jorkata wrote:
That's why in my previous posts I'm suggesting that Canon should make the 7D in two flavors - one with lots of megapixels for those who like that, and another one will less megapixels and better ISO.
Like I said, you still don't understand this issue at all. "Better ISO", as you call it, isn't brought about by marginal reductions in pixel numbers.
If you took the G11 sensor and extended it to full-frame it would contain 292 Mpixels. A 100 Mpixel full-frame sensor wouldn't necessarily have worse low light performance than the 7D, but it would have latitude more like a very good point+shoot.
4-to-1 pixel binning would make possible a 40 megapixel camera with very crisp 10 megapixel sraw files for those who don't always want big files.
jorkata wrote:
I'm suggesting that Canon should make the 7D in two flavors - one with lots of megapixels for those who like that, and another one will less megapixels and better ISO.
A thread about 800 ISO which just happens to favour the 1D as a response to a question about high ISO performance?
You just can't help yourself, can you?
And - as responses in that thread clearly spell out - it's an unfair comparison anyway.
Enough. It's the Kill File for you.
Bobby, the 40D is the best part of a stop better than the 1D2n at high ISO and from what I can see the 7D approaches another stop over the 40D - plenty of posts to that effect, and posts like this tend to support it:
Just so you know where I'm coming from, I own and really like the 40D, and have used a 1D2n a fair bit, and have a pal who has provided me with a good few high ISO 1D2n files from when I was looking to buy one.
And I will be having a 7D: high ISO performance is a big part of the reason why.
Okies, redid the ADU/Sdev plots - without pixel-doubling... :-) Results are fairly consistent with my first thoughts, but I'm having trouble with my testshots. I can't redo them right now (since it isn't my camera! :-) ) - but I shot the charts/ISO-ranges too well focused. Not very often that you hear those words and "Canon" in the same sentence, huh...? :-) I didn't want the border of the patches blurring too far into the patches themselves (as colour was my main concern), so I aimed focus on the front table-edge - the charts were placed around 70cm back in the inspection cubicle. Light was D50-filtered Halogen, btw.
Either that, or there was some sort of dust/specks gathering on the CC24 - I'm getting slightly too incisistent Sdevs. Anyways, here are the results - first up, Sdev compared to ADU level. One channel is very much more efficient than the other, around 1.6 times according to my (poor accuracy) readings. Error margin +/-10%!
And then, as per request, the scatterplot of the (Green1-Green2) vs Red/GreenAverage saturation following Emils suggestion of a simple
[ (Gr1-Gr2)/GrAverage vs (Gi-R)*(Gi+R) ] to get the correlation between Red channel saturation compared to average green and green channel difference signal.
Same for blue channel saturation.
There's a strong correlation, the only thing messing things up is the fact that the red channel swings back in efficiency as you get colser to the other end of the spectrum, near blue-violet. Therefore, the blue channel (which is a single peak, no swingbacks or plateaus) shows a much stronger correlation. In fact, a VERY strong correlation.
So, there's very strong evidence that:
1) the two green channels are not the same efficiency (one has a "diluted" colour filter) and:
2) one green channel is significantly more yellow than the other.
Not of interest to most people I guess, but some of us really want to know the "why" in some of these things...
Can someone with a 7D post comparative shots with RAW,sRAW,mRAW at 100% and at same size. Perhaps something with shadows at a few different ISO (200,800,3200).
I'm interested in general sharpness and noise. Is that asking too much?
brainiac wrote:
Like I said, you still don't understand this issue at all.
Here's a quote from you essay: Is there a downside to jamming in more sensels?
Yes - probably: the smaller a well, the fewer photons it takes to saturate it.
So, let me put it straight to you:
I want a 1.6x sensor with larger wells than the 7D, that saturate with more photons.
I repeat, larger wells is what I'm after.
If this requires the number of megapixels to be reduced copmpared to the 7D, I'll happily accept it - as long as the megapixels count is not too low.
In fact, 12mp with larger wells is what I'd like and pay for.
The current 7D performance (as well as image downsizing) does not work for me.
I'm glad that it works for you, guys, and totally respect your decision to buy the 7D.
I don't get it. You post actual pictures? But who cares? We want pixels, facts, graphs and math, not pictures. What's the point of using the camera this way?
theSuede wrote:
Okies, redid the ADU/Sdev plots - without pixel-doubling... :-) Results are fairly consistent with my first thoughts, but I'm having trouble with my testshots. I can't redo them right now (since it isn't my camera! :-) ) - but I shot the charts/ISO-ranges too well focused. Not very often that you hear those words and "Canon" in the same sentence, huh...? :-) I didn't want the border of the patches blurring too far into the patches themselves (as colour was my main concern), so I aimed focus on the front table-edge - the charts were placed around 70cm back in the inspection cubicle. Light was D50-filtered Halogen, btw.
Either that, or there was some sort of dust/specks gathering on the CC24 - I'm getting slightly too incisistent Sdevs. Anyways, here are the results - first up, Sdev compared to ADU level. One channel is very much more efficient than the other, around 1.6 times according to my (poor accuracy) readings. Error margin +/-10%!
I'm confused. On a log-log plot, the slope is the power law p in
y = c * x^p
so different slopes mean different power laws. So the different slopes must be experimental uncertainty -- the slope should always be 1/2 for photon shot noise. If the lines were parallel, then the vertical offset of the two lines would be the log of the ratio of the two efficiencies.
Not of interest to most people I guess, but some of us really want to know the "why" in some of these things...
If true, the deviation from standard RGGB Bayer color filters will make 3rd party RAW converter support for the 7D much more difficult. For instance, it really messes up the interpolation algorithm I have developed.
That's great, but you keep stating that larger wells will give you "better iso" by which presumably you mean better high iso performance in low light, i.e. at the underexposed end of the range. It won't. If instead you said you would prefer larger wells for the sake of higher latitude you would show that you had understood what I and others have been trying to put to you.
Shot with polarizer? That's a nice blue tone for a NewMexico sky... The sun seems to be a fair bit up in the sky - if the EXIF is ok, then the shots are taken around 9 to 11 in the early noons... :-) One of the things the 7D is/will be very good for... - Birding!
Here's a good piece of information, copied from the Adobe LR2.5 release info. So they will fix it when the "non-beta" 7D profile hits the market (which will be in two weeks according to my ninformation)
Release Notes:
* Camera Raw 5.5 and Lightroom 2.5 include a correction to the demosaic algorithms for Bayer sensor cameras with unequal green response. Olympus, Panasonic and Sony are among the more popular camera manufacturers affected by this change. But the demosaic correction provides only a subtle visual improvement to the processing of those raw files.
brainiac wrote:
That's great, but you keep stating that larger wells will give you "better iso" by which presumably you mean better high iso performance in low light, i.e. at the underexposed end of the range. It won't.
With larger wells, S/N ratio is better as well - at the pixel level. Hopefully you are not going to argue about that.
Not interested and don't care about image level comparisons at all.
brainiac wrote:
That's great, but you keep stating that larger wells will give you "better iso" by which presumably you mean better high iso performance in low light, i.e. at the underexposed end of the range. It won't.
jorkata wrote:
With larger wells, S/N ratio is better as well - at the pixel level. Hopefully you are not going to argue about that.
Not interested and don't care about image level comparisons at all.
That's fine, just don't call it "better ISO" since that is a misleading and inaccurate characterization of what you are after.
dwweiche wrote:
I think this means people should start posting more images for the rest of us to look at
Although the little web-size images people have been posting could have been taken by the 20D. Not much IQ difference if one is going to downsample that much.
