My 5D Mark III vs. 1D Mark IV high ISO comparison

[Doug Brown]

Hi everyone. I've just posted a comparison of Canon's 5D Mark III and 1D Mark IV at high ISOs. As opposed to other tests that use a bunch of studio shots, I devised my test to be relevant to bird photographers. Click on this link to read the comparison. The results may surprise you!

[Robert Hardy]

Many thanks Doug for takeing the time to do this test ,iv been back and forth over should i get the 5dmk3 or a 1dmk4 as my next camera this makes the mk4 the better camera for me here in the uk were we often carnt get close enough to the birds.
It will be very intresting to get your views on the AF once you have the time to compare them.
Thanks
Rob.

[Roger Clark]

Hi Doug,
Nice comparison. But I do not understand your comparison 2. The 5DIII and 1dIV images are the same field of view and same size. But if identical focal length and distance, the 1DIV images should be have the bird larger as the pixel density is higher assuming your are presenting 100% sizes. Did you resample the 1DIV images to match the size of the 5DIII (or visa versa)?

In case 1, the ratio of the lens apertures is (300 mm at f/7.1 = 42.25 mm diameter / (221 mm at f/7.1 = 31.1) = 1.36, so the lens area ratio is 1.36*1.36 = 1.85 or 0.9 stop, which is close to what you said the advantage of the 5DIII was, but the test gave the 5DIII the 0.9 stop more light. It is not the larger pixels, but the the change in the lens diameter that delivered more light producing a lower noise image.

In your closing statements: "And if you are focal length limited and will need to crop, the Mark IV’s 1.3 FOV and higher pixel density give it the edge, even in low light conditions." I would drop the 1.3 FOV as it is not relevant and just feeds the crop factor myth.

I would also be interested in lower ISO comparisons that bird photographers more commonly use with some shadows and then lift the shadows to try and reveal shadow detail. My 5DII shows fixed pattern banding while my 1DIV is much better. Initial reports on the net said the 5DIII had about the same fixed pattern noise. (This fator is also important for night photography.)

Roger

[Doug Brown]

Hi Doug,
Nice comparison. But I do not understand your comparison 2. The 5DIII and 1dIV images are the same field of view and same size. But if identical focal length and distance, the 1DIV images should be have the bird larger as the pixel density is higher assuming your are presenting 100% sizes. Did you resample the 1DIV images to match the size of the 5DIII (or visa versa)?

In case 1, the ratio of the lens apertures is (300 mm at f/7.1 = 42.25 mm diameter / (221 mm at f/7.1 = 31.1) = 1.36, so the lens area ratio is 1.36*1.36 = 1.85 or 0.9 stop, which is close to what you said the advantage of the 5DIII was, but the test gave the 5DIII the 0.9 stop more light. It is not the larger pixels, but the the change in the lens diameter that delivered more light producing a lower noise image.

In your closing statements: "And if you are focal length limited and will need to crop, the Mark IV’s 1.3 FOV and higher pixel density give it the edge, even in low light conditions." I would drop the 1.3 FOV as it is not relevant and just feeds the crop factor myth.

I would also be interested in lower ISO comparisons that bird photographers more commonly use with some shadows and then lift the shadows to try and reveal shadow detail. My 5DII shows fixed pattern banding while my 1DIV is much better. Initial reports on the net said the 5DIII had about the same fixed pattern noise. (This fator is also important for night photography.)

Roger

Hi Roger. In Comparison 2 the bird is larger in the frame in the Mark IV image; you can see that in the first two image posts showing the entire frames I used for the comparisons. The second FF image is the Mark IV frame used for Comparison 2. In Comparison 1 I downsized the 5D III crop to match the size of the 1D IV, and in Comparison 2 I downsized the 1D IV crop to match the size of the 5D III.

I don't agree with your statement about giving the 5D III 0.9 stops more light, but you and I have discussed that previously.

Good point about dropping the 1.3 FOV from the conclusion.

[arash_hazeghi]

Doug tests are great, IMO and it closely resembles how photographers actually make images in the field. It is not the larger pixels that give the 5D3 its advantage but the larger sensor area which collects more light. That's why manufacturers make full frame sensors at first place. In the field photographers don't use lens diameter as a parameter when they make photographs, they use f-number. f-number and shutter speed (exposure) are fixed by the scene and subject, just as Doug did in his test.


As for 5D3 vs. 5D2

I think the 5D3 FPN has improved a little bit but it is still more than the 1D series and it will limit the low ISO DR unfortunately. Overall the improvements of 5D3 over 5D2 are limited to ISO 12,800 and ISO 25,600. In EOS 5D Mark II these ISOs are called "High1" and "High2" They are achieved by software (i.e. numerical scaling of RAW values) and therefore lead to very poor results. In EOS 5D Mark III these ISOs are achieved by on-chip amplification and will thus result in better output. Canon claim about 2-stops advantage also refers to the ISO range (true 25K vs. expansion) and it is ONLY for JPEG and EOS movie output as indicated by Chuck Westfall.

