Nikon Announces D300 and D3(YES FULL FRAME)

[DarkDTSHD]

NHL wrote:

DarkDTSHD wrote:

Yeap - The "Pro" models are $4000+ (and low volume) - Would you pay that price? :G

ISO-9001 and ROHS (lead-free) are requirements for Europe - i.e. you don't get to do business in the EU otherwise...

Ummmm the D3 does look very attractive. At least on paper. And the pics we've seen from Nikon USA are darnn impressive. Especially the fireman pic. Hmmm...

Kidding. Probably also go with a D300. Assuming I like what I read in the test reports. Keep us posted.

Regarding ISO-9001 and Europe...I see.

[DarkDTSHD]

Here are some recent D3 high ISO sample pics. There are a few pics with dead pixels but those were shot using a preproduction D3 (David Einsel's...the others were probably shot with preproduction D3's too but don't say)...

http://www.robgalbraith.com/bins/mul...id=7-8745-9153

Hope I don't get a D300 with dead pixels.

[NHL]

DarkDTSHD wrote:

Hope I don't get a D300 with dead pixels.

If anything you are guaranteed to get a dead pixel... :-)

If a manufacturer can get a 99.999999% yield on a 12Mp which is rather unlikely that's still 12 dead pixels per sensor. With smaller dies on a large wafer they can throw it out, but a sensor is rather large and a full-frame has even a bigger area (i.e. more chance for bad pixels) than a cropped one - They'll just map it out
On a Bayer based sensor you are going to interpolate the result anyway! :idea:

-> Think about your flash card, it's the same thing, the defects are built-in to the design of the cards and there's already error/detection/correction algorithm in them so you don't get a corrupted picture...

[DarkDTSHD]

NHL wrote:

DarkDTSHD wrote:

If anything you are guaranteed to get a dead pixel... :-)

If a manufacturer can get a 99.999999% yield on a 12Mp which is rather unlikely that's still 12 dead pixels per sensor. With smaller dies on a large wafer they can throw it out, but a sensor is rather large and a full-frame has even a bigger area (i.e. more chance for bad pixels) than a cropped one - They'll just map it out
On a Bayer based sensor you are going to interpolate the result anyway! :idea:

-> Think about your flash card, it's the same thing, the defects are built-in to the design of the cards and there's already error/detection/correction algorithm in them so you don't get a corrupted picture...

That's reassuring. Usually when you see bad pixels on your monitor it's time to exchange it. So, I suppose with digital cameras and the pictures they produce, as long as you don't notice it...I didn't see the dead pixels on the sample D3 pics...

When you said we're going to "interpolate the result anyway" did you mean when PP?

[JimC]

Those pixels are hot (too bright), not dead.

If you have a pixel that is always hot (bright), even on shorter exposures in good light, then it's referred to as a "Stuck" pixel. A pixel that's always dark is referred to as a dead pixel.

In most cases, when a camera develops a stuck pixel, a trip back to the manufacturer is needed. The manufacturer then updates a bad pixel table in EEPROM. When a photo is taken the image processing automatically checks the camera's bad pixel table and maps the pixel out (replaces it with values determined from adjacent pixels using sophisticated interpolation algorithms.

With most Nikon models, a trip back to the manufacturer is needed. But, with the D100 you can find software to do it yourself. The manufacturers don't normally release service software to the public though.

The idea is to make sure there is an acceptable number of defects in the conditions you're using the sensor in (light, temperature, ISO speed, shutter speed).

For the defective photosites that are in a camera's sensor at typical conditions, the factory has already mapped them out. So, while you may think your sensor is perfect, chances are, it's not. When users of cameras that develop a bad pixel use software to update the table in EEPROM, they usually see that other pixels were also found (these are the ones that were already mapped out by the factory).

Keep in mind that you've got millions of photosites, and each one is only sensitive to one color anyway (Red, Green or Blue in most Bayer Pattern designs). So, the demosaic/interpolation algorithms are combining values from multiple photosites to store at each pixel location.

As a result, photosites that are not linear in their sensitivity (either too bright or too dark compared to the way they should be), don't have as much impact on a final image when they are remapped.

It's also not uncommon for a CCD to develop more bad pixels as a camera ages (and temperature plays a role, too).

With older camera models, when you had a bad pixel (either dead or stuck on) the manufacturer's typically ran a service program to update the bad pixel map. Most consumers thought the CCD was being replaced, when the camera's processing is just interpolating to replace the bad ones. ;-) Some models only do this with jpeg data, and some also replace data in raw files.

You can find software to update some models yourself now (for a number of consumer cameras made by Nikon, Olympus and others). You can even find software to update the bad pixel table in some DSLR models (for example, the Nikon D100).

