 side does show cleaner shadows and deeper blacks.

From all info ive been able to gather on the net, the tests, reviews, commentaries regarding the D800/e image quality and from using the camera myself, at least for Landscape, the Nikon is about equal and in some cases better than the lower MP backs (if you dont need the movements of a tech camera). To really Wipe the Floor with a D800/e IQ, again, in landscape situation, low iso, workable light and using wide angles, you need one of the newer 60MP or 80MP backs and a tech camera with good lenses. The Nikon just cannot touch those rigs.

The larger sensor backs (no matter the pixel count) do offer a very different look than the Nikon when shooting People. Mainly due to them being a larger format but also to sensor tech and processing. Also, lenses have a pretty big effect on the way a scene is rendered.

Even in a landscape scene, like the links posted, the lens will have a big effect on the final result, not only in sharpness (the most popular concern), but in color and contrast. A very contrasty lens will render deeper shadows and when you go and try to bring them up in processing you will get more noise than if you used a lens that has a bit lower contrast (given equal sharpness). (There is a reason cinema lenses are purchased in sets most of the time)
 

Nectar

New member
As someone who was looking to dip their toe into DMF, with this model in particular, you make very interesting posts Tim. Thanks for taking time to explain it all too. I subscribed to your magazine for a year, great to read on the ipad.
 
Top