7D or 1DIV: better noise?

[Doug Brown]

Hi Doug,
Is it possible to include SS, Fstop, and subject distance for the two images?
Also, can we assume the same exposure adjustment for both in LR?
thanks!
Enrique

See my two posts below for the information you requested. Subject distance is not relevant IMO because as I stated earlier, I try to have my subject bird occupy a relatively constant percentage of the frame regardless of focal length and crop factor. For flight I frame a little loose as a general rule.

[Doug Brown]

Doug Brown, I believe what Roger is implying that if you are the same distance from the bird using both cameras then to get the bird to fill the same same of the final frame you will be adding an extender, increasing the f-stop by 1 stop and then increasing ISO by 1 stop also to create the same shutter speed.

Your scenario assumes that I am shooting wide open before I put on an extender. If I'm shooting f/5.6 or smaller (which is most of the time), I don't have to change my ISO at all when I put on an extender.

[Alan Stankevitz]

At least in my world, I am focal-length-limited during many occasions. Here are three images taken with a 5DM2, a Mark IV and 7D. All taken at 40-feet distance of a pheasant feather:

5DM2:


Mark IV:


7D:


These 3 images are all at ISO 800 taken with my 600mm f/4 Canon lens.

In post processing I typically apply noise reduction (Topaz Labs) and then crop the image in Photoshop typically to 300 dpi which is usually what my publishers are requesting. I crop the image to frame the bird that is pleasing to the eye. Given the three images above, I perform post processing on all three. Obviously the 5DM2 image will require a lot more cropping to obtain the same field of view. The heavier that you crop an image, the more noise will be revealed in the image. The Mark IV's image will require moderate cropping and the 7D's image will require very little cropping at all.

In the following image, I have taken a screen shot showing all three images cropped identically. The end results are very, very close:


The 7D's image did require more noise reduction, but it also has higher resolution to start with and this equalizes out the final image when compared to the Mark IV and 5DM2.

Please note: I am not trying to convince anyone that the 7D is as good of a camera as the Mark IV. I am just stating that (at least in my world) I am photographing quite often from a fixed distance to the bird and in such instances, the image quality is very similar after post-cropping the image to the same field of view.

Alan

[arash_hazeghi]

Arash,

Examples of focal-length-limited scenarios in which I cannot move closer to the bird:

Example 1: I am in Florida photographing birds. Yes, the birds may be tame but I am photographing a bird from the boardwalk at Wakodahatchee Wetlands. I cannot move any closer to the bird or I will end up in the drink.

Example 2: I am photographing a Blue-Winged Warbler that is moving from branch to branch 30' off of the ground. I can only move closer if I climb the tree.

Example 3: I am photographing a Bald Eagle from Lock and Dam 14 in Le Claire, IA. I have my gear set up on the I-wall of the dam. I cannot move closer to the eagle unless I end up in the lock.

Example 4: I am photographing Short-Eared Owls hunting over a grassland. Although I may attempt to move closer, I am at the mercy of where the owl is flying at the time that I shoot. By the time I would attempt to move closer, I would miss my shot.

I could continue with examples of focal-length-limited scenarios, but I'm sure everyone is already bored with my post. There are plenty of times in which you are chained to a tree as you put it.

Alan

Thanks Alan,

I don't know about your technique and how you choose your locations but I have photographed in Florida too, last time I was with Doug Brown, Jim Neiger and Jim Shadle. We were never focal length limited. Perhaps we put more effort into choosing our locations and positions. In fact I got many shots like these with a 400mm lens and MK4.



Osprey, Florida 1D4 420mm taken from Jim Neiger's boat. almost FF horizontal cropped to vertical.

As for my own photography I am rarely focal length limited, I hand hold my gear and somehow I manage to get close enough to the birds and take the picture that I want So what if the bird is too far? I just don't take a picture. I think it depends on one's style and technique. At the end of the day what matters is the results. One picture is worth a thousand words.

So for me, bird photography is not focal length limited. It also happens that all of the photographers that I personally know and admire don't think they are FL limited and everyone one of them uses a MK4 or a FF Nikon.

I am not sure if anyone on this planet can even remotely come close to the level of perfection Alan Murphy has achieved in bird photography, Alan's award winning images are made with Nikon FF cameras that are "only" 12 Mpixels so they have the minimum "reach" but his images are miles better in aesthetics and quality than many of the photographers who think they should use a 7D and a 800mm lens to "fill" the frame with the bird from 200 yards away ;)

So I don't think anyone can generalize bird photography as a "focal length limited situation".

Please stay tuned for the article that we will publish soon, maybe you will find it useful.

[arash_hazeghi]

A These 3 images are all at ISO 800 taken with my 600mm f/4 Canon lens. In the following image, I have taken a screen shot showing all three images
cropped identically. The end results are very, very close:
The 7D's image did require more noise reduction, but it also has higher resolution to start with and this equalizes out the final image when compared to the Mark IV and 5DM2. Alan

Good example Alan, now which of these three crops is your personal pick? You have shown that the visual difference is very small, right? and I agree they all look good and your processing is good too, great job on that. You had to spend more time on the 7D file to get it there and time is money, but let's ignore that for now :)

Now let's change things a little bit in your FL-limited scenario. You are shooting BIF so you need fast SS, and the light is not great like your example. You will have to bump up your ISO to 1600 and you will have shadows in parts of the image, you need to raise those shadows and apply NR to those areas and then sharpen up as well. If you did your optimized processing which camera do you think would have the best output? Here is an example for the condition I just described


Attachment 107445

1D4 ISO 1600 1/1600sec and here is a high def version to see the IQ I expect.

Here is the original to see how much I have pulled those shadows

Attachment 107446

Hope this helps

[Roger Clark]

And here is the Mark IV full frame image. I pulled the exposure down 1/3 stop in Lightroom on this frame to make the exposures comparable.


Canon 7D, 600mm, f/5.6, 1/1000, ISO 800
Canon 1D Mark IV, 600mm, f/4, 1/1000, ISO 1000

Doug,
These two situations are not at all comparable. The light had changed a lot. First, changing ISO does not change sensitivity nor the amount of light collected by the sensor. If you reversed the two cameras, it would be the 7D image that looked less noisy and much better than the 1DIV.

Here is why: The shutter speeds are the same, but the f/ratio is 1 stop darker than the 7D image, so the light recorded in the 7D image is down a factor of 2, so of course the image will look noisier. Second, the perception of the noise in the 7D is made stronger by the slightly lower than mid-level gray background. Again switching the scenarios with the 7D image with the bright background would have made the 1DIV look worse. Finally, the bird in the 7D image is further away, which means you have less light for a given amount of detail. So your example is not only not equal, but biased against the 7D in 3 major ways. Not a fair test at all.

