Moderator: E.J. Peiker

All times are UTC-05:00

  
« Previous topic | Next topic »  
Reply to topic  
 First unread post  | 73 posts | 
by ColorChange on Mon Dec 01, 2008 10:31 am
ColorChange
Forum Contributor
Posts: 593
Joined: 30 Jun 2005
Hmmmmmm, I lean toward Emil and what I see, but how can we get further proof/evidence? I'd really like to track this down.
Tim
 

by Bill Lockhart on Mon Dec 01, 2008 10:37 am
User avatar
Bill Lockhart
Lifetime Member
Posts: 3058
Joined: 29 Sep 2003
Location: Safety Harbor, Florida
Member #:00215
Emil,

You imply that your own measurements found the LX3 sensitivity to be higher or near what Panasonic states it to be. Thus negating the findings published by DxO. Bottom line for me is whether the LX3 when set to ISO 400 is identical to the G10 set to ISO 400. Are they the same?

Best regards,

Bill
Bill Lockhart
[url=http://www.phototravelreview.com]Photo Travel Review[/url]
[url=http://www.bill.lockharts.com]Personal Website[/url]
 

by ejmartin on Mon Dec 01, 2008 11:29 am
ejmartin
Forum Contributor
Posts: 2693
Joined: 22 Oct 2005
Location: Chicago, IL
You could run the same test I had the others do for the LX3 and I can analyze the results:

I'd like shots of the color chart from each camera (DLSR and G10), taken at the same ISO (say ISO 100), at the same shutter speed and aperture and effective focal length in identical lighting. Probably good to stop down from wide open a stop or two to reduce the influence of vignetting on the results.

And just to satisfy my curiosity, although you are going to be setting the shutter speed and aperture manually, could you also note down what the camera's meter is reading (ie how many EV off from correct exposure the camera thinks is occurring)?

It would be sufficient to work indoors with tungsten light, or outdoors on a cloudless day, so long as the light is constant. Either way, use a tripod to ensure consistency.
emil
 

by E.J. Peiker on Mon Dec 01, 2008 11:35 am
User avatar
E.J. Peiker
Senior Technical Editor
Posts: 86761
Joined: 16 Aug 2003
Location: Arizona
Member #:00002
I would love to see that Emil - still on the fence whether or not to get one of these as my toss it in the briefcase camera.
 

by Eric Chan on Mon Dec 01, 2008 12:22 pm
Eric Chan
Forum Contributor
Posts: 1945
Joined: 10 Sep 2004
Location: Boston, MA
Member #:01107
emil, the saturation-based metric that DxO uses based on ISO 12232 has nothing to do with how the camera meters. It is only based on the actual exposure (measured in lux-seconds) and the recorded raw level.

One possible reason for a discrepancy is that for pocket cameras, they don't remove the lens (i.e., they don't disassemble the camera), so they throw in a fudge factor to compensate for estimated lens transmittance, effective aperture, etc. So a lot more variables are involved in the case of pocket cameras, and perhaps one of their fudge factors is a bit off.

Also note that measured results will depend significantly on which illuminant you use. A typical D55-like sunny day (one of the two standards in ISO 12232) has a very different spectrum compared to a tungsten-filament bulb (the other standard in ISO 12232). The fact that two different camera sensors have different RGB filters complicates the comparison. I'm not sure what illuminant DxO Labs is using.

Finally, where the camera decides to meter is a completely separate and independent topic from the numbers DxO reports. DxO's ISO measurements are really more about the choice of gain in the amplifier, rather than the choice of where to meter.
Eric Chan
[url=http://people.csail.mit.edu/ericchan/photos/]MadManChan Photography[/url]
 

by ejmartin on Mon Dec 01, 2008 1:04 pm
ejmartin
Forum Contributor
Posts: 2693
Joined: 22 Oct 2005
Location: Chicago, IL
madmanchan wrote:emil, the saturation-based metric that DxO uses based on ISO 12232 has nothing to do with how the camera meters. It is only based on the actual exposure (measured in lux-seconds) and the recorded raw level.


