Mk3 vs Mk4 vs 7D - IQ:Noise

[Peter Hawrylyshyn]

Got lucky and managed to pick up one of the first Mark 4's released here in Ontario this week. Too cold to find any birds, so set up some comparative test shots. All three cameras were re-calibrated for AF, used a 400mm DO on a tripod with timed shutter release, focus set/locked to manual after 1st picture, same exposure settings in all 3 cameras based on settings from Mk3. The images are at 100% (250px by 250 px) with NO sharpening applied.

Initial reactions : 1) the Mk3 and Mk4 do not have the same crop ratios! 2) Unlike suggestions in prior thread, noise levels on the 7D are NOT comparable to the Mk4. The differences between the 3 cameras are readily apparent when viewed at 300% . 3) Images on the Mk4 and 7D are slightly "softer" than the Mk3. 4) Not as readily evident on this comparison, there's also a slight exposure loss with the Mk4 and 7D.

Hope this helps anyone still undecided between the 3 cameras.

[Peter Hawrylyshyn]

here are the images at the higher ISO's

[Roger Clark]

Peter,
Thanks for posting these. Could you please give a little more information.
What is the origin of the images (e.g. in-camera jpegs, raw conversion in XXX program)?
It looks like the ISO 400 and 800 images are blurred due to mirror shake, correct?
The contrast on the 1D4 appears higher so the tone curve seems different, correct?

Roger

[Peter Hawrylyshyn]

Roger -
Sorry. the original image is part of a floral print on a Kleenex box. I'd taken various pic's of indoor items (dollar bills, wine labels, ...) .
All were taken in RAW, used latest version of ACR for the conversion. I made NO changes to any settings (tone curves, sharpening,white balance, contrast, ...) in the ACR conversion or subsequent PS processing which just consisted of taking 250px by 250px crops, merging into one new photo and saving for web at 200K. What's shown is essentially straight out of each camera.
Yes i was surprised by the slight blur at ISO 400 and 800 which must be due to mirror shake at slower shutter speeds given cameras were on a tripod using timed shutter release
Can you explain the different Mk3 vs Mk4 crop factors?
Here's another image taken on a color checker with a 70-200m f2.8L at 200mm which perhaps better compares the Mk3 and Mk4. Used auto WB "as shot" in ACR conversion which showed temp for Mk3 was 4500 and for Mk4 was 4650

[Axel Hildebrandt]

Can you explain the different Mk3 vs Mk4 crop factors?

Both cameras have the same 1.3 crop factor, the Mark IV has about 20% more pixels, though.

[John Chardine]

Both cameras have the same 1.3 crop factor, the Mark IV has about 20% more pixels, though.

16.1mp - 10.1mp = 6.0mp
6.0/10.1 = 0.59 = 59% more pixels

The mk IV sensor is about 0.18 sq cm smaller than the mk III.

[Axel Hildebrandt]

16.1mp - 10.1mp = 6.0mp
6.0/10.1 = 0.59 = 59% more pixels

The mk IV sensor is about 0.18 sq cm smaller than the mk III.

I was referring to the horizontal pixels only, Mark III 3888 v. Mark IV 4896.

[Roger Clark]

Peter,
The sensor sizes are:
1D Mark III: 28.1 x 18.7 mm, 10.1 megapixels, 7.2 micron pixel pitch, 1.28 crop factor
1D Mark IV: 27.9 x 18.6 mm, 16.1 megapixels, 5.7 micron pixel pitch, 1.29 crop factor

So the sensor size difference (crop factor) is minor. The reason you see a large
difference in image sizes is due entirely to the pixel pitch. You will see
7.2/5.7 = 1.26 times more pixels (measuring linearly) on a subject with the 1D4
than with the 1D3. This is another good example of how crop factor is irrelevant
when you use the same lenses on different bodies and there is no "telephoto reach" from
crop factor. (Peter,I know you know this, I'm just writing it again for others who
may read this as it seems to come up again and again, even in recently published magazine
articles and photo books).

Roger

[arash_hazeghi]

The active sensor area for MKIII and MKIV is identical giving the same crop factor. The slight difference in die size is due to difference in number of refrence pixels which do not contribute to image capture

Peter, as Roger mentioned it is the pixel size that determines resolution and the final image image size not crop factor.

[Alan Stankevitz]

I don't mean to stir any mud here, but since this is a bird photographer's forum, I believe a test between all three cameras should be cropping a subject identically. This to me, is a better comparison since most of the time when we photograph birds we do so in a focal-length-limited scenario in which we cannot walk up to the bird and photograph it. We photograph the bird from a set distance.

Cropping a subject identically with all three cameras, should reveal a more relevant test for photographing birds/animals in the wild.