I have 1DIV and 7D cameras and photograph with both of them. When I first started with the 7D, I thought it had more noise. But then I thought about what the differences were and realized the situations were different. When I equalized conditions, I found noise very very close, and in measurements, the 7D came out slightly better. The physics does not change from lab to field, but perceptions do.

If one is able to move closer to the subject to fill the frame, then the larger frame sensor will do better, period. In that case the new 1DX will be wonderful for you.

I do agree that there are situations where one can get close to one's subject, but there are also situations where one can't. It is fine to limit one's photography to such ideal situations. But there are situations where one can't get close to the subjects, and knowing one's equipment can make the best of the situation. While some situations have been cited, I'll add one that is significant for me: Africa. There are some locations where off road driving is not allowed and one is not allowed to exit the vehicle (too dangerous). Thus one is limited to how close a subject may be. Yes, there are many situations where it is too far. But a rare subject often occurs in focal-length limited situations.

But this thread was about whether or not the 7D was noisier than the 1DIV. Sure one can make up scenarios where either camera fails, but that doesn't answer the real and basic question of noise performance. The noise performance when the systems are made under equivalent conditions is clearly given by the data I and Alan presented in the thread.

Roger

[Doug Brown]

The shutter speeds are the same, but the f/ratio is 1 stop darker than the 7D image, so the light recorded in the 7D image is down a factor of 2, so of course the image will look noisier.

I'm confused. The two frames, which BTW were taken 2 years apart, have nearly identical exposures on the subject (off by 1/3 of a stop give or take). There was more ambient light when the 7D frame was taken so I was able to stop down. If the exposures are essentially the same, how is it that the light recorded in the 7D image is down by a factor of 2?

[Alan Stankevitz]

Arash,

I have both cameras, so I have no bone to pick with either one. And as far as post-processing, the amount of time I spend adjusting the 7D vs a Mark IV's image is the same. I just apply a slightly heavier noise reduction level using Topaz Labs DeNoise 5. As an example of a 7D, ISO 1600 image, here's a before and after of a SE Owl taken after sunset. The first pick is uncropped, in its original field-of-view:



The second image, is a cropped version of the first with the shadows brought out as well:



If I would have taken this same shot with the Mark IV on my lens, I would have a lower noise image to start with, but I would have had to crop the image to a greater extent to obtain the same field-of-view.

On another note, here's a link to my best of 2011 bird images. Most of these images were taken with the Mark IV, but near the bottom of the page are a number of SE and Snowy owl images that were taken with the 7D. I love both of these cameras dearly for bird photography:

http://iwishicouldfly.com/iwishicoul...ml/123111.html

Cheers,

Alan

[arash_hazeghi]

your owl shot looks great at 1024 pixels, but can you post 2000 pixel image so we can see how details hold?

BTW, it's not quite like my example here's why: Your lightning was uniform, in my example light was coming from the side so in order to protect the whites I had to expose for the whites and thus the deep shadows. The difference in shadows and highlights is about 6-7 stops or more in my example so it had much higher DR. Your photo was underexposed by maybe 1.5 stops but there wasn't much DR needed :)

it's a nice photo for sure, good wing position and excellent HA! consider moving the owl to the right and bit more space on the bottom and pulling up by another half stop ;) sorry I really liked it so I came up with some suggestions.

[arash_hazeghi]

Oh, and nice snowy shots my pick is the first snowy and the photo right above that.

Please do share your images with is in the avian forum.

[Roger Clark]

I'm confused. The two frames, which BTW were taken 2 years apart, have nearly identical exposures on the subject (off by 1/3 of a stop give or take). There was more ambient light when the 7D frame was taken so I was able to stop down. If the exposures are essentially the same, how is it that the light recorded in the 7D image is down by a factor of 2?

Hmm, you've added yet more variables making comparison more difficult. If you send me the two raw files I can analyze them and tell you how close the exposure actually was. Having the two images two years apart makes the test pretty uncertain. A linear conversion of the raw file is necessary (e.g. with DCRAW). But regardless of the exposure, assuming the birds are close to the same size, the 7D image was further away. That changes the light.

Roger

[BenBotha]

A competent bird photographer does not chain themselves to a tree and photograph a far subject and then try to crop the image. Heavy crops and farming noisy pixels is the number 1 reason for those lousy IQ images that we often see posted. To think bird photography is generally "focal length limited" is ridiculous IMO and it indicates a short coming on the photographer side. If the birds are too far and you can't get close enough you should not waste those electrons taking a picture at first place...We will soon put together an article that not only scientifically compares noise based on RAW data and spectral analysis it will also provide field-relevant visual examples to see what differences there are. Please stay tuned, the article will be posted soon and many of these issues will be addressed.

In my opinion a slightly simplistic view.
Where I am from:I am not allowed to get out of the vehicle.
If I do get out of the vehicle I will either be eaten by a lion or crocodile, trampled by an elephant or crushed by a hippo.
I am also cash limited.
Thus, being cash limited, focal length limited and probably talent limited I have to disagree with some of your comments.
Going on a once of a lifetime trip and taking photographs IMHO is not a waste of electrons.I do agree that cropping, in my situation, is the most obvious reason for the "lack" of IQ in my photos. But then again, I also gave a list of reasons that would permanently limit my lifespan if I try to to get closer to some of my subjects.

Ben

[arash_hazeghi]

In my opinion a slightly simplistic view.
Where I am from:I am not allowed to get out of the vehicle.
If I do get out of the vehicle I will either be eaten by a lion or crocodile, trampled by an elephant or crushed by a hippo.
I am also cash limited.
Thus, being cash limited, focal length limited and probably talent limited I have to disagree with some of your comments.
Going on a once of a lifetime trip and taking photographs IMHO is not a waste of electrons.I do agree that cropping, in my situation, is the most obvious reason for the "lack" of IQ in my photos. But then again, I also gave a list of reasons that would permanently limit my lifespan if I try to to get closer to some of my subjects.

Ben

Ben,
I am not sure if I buy this argument, the 7D pixel size is 4.3um 1D4 is 5.3um the difference in "reach" is 30%. So are you saying being 30% farther (say 39 ft instead of 30 ft) saves your life? Like a lion can't attack from 39 ft but it can attack from 30ft? Am I missing something :)
If you have a look at the wildlife forum many of the most impressive wildlife photographs are taken with the 1D4 or Nikon FF cameras so the photographers were not eaten My friend Oliver Klink who has one of the best grizzly portfolios uses FF cameras to get his bear shots. Chas also uses FF and 1D cameras for his work...see his comments on a similar thread...

have a look at these amazing wildlife portfolios:

http://www.oliverklinkphotography.com/
http://www.shootthelight.com/#/Portfolio/Wildlife/6/


Also I was mainly talking about professional wildlife/bird photography, of course some people take snapshots of wildlife with point and shoot cameras, so obviously to them is not a waste of electrons...