The EOS 5D Mark II and Mark III RAW files are very similar at ISOs 6400 or lower.

Very nice and thanks for sharing!!!

BTW, the first comparison (i.e. normalized FOV) is somewhat similar to DxO benchmark except for they do it with calibrated targets.

There is one more advantage of 1D4 and that is because of 1.3 crop viewfinder is more magnified, so when you look through the finder you can closeup details better when the subject cannot be approached.

[Greg Basco]

Doug, great stuff. Thanks so much for sharing something that interested me in particular for rainforest shooting and I'm sure many people out there for their general bird photography.

Cheers,
Greg Basco

[Roger Clark]

Hi Roger. In Comparison 2 the bird is larger in the frame in the Mark IV image; you can see that in the first two image posts showing the entire frames I used for the comparisons. The second FF image is the Mark IV frame used for Comparison 2. In Comparison 1 I downsized the 5D III crop to match the size of the 1D IV, and in Comparison 2 I downsized the 1D IV crop to match the size of the 5D III.

Hi Doug,
You should state the resizing and how it was done in the samples shows. Any resizing will bias the noise impression and the image detail.

I don't agree with your statement about giving the 5D III 0.9 stops more light, but you and I have discussed that previously.

It is simple: you used a zoom lens at two settings. 300 mm f/7.1 and 221 mm at f/7.1. The clear aperture of the first is 300 /7.1 = 42.25 mm and the second is 221/7.1 = 31.13 mm diameter. A larger lens collects more light. The ratio of the aperture areas is 42.25 squared / 31.13 squared = 1.84. So 1.84 is the ratio of the amount of light collected by the two lens settings.

You can demonstrate this by aiming your camera at the north star in a dark sky (wait a couple of nights and do it after sunset and before moon rise). Make 30-second exposures at 300 mm f/7.1 and 221 mm at f/7.1. You can try it with both cameras or a clearer result is if you use only one camera (say the 1DIV). Which setting (300 f/7.1 or 221 f/7.1) shows fainter stars? It will be the 300 f/7.1.
The larger lens collects more light.

Another factor in your test is the image quality at the two zoom settings. Changes in that quality could bias the test too. Another way to avoid that problem is simply change distance, then the lens remains the same quality and the same lens diameter. But also moving closer, one gets more light (remember that inverse square law).

I agree with most of your conclusions (but am confused by the resizing). But the general impression today is that pixel size is the factor that controls the noise. But it is the lens/distance that actually controls the noise by delivering the light to the pixel. Using a full frame sensor simple allows one to use a larger aperture lens to collect more light and the larger pixel simply enables the pixel to collect the light that the lens delivers. It is the lens and/or distance that determines the light collected (along with the exposure time) and not the pixel (or ISO). So effectively the larger sensor wins because of the lens and position it enables for framing the scene that collects more light producing a higher signal-to-noise image, and not the sensor itself collecting more light.

Roger

[Doug Brown]

Hi Doug,
You should state the resizing and how it was done in the samples shows. Any resizing will bias the noise impression and the image detail.

Hi Roger. From the article:
"I uploaded the RAW files from Lightroom 4.1 RC into Photoshop CS5, making no adjustments other than the Lightroom defaults. For Comparison 1 I took a 100% crop of the 1D Mark IV image taken using the same FOV as the 5D Mark III; this crop measured 1024 x 683. I cropped the same portion of the image from the 5D Mark III file and then downsized it using Photoshop’s bicubic method to 1024 x 683. For Comparison 2 I took a 100% crop of the 5D Mark III image measuring 1024 x 683. I then cropped the identical area in the Mark IV image taken at the same focal length and distance to subject and downsized it to 1024 x 683."

[Kaustubh Deshpande]

Doug, thanks for sharing.

[Alan Stankevitz]

Thanks for doing the tests. I appreciated the fact that you tested them for a focal-length-limited scenario. Your results do not seem that surprising to me. Canon's claim of 2-stops improvement is a bunch of snake oil only found in jpg's not raw.

What I am really curious about is your assessment of the autofocus compared to the Mark IV in normal light conditions and in low light conditions.

Thank you,

Alan

[Charles Glatzer]

Roger,

"But also moving closer, one gets more light (remember that inverse square law)."

The mean distance to the sun from earth is approx 92 million miles... as the light to subject distance remains fixed I do not see how the inverse square law is applicable? Could you explain further.
In outdoor ambient conditions I have never had to alter my manual exposure when moving a few meters closer to a subject when all other parameters remain constant. Macro is another animal altogether.


Best,

Chas

[Lorant Voros]

Thank you for this great test, Doug. I'm looking forward to hear about the AF performance of the new body.

