Will the owners of Canon SD600 please stand up?

[DougPaul]

But then if the camera is doing the right thing, pictures shouldn't be underexposed, after all I did use the Snow mode for snow scene. I'll try again this weekend, using both Manual and Snow mode, and overexpose the pictures. Also avoid using wide angle to reduce the vignetting effect.

Proper exposure can be difficult--ideally one should use the maximum exposure that just barely misses saturating ("blowing out", over exposing) the brightest pixel. But what if there is a very small very bright object? Then the main part will be underexposed, so under certain situations, saturating a few pixels may give the best overall effect. In difficult situations, you need to take the picture in order to find the exposure setting necessary to take the picture... (This is where raw image formats are helpful--they have a greater dynamic range than JPEG so one has a better chance of setting the effective exposure correctly in postprocessing.) Practical cameras use much simpler schemes to estimate the exposure and may make poor choices and no automatic procedure can read the photographer's mind to incorporate the artistic judgements. Many digital cameras can display a histogram of the pixel brightnesses which can help you to judge the exposure of a pic that you have just taken. See http://www.luminous-landscape.com/tutorials/understanding-series/understanding-histograms.shtml for more info on histograms.

Doug, the Snow mode do not allow manual ISO setting.

The trade-off between automatic mode groupings and control... There is a table in my Canon A75 manual that shows what the mode settings do, perhaps also in your SD600 manual. I would guess that it would assume that snow scenes are relatively bright and tend to choose a fairly low ISO setting if possible.

Hmmm... for a point and shot, I actually think Canon SD series are not cheap.

I was comparing to DSLRs... In general most P&S camera designs trade-off image quality and control for small size and lower cost. Most have fairly small image sensors which tends to increase the quantum noise in the images. (Some DSLRs have a full 35mm film sized sensor.)

The color shift and increased blue noise towards the sides of the image is somewhat puzzling to me. Vignetting is a function of how far off-center the pixel is (ie left-right and vertical), not just left-right which suggests that it is not the underlying cause of the problem. Most optical (lens) problems would not cause this pattern of changes--my guess is that it is more likely to be sensor and/or shutter related. I would have to know more about the detailed construction of the image sensor to hazard a guess whether it might have a problem. It is also possible that the raw-to-image conversion algorithm changes somewhat as a function of left-to-right position on the sensor to compensate for something and is mismatched to the problem giving the net blue-red-blue cast that we are observing. (If so, the problem might be inherent in the design, but might show more in some units than others.)

Doug

 

[Grumpy]

... Vignetting is a function of how far off-center the pixel is (ie left-right and vertical), not just left-right which suggests that it is not the underlying cause of the problem. Most optical (lens) problems would not cause this pattern of changes--my guess is that it is more likely to be sensor and/or shutter related. ...

To me, the arched (semi-circular) pattern of the vignetting is an important clue.

Most optical-mechanical vignetting due to lens shades or filter rings (or construction of the lens internally) is circular, impinging most heavily in a square image on the four corners and then less so on top-bottom and left-right edges. But we are not dealing with a square image here; we have one that is slightly rectangular (1 : 1.33). A circular shade used with a rectangular format oriented horizontally well might vignette noticeably, left-right yet hardly at all, top-bottom.

That most of our shooting formats are rectangular gives rise to those new fangled fancy scalloped lens shades, especially for zoom lenses. Note that the top-bottom scallops are longer than the left-right ones when the shades are installed to correspond with the image format's long side and short side.

I am not familiar with the type of shutter used in this particular camera. However, over the years I have used leaf shutter cameras and lenses which can, indeed, cause a circular vignetting as the iris mechanism opens and closes. A bit of gunk on the leaves is all it takes. But the high end shutter speed (1/1500 sec) for this camera seems somewhat out of the range for most leaf shutters (which tend to top out at 1/500 sec, or so).