You're right, and that use of actual exposure takes away the uncertainty of the camera's metering. What I had in mind is that if the camera's meter is accurately calibrated to the ISO standard, then it places middle grey somewhere in the RAW data. For practical use it is relevant to know where that point is relative to saturation. That's certainly what I want to know, since that tells me how much headroom I have between the metering and the saturation point. Of course, for the test outlined above, one is simply trying to compare for a given ambient illumination the relative sensitivity between two cameras, and the issue of metering doesn't enter.
One possible reason for a discrepancy is that for pocket cameras, they don't remove the lens (i.e., they don't disassemble the camera), so they throw in a fudge factor to compensate for estimated lens transmittance, effective aperture, etc. So a lot more variables are involved in the case of pocket cameras, and perhaps one of their fudge factors is a bit off.
Ah, that's good to know. I was wondering how they took that into account for digicams, since the test procedure is necessarily different.
Also note that measured results will depend significantly on which illuminant you use. A typical D55-like sunny day (one of the two standards in ISO 12232) has a very different spectrum compared to a tungsten-filament bulb (the other standard in ISO 12232). The fact that two different camera sensors have different RGB filters complicates the comparison. I'm not sure what illuminant DxO Labs is using.
OK, you're the expert here. How variable are the spectral responses of CFA filters between different cameras (say in green)? I have been assuming that choice of illuminant wouldn't strongly affect my proposed test, but perhaps that's not the case.
Finally, where the camera decides to meter is a completely separate and independent topic from the numbers DxO reports. DxO's ISO measurements are really more about the choice of gain in the amplifier, rather than the choice of where to meter.
Are these two all that different? Assuming that the camera manufacturer knows how to make an accurate light meter, then how many stops the camera puts between metered middle gray and saturation in the RAW data is essentially the choice of how much gain is applied to the signal. If the camera's meter is poorly calibrated, then of course metered middle grey need not be closely tied to a particular light flux.
emil
 

by prashant on Tue Dec 02, 2008 8:37 am
prashant
Forum Contributor
Posts: 910
Joined: 25 Aug 2003
Location: UK
after reading part of this post, I'm hoping that SIGMA comes up with a DP3 or whatever with 28-35-50 kind of leica lens, darn am ready to pay even 1000.
[b]PrashanTeju Khapane[/b]
[i]Photography, Paintings & Travelogues [/i]
http://www.prashanteju.de
 

by Eric Chan on Tue Dec 02, 2008 11:45 am
Eric Chan
Forum Contributor
Posts: 1945
Joined: 10 Sep 2004
Location: Boston, MA
Member #:01107
How variable are the spectral responses of CFA filters between different cameras (say in green)? I have been assuming that choice of illuminant wouldn't strongly affect my proposed test, but perhaps that's not the case.
In general, they are quite different (enough to warrant different color profiles for different cameras, for instance). For example, the 1Ds Mark III and 5D Mark II have similar sensor and pixel sizes, but different CFA spectral transmittance functions.

As another example, the 1D Mark II and (original) 5D have similar pixel sizes but different CFA spectral transmittance functions. If you photograph a spectrally-flat gray card in shade (which has a lot of short wavelength energy, i.e., "blue" light, with correlated color temperature > 7500 K), you will find that the 5D and 5D Mark II clip the green channel first, but the 1D Mark II clips the blue channel first. In contrast, if you photograph the same gray card under a tungsten-filament bulb, all 3 cameras clip the green channel first. Thus, if one is considering a saturation-based ISO metric, the choice of illuminant plays a role, which is why ISO 12232 requires reporting the measurement illuminant used (and also recommends two specific illuminants).

