It might well be the case that any impression I have of chroma noise reduction already having been applied, one way or another, might be from in-camera jpegs or tiff images processed in Nikon Capture NX, that I've come across.

Nevertheless, comparing 5D RAWs with 12 bit D3 RAWs, both converted in ACR with no noise reduction or sharpening applied, it seems clear to me that any noise advantage of the D3 at high ISO is very marginal.

I'll be interested to see how much difference 14 bit A/D conversion makes as well as conversion in Nikon Capture NX.

Below is an example comparing a 5D shot with one stop more exposure than the D3 shot of the same scene, ie. the 5D at f8 and 1/40th compared with the D3 at f11 and 1/40th. The D3 is a full ETTR at ISO 25,600. The 5D is about 1 1/2 stops underexposed at ISO 3200 (actually ISO 4000).

You might expect that the 5D shot, with one stop more exposure, would exhibit less noise. And it certainly does, thus demonstrating that any noise advantage of the D3 is certainly considerably less than one stop.

Comparisons of other RAW images lead me to believe the D3 noise advantage, in 12 bit mode at least, is in the order of 1/4 of a stop.

A word of explanation about the file sizes

Ray,

I think it's time for you to stop discussing this and do the tests correctly. Just go and buy three sets each of a D3 plus your preferred lens, and three sets of the 5D with the closest matching lens (or just two more sets, since I guess you already have one 5D). Do the tests exhaustively with all six camera-plus-lens pairs (or better yet, each camera matched with each of its maker's lenses, for eighteen combinations in all). Then you can write us a definitive, thirty-page report on the high ISO noise characteristics of each.

Since you will own and have paid for all six sets, you will remove any hidden bias for or against rental equipment.

And, when you are finished, of course, I'll be happy to take the second best pair from each maker off your hands for free. I'll even pay shipping. How's that for an offer you can't refuse?

Eric

Eric,
You're probably right. There's not only variation amongst different converters and variation amongst different lenses but also variation amongst camera bodies of the same model. All three working additively could produce too much variation to be sure of anything.

It's probably not worth the trouble. After all, the camera doesn't really matter, does it!

ps. I thought I'd cancelled the previous post. The images were taking too much time to upload on my dial-up connection.

Ah, but you haven't quoted KR quite correctly. It's "Your camera doesn't matter." And, as I said in another thread, "but my camera does matter."

Ah, the good old days when some people often claimed to like grain. You don't hear many folks these days saying, "I was going to buy a D3, but I settled for a Canon S60 because I was afraid the Nikon couldn't provide the noise I need."

Eric

P.S. Seriously (well, a little more seriously anyway), I will be curious to see your results if you ever do rent a D3 and do careful comparisons with the 5D. I haven't done definitive tests, but I'm convince my 5D has much better (i.e., less) noise at every ISO than my S60 P&S.

Eric,
I have some reservations about this project because I think it's unlikely I'll ever buy a D3, not because it's not a superb camera; it is, no doubt about it.

But because the 5D upgrade must be on its way and when it arrives, it's almost inconceivable that it will not match or exceed the D3 with regard to all essential image quality parameters.

There's also the future Sony 24mp sensor which will soon appear in a body with a Minolta mount. I've got a few Minolta lenses.

If I do this comparison, it will be purely for the fun, the interest and the curiosity.

I'm beginning to wonder if the same amount of money I might spend on hiring a D3 and Nikkor lens would be better spent on converting my now redundant 20D (because I now have a 40D) to an IR camera.

In fact, it seems that replacing the IR filter with a clear glass filter, would not only remove the AA filter, but might improve low light performance to a degree that would knock the socks off the D3 .

For what they're worth, here are the images which I intended to upload in my earlier post. The FoVs have been precisely matched with cropping. The full images before cropping (in 16 bit) are 72.8mb for the 5D and 69.0mb for the D3. Any slight discrepancy in FoV is immaterial, in my opinion.

A major motivation in my redoing these tests would be to find out what difference 14 bit mode makes with the D3. It seems to make little difference in the 40D.

Click to view attachment Click to view attachment

Hi,

Base sensivity is a property of the sensor. There is no way you can go below, except for overexposure. You can overexpose and compensate in "raw-development" but this may lead clipping of highlights.

Check here for some insight:
http://www.clarkvision.com/imagedetail/dig...rmance.summary/

Hi,

I just checked at: http://www.clarkvision.com/imagedetail/dig...rmance.summary/

and R. N. Clark found that Low Light Sensitivity Factor is about 420 for 5D and 670 for the D3. That figure is quite close to your observation. (Should not be discussed out of context, but still interesting.)

Best regards

Erik

We should also bear in mind that low light sensitivity might vary amongst different copies of the same camera model. Perhaps my 5D is above average in that respect. I returned my first 5D because of excessive banding in shadows.

I still like grain. And I am not that keen on the plastic look that skin often has with so many grain free digital images. Having said that, I still buy a decent camera and then add grain back in if I need/want it.

My impression is that the techniques for adding simulated grain are much more attractive than simply having excess digital noise. But it is nice that we can choose to have grain or not (or noise or not) to a great extent now.