Sensor blooming and CA, effected by sensor resolution?

[timmciglobal]

So, as some of the canonites know, I bought a D200 and got rid of my 20D. Since then I've been looking over images I shoot and notice that I don't have as much perceived CA as I was having in certain situations (be it CA or blooming, I don't know) and it led me to a theory..

If "CA" or "bloom" is say, 1 pixel on image boarders, your far less likly to notice 1 pixel edge on a 10.2 megapixel shot then an 8,and an 8 to a 6..

Make sense to anyone else?

I'm wondering if sites like photozone take that into account when calculating CA on tests, since the bloom or CA per pixel would be difference per "physical size" given a 6megapixel shot @ 8x10 and a 10 megapixel shot @ 8x10.

Make sense? Off base?

Tim

[coldrain]

Totally offbase.

Since CA is lens induced, not sensor induced, CA will be worse with smaller sensor diodes. The CA that is in the light reaching the sensor will encompass more pixels, just because the pixels are smaller but the CA will have the same width wen using the same lens and aperture.

And the same goes for purple fringing, the same width of purple fringing will be on the sensor, and thus more pixels will be showing it when the photo diodes of the sensor are smaller. Purple fringing is also a function of the lens, not the sensor... even though it is the sensor that makes it happen. It is light reflected from the sensor, and then bouncing back via the lens. SOme say that it is just the sensors fault, but then they should know better, since it is very lens dependable if, and then how much purple fringing shows up.

[timmciglobal]

See coldain, but the data doesn't seem to support that, OR canon lenses have a FAR bigger Ca issue then nikon lenses.

Look at the CA #'s on photozone reviews, that would mean that being as the D200 (the test camera for most of the nikon ones) vs the XT that the D200's pixels are in fact "more in x area" thus leaving me to conclude if it's a "lens not resolution" issue in fact the CA is 25% LESS compared directly to an XT sample, since 1 pixel on an XT takes up a much bigger physical area.

Otherwise what would explain the difference?

If CA is "1/100th of a MM" for an example number, a D200 should in theory have MORE ca in pixels then a 20D of same lens, since in 1/100th of a MM area the # of pixels is greater on D200 due to smaller pixel size... yet that's not true it seems.

Tim

[coldrain]

What data, Tim. What lenses are you referring to?
Of course some Canon lenses have a lot of CA, others show none. And then some Nikon lenses are CA monsters, and then others show none.

Take a look at the CA figures for the Nikkor 20 and 24mm primes for example. They have quite some, especially for a prime design.

So.. Allow me to be confused about what you actually are comparing. Photozone does not have CA figures for EOS 350D, they have CA figures for different lenses measured on a 350D. And so it would help to know what you are actually comparing. (CA really is a function of the optics, not the sensor).

And the pixels of the D200 do not take a much smaller physical area than on the 350D. It is only a 10mp sensor, a very small difference with 8mp. And the Canon sensor is a 1.6x crop sensor, the D200's Sony sensor is a 1.5x crop sensor. This means that the pixels are about the same size.
EOS 350D/XT: 3456 pixels wide, on about 21.9mm. Around 158 pixels per mm.

D200: 3872 pixels wide, on about 23.3mm. About 166 pixels per mm.

Not a big difference at all, wouldn't you say? The D200's pixels are 95% of the size of the pixels from the 350D.

Nikkor AF 20mm f2.8D:
http://www.photozone.de/8Reviews/lenses/nikkor_20_28/ca.pngquote from Photozone: Chromatic aberrations (color shadows at harsh contrast transitions) are pretty bad especially for a fix-focal. With an average CA width of around 2x at the image borders the Nikkor is among the worst fix-focals tested so far.

Canon EF 20mm f2.8 USM:
http://www.photozone.de/8Reviews/lenses/canon_20_28/ca.gif
quote from Photozone: CAs (visible as color shadows at harsh contrast transitions) reach an average CA pixel width well below the 1 pixel barrier which is very good for a lens in this focal length class.

[timmciglobal]

I was looking at the 80-200 vs 70-200 but my argument wasn't lens vs lens but "Resolution" vs "Resolution" on a lens.