The arched pattern in the vignetting is not consistent with what I've seen in uneven exposure problems arising from various types of focal plane shutters -- even the new multi-leaf designs as compared to the old fabric curtain types. Those more typically do not leave an arch pattern, and the new ones often travel top-bottom rather than left-right (meaning we'd see the effect as a horizontal band rather than vertical band phenomenon).

I am a bit puzzled by the evident tendency to vignette more to the right than to the left, although that could be caused by a slightly off-center mechanical element in the system

So I am left to still think this whole business -- color shift and all -- has something to do with mechanical-optical vignetting. Call me stubborn.

G.

 

[DougPaul]

To me, the arched (semi-circular) pattern of the vignetting is an important clue.

Looks to me like classic off-axis vignetting from the lens. (consistent with your opinion)

I don't think she is using any lens shades or filters.

I am not familiar with the type of shutter used in this particular camera.

<speculation>
I'm guessing that it is a horizontal focal plane shutter, but it is only a guess. If we assume a horizontally moving fabric curtain, it is possible that its speed varies as it moves across the sensor. (Perhaps as a compromise to make the camera smaller.) This might result in a reduced amount of light on the left and right edges of the sensor which could then be compensated for in the raw to image conversion process. The net effect could be higher ISO at the sides. Add some non-linearity (more red at higher intensities, more blue at lower) and imperfect compensation and one could get the observed effect. If the compensation does nothing about the lens vignetting, then it would likely be observable in the corners of the image at short FLs.
</speculation>
The above is a speculation, but it is a hopefully not too implausable guess for a mechanism which might cause the observed effect.

The sensor is a CCD which could also give horizontal gradients if the bucket transfer efficiency were sub-par and the charge shfit were horizontal. If the physical layout were correct and the compensation were a bit off, it could also accout for the observed effects.

I am a bit puzzled by the evident tendency to vignette more to the right than to the left, although that could be caused by a slightly off-center mechanical element in the system

My speculated mechanism could account for any left-right asymmetry. I'd like to analyze an image of a uniform white target before I tried to decide how asymmetric the problems are.

So I am left to still think this whole business -- color shift and all -- has something to do with mechanical-optical vignetting. Call me stubborn.

I think we are both in educated guess land...

Doug

 

[BorealChickadee]

.

BTW, the standard color temp of digital cameras is 6500K. Most computer displays are set around 9000+K--far too blue. Setting a proper color temp (6500K) on your display (if you can) might help.

Doug

Does film have a standard temperature?

I'm wondering if that has anything to do with why someone's scanned prints never look as good on my LCD monitor as on their CRT monitor.

 

[Grumpy]

Doug, if you look at the images you linked us to above – the ones in which you ran up the color saturation – you will observe that we probably are not looking at an asymmetrical phenomenon. The vignetting pattern appears quite consistent left-right and top-bottom, although it is significantly masked by subject details on the left in the first image.

I don’t dismiss the possibility that something may be going on with the sensor.

But my speculation still leans heavily toward an optical-mechanical quirk that evidently shows up in more than one sample of this camera model. While it has been stated that no lens shade or filter was attached (and so those must be ruled out as factors), this still looks like a fairly simple case of underexposure by vignetting wrought by some component in the optical-mechanical system.

My preference is to test simpler, classic explanations first. We are assured this will be done, and I eagerly await the results. If we dismiss those simpler explanations by testing them we can get on to examining the more complex explanations involving sensor design and manufacture.

In short, I look forward to seeing what happens in scenes like this at longer focal lengths in the zoom range (primarily), and at different apertures (secondarily). That should provide a good start.

If the vignetting pattern changes throughout the zoom range and/or at various f/stops we can look at patterns and perhaps isolate the problem in the optical-mechanical system; if it remains stubbornly constant (like Yours Truly) we probably should be looking for sensor related issues.

G.

 

[DougPaul]

Does film have a standard temperature?

Yes, several. Eg. indoor film and outdoor film.

Wet darkroom printing of color pics involves the use of color filters to compensate for the white balance.