Here is one example that can work with some of the Nikon and Olympus consumer models (and I've got software to do it for the Nikon D100, too; and I know of someone that has the factory service software for some older Canon DSLR models).

http://e2500.narod.ru/ccd_defect_e.htm

With some newer cameras, the manufacturers started finding a way to let the camera do it without the need for separate service software. Olympus started it first (AFAIK), beginning with their Olympus E10 Model (this 4MP 2/3" Olympus desginedCCD was very bad for getting stuck pixels, so they came out with a firmware upgrade designed to check for them and map them out). Many newer models from Olympus also have a menu choice to remap bad pixels (even though they're not using Olympus designed sensors anymore).

Konica Minolta put in an Automatic routine to check for bad pixels and map them out monthly in some of their newer models (for example, models like the KM DiMAGE A200 and my Maxxum 5D do this). Sony retained this feature with the DSLR-A100.

I haven't checked my A700 to see if it does the same thing yet (it probably does though). You can force a remap with these models by setting the date up one month, turning it off and letting it go through the remap routine, then seting the date back correctly. It' s an undocumented feature that's been shown to work with these models (and you can see the busy light stay on much longer while it's going through the remap routine after a month change and a power off).

[DarkDTSHD]

Hey JimC,

Thanks for clearing that up. "Bright" versus "dark" pixels. But as I said I didn't see any dark pixels in the D3 pics. So maybe it's not a big deal. At least for "inexperienced eyes". Perhaps you'd have to make a print to see them? (shrug)

Btw...can you explain what is mean by "the second version has been lightened slightly in Photoshop and shows how well a D3 file can handle this type of adjustment, even when shot at a higher ISO setting."?

[JimC]

I haven't read the article.

But, if you brighten an image later, you're "pushing" the exposure. So, you have a higher effective ISO speed when you use that technique (because you would have normally required a slower shutter speed or higher ISO speed for the same exposure at a given lighting level and aperture). You'll see a visible increase in noise levels when you use that technique.

That's also a problem I'm seeing when users start to compare various camera models to see how noise compares. They sometimes compare totally different exposures (one model is exposing brighter than the other, or totally different aperture/shutter speed combinations are being used for the same exposure).

For example, I looked at one recent comparison of the Sony DSLR-A700 and Nikon D300 (pre-production) where the D300 appeared to have lower noise levels at higher ISO speed settings for the same subject in the same lighting.

But (big but), the Sony had shutter speeds exactly *twice* as fast as the D300 for the same subject, lighting, aperture and ISO speed. So, in that comparision, you'd really need to compare the D300's ISO 3200 with the A700's ISO 1600 images for a fair comparison (because the Sony was able to get the same exposure at ISO 1600 as the D300 required ISO 3200 to get from a shutter speed/aperture perspective).

We'll have to wait for reviews in controlled conditions to see how they really compare (and to make sure there are not some significant sensitivity differences between them for a given set ISO speed).

Some of that can boil down to metering differences if the images end up with different brightness (and noise tends to be worse in underexposed areas). Some of it may be sensitivity differences (cameras more or less sensitive than set). So, you have to be careful jumping to conclusions when comparing cameras, unless the conditions are *very* well controlled (preferrably with manual exposure, identical subjects/lighting, ISO speed settings, WB settings, etc.).

Another issue I'm seeing with some newer cameras is more sophisticated image processing (like Nikon's D-Lighting Feature and Sony's Dynamic Range Optimization).

These features are typically just boosting shadow areas or using different tone curves designed to decrease or increase brightness of parts of the image. If you brighten a darker ISO 3200 image by one stop, for all practical purposes, you're shooting at ISO 6400. ;-) IOW, make sure any comparisons you look at have these types of features disabled, using the same camera settings for apeture and ISO speed, same lighting, subjects, ISO speed, white balance, etc. to get a better idea of how the camera models really compare.

[DarkDTSHD]

Nikon Japan just posted a few more samplepics using a D300. But unfortunately they are at low ISO again. I think there is one at ISO 400. Not sure. But here's the link...

http://www.nikon-image.com/jpn/produ...300/sample.htm

Also in the news...Nikon UK is planning to include a copy of Capter NX with their D300/D3 cameras. Limited time offer. Hope we North Americans will get the same offer (if I end up going with a D300 heh).

http://www.dpreview.com/news/0711/07111604capturenx.asp

[NHL]

DarkDTSHD wrote:

When you said we're going to "interpolate the result anyway" did you mean when PP?

No - A camera sensor do not capture the whole color image, but only a 1/3 of its resolution (let's keep it simple) in each primary color: Red, Green, and Blue. The final image you get from the camera is in full-color, but 2/3 of of the color data are missing from each photosite - for example a red pixel can not capture the green or blue color and vice versa...

-> Theses colors info have to be interpolated from nearby pixels: http://en.wikipedia.org/wiki/Bayer_filter

[DarkDTSHD]

New D3 news...for Canadians...but the release date could be the same in the US too...

http://www.robgalbraith.com/bins/con...id=7-8745-9167