I'd really love to see 2 CA readings on a d50 vs D200 because in theory the d50 should show much less "ca" on a per pixel basis since its pixels cover more physical area if it's a purly lens related issue, where as if it's a sensor related issues the CA #'s will be very similar.

Having said that however, the logic would still to my mind mean that the nikon should show "more CA" since more pixels are in a particular area then a canon camera of the same lens.

And in fact if it doesn't, 1 pixel CA on a 10.2 or a 12 or a 16 megapixel image would appear to be "less" CA then on a 6 megapixel image per say... seems logical.

Since the "lens" may cause it but the actual result, especially in fringing is usually bloom not CA (IE: distortion caused by bayer patern and blown highlight recomposistion)

Tim

[cdifoto]

For the sake of argument and easier math, my post assumes (incorrectly of course) that the D200 is a 12MP camera and the D70 is a 6MP camera.

I know what you're saying Tim. You're basically asking....is the visibility of the CA relative to the sensor size, or is it a hard number. In other words, 1px of CA on a 6MP sensor will be a "bigger deal" than would 1px of CA on a 12MP sensor. Absolute figures are the same (1px), but it's 50% less visible on the bigger sensor.

I think, however, this would also have to assume the sensor is the same size, because pixel density is a factor. If the D200 and the D70 use the same sensor size, then you're cramming more pixels on the D200 sensor since it has 2x as many, thus the pixels would have to be 50% smaller. As a result the 1px of CA on the D200 would be less noticable than the same 1px on the D70.

Am I making sense Tim?

[timmciglobal]

Aye, IF that is true though.

See, I'm wondering if it is.

If it is, then for example the 10D would have "less" CA measured in pixels then a 20D as the 20D has "more pixels" in a given area to which CA would be viewed on the image.

IF however CA is dependant not only on lens but ALSO on photosite #'s then the difference would NOT be there and thus more megapixels on a CA basis wouldn't = more CA in a per pixel basis.

This is the question which I wonder.

As far as "Real" ca as in, CA caused by transmission rates of glass, I'd think the pixel count would be irrelvant and "more pixels in an area" would show higher "pixel" #'s for CA (ex: 2 pixel Ca vs 1 pixel CA on a 50% less pixels same area sensor) where as in sensor blooming (IE: highlight bleeding effect from bayer paterns and blown highlights) the effect would be LESS noticable and "per pixel" would be the same.

I'm just wondering if in fact this is or isn't true.

I'm really tempted to borrow a D50 and test this out on some standard chart to see.

Edit: why do I care? Simple, if the effect is sensor size dependant (ie: density dependant) CA figures are important to look at on a "Sensor size" basis, so that, a 1.2 pixel CA on an XT shot then compared to say a 50D with 12 megapixels comes out that 1.2 pixels would not be almost 2 pixels in same physical area thus having a 2 pixel "ca fringe" around say a tree edge, instead of a 1.2 pixel fringe on average. Because if it's a physical area "Ca" (IE: 1/100th of a mm of sensor area has CA in this spot with this lens at this focal length) then 1/100th of the sensor area is "1 pixel" on a XT but could be 3 pixels on another camera.)

Tim

[coldrain]

Fringing has nothing to do with purple fringing, and that of course is very confusing. And CA, as you state, again has nothing to do with "fringing" that you sometimes can see.

And as I typed in my post above, the D200 does not have more pixels in an area than the 350D. The pixels are almost the same size, D200 -> 350D = ~95%... pretty close!

Again... CA has nothing to with other types of fringing. CA is when the lens system bends certain wave lengths of light more than other wave lengths of light, resulting in for instance a bigger image in red than in green or blue. The image will differ more in red the more you go from the center of the lens, making the CA get worse to the corners. CA is correctable to shrinking the channel of the picture that differs the most, keeping the center of the image in place.