[Roger Clark]

All,
This really shouldn't be that difficult. But I see a lot of emotion here. So I put together one more test. This time 1DIV versus 7D. I set up a bird model in shade against a bright blue sky background. I imaged the bird first with a 1DIV with 300 mm lens + 1.4x TC. Manual mode, I chose the exposure that did not overexpose the blue channel on the sky. This made the bird underexposed so it would enhance any noise and dynamic range differences between the two cameras. This setup, if anything, biased for the 1DIV. Then I put the 7D on the lens without a TC, but kept the aperture the same diameter, and kept the shutter speed the same. ISO is irrelevant (remember, the amount of light captured does not change with ISO). The ISO was set to 800 on the 1DIV. The 7D having slightly more efficient pixels would overexpose the sky, so I moved back to ISO 400. Remember, there is no change in aperture diameter or shutter speed, so the light per pixel is as identical as one can make it (to better than about 0,1 stop). Finally, because the pixel pitch ratio is 5.7/4.3 = 1.33x, and I had a 1.4x TC with the 1D4, it meant the 1D4 pixels on subject would be more by 1.4/1.33 = 1.05 or 5%. To compensate for this, I moved 5% closer for the 7D images. Images were converted with photoshop CS5 with all the same settings, except chromatic aberration corrections were optimized independently for each image (a very small correction for a 300 f/2.8 L IS lens). Because of gain and pixel efficiency differences, I did a slight post processing curves adjustment to eqiualize the intensities (this was less than about a 5% correction, or 0.07 stop). Such a correction does not change signal-to-noise levels, nor total recorded dynamic range.

First image is here showing the bird as seen by the two cameras, reduced by about a factor of four from the original image. Which one is the 7D and which the 5DIV?

Roger

[Roger Clark]

Second image: A full pixel crop of the region around the head. Which is the 7D and which is the 1DIV? Which image has more noise? Which image has more dynamic range?

[Roger Clark]

Third image: A full pixel crop of the region around the leg. Which is the 7D and which is the 1DIV? Which image has more noise? Which image has more dynamic range?

[Roger Clark]

Fourth image: A full pixel crop of the region around the head, stretched to show the shadow noise. Both images were stretched identically from 16-bit images. Which is the 7D and which is the 1DIV? Which image has more noise? Which image has more dynamic range?

[Roger Clark]

Fifth and final image: A full pixel crop of the region around the leg, stretched to show the shadow noise. Both images were stretched identically from 16-bit images. Which is the 7D and which is the 1DIV? Which image has more noise? Which image has more dynamic range?

[Roger Clark]

Some observations: I set the black point to zero in the raw converter (I always do this). The fact that the blacks are not as rich in one camera is mainly due to a different offset (most cameras have a built-in offset so the noise does not hit zero, clipping shadow detail). One could set the offset differently and the two images would look closer. This is biasing the perception of image quality. Try and ignore that and see which image shows more detail in the shadows. The noise levels, shadow detail, and dynamic range are very close. The resolved detail is slightly different in each and I'll give my assessment after I reveal which camera is which. The color is slightly different between the two cameras.

Roger

[Enrique Patino]

I'll play...

Image on the left is from a 7D. It appears to have more DR, and a bit more noise.

[arash_hazeghi]

The above test is invalid, 7D was ISO 400 and 1D4 was ISO 800. The higher ISO means the light was weaker for 1D4 and therefore signal had to be amplified to match exposure and result in identical histogram. Identical exposure means identical shutter speed, aperture and ISO. ISO is the amplifier gain, it affects the results. If light was equal ISO should have been exactly the same, so something was wrong...the light captured by sensors was not identical.

The images are not equal in processing and contrast either.

There is no significant mismatch between 7D and 1D4 gain calibration according to DxO measurements so ISO should match. If anything 7D ISO is overrated, so the correct ISO for 7D should have been 1000 not 400.

1D4
http://www.dxomark.com/index.php/Cam...#measuretabs-1

7D
http://www.dxomark.com/index.php/Cam...#measuretabs-1


why not crank up the 1D4 ISO to 6400 and compare with ISO 100 on the 7D, if that's irrelevant? I guess that will prove 7D is a better camera, right

[Luis Villablanca]

I agree, the images on the left are from the 7D. If I trust my eyes, they seem to have some banding, which is something I usually find on the deep shadows of my 7D, but not on my 1D.

[arash_hazeghi]

For the record, ISO definitely affects SNR at the output, because any MOSFET amplifier has more noise at higher gain which degrades SNR further. According to DxO measurements (they are careful and thus industry standard) 7D real ISO at 400 is 325 with SNR for 18% grey = 33.5dB, at ISO 800 (real ISO 635) 18% grey SNR drops to 31.5dB that is 2dB.

you can see SNR and DR measurements on DxO website.

[arash_hazeghi]

I agree, the images on the left are from the 7D. If I trust my eyes, they seem to have some banding, which is something I usually find on the deep shadows of my 7D, but not on my 1D.

Yup, the 7D suffers from heavy FPN, that is because they use a cheaper readout scheme with fewer stages to compensate for sense amplifier mismatch.

here is an example 7D FPN from dark frame. Another reason why 7D performs poorly when raising shadows compared to the 1D.



7D FPN at low gain (ISO 100) from dark frame. You can almost map out the readout layout

[Roger Clark]

The above test is invalid, 7D was ISO 400 and 1D4 was ISO 800. The higher ISO means the light was weaker for 1D4 and therefore signal had to be amplified to match exposure and result in identical histogram. Identical exposure means identical shutter speed, aperture and ISO. ISO is the amplifier gain, it affects the results. If light was equal ISO should have been exactly the same, so something was wrong...the light captured by sensors was not identical.

Absolutely incorrect, ISO has NO effect on the number of photons collected. The shutter speed, and aperture were the same. Exposure is the same shutter speed and aperture at the same equivalent magnification. ISO only changes post sensor gain. If anything, the 7D would show another disadvantage as A/D converter and downstream electronics noise would be a greater factor in the shadow noise results. So the test should favour the 1DIV.

The images are not equal in processing and contrast either. .

Incorrect. I explained how I did the same processing steps except for chromatic aberration correction (which were small and nearly the same).

There is no significant mismatch between 7D and 1D4 gain calibration according to DxO measurements so ISO should match. If anything 7D ISO is overrated, so the correct ISO for 7D should have been 1000 not 400.