Alan

[Roger Clark]

Alan,
That is what Peter did. Each camera will give a different number of pixels on subject because the pixel spacing is different in each.

Roger

[Robert Hardy]

i have the mk3 and 7d but it does not make me want to rush out to get a mk4 .
Rob.

[Alan Stankevitz]

Alan,
That is what Peter did. Each camera will give a different number of pixels on subject because the pixel spacing is different in each.

Roger

It doesn't have to do with the number of pixels...at least when interpolation is used with Photoshop for example. This may not seem like a fair test, but I believe that it is since in most cases when sending an image to a publisher, they request images at 300 ppi (typically). Cropping the images identically @ 300 dpi for example, will produce results that are framed identically.

The reason why I bring this up, is that at least in my work flow I am commonly cropping images within Photoshop at 300 ppi. I would assume this would be part of any photographer's work flow for processing images for publication or print. 300 ppi is just an example...it could be 400 ppi if need be for certain printers.

Wouldn't this be a normal work flow for most photographers? Cropping images at x ppi to crop the subject in a artful way? That is why I would like to see comparisons to what I consider more-of-the-norm for most wildlife photographers.

Alan

[Peter Hawrylyshyn]

Roger,Axel, Arash - thanks for the explanation , my math was slow that morning

R0bert - i'm not making any claim on which camera (MK3 v3 Mk4) is better. I would have been curious to hear from others who own/have owned both cameras as to comparative IQ/noise levels.

Alan - each image is effectively being shown at 100ppi. If i had interpolated the images to identical sizes, this could have introduced artifacts negating the comparisons. You can easily download the jpeg, resize to 300ppi and it will give you the view you want to see

[John Chardine]

My own take on this is that:

1. Noise in photography is a fact of life.
2. Therefore you have to deal with it.
3. Dealing with it is easy.
4. Assessment of noise between cameras should be done on the processed final product rather than on the RAW, unprocessed image at 100%. The latter is fine for scientific assessments but in the real world, we process our images for what might be their final outcome (web, print, whatever), and that's when noise should be evaluated.

So to really evaluate say the FF 12mp D3s and the 1.3 crop 16 mp mark IV or the mk III or the D300s or whatever:

1. Set up exposure conditions in each camera to produce roughly the same histogram- note that this might mean that actual exposures on each test camera are different.
2. Set high ISO noise reduction in the camera (e.g., Standard setting) and shoot RAW images with standard high-ISO settings and process the images with the proprietary software (DPP etc), or turn off NR in camera and process using ACR's NR.
3. Process images as usual which may include cropping, resizing (for web), running NR on background etc.
4. Compare results.

Some might say "well that's not a real test because you have dealt with the noise" Exactly grasshopper!

[Alan Stankevitz]

"2. Set high ISO noise reduction in the camera (e.g., Standard setting) and shoot RAW images with standard high-ISO settings and process the images with the proprietary software (DPP etc), or turn off NR in camera and process using ACR's NR."

Setting the noise reduction levels in cameras is for jpg output only. When shooting RAW, noise reduction is not applied until afterward while processing the RAW image in ACR or other software. At least, this is the way it works with Canon cameras.

Alan

[John Chardine]

"2. Set high ISO noise reduction in the camera (e.g., Standard setting) and shoot RAW images with standard high-ISO settings and process the images with the proprietary software (DPP etc), or turn off NR in camera and process using ACR's NR."

Setting the noise reduction levels in cameras is for jpg output only. When shooting RAW, noise reduction is not applied until afterward while processing the RAW image in ACR or other software. At least, this is the way it works with Canon cameras.

Alan

I agree Alan but it is my understanding that only the proprietary processing software like DPP understands and uses as a starting point the NR settings for RAW images set in-camera. If you set High ISO NR to Standard and process via ACR, ACR does not "know" these values, DPP does. This is pretty standard with many of the in-camera settings such as Picture Style etc when shooting RAW. Not sure if the Nikon software works the same way.

[Roger Clark]

My own take on this is that:

1. Noise in photography is a fact of life.
2. Therefore you have to deal with it.

John,
I agree.

3. Dealing with it is easy.
4. Assessment of noise between cameras should be done on the processed final product rather than on the RAW, unprocessed image at 100%. The latter is fine for scientific assessments but in the real world, we process our images for what might be their final outcome (web, print, whatever), and that's when noise should be evaluated.

I disagree with these two. If you want to understand the noise of the camera, you shouldn't be processing the data through ANY raw conversion software that modifies any original data values. To do so means you are really testing the conversion software and not the camera capabilities. That is two very different things.

