Spin off thread to keep Alex's Canon lens thread on topic.

Wow, check out Phil's curve for the 20D, S3pro, Minonta 7D and D70:

The noise curve for the S3 is in fact flat (well, at least according to Phil). Very interesting!

And you can see that the base ISO for the Sony sensor used in the D70 and the 7D really does have a base ISO of 200... But the visible noise from the Sony sensor is knocked down MUCH more by the Minolta firmware.

Cheers

James

Sorry for taking so long to get back to this. I went on a short ski trip and returned to find a minor crisis( more on that on another thread ).

Craig, Phil does say what the test is. Here's his description of it again:

To measure noise levels we take a sequence of images of a GretagMacBeth ColorChecker chart (controlled artificial daylight lighting). The exposure is matched to the ISO (ie. ISO 200, 1/200 sec for consistency of exposure between cameras).

 

The image sequence is run through our own proprietary measurement tool which measures the standard deviation (normalized) of the middle gray patch (indicated by the red rectangle above)....

The numbers plotted are the standard deviation of the pixel values which is well defined. What he means by (normalized) can be ambiguous, but it's not important in fitting the measurement to noise models of the sensors, as I will try to explain.

For the low ISO points, the shot noise is equal to the square root of the number of electrons in the bucket, and that number is proportional to the exposure. So in this region the noise should go up sqrt(2) for every stop and doubles for every two stops.

Here's that plot again:

Here's some numbers that fit the shot noise model:

For the 1D and 10D

ISO Noise

100 1

200 1.4

400 2

800 2.8

For the 1D-Mk2

ISO Noise

50 0.55

100 0.8

200 1.1

400 1.5

800 2.2

1600 3.1

pretty close....

At higher ISO the fixed, exposure independent part should make the curves go up faster than the square root relationship unless a Noise Ninja like noise reduction algorithm kicks in. I think this is what happens in the 20D at ISO 1600 and 3200. The 1D and 1D-MK2 dose not have much of this type of noise reduction, but the 1Ds and D2H has it even at low ISO.

Alex, you got it right. A poor high ISO performance dose imply noisy dark regions at lower ISO. Because some cameras have strong noise reduction a low noise high ISO does not imply that you'll have cleaner dark places. You need to find out if the better high ISO performance is due to a lower noise sensor or stronger noise reduction.

And you can see that the base ISO for the Sony sensor used in the D70 and the 7D really does have a base ISO of 200...  But the visible noise from the Sony sensor is knocked down MUCH more by the Minolta firmware.

Yes. At the cost of image details according to Phil.

But who use ISO higher than 100 in the most cases (OK Nikon users). By the way Minolta is probably better camera for underwater purpose than the others using the same sony sensor (large viewfinder, excellent manual control, no needs to browse the menu, anti-shake etc.). Unfortunatelly, due to the wrong marketing (high prize), it won`t be a bestseller and nobody will fabricate any UW housing for it.

Marcell

Marcell,

I don't know much about Minolta's system. What lenses do they have? I think that this is an equally important issue as price regarding whether anyone would consider making a housing?

For UW use I would want at least 2 macro lenses capable of 1:1

35mm equivalent: 50-60mm and 100-150mm

And a fisheye wide angle (180 degrees coverage) and a wide rectilinear (18-35mm equiv. on 35mm) although personally I could live without the rectilinear lens.

And a mid range zoom. 28-70mm ish. Which will fit in a macro port!

Alex

p.s. As a Nikon user I would say this - but I think that many people get overly hung up on noise because it is one area of picture quality that can be understood and evaluated easily. Issues like colour are less often discussed - which I think is a pity.

I think minolta has a pretty wide selection of good quality lenses (not as good as nikon and canon has). Third party companies (sigma, tamron) offer the same selection of lenses as for nikon or canon.

I agree that colour is overlooked somehow, but a very important issue.

(Minolta also has good colour reproduction according to the reviews)

To digress and talk about the Minolta a bit, have a look at Phil's comparison photos, as compared to the 20D. Somehow Canon knocks down the noise but keeps the detail. Like Herb said, Minolta knocks down the noise, but the image ends up smudged and loses the detail.

This has been a great thread about noise and how it's handled by the different cameras. Thanks guys!

Cheers

James

p.s. As a Nikon user I would say this - but I think that many people get overly hung up on noise because it is one area of picture quality that can be understood and evaluated easily. Issues like colour are less often discussed - which I think is a pity.

With the amount of post processing we do, I don't see color as a camera issue. I think they all cover a larger gamut than the display devices we have. The results depend more on the post processing software/workflow than the camera hardware or firmware. Also, for underwater shots, our sense of what is the "correct color" is not as well calibrated.

For underwater shots that include direct sun light, the dynamic range of the sensor is very important in both getting a low enough exposure for the sun and at the same time rendering clean shadow details. This is the point I've been trying to make in examining these noise tests in detail.