Then DXO is missing something. Attached is a yet another demonstration that the 7D produces a brighter image for the same exposure. The attached moon images are out of camera jpegs taken within a couple of minutes of each other with the same lens, same exposure, same ISO, same aperture, and the raw show the same story. With the same lens same settings (aperture and exposure) the 7D produces a slightly brighter image than the 1DIV.

why not crank up the 1D4 ISO to 6400 and compare with ISO 100 on the 7D, if that's irrelevant? I guess that will prove 7D is a better camera, right

It is not relevant as the A/D converters do not have the range. Such a test would show A/D converter and downstream electronics noise dominating shadows in the 7D, but the 7D would have a much greater dynamic range.

Arash, I'm surprised that as an engineer you do not seem to know about Etendue and the effects on system performance.

Roger

[Roger Clark]

And the leg area better balanced.

[arash_hazeghi]

Roger,
I did not say ISO has anything to do with number of photons collected. I said it changes the gain. Does it change the gain or not? Please answer yes or no

Does ISO change the gain and output SNR? yes or no?

The DxO SNR data is accurate, please provide your own measurements if it contradicts and explain how you measured it.

A) What is the 7D SNR for 18% grey at ISO 400?

B) What is 7D SNR for 18% grey at ISO 800?

Is A larger or smaller than B in your measurements?

Please these are short answers.

Thanks

[Roger Clark]

Roger,
I did not say ISO has anything to do with number of photons collected. I said it changes the gain. Does it change the gain or not? Please answer yes or no

Arash, please take a look at what you wrote: "The higher ISO means the light was weaker for 1D4..."

ISO changes post sensor gain, not the amount of light the sensor collected.

Does ISO change the gain and output SNR? yes or no?

Basically no (change in SNR) (but there is a gain change)! There is a small change in shadow noise due to varying amounts of read noise + A/D and downstream electronics noise.
What photographers observe as a reduction in image quality (reduction in signal-to-noise ratio) as one increases ISO is a decrease in shutter speed so less light is delivered to the sensor.

For those following, the dominant noise source in all but the deep shadows in digital camera images is photon noise: the square root of the number of photons collected. Longer exposure gives more light and better S/N. Increasing ISO simply instructs the camera (when not in manual mode) to use a shorter exposure time (less light) and then boosts the signal after the sensor is read out.

The DxO SNR data is accurate, please provide your own measurements if it contradicts.

I just did. See the moon images. The model bird images show the same thing.

A) What is the 7D SNR for 18% grey at ISO 400?

B) What is 7D SNR for 18% grey at ISO 800?

Is A larger or smaller than B in your measurements?

Please these are short answers.

Thanks

If the shutter speed is the same, when you change ISO the photons collected are the same so the S/N stays the same (with the read +A/D noise caveat).

Specifically, if you set the exposure on the 7D 18% gray card at ISO 800 the S/N (linear, single pixel before Bayer demosaicing) is:
Full well ~ 4080 electrons, read + electronics noise = 3.8 electrons, therefore 18% gray = 4080*.18 = 734 electrons with photon noise = square root 734 = 27.1, so the image noise per pixel = sqrt (734 + 3.8*3.8), and S/N of 734/27.4 = 26.8

Now do the same exposure at ISO400, one gets the same photon and electron count and read + electronics noise = 4.9 so the S/N = 734/27.5 = 26.7.

Now, if you let your light meter change the exposure with different ISOs, then the amount of light delivered to the sensor changes and the S/N changes.

In my test, I kept exposure time constant, so the S/N stays basically the same as ISO is changed. One just does not want to push too far as fixed pattern noise at lower iSOs can become a factor.

Roger

[arash_hazeghi]

Roger,

I did not say ISO changes number of photons collected, I said because you had to use lower gain for 7D it means that you had collected more photons to begin with for some other reason. Maybe wording was bad? I understand ISO does not change the number of photons

Let me rephrase this:

We assume you were careful with your test so that the number of photons per unit area striking the sensor was the exactly the same. In order to get the same exposure in the final image you had to use 2X more gain (2.33 given the overrated 7D ISO) for the 1D4 as opposed to 7D (You agreed ISO = gain). This means the 7D pixels have two times the quantum efficiency of 1D4 pixels.


Is that what your test tells us? Where did this factor of 2 come from?



On the side the ADC convertor doesn't add much noise in the modern sensors, its mostly the amplifier noise but that's the details....

[arash_hazeghi]

So Doug is sending me some samples to do measurements myself. In the meanwhile I found these two measurements online, they are not mine but they seem reasonable

1D4 http://www.sensorgen.info/CanonEOS-1D_MkIV.html

7D http://www.sensorgen.info/CanonEOS_7D.html

They measured a QE=41% for 7D and 44% for 1D4, so very slightly higher for the 1D which is consistent with the pixel scaling trends of FF for a given a technology generation a

[Roger Clark]

Roger,

I did not say ISO changes number of photons collected,

Arash,
You said the light was weaker, which can only mean fewer photons. I know of no other way to interpret weaker light since we are talking about the same wavelength light. At least that is the only way I see to interpret your statement.

I said because you had to use lower gain for 7D it means that you had collected more photons to begin with for some other reason. Maybe wording was bad?

But I said the shutter speed was the same. It was 1/800 second for both the 1DIV and 7D. The diameter of the entrance pupil was the same. So the lens collects the same amount of light per unit time, and since the exposure time was the same the same total amount of light was delivered to the focal plane.

I understand ISO does not change the number of photons

Let me rephrase this:

We assume you were careful with your test so that the number of photons per unit area striking the sensor was the exactly the same. In order to get the same exposure in the final image you had to use 2X more gain (2.33 given the overrated 7D ISO) for the 1D4 as opposed to 7D (You agreed ISO = gain). This means the 7D pixels have two times the quantum efficiency of 1D4 pixels.

Is that what your test tells us? Where did this factor of 2 come from?

I do not know what factor of 2 or 2.33 you are talking about. The adjustments I did were a few percent. The in-camera meter reading was almost the same between the two cameras reading within 1/3 stop.

I made several things equal:
1) exposure time
2) clear aperture that collects the light,
3) angular area pixel size

This equalizes the Etendue, then one can properly compare noise. The photons per pixel were identical in the two cameras (ignoring the few percent difference in system throughput between the pixels in the two cameras). If you change any one of the above 3, you change the amount of light delivered to the pixel, and the observed S/N. When we equalize the Etendue, the 7D has about the same S/N as the 1DIV.

I agree the fixed pattern noise is a little worse in the 7D but the 1DIV has some fixed pattern noise too and the 5D2 is even worse. But these are at a very low level and one has to really push things to see it, And one can record a few bias frames and subtract most of it out on the few images where it does show.