So, in conclusion, CA is a result of light being bent through the lens system when entering and exitting the different media (air, different glass elements), where different colours of light can have different breaking indexes. CA has nothing to do with purple fringing or other kinds of fringing, sensor related or not.
And yes, IF a camera would have 3 times more pixels in the same area, you would see 3 times more pixels being affected by the CA colour shadows.

And I do not know what Photozone actually tests. It depends a lot on where you look in a photo how wide the CA can get, in the center you will never see CA (unless the lens is very badly centered :P) , and to the edges it will get worse and worse. They talk about average CA (that is how they can end up with measuring CA smaller than a pixel), which will mean that then still you CAN see CA at the border pixels, just less than with lenses with a higher figure. So... when they say a lens has CA smaller than a pixel... it does not mean you will not see any CA at the borders.

[timmciglobal]

Actually the D200's pixel pitch is 6.05 um where as the pixel pitch of 20D is 6.44 um.

Difference of 7%.

But my question isn't "is XYZ better" it's simply a logic question, for say, a "magic" 50D that has a 16 megapixel Sensor.

Would the CA numbers then assuming same sensor size be "double" the CA shown since it has "Twice" the number of pixels per area?

Tim

[timmciglobal]

Coldrain: The question then becomes what most people see, which imho, is sensor blooming, "ie: purple fringing along harsh areas of contrast" which is in fact related to pixel count more so, this having lower "blooming" on a 10 megapixel vs a 6 since 1 pixel/area bloom vs 1/area will appear to show less though actually having same bloom.

Tim

[coldrain]

No, purple fringing is NOT CA, NOT what photozone is measuring. Purple fringing is not in high contrast areas, but in TOO bright areas. So bright, that light gets reflected from the sensor, and gets bounced back again, and you see the light (most often a purple-ish ultra voilet-ish bloom into the surrouding area).

CA on the other hand is just noticable more around high contrast areas, but is all over the photo. Which you actually can see when you study CA carefully.

Two different things, two different ways they show up. I know that ofthe people tend to think CA = purple fringing. But they are not related. And that they are confused/mixed up all the time does not help understanding what is what.

And again, YES, if you have double the pixels in an area, you have double the CA pixels.

[cdifoto]

No, purple fringing is NOT CA, NOT what photozone is measuring. Purple fringing is not in high contrast areas, but in TOO bright areas. So bright, that light gets reflected from the sensor, and gets bounced back again, and you see the light (most often a purple-ish ultra voilet-ish bloom into the surrouding area).

CA on the other hand is just noticable more around high contrast areas, but is all over the photo. Which you actually can see when you study CA carefully.

Two different things, two different ways they show up. I know that ofthe people tend to think CA = purple fringing. But they are not related. And that they are confused/mixed up all the time does not help understanding what is what.

And again, YES, if you have double the pixels in an area, you have double the CA pixels.


Hmmm.....

http://www.dpreview.com/learn/?/key=chromatic+aberration

http://en.wikipedia.org/wiki/Purple_fringing

Both seem to say that Purple Fringing IS a type of CA. I dunno though Coldy. I tend to shy away from this technical mumbo-jumbo and just shoot with good lenses, not worrying about the hows, whys or semantics.

[coldrain]

As you can see the person writing that wiki has not really an idea what purple fringing is, or what causes it. Look up chromatic aberation and you will see that it is that different colours end up in different places (chromatic = colour).

It really is not a big secret that purple fringing is just a light source fringing into surrounding areas, and that it has nothing to do with CA (in what actually is meant by CA, colours ending up in different spots).

[cdifoto]

As you can see the person writing that wiki has not really an idea what purple fringing is, or what causes it. Look up chromatic aberation and you will see that it is that different colours end up in different places (chromatic = colour).

It really is not a big secret that purple fringing is just a light source fringing into surrounding areas, and that it has nothing to do with CA (in what actually is meant by CA, colours ending up in different spots).

Vincent Bockaert has no idea what Purple Fringing and CA are either then huh? Just you? I guess Phil Askey has no idea either since he posted Vincent's article on his website...

