p.1 #1 · High-Iso Performance vs good colour rendition
Dear all,
over the course of a few months of (mostly) reading on this forum I've seen a few posts in various threads that seem to suggest that there is a trade-off when manufacturing sensor chips between achieving good (i.e. low-noise) high iso performance and accurate colour rendition at lower iso.
As an example, I think I remember some posters suggesting that the 5D classic has better colour rendition than the 5d II at ISO 100.
Personally, it is not obvious to me whether this trade-off exists and if it indeed exists what the technical/physical reason for it is.
If anyone can enlighten me on this issue that'd be great.
Another possibility is that I'm simply mis-interpreting some of the posts.
p.1 #2 · High-Iso Performance vs good colour rendition
I think you are right. The Bayer color filter array can of course be stronger or weaker, with more or less "overlap" between R, G and B wavelengths. A weak filter will transmit more light (for example, you lose ~2 stops when using an A25 red filter on film) per pixel, but then every single pixel in the RGGB pattern will have a broader wavelength bandwidth, and less separation from the neighbouring pixels.
So, if the manufacturers are using weaker Bayer filters to let more light in, it's obvious that it comes at the price of lesser color accuracy. The question is if they do that. I have no idea.
p.1 #3 · High-Iso Performance vs good colour rendition
The reason for it is the density of the Bayer Colour Filter Array in front of the sensor sites. These colour filters must be dense to provide good colour resolution, but dense filters cut the amount of light hitting the actual photosite and thus decrease the system SNR and thus reduce noise performance and dynamic range
So camera makers must balance noise & DR performance with colour resolution. Canon's come solidly down on preferring better Noise/DR performance while Sony did the opposite. Nikon's been balancing in the middle but seems to be headed in Sony's direction as newer sensor tech has granted improvements to the noise/DR side of things which offsets the losses from the denser CFA.
p.1 #4 · High-Iso Performance vs good colour rendition
I've had the 5D classic, 1DsII and now 5DII. I think the 1DsII has better color than 5D classic. I also feel that the colors from 1DsII and 5DII are similar and thus tiny bit better than 5D classic. ...but that's just a non-scientific feeling. But I'm definite about the 1DsII having better color than 5D classic.
A friend of mine has been a Nikon D3s and D3x shooter and got the D800 few days ago. He says he's going to keep the 3Dx (alongside the D800) because he thinks it has better color than the D800.
p.1 #8 · High-Iso Performance vs good colour rendition
CFA is either made for high color accuracy (A900) or high ISO (Canon 7D is probably best "worst" example). Usually its paired with tuned AA filter aswell.
As logic says, chroma noise will be more visible, when CFA is accurate (cause you can easily distinguish between noise and color) and less visible when everything is bit more "flat" and look about same colorwise.
Same goes with noise vs AA filter. More you smear details, less visible is noise. Another thing is that good AA is quite expensive and CFA too (altough Sony apparently doesnt have issues with puting pretty good CFA into low end models).
Bad CFA is noticable quite easily, cause its very hard to get good skin colors if CFA sux. Its somewhat fixable with having ICC profiled camera and model holding GretagMacbeth table. But only to certain degree. It does have effect on landscapes, when you see that all greens in woods look same, then thats why (another thing might be that author doesnt know difference between vibrance and saturation ).
1Ds MK2 indeed has better colors than 5Dc or 5DMK2. Most 1D/s are like that. Tho difference between 5Dc and 1Ds MK2 in lab metering is really small (exactly 1 point of metamerism index under daylight simulation and 0 point difference under artificial light). 5DMK2 is worse than both (but you have 21 mpix etc.).
Colors vs ISO, well higher is gain, lower is color accuracy and more important is having exposure right (when you shift exposure you shift colors aswell, quite visible at high ISO, much less at low ISO as theres more headroom).
Best overall camera for high ISO color "accurate" is probably D700, which has decent metamerism index at base ISO - 83/83 and colors degrade very slowly. D3s is decent too, but thanks to being improved for high ISO, colors are bit worse right from base ISO, D4 is no different.