Since people noticed the fixed pattern noise in the heavily stretched images, there is one other effect to be noticed. The 7D shows a few hot pixels (e.g. notice the red pixels). I have noticed (statistics of small numbers of cameras) that 1D series cameras have fewer hot pixels than the lower priced cameras.

But in any case, this test shows the 7D is a formidable camera and does not deserve the downgrade as a noisy camera. For those who can not afford a 1D series camera, or bigger lenses, it is a great camera. I use my 7D for macro more than anything as it gives higher resolution with a macro lens.

On my previous trips to Africa, I took 1DIV and 5D2 cameras. Next trip I'll probably take 1DIV and 7D.

Roger

[arash_hazeghi]

Roger,

let's focus one more time :)

you collect the same number of photons because aperture and shutter speed were the same and light did not change.

you used ISO 400 for the 7D and ISO 800 for the 1D4 to get identical output.

ISO 800 gain = 2X ISO 400 gain.

You amplified the 1D4 signal 2X more than the 7D, correct? this implies 7D QE is a factor of 2 more than the 1D.

why?

I think my question is clear.

Thanks

[Roger Clark]

So Doug is sending me some samples to do measurements myself. In the meanwhile I found these two measurements online, they are not mine but they seem reasonable

1D4 http://www.sensorgen.info/CanonEOS-1D_MkIV.html

7D http://www.sensorgen.info/CanonEOS_7D.html

They measured a QE=41% for 7D and 44% for 1D4, so very slightly higher for the 1D which is consistent with the pixel scaling trends of FF for a given a technology generation a

And for comparison, here are my results for the two cameras (pretty close on the 1D4, different by 20% on the 7D):

http://www.clarkvision.com/articles/...on-canon-1div/

http://www.clarkvision.com/articles/...tion-canon-7d/

I'm surprised they would specify QE, as one would need to know the spectral transmission of the blur filters, the IR filter, and the Bayer filters, which would be very very difficult to do. At best it would be accurate to 10 to 20%. It would take your stanford lab to do better.

I've now examined three 7D cameras, and the produce the same result as shown in my Moon comparison.

But relative exposure between two cameras is simple: same exposure time, same lens, same subject. How do the images look? The moon images clearly show the 7D is brighter.

So how does DXO mark get higher S/N than the web site you referenced, and my tests (one can't do better than S/N from photon noise)? It can only be because they are working with demosaiced data, which means the results are biased by the demosaicing algorithm.

Roger

[arash_hazeghi]

For now let's just focus on your test and my question

The difference in QE between 7D and 1D4 is small either way, it cannot be a factor of 2 as implied by the ISO difference.

Can you upload the original CR2 files?

We are repeating your test and so far we do not need to change ISO to achieve identical output.

Alan Stankevtiz did no not need to change ISO between the three cameras in his tests either.

[Roger Clark]

Roger,

let's focus one more time :)

you collect the same number of photons because aperture and shutter speed were the same and light did not change.

you used ISO 400 for the 7D and ISO 800 for the 1D4 to get identical output.

ISO 800 gain = 2X ISO 400 gain.

You amplified the 1D4 signal 2X more than the 7D, correct? this implies 7D QE is a factor of 2 more than the 1D.

why?

I think my question is clear.

Thanks

Arash,
No it does not mean the QE is 2x higher. I measured system pixel efficiency on my 7D just 3% higher than my 1D4. So the data numbers are cut in half in the ISO 400 data (actually 0.5*1.03, along with how accurate the aperture and shutter are). I figure accuracy to about 10%, ~ 0.1 stop. So the photons collected are the same to within about 10%, the S/N the same to similar accuracy. Moving the 7D to ISO 400 increased the read + downstream electronics noise, which impacts shadow noise, but virtually no effect on higher signal levels. This again puts the advantage to the 1DIV, yet the 7D comes out almost identical in S/N.

Roger

[arash_hazeghi]

Arash,
No it does not mean the QE is 2x higher. I measured system pixel efficiency on my 7D just 3% higher than my 1D4. So the data numbers are cut in half in the ISO 400 data (actually 0.5*1.03, along with how accurate the aperture and shutter are). I figure accuracy to about 10%, ~ 0.1 stop. So the photons collected are the same to within about 10%, the S/N the same to similar accuracy. Moving the 7D to ISO 400 increased the read + downstream electronics noise, which impacts shadow noise, but virtually no effect on higher signal levels. This again puts the advantage to the 1DIV, yet the 7D comes out almost identical in S/N.

Roger

Roger,

I don't follow, are you just repeating the same thing? I understand the number of photons was the same.

Why did you need ISO 800 on the 1D to match the output of the the 7D at ISO 400 if the number of photons was the same? Please answer this question.

Can you please provide the CR2 files?

you can email them to me, ahazeghi at gmail dot com so I can see what the RAW files looked like. We need to get this clear before we continue the discussion.

Thank you.

[Roger Clark]

For now let's just focus on your test and my question

The difference in QE between 7D and 1D4 is small either way, it cannot be a factor of 2 as implied by the ISO difference.

Can you upload the original CR2 files?

We are repeating your test and so far we do not need to change ISO to achieve identical output.

Alan Stankevtiz did no not need to change ISO between the three cameras in his tests either.

Here are the raw files:

Before you claim you found an error (I'm not saying I don't make errors), check the three factors of Etendue and make sure they are equal between the two cameras:

1) exposure time (should be 1/800 second in this test)
2) lens clear aperture (that is NOT f/ratio) (should be 37 mm diameter in this test)
3) pixel angular size. (should be 2.8 to 2.9 arc-seconds)

The above #1 and #2 should be accurate to within what the camera can do: 1/3 stop. #3 is limited by pixel pitch ratio (1.33) and available TCs (1.4x) and slight adjustment (7%) change in distance to subject to equalize pixels on subject.

The raw files are here:

http://www.clarkvision.com/tmp/bpn/r...4-C45I3003.CR2

http://www.clarkvision.com/tmp/bpn/r...d-IMG_0015.CR2

Roger

[arash_hazeghi]

Great let me have a look!

[Roger Clark]

Roger,

I don't follow, are you just repeating the same thing? I understand the number of photons was the same.

Why did you need ISO 800 on the 1D to match the output of the the 7D at ISO 400 if the number of photons was the same? Please answer this question.

Well, if I would have known that people would get hung up on what I consider irrelevant, I would have done the test differently.

I set the exposure for the 1DIV, expose to the right. That gave the 1/800 second exposure. When I put the 7D on the lens, it was a little bright saying some of the sky would be clipped (on the histogram), so I reduced the ISO to be able to digitize a slightly higher signal. I didn't want to be accused of clipping.