If you are just trying to evaluate two cameras against each other, then you don't need a lux meter but it is still worth noting which illuminant you used for the measurements (even roughly). Even just shooting blue sky on a sunny day or a patch of gray clouds on an overcast day gives you a very good estimate of the actual spectrum (roughly D55).
Finally, where the camera decides to meter is a completely separate and independent topic from the numbers DxO reports. DxO's ISO measurements are really more about the choice of gain in the amplifier, rather than the choice of where to meter.
Are these two all that different? Assuming that the camera manufacturer knows how to make an accurate light meter, then how many stops the camera puts between metered middle gray and saturation in the RAW data is essentially the choice of how much gain is applied to the signal. If the camera's meter is poorly calibrated, then of course metered middle grey need not be closely tied to a particular light flux.
They are certainly different (or can be). As you know from your own studies, the gain of the amplifier (i.e., the conversion factor between electrons and raw levels, or data units) is a property of the sensor design, rather than the meter design. Two cameras with the same gain could meter differently. Even the same camera with a known gain can meter slightly differently, under different occasions. For example, one of my cameras consistently spot meters 1/3 stop brighter in Manual mode vs Aperture Priority mode. That is, if I use Av mode and spot meter a given patch (with no exposure compensation), I get one reading. If I use Manual mode and spot meter the same patch, I get a reading that recommends adding 1/3 stop more light. In other words, Av mode on this camera will underexpose relative to Manual by 1/3 stop, leaving 1/3 stop more highlight headroom.
Eric Chan
[url=http://people.csail.mit.edu/ericchan/photos/]MadManChan Photography[/url]
 

by ColorChange on Tue Dec 02, 2008 12:29 pm
ColorChange
Forum Contributor
Posts: 593
Joined: 30 Jun 2005
Eric, enlightening. I wouldn't have thought the differences (especially within manufacturer's lines) would be significant. It would be interesting to test.
Tim
 

by ejmartin on Tue Dec 02, 2008 1:30 pm
ejmartin
Forum Contributor
Posts: 2693
Joined: 22 Oct 2005
Location: Chicago, IL
OK, but I'm interested in a slightly different question -- not the relative transmittances of different color filters in the CFA between different cameras, but the differences in spectral response function for a given color between different cameras, and how that interacts with the spectral distribution of the illuminant. That's what is relevant if I run a test comparing different cameras by working on only the green channel, but I have a variety of side-by-side (DSLR vs digicam) samples taken under different illuminants.
emil
 

by Eric Chan on Tue Dec 09, 2008 8:09 pm
Eric Chan
Forum Contributor
Posts: 1945
Joined: 10 Sep 2004
Location: Boston, MA
Member #:01107
Oops, I missed seeing this earlier ...

... anyways, Emil, yes, given a certain color channel (green, in your case), the spectral transmittances curves will be somewhat different from sensor to sensor. They don't tend to vary that much within a generation of sensors used by a given camera maker (e.g., I'd expect the spectral transmittance curve for the green filter of a Canon 1D Mark III to be similar to that of a 1Ds Mark III). But it's certainly enough to introduce a dependency on the lighting.

For example, if we just analyze the green pixels for a 1D Mark II and 5D Mark II (a few generations apart), we see that for a given exposure (e.g., ISO 100, f/8, 1 second), the 5D Mark II underexposes by 1/3 of a stop relative to the 1D Mark II under cloudy daylight (roughly D65, ~6500 K) but has the same exposure as the 1D II when the same subject is photographed under a tungsten-filament bulb (roughly illuminant A, ~2850 K). This is why ISO 12232 recommends 2 specific illuminants to report, with quite different spectral characteristics, and asks designers to report the actual illuminant used in case it deviates from the two recommendations.

That said, picking the green pixel is probably the safest thing to do for your studies, since there is less variation in its response curve among sensors compared to the other two filter types.
Eric Chan
[url=http://people.csail.mit.edu/ericchan/photos/]MadManChan Photography[/url]
 

by ejmartin on Tue Dec 09, 2008 10:41 pm
ejmartin
Forum Contributor
Posts: 2693
Joined: 22 Oct 2005
Location: Chicago, IL
Thanks, Eric.
emil
 

by fr0z on Tue Dec 09, 2008 10:48 pm
fr0z
Forum Contributor
Posts: 127
Joined: 18 Jan 2005
Location: Finland
Here is my noise comparison pics -> http://www.janneheimonen.net/blog/tekni ... atiloissa/ , text is Finnish, but there is quite a little information there =)
Pictures are converted from raw with newest adobe camera raw.
Janne Heimonen
http://www.janneheimonen.net/
http://www.janneheimonen.net/blog/
 

Display posts from previous:  Sort by:  
73 posts | 
  

Powered by phpBB® Forum Software © phpBB Group