Examples:
See Peter's does raw converter matter examples:
http://www.birdphotographers.net/for...ad.php?t=56325

Conversion software and algorithms change with time. If you have some raw files from 6+ years ago, try processing them with modern converters and your result will likely be much better. New algorithms are invented all the time, so compared to conversions done today in a couple of years converters will likely do a better job.

Roger

[Peter Hawrylyshyn]

John -
You're not alone, other BPN moderators ( the Forns, even Artie i suspect) have expressed a similar opinion that the end product is all that matters.
It was curiousity and wanting to know "what was in the box" that prompted my making the copmparisons.
The Mk3 has an excellent reputation for its IQ and low noise. It would be nice to know and see that the Mk4 has added 50% more MP and further improved the native IQ:noise without having to use any postprocessing(PP). I was always under the impression that the less PP you apply to an image, the better.
PH

[Roger Clark]

John -
You're not alone, other BPN moderators ( the Forns, even Artie i suspect) have expressed a similar opinion that the end product is all that matters.

I'll also agree that the end product is all that matters. But depending on your starting point, it is easier/harder to get to the better end product. For example, people here seem concerned with lens sharpness. After all, the shaper the lens to start with, the sharper the final image. Same with image noise. The less it is to start with, the less there will be in the final image.

It was curiousity and wanting to know "what was in the box" that prompted my making the copmparisons.
The Mk3 has an excellent reputation for its IQ and low noise. It would be nice to know and see that the Mk4 has added 50% more MP and further improved the native IQ:noise without having to use any postprocessing(PP). I was always under the impression that the less PP you apply to an image, the better.
PH

We'll soon have the sensor data on the 1D4. Peter sent me a disk with test data that I'll analyze.
Peter: it is not here yet; hopefully today.

Roger

[Robert Hardy]

i like tests like these without pp after all my pp work is not going to be the same as someone elses.
One point though
at inbetween iso's i.e iso 320 /2000 etc the 7d is disavantage against the other 2 as it is not a true iso on the 7d .
Rob.

[Emil Martinec]

One point though at inbetween iso's i.e iso 320 /2000 etc the 7d is disavantage against the other 2 as it is not a true iso on the 7d .
Rob.

Actually, no. ISO 160/320/640/1250 etc are 200/400/800/1600 etc pulled by 1/3 stop on the 7D, while on the 1D3 at least (don't know yet for the 1D4) the intermediate ISO's of this family are as noisy as the next lowest "main" ISO. So for these ISO the 7D actually has a better ISO implementation than the 1D3 as far as noise is concerned (at the expense of a loss of 1/3 stop of highlight headroom).

For ISO 125/250/500/1000 etc, the 7D uses ISO 100/200/400/800 etc pushed 1/3 stop. But the 1D3 amplifier that implements these ISO's is so noisy it is no better than using the next lowest ISO and pushing, so it's a wash.

Bottom line is that the software ISO implementation in Canons is not really worse than the hardware implementation in the 1D3; neither is particularly well done.

Also, the biggest difference between the 7D and 1 series is the sensor size; the 1 series collects about 2/3 stop more light for a given exposure.

[Robert Hardy]

Very intresting Emil
something i did not know or should i say had the wrong ider .
Thanks for that info.
Rob.

[Roger Clark]

Also, the biggest difference between the 7D and 1 series is the sensor size; the 1 series collects about 2/3 stop more light for a given exposure.

That 2/3 stop is only summing up all the light over the whole sensor. For focal length limited subjects (e.g. most bird photos), it will come down to the efficiency of the pixels, which will be similar. The 7D will have slightly higher resolution but noisier pixels than the 1D4.

Roger

[Emil Martinec]

That 2/3 stop is only summing up all the light over the whole sensor. For focal length limited subjects (e.g. most bird photos), it will come down to the efficiency of the pixels, which will be similar. The 7D will have slightly higher resolution but noisier pixels than the 1D4.

Roger

2/3 stop is a fair assessment of the capability of the sensor. If however the image is cropped to a given field of view, then the assessment should of course be made on the part of the image that is not thrown away. I would be surprised in that case if there is much to choose from between the 1D4 and the 7D; both employ the same generation of sensor technology. Are you saying that read noise per unit area, QE or level of pattern noise is substantially different?

[Christopher C.M. Cooke]

Well, the maths went right over my head (would come as no surprise to my teachers over the years) however I have used all three (MKIV was borrowed) cameras and all I can say is that I have recently purchased a second MKIII and 5DMKII for less than the MKIV and I am very happy.

The MKIV is a very nice camera but it is no 5DMKII and I could notice no autofocusing advantages over the MKIII on BIFs and little difference over the 7D, so nice camera yes, worth the money NO.

PS I still take out my Rebel 350XT with 120,000 shutter rotations (verified by Canon) and as long as you stay under ISO 400 it still takes stunning and nicely small images.