[Roger Clark]

Roger,

"But also moving closer, one gets more light (remember that inverse square law)."

The mean distance to the sun from earth is approx 92 million miles... as the light to subject distance remains fixed I do not see how the inverse square law is applicable? Could you explain further.
In outdoor ambient conditions I have never had to alter my manual exposure when moving a few meters closer to a subject when all other parameters remain constant. Macro is another animal altogether.

Hi Chas,

Let's look at two situations.

The first addresses the above. Think of a small spot on your subject. That small spot is reflecting light in all directions, like a candle. Next think of the scene made of a lots of little spots, each reflecting light in all directions. As you move closer, the spot gets brighter (inverse square law). But you are correct that your exposure does not change. But it doesn't change because as you move closer each pixel sees a correspondingly smaller area and the decrease in area compensates for the more light one gets from the small spot. Thus one sees finer detail with plenty of light.

Now let's look at the alternative to image that smaller detail if one doesn't move closer. The alternative is to increase focal length (e.g. add a TC). By adding a TC the scene is magnified more, so there is less light per pixel in a given exposure time to see that same detail in a pixel. So we see less light from that small spot than we would by moving closer.

Another way to look at it is when you move closer the lens appears to cover a larger angular area from that spot, so gathers more light.

So moving closer gets more light from the fine detail.

It is probably simpler to envision this by approaching a light, like a candle. The candle certainly gets brighter as one moves closer. Or approaching car lights at night. Each tiny detail is like a little light and as we get closer it gets brighter.

Roger

[Ed Cordes]

Thanks for doing this test Doug. I have loved my 1D Mark 4 a lot since getting it 2 years ago. The new 5D Mark 3 seems to be a good alternative as a second body if you can afford it. However, I will take the pixels on the subject approach anytime when the noise characteristics are so close as in this comparison.

[Greg Basco]

Hi, Doug. I hope it's OK if I post this follow-up query to one of Arash's comments here on this thread. I think it will still be of interest to some.

Arash. I'm currently considering adding a 5DII to my kit for my landscape work. I do quite a bit at night so I was wondering about the high ISO performance. I noted with interest your comment that the 5DII and 5DIII have very similar ISO performance in the RAW files through 6400. I was hoping you could answer two questions for me.

First, do you think this will hold true for nocturnal landscape photography as well as for wildlife in low light daylight?

Second, do you think ISO 12,800 on the 5DIII is actually useful for a nature photographer shooting RAW for a coffee table book or is it just improved over the 5DII but still not really useable for high-quality work?

I really appreciate your advice!

Cheers,
Greg

[arash_hazeghi]

Hi, Doug. I hope it's OK if I post this follow-up query to one of Arash's comments here on this thread. I think it will still be of interest to some.

Arash. I'm currently considering adding a 5DII to my kit for my landscape work. I do quite a bit at night so I was wondering about the high ISO performance. I noted with interest your comment that the 5DII and 5DIII have very similar ISO performance in the RAW files through 6400. I was hoping you could answer two questions for me.

First, do you think this will hold true for nocturnal landscape photography as well as for wildlife in low light daylight?

Second, do you think ISO 12,800 on the 5DIII is actually useful for a nature photographer shooting RAW for a coffee table book or is it just improved over the 5DII but still not really useable for high-quality work?

I really appreciate your advice!

Cheers,
Greg

Hi Greg,

Yes, 5D2 and 5D3 RAW files are almost indistinguishable in all conditions up to ISO 6400 since the sensor performance is identical. The pattern noise seems to be slightly less but not much to make a practical difference for wildlife photography in rain forest or landscape work.

No, ISO 12800 in 5D3 does not meet my personal IQ standards.

The only reason to upgrade to 5D3 is the new AF system and slightly faster continuous shooting, so for landscape I'd skip the MKIII and get the MKII. I got a MKIII to test the new AF system for flight knowing that there was no advantage in IQ vs. MKII.

Hope this helps

Cheers,

Arash

[Greg Basco]

Hi, Arash. It definitely does. Thanks for your answer -- spoke right to my specific situation :-)

Cheers,
Greg

[Arthur Morris]

Good job Doug I think. I am not smart enough to understand all the variables and the science and suspect that all tests are somewhat flawed. I never do tests and I never do comparisons. I had a chance to do some actual bird photography with the 5D III at ISO 1600 and ISO 3200. You can see the images here.

BTW, B&H was able to get 5D Mark III bodies to dozens of BAA folks out of their first two shipments.

[Doug Brown]

Thanks Artie. It seems like all tests are flawed, but I gave it my best shot. The one thing I may do is upres the smaller files instead of downsizing the larger files, and see how the results compare.

[Arthur Morris]

You are welcome. Still a help and worth looking at especially for those who like tests. I am not smart enough to understand most of them :)