If you are sucker for colors, then its pretty much Sony A900, 1DsMK3 or 1DMK3 (last gives quite nice results even at higher ISO).
p.1 #9 · High-Iso Performance vs good colour rendition
ken.vs.ryu wrote:
would like to *see* how the Sony a850/a900 does against the 5diii at low iso.
thanks ahamb lol
A850 is slightly worse than A900 (tho difference is really hard to notice). 5DMK3 .. probably similar to 5DMK2. Colors seem ok from samples and it can use custom ICC profiles as most Canon dSLRs (even tho users are not aware of that too much).
I can bet that A900 will have more accurate colors than 5DMK3.
p.1 #12 · High-Iso Performance vs good colour rendition
I think it would look bit different if it was Minolta tuned, especially JPEG output. But I guess its tuned right, which is important. With good RAW developer, colors are just where they should be.
Minolta had 7D calibrated that its color output was fitted exactly into sRGB or aRGB color space (tho they actually had own sRGB/aRGB modified profiles, kinda like own standart).
Interesting thing is that original 7D isnt exactly metamerism wonder, still it produces pretty amazing colors.. I guess thats cause output is matched to sRGB.
Nice thing about having good colors to start with is that it makes color manipulation output (crossprocess, film simulation etc.) much more pleasing (and easier to do).
p.1 #13 · High-Iso Performance vs good colour rendition
Interesting discussion. What to make of the NEX-5n and the Nikon D5100/D7000? They are all best of class for high ISO performance on a crop sensor, and they also have exceptional color at low ISO. I prefer their low ISO color to just about any other crop body, anyway (short of the NEX-7, which is also great).
I would agree that the D700 is probably the most consistent all around from ISO 200-1600.
p.1 #15 · High-Iso Performance vs good colour rendition
FlyPenFly wrote:
I hope some day someone will make sensor that gives me 90% color performance at ISO1600 as it does at base ISO.
The real problem is that when you're shooting at ISO 1600, your light sources tend to be pretty bad. And if you have bad light sources, you cannot possibly have good color.
p.1 #17 · High-Iso Performance vs good colour rendition
Everyone,
thanks for the replies so far. Makes sense that the density of the Bayer filter would affect ISO Performance and colour rendition and that different manufacturers make different trade-offs there.
p.1 #18 · High-Iso Performance vs good colour rendition
Agree with the general sentiments expressed here. Sony 'inherited' Minolta's deep interest in colour and it is to be hoped they continue on with the next full frames in that respect.
The best alt lenses have their own colour character and you need high quality colour tonal separation to let them fully express it.
And for many types of photography colour is also a key contributor to apparent image depth - if the image does not signal a near to correct colour signature to the viewer, this quality is diminished. If separation is good, less of the colour is likely to nudge the gamut wall, another bonus.
Also best to be aware this is a highly individual area. I am almost certain that people have different colour perception, even on a cultural basis - the Irish with greens and the Inuit with tones of snow and ice, maybe the British with grey (sorry, last one is a in joke). And there are many who simply cannot register colour properly, mostly males. Many more see it as less significant.
p.1 #19 · High-Iso Performance vs good colour rendition
Douglas, that is an amazing table, thanks. I'm not much au fait with DxO, probably been turned off by people here, lol.
But it does show clusters of brands at different levels, and there appears to be no obvious correlation by cost or quality level of camera.
Nikon are widely regards as making superior 'toppings' for their Sony-sourced sensors for general photography, but we see the D7000 (same base sensor as A55 and NEX5N) and D3X (~ for A900/850 sensor) and D800 (Sony eq. camera yet to see light of day). Correct me if I'm wrong about this. They are all on 78-79, and the Sony equivalents are on 87-84-82.
The older pro Canons and 5D are rated very high on this index also, and Canon have cameras right through the range.
Thanks for all that information, Mescalamba, very valuable. I'm inclined to put some credence in this table's data, and will have eyes on this measure when Sony's next FF camera arrives, for sure.
p.1 #20 · High-Iso Performance vs good colour rendition
Hi, Philip. Keep in mind that I'm not sure exactly how much difference there is between a score of, say, 85 and 79 in that chart. DxO Mark's tests are considered legitimate, but interpreting the numbers is the trick. We probably need theSuede to talk us through all of this.
).