I could have set the 1DIV exposure at 1/1000 second then I would not have needed to reduce the 7D ISO. But all of that is irrelevant because the light is the same, and if anything biases the test in favour of the 1DIV.

Roger

[arash_hazeghi]

Here is the EXIF from Roger's photos

ad4_C45I3003.CR2:

420mm f/11 1/800sec ISO 800

7d-IMG_0015.CR2:

400mm f/8 1/800sec ISO 400.


the number of photons incident on the sensor depends on the physical aperture AND the focal length. It depends on the F number of the lens. For example we all know exposure is the same for two f/2.8 lenses regardless of focal length, despite the fact that the physical aperture diameter is different. This was pointed out by captain Shadle too (sunny f/16 rule).

Here is from wikipedia
http://en.wikipedia.org/wiki/F-number

"A 100 mm focal length lens with an aperture setting of f/4 will have a pupil diameter of 25 mm. A 200 mm focal length lens with a setting of f/4 will have a pupil diameter of 50 mm. The 200 mm lens's f/4 opening is larger than that of the 100 mm lens but both will produce the same illuminance in the focal plane when imaging an object of a given luminance."




In the test above Roger used f/11 for 1D4 and f/8 for 7D therefore the number of photons was not the same . The 1D4 collected half the number of photons and therefore ISO had to be bumped up to 800.


Thank you for providing RAW files.

[Roger Clark]

Here are the raw files:

Before you claim you found an error (I'm not saying I don't make errors), check the three factors of Etendue and make sure they are equal between the two cameras:

1) exposure time (should be 1/800 second in this test)
2) lens clear aperture (that is NOT f/ratio) (should be 37 mm diameter in this test)
3) pixel angular size. (should be 2.8 to 2.9 arc-seconds)

The above #1 and #2 should be accurate to within what the camera can do: 1/3 stop. #3 is limited by pixel pitch ratio (1.33) and available TCs (1.4x) and slight adjustment (7%) change in distance to subject to equalize pixels on subject.

The raw files are here:

http://www.clarkvision.com/tmp/bpn/r...4-C45I3003.CR2

http://www.clarkvision.com/tmp/bpn/r...d-IMG_0015.CR2

Roger

Another point when you look at the data numbers (DN) in the raw files.

Lets say the 1DIV collected 6000 photons. The ISO 800 gain is 0.52 electrons/DN, so the 6000 photon signal would be 6000/0.52 = 11538 (180 on a 0 to 255 scale)

On the 7D, the ISO 400 gain is 0.61 electrons/DN. Where the 7D collected 6000 photons, the signal would be 6000/0.61 = 9836 (153 on a 0 to 255 scale).

Roger

[Enrique Patino]

I have never used a 1D4, but I for one am very impressed by these comparisons Roger. I have seen the 7D perform very well under certain "conditions." But not all the time, so I am very interested in this information because I believe it can lead me to be able to make better images with my camera more often. Thanks for the discussion here...

Roger, what do you mean when you say... "And one can record a few bias frames and subtract most of it out on the few images where it does show".

Thanks!

Enrique

[Roger Clark]

I have never used a 1D4, but I for one am very impressed by these comparisons Roger. I have seen the 7D perform very well under certain "conditions." But not all the time, so I am very interested in this information because I believe it can lead me to be able to make better images with my camera more often. Thanks for the discussion here...

Roger, what do you mean when you say... "And one can record a few bias frames and subtract most of it out on the few images where it does show".

Thanks!

Enrique

Hi Enrique,

I'm glad someone sees value. This is all a prelude to a new article on Etendue and telephoto reach that I'm working one. The results are very interesting and will probably turn some more heads.

A bias frame is a very short exposure that is subtracted from your image. Take a bunch of 1/8000 second exposure in a dimly lit room at each ISO and average each group. Then subtract that average from your normal exposures. I think you can do that (subtraction) in photoshop; I do it in imagesplus or my own custom code. That should reduce the fixed pattern noise by a factor of 5 to 10 or more.

Similarly, if you are taking long exposure images, many seconds to minutes, make some dark frames (say 3 to 9, or even 16 if you have time). On long exposures, dark current adds to the signal. Average the dark frames and subtract from your regular images. Dark current is temperature dependent, so take them at the same temperature as your other images. Those dark current images also have the bias included, so do not subtract dark current and bias. In theory (and works well in practice), a set of darks at a given temperature can be saved and used for many months, if not years. But you need a set of darks for every few degrees change in temperature. Fortunately, modern cameras like the 7D and 1DIV have pretty good on-chip dark current subtraction, so it is not as much of a problem as it used to be, but subtracting darks helps improve image quality.

The bias and dark frame suppression methods are being applied by amateur astronomers with stunning results. It is a little work but makes a good image even better.

Here is an article that shows dark frame subtraction on 1DII images:
http://www.clarkvision.com/articles/...t.photography/

Roger

[BenBotha]

Ben,I am not sure if I buy this argument, the 7D pixel size is 4.3um 1D4 is 5.3um the difference in "reach" is 30%. So are you saying being 30% farther (say 39 ft instead of 30 ft) saves your life? Like a lion can't attack from 39 ft but it can attack from 30ft? Am I missing something :) If you have a look at the wildlife forum many of the most impressive wildlife photographs are taken with the 1D4 or Nikon FF cameras so the photographers were not eaten My friend Oliver Klink who has one of the best grizzly portfolios uses FF cameras to get his bear shots. Chas also uses FF and 1D cameras for his work...see his comments on a similar thread...have a look at these amazing wildlife portfolios:http://www.oliverklinkphotography.com/http://www.shootthelight.com/#/Portfolio/Wildlife/6/Also I was mainly talking about professional wildlife/bird photography, of course some people take snapshots of wildlife with point and shoot cameras, so obviously to them is not a waste of electrons...

I simply said that in the areas I take photographs, we are not allowed to get out of our vehicles. The lion story, we are in African game reserves and not in zoos, where lion and other predators roam freely and several people are attacked on a regular basis. Game rangers are carrying weapons because predators are unpredictable-in a lion or leopard attack 1-2 feet could save your life!! On a personal level, I personally treated an Australian tourist whose hand was ripped off by a hayena when he left his vehicle to relieve himself. Most professional photograhers make special arrangements that we as normal tourist can only dream of. So, I still maintain that every mm of reach, in my case, is of tremendous help.

Ben

[Chris Warren]

Here is the EXIF from Roger's photos

ad4_C45I3003.CR2:

420mm f/11 1/800sec ISO 800

7d-IMG_0015.CR2:

400mm f/8 1/800sec ISO 400.


the number of photons incident on the sensor depends on the physical aperture AND the focal length. It depends on the F number of the lens. For example we all know exposure is the same for two f/2.8 lenses regardless of focal length, despite the fact that the physical aperture diameter is different. This was pointed out by captain Shadle too (sunny f/16 rule).