[coldrain]

Yes, the wikipedia writer has no idea. The article in wikipedia actually states that... it no where says purple fringing is CA, and states it is attributed to a number of causes by different people, but that the writers are not really sure what actually is the cause. And no, it is not just me that actually has an idea.
Just because some guy writes something in wikipedia (in this case, that PF is unknown to what it is caused by) does not make it unknown, the big problem of wikipedia.

And if you actually READ the Vincent Bockaert piece, that would help. He actually explains the difference between CA and PF! No where he says they are the same phenomenon!!!

CA is just what it is, problems on getting the different colour wavelengths ending up on the same spot of the film/sensor. And when that does not go perfect, you see colour shadows. That really is what chromatic aberation means. I can not make it mean something different.
Here a (real) description of what CA is:
http://www.chemistry.adelaide.edu.au/external/soc-rel/content/lenses.htm

And that people tend to call EVERY fault in an image chromatic aberation is not my fault either. Do I really have to type again the difference to what it called CA and what is called purple fringing? One is just an effect of the lens system bending light, and the other is just an effect of very bright light spots spilling over into a surrounding area. They are not related, nor the same.

Wikipedia or not, you can understand for yourself that they are not the same.

[cdifoto]

Yes, he has no idea. And no, it is not just me that actually has an idea.
Juts because some guy writes something in wikipedia does not make it true, the big problem if wikipedia.

CA is just what it is, problems on getting the different colour wavelengths ending up on the same spot of the film/sensor. And when that does not go perfect, you see colour shadows. That really is what chromatic aberation means. I can not make it mean something different.

And that people tend to call EVERY fault in an image chromatic aberation is not my fault either. Do I really have to type again the difference to what it called CA and what is called purple fringing? One is just an effect of the lens system beding light, and the other is just an effect of very bright light spots spilling over into a surrounding area. They are not related, nor the same.

Wikipedia or not, you can understand for yourself that they are not the same.


I'm not talking about the Wikipedia link Coldy.

[coldrain]

Sorry, to add to the confusion I editted my post even before you posted your reply. :D :eek: :rolleyes:

Sorry again, this makes for nicely confusing threads.

[cdifoto]

Sorry, to add to the confusion I editted my post even before you posted your reply. :D :eek: :rolleyes:

Sorry again, this makes for nicely confusing threads.

He does say "Chromatic aberration or "color fringing" is caused by the camera lens not focusing different wavelengths of light onto the exact same focal plane (the focal length for different wavelengths is different) and/or by the lens magnifying different wavelengths differently."

Is purple not a color? Or is purple the ONLY color that is not CA? (Seriously. I don't know.)

[coldrain]

Purple fringing is a name giving to a different phenomenon, scroll down on that page. The reason for this name is that THAT kind of anomality is most often purple-ish (white/violet--ish) (but can actually sometimes appear to be red-ish or even blue-ish).

So... PF is a NAME given to that anomality... and that does not mean that CA can not give purple colour shadows. it can, of course... but then it still is not the same as what we call "purple fringing". CA is not fringing anyway, actually... it is a colour shadow because one or more colour wave lengths produce a bigger image. That it appears to be fringing to the casual observer adds to the confusion though.

[cdifoto]

Purple fringing is a name giving to a different phenomenon, scroll down on that page. The reason for this name is that THAT kind of anomality is most often purple-ish (white/violet--ish) (but can actually sometimes appear to be red-ish or even blue-ish).

So... PF is a NAME given to that anomality... and that does not mean that CA can not give purple colour shadows. it can, of course... but then it still is not the same as what we call "purple fringing". CA is not fringing anyway, actually... it is a colour shadow because one or more colour wave lengths produce a bigger image. That it appears to be fringing to the casual observer adds to the confusion though.

I guess I'm a casual observer then because it looks close enough to me to be the same, and either way it sucks. :eek:

[coldrain]

I guess I'm a casual observer then because it looks close enough to me to be the same, and either way it sucks. :eek:
Yes, it does suck. And to go back to the original "question", purple fringing you can not measure, since it depends on the scene. CA can be measured since it is always the same, no matter what scene you photograph.

So, in the photozone reviews what is measured is CA, not PF.