I'm wondering if that has anything to do with why someone's scanned prints never look as good on my LCD monitor as on their CRT monitor.

Many of the older LCD monitors are significantly inferior to CRTs. My CRT is much better than my laptop LCD (older). On my laptop I can "adjust" the gamma by looking at it from a higher or lower angle. I've read that some the newer LCDs are pretty good. Certainly you want the image to look good over a range of viewing angles and heights.

Go to a TV store and look at the LCD displays--some vary with viewing angle/height, some do not.

Doug

 

[HuiYeng]

So I went out yesterday to take some pictures. The first few were taken in pair, one zoom all the way out, the other all the way in. I think the picture quality is pretty good for a p&s, I guess I'll just make all my winter shots B/W instead of pulling my hair over the problem...

BTW, do you think it is possible that my big fat hand casts shadow on the lens? I do most of the time have my gloves on when I took those winter shots. That'll be very funny if that's the case .

 

[DougPaul]

BTW, do you think it is possible that my big fat hand casts shadow on the lens? I do most of the time have my gloves on when I took those winter shots. That'll be very funny if that's the case .

I wouldn't worry about your hand small, big, skinny, fat, bare or gloved. If it is in your picture, it will be obvious. The problems that we are discussing here are not caused by your hand.

Haven't looked at the new pics yet--unfortunately the scenes make analysis harder. (The snow pictures are "easy" scenes to analyze because of all the pure white in them.)

I just took my P&S and DSLR and took some pictures of a flat white target. (4 indoor pics (ie wide open apertures): high vs low ISO * wide-angle vs telephoto). The pics all look fine with normal color saturation. I then increased the color saturation as I did with the snow scenes. I can see color noise and shift which varies across the image, but looks like it is due to uneven lighting of the target and vignetting. (The apparent color shift could be due to noise and nonlinearities.) Nothing like the pattern in your SD600--mine are smooth and slowly varying across the images.

I'm considering carting the gear outside (a bit difficult with my bum leg) and shooting another 4 in sunlight (small apertures). Also a sequence with different exposure compensations (see below).

7summits:
If you take a similar set of 8 pics, it would make analysis much easier. All combinations of:
* indoor (low light, wide aperture) vs sunlit (bright light, small aperture)
* wide angle vs telephoto
* max ISO vs min ISO
You can use the P setting on the camera. The target can be a stack of several sheets of white paper, lit as evenly as possible. The target doesn't even have to be in focus--it just has to fill the entire frame. (You might need to take the wide-angle shots out of focus to fill the frame. Just be careful not to block the light on the target if you hold the camera close to the target.)

Note: all of the above pics will look grey because the metering system sets the exposure so that the average of the picture is grey.

If you have exposure compensation, a series of pics of the target (telphoto, sunlit) at {-2,-1,+0,+1,+2} stops at both max ISO and min ISO would also help.

Doug

 

[Grumpy]

... BTW, do you think it is possible that my big fat hand casts shadow on the lens? I do most of the time have my gloves on when I took those winter shots. That'll be very funny if that's the case .

It is entirely possible for a sort of vignetting to be caused by your hand, a finger, or whatever. Your hand would have to do more than cast a shadow on the lens; it actually would have to impinge on the lens' field of view (which is cone-shaped, of course). The closer your hand or whatever is to the lens, the more it will appear as a soft vignetting rather than a (possibly) recognizable object.

But the pattern of vignetting I've seen in your photos is too consistent, and its shape is such that it suggests something other than a hand or finger or glove.

Of the latest photos you did, the ones of the flag probably have greatest potential to help sort this out.

G.

 

[DougPaul]

I took a series of pics of a white target in sunlight (similar to what I described in my previous post) a little earlier this afternoon--I'll analyze them a little later. Perhaps I'll repeat my earlier indoor series with better controlled lighting. They will make a good baseline if 7summits can shoot a similar series with her camera.

Doug