Here is from wikipedia
http://en.wikipedia.org/wiki/F-number

"A 100 mm focal length lens with an aperture setting of f/4 will have a pupil diameter of 25 mm. A 200 mm focal length lens with a setting of f/4 will have a pupil diameter of 50 mm. The 200 mm lens's f/4 opening is larger than that of the 100 mm lens but both will produce the same illuminance in the focal plane when imaging an object of a given luminance."




In the test above Roger used f/11 for 1D4 and f/8 for 7D therefore the number of photons was not the same . The 1D4 collected half the number of photons and therefore ISO had to be bumped up to 800.


Thank you for providing RAW files.

Yes hi Arash,

When Roger equalized entrance pupil diameters between 7D and 1DIV, this ensures that the same # of photons was delivered to each pixel. If we go by F#, this does mean that the illuminance, or irradiance, or #photons/sq. micron at the focal plane is higher for the 7D than the 1DIV, but to get to # of photons, we have to multiply this #photons/sq.micron by each pixel's area, thus the 1DIV's higher pixel area compensates for this difference.

All Roger is saying is the following from radiative transfer:

P = L * A * a / (f^2) , where P is light power collected by a pixel (# photons in/sec) , A is the area of the entrance pupil, a is pixel area , f is focal length, and L is the radiance (luminance) or brightness of the object that we're taking a picture of. This takes the form of :
P = L * A * omega , where omega is a solid angle. Roger said that he is holding the number of pixels on the subject as constant, so he is holding (p/f)^2 constant (where p is the pixel pitch and a = p^2), which just says that p/f , which is the FOV that a pixel sees, is constant between the 7D and 1DIV. If that quantity is held constant, then the only quantity that determines P is A, the entrance pupil area, and Roger is holding that constant as well between the 2 cameras. So he is sending in the same amount of photons/sec, if using the same shutter speed, then the # photons incident is the same between the 2 pixels. Okay so this just leaves the QE, and from both of Roger's tests and from the Sensorgen data, these are very similar, giving a similar # of electrons collected in an exposure time. (Sensorgen data, btw, are done from curve fitting and data reduction of the Dxomark full SNR and measured ISO data).

Apart from read noise and PRNU, which are small effects at these ISOs, these 2 scenarios will give very similar SNRs as Roger has shown. Let's assume part of the picture has about 700 e-. For 7D at ISO 400, this would amount to about 1148 DN (using Roger's measured camera gain constant of 0.61 e-/DN). For 1DIV it would be about 1346 DN. Both of these are pretty similar "brightness" for the 14-bit system. (I use "brightness" in quotes because these are in units of DN).

Anyway, in short, Roger's tests showing the utility of etendue considerations, sound good to me.

Chris

[Roger Clark]

Here is the EXIF from Roger's photos

ad4_C45I3003.CR2:

420mm f/11 1/800sec ISO 800

7d-IMG_0015.CR2:

400mm f/8 1/800sec ISO 400.

the number of photons incident on the sensor depends on the physical aperture AND the focal length. It depends on the F number of the lens. For example we all know exposure is the same for two f/2.8 lenses regardless of focal length, despite the fact that the physical aperture diameter is different. This was pointed out by captain Shadle too (sunny f/16 rule).

Arash,
Chris is correct.

You have 2 errors in your above numbers: the 1D4 was at 11.3 not 11, and the 7D was at 300 mm not 400 (it was a 300 mm f/2.8 L IS lens).

Lets look at the parameters:

1D4: 420mm f/11.3 1/800sec, 5.7 micron pixels (0.0057 mm), 1/800 second exposure:
clear aperture = 420/11.3 = 37.2 mm diameter
pixel field of view = 206265 * 0.0057/420 = 2.8 arc-seconds

7D: 300mm f/8 1/800sec, 4.3 micron pixels (0.0043 mm) 1/800 second exposure
clear aperture = 300 /8 = 37.5 mm diameter
pixel field of view = 206265 * 0.0043/300 = 2.95 arc-seconds.

Thus, exposure identical, clear aperture 0.8% different diameter (1.6% area), field of view 5.3% different (11% angular area different).

Here is from wikipedia
http://en.wikipedia.org/wiki/F-number

"A 100 mm focal length lens with an aperture setting of f/4 will have a pupil diameter of 25 mm. A 200 mm focal length lens with a setting of f/4 will have a pupil diameter of 50 mm. The 200 mm lens's f/4 opening is larger than that of the 100 mm lens but both will produce the same illuminance in the focal plane when imaging an object of a given luminance."

As Chris notes, the same f/ratio produces the same illuminance, e.g. photons per square micron in the focal plane. But since the two cameras have different area pixels , the photons per pixel are different. The 7D pixel area is 18.5 square microns and the 1D4 32.5 square microns. So for the same f/ratio lens, the 1D4 gets 1.76 times the photons per pixel that the 7D does, so equalizing f/ratio does not equalize signal. And if using the same focal length, then the detail on subject is different too.

And the f/ratio rule only applies to extended sources. For point sources, e.g. stars which are diffraction disks in the focal plane, it does not apply at all.

In the test above Roger used f/11 for 1D4 and f/8 for 7D therefore the number of photons was not the same . The 1D4 collected half the number of photons and therefore ISO had to be bumped up to 800.

Thank you for providing RAW files.

As both Chris and I have shown, you conclusion is incorrect.

The beauty of Etendue is that when one equalizes the Etendue, the photons/pixel/time is equalized, and it works for both extended and point sources, and the depth of field is equalized too. Simply elegant!

Photographers have been accustomed to thinking f/ratios, but that is a simplification that does not apply in all cases, and the variety of digital cameras have exposed the flaws in the f/ratio applies to all idea.

Roger

[Doug Brown]

I am not a scientist, so much of this back and forth is lost on me. But Roger, your sample frames do not pass my sniff test. You stopped down a full stop for the Mark IV image and then doubled the ISO to match the exposure of the 7D. Without looking at the sensor under a microscope, I would say that the Mark IV's exposure let in 1/2 the amount of light compared to the 7D's exposure. I'm not particularly interested in the per pixel number of photons collected; I have a much more simplistic understanding of sensors and exposure. When comparing noise in two cameras, the gold standard test for me is this:
1. Use manual exposure and identical f/stop, SS, and ISO on both cameras.
2. Use the same lens on both cameras.
3. Adjust your position relative to the subject so that it fills a comparable portion of the frame with each body.
4. Take photos with each body in rapid succession.
5. Compare noise.
I understand that this testing methodology may have some shortcomings in your eyes, but this type of test produces results that are relevant to me in the field. Using my methodology, the Mark IV has a clear advantage over the 7D. Which is not to say that the 7D isn't an excellent camera body in its own right, because it is. It offers excellent features and image quality at a reasonable price. But IMO for those who can afford one, the Mark IV is superior in the areas that I outlined in an earlier post.

[Roger Clark]

I am not a scientist, so much of this back and forth is lost on me. But Roger, your sample frames do not pass my sniff test. You stopped down a full stop for the Mark IV image and then doubled the ISO to match the exposure of the 7D. Without looking at the sensor under a microscope, I would say that the Mark IV's exposure let in 1/2 the amount of light compared to the 7D's exposure. I'm not particularly interested in the per pixel number of photons collected; I have a much more simplistic understanding of sensors and exposure. When comparing noise in two cameras, the gold standard test for me is this:
1. Use manual exposure and identical f/stop, SS, and ISO on both cameras.
2. Use the same lens on both cameras.
3. Adjust your position relative to the subject so that it fills a comparable portion of the frame with each body.
4. Take photos with each body in rapid succession.
5. Compare noise.
I understand that this testing methodology may have some shortcomings in your eyes, but this type of test produces results that are relevant to me in the field. Using my methodology, the Mark IV has a clear advantage over the 7D. Which is not to say that the 7D isn't an excellent camera body in its own right, because it is. It offers excellent features and image quality at a reasonable price. But IMO for those who can afford one, the Mark IV is superior in the areas that I outlined in an earlier post.

The test you propose is not an equal test. You have added two variables that change conditions:

By requiring the same lens and same f/ratio, the light per pixel is different in favor of the larger pixels.

By changing position you are changing detail on the subject that depends on the crop factor. Try that test with a 5D mark II and a 30D. Your results may be surprising.

In your test, if you can fill the frame to the desired level, the larger sensor will provide the better signal to noise ratio, with additional complications of number of pixels on the subject.

Roger

[Doug Brown]

By requiring the same lens and same f/ratio, the light per pixel is different in favor of the larger pixels.

Based on the information you have provided in this thread, that statement appears to be correct; it also highlights why the Mark IV produces cleaner images at high ISOs IMO. But I don't set my exposure based on the light per pixel. You are trying to control all variables and look at sensor noise under a microscope. I am trying to devise a test that has relevance in the field. My test answers the simple question "At a given ISO, SS, and aperture, which camera body produces a cleaner image of a subject occupying a constant percentage of the frame." Let's say that I'm at Bosque del Apache photographing Sandhill Cranes at the morning fly out. I have a Mark IV, a 7D, and a 500mm lens. I like the birds to occupy a certain percentage of the frame. In my test I set both cameras up with identical exposure settings. And to accommodate the 1.6 crop and the 1.3 crop, I focus on closer birds with the Mark IV than I do with the 7D. After a morning of shooting I download the images onto my computer and compare the results at a variety of ISOs. Not a perfect test, but it answers the question that I set out to answer.

In your test, you attempt to normalize the light per pixel. How do I go about doing that in the field? And more importantly why would I want to? Why would I use a test for noise that compares one camera at f/8, 1/800, and ISO 400 to another camera at f/11, 1/800, and ISO 800?

[Roger Clark]

Based on the information you have provided in this thread, that statement appears to be correct. But I don't set my exposure based on the light per pixel. You are trying to control all variables and look at sensor noise under a microscope. I am trying to devise a test that has relevance in the field. My test answers the simple question "At a given ISO, SS, and aperture, which camera body produces a cleaner image of a subject occupying a constant percentage of the frame." Let's say that I'm at Bosque del Apache photographing Sandhill Cranes at the morning fly out. I have a Mark IV, a 7D, and a 500mm lens. I like the birds to occupy a certain percentage of the frame. In my test I set both cameras up with identical exposure settings. And to accommodate the 1.6 crop and the 1.3 crop, I focus on closer birds with the Mark IV than I do with the 7D. After a morning of shooting I download the images onto my computer and compare the results at a variety of ISOs. Not a perfect test, but it answers the question that I set out to answer.

Doug,
In your example, you changed the subject between the two cameras! That's not an equal comparison. To equalize the comparison, to image the same subject, and change lenses instead of distance.

What if there weren't any close birds? Say all the birds were on the opposite side of the lake and you can't go over there. So you need to put on a telextender on the 1D4. But the smaller pixel of the 7D, by understanding Etendue, you can go with the 500 bare. And knowing the sensor characteristics, work in manual, one can optimize the signal and produce equal quality images. Instead of thinking in f/ratios, think in terms of aperture diameter. So if you switch to a shorter focal length lens for close birds, think in terms of keeping the aperture diameter the same, not the f/ratio.

Crop factor is the most confusing factor in digital photography, but depth of field is a close second, and no I'm putting f/ratio up there as another most confusing factor.

In your test, you attempt to normalize the light per pixel. How do I go about doing that in the field? And more importantly why would I want to? Why would I use a test for noise that compares one camera at f/8, 1/800, and ISO 400 to another camera at f/11, 1/800, and ISO 800?

First, why would you want to? Answer: to get the best image quality. For example in my test, I used a 1/800 second exposure at f/11.3 at 420 mm on the 1D Mark 4. A quick and simple computation says the aperture is 37 mm. So I keep 37 mm if I change lenses and cameras, or lenses +TC and a different camera. If I switch to another camera with smaller pixels, I can drop a TC, or with a zoom lens, change focal length. But keep the aperture diameter the same, not f/ratio. For anyone having slow AF at f/8: change to a TC that gives f/5.6 and a camera with smaller pixels (we are ignoring differences in camera performance, like frames per second). Then knowing ISO does not change real exposure, simply work in manual and and set the ISO and exposure for the dynamic range you need.

Now, as you already have a 1DIV, which is the top performing canon action camera, there is no need to switch to a 7D. But for those who can't afford a 1D series, knowing 7D is a good performing camera and one can get equally good S/N images as those with 1D series cameras, if one uses it properly. Another way to look at it, is if you were to suddenly lose your 1DIV (say it broke or was stolen), and if you had a 7D backup, then knowing Etendue, you could produce equally high quality images (noise levels and detail on subject) as you did with your 1DIV with of course some lose due to FPS, and AF speed and accuracy. But properly used, the difference is smaller than it may appear in the limited box of f/ratios and ISO.

The ISO, f/ratio, focal length box is a very narrow box and is confusing in the digital camera world where sensor size and pixel size have come into the fold. Then the fact that digital cameras are photon noise limited sets interesting properties and constraints on our images. The old box is too confining. Understand the larger picture (pun) enables one to produce equal or better images in more conditions with different equipment.

Roger