Yikes! The rumored Nikon D800.

Started Oct 5, 2011 | Discussions thread
orpheo Forum Member • Posts: 81
Re: Yikes! The rumored Nikon D800.

Thanks for your answer.

bobn2 wrote:

There are at least two reasons. First, the source of the pixels is different, d3s is Nikon, D7000 is Sony. Second, the pixel count of the D3s was limited by the EXPEED processor - they could have doubled the pixel count, but then it would have been a 4.5FPS camera, and wouldn't have competed in its intended marketplace.

As it was, putting this excellently efficient sensor in the existing D3 allowed Nikon to build a 'mid life kicker' which stole much of the 1DIV's thunder.

That sounds reasonable (although I thought D7000 is a Nikon-sensor as well?)

The pixel size of the D3s was way above what was needed for that technology, it was dictated by the speed consideration, above.

So you say they ended up with way the best low-light-performance so far ONLY as a byproduct? I doubt that a little...

The Nikon 1 series has pixels with 1/4 of the area and much the same performance.

I suppose you're referring to quantum efficency and read noise NOT the low-light-performance of the finished J1/V1cameras? And isn't pixelsize 1/6th of the D3s?

Give an example - I doubt you'll actually find one,

I look at things more in a consumer-way, yet one that doesn't want the most MP but the best low-light-performance in RAW-Format. So I look at tests of the RAW-Formats, for example at DxO. There I look at the "Sports"-Score or at the SNR18% measurement computed for "Print". Furthermore as a consumer I can only guess that Cameras launched at aprox. the same time and in the same league are also technologically cousins...

So first example: Canon 40D - Nikon D300 (not exactly the same league alltogether). Pixelsize very close and the smaller ones of the D300 end up with a little lower score.

Next came the Canon 50D - Nikon D90 pair: the Nikon with the bigger pixels about half a stop better.

And so on, the Nikon Models always have the bigger pixels and the better low-light-performance. Up to the D7000 - 60D pair.

Now you will propably say I can't compare these like that because of sensor-tech etc. But I can't help wondering what low-light-performance a 60D would have with the pixelsize of a D7000. From what you say, I guess it would still be worse, but wouldn't it come closer?

Or let's compare closer to "sensor-home": D700 and Sony Alpha 900. If pixelsize wouldn't matter at all, they should be about the same in low-light-perf, or does Sony not use its best sensors for their own cameras??

Do you have an example of that sort?

Simply, your conclusion is wrong and based on a false observation. Your guess is pessimistic. We can accurately predict the performance of the three sensors from the measurements DxO has made of the pixels.
D700 pixel has 38% quantum efficiency and 5.3 e- read noise
D3s pixel has 57% quantum efficiency and 2.8 e- read noise
D7000 pixel has 48% quantum efficiency and 2.5 e- read noise (4.3 e- at 12MP FX)
So for the same exposure, the photoelectrons collected compared to D3s, will be

D7000 FX 0.84x, giving a highlight/mid-tone noise of 1.1x, basically 0.25 stop worse.

D700 0.66x, giving a highlight/mid tone noise of 1.25x, basically 0.6 stop worse
In the shadows, again based on the D3s we'd get

D7000 FX, noise is 1.54x, D700 noise is 1.9x. One way of looking at the combined effects is 'tonal range', as defined by DxO. In this case we get, at 25k ISO
D3s - 5.5 bits
D7000 FX - 5.1 bits
D700 - 5.09 bits
Thus this will be a little better than the D700, a bit behind the D3s.
This is based on measurements and analysis.

I can't quite follow all of your math (by the way, where do you find that data?) but I'd translate it to that: enlarge the D7000s-Sensor to FF and you end up with about 37MP while pixelsize stays the same. The low-light-perf will rise according to the larger total light collecting area and the result will still be better than with the 3x bigger pixels of the much older D700. Is that correct?

The pure pixelsize-is-all-theory would propably say it stays the same and only rises a little when the image is downsized. My guess is, it will be somewhere in between. Because digital pictures are made up of individual pixels containing more or less information. Something like pixel-quality, which is what we see at 100%. The loss of the single pixels information can be compensated for by more sources of information and then downsizing the result to a useful size. But only to a certain degree, because every pixel is first processed individualy and in the extreme we would have a huge number of very tiny pixels hardly capturing any relevant (true to the light situation that is captured) information anymore. How are you gonna make a real high-resolution-picture out of that?

So my question is, where is the tipping point? Or back to Nikon 1: that would result in a 74MP FF-Sensor with the same DxO-sports-score as the D3s? And they only don't build that yet because of speed-considerations or production-cost or something like that? But then surely there would exist some high-tech-camera maybe for astronomy-uses that already has such a sensor - do you know of any example?

At the end of the day, I'm interested in understanding technical aspects to a certain degree and that is why I read and discuss in threads like this from time to time.

But as a consumer I just want a camera to fit my Canon-lenses, which at least equals the low-light-perf of a Nikon D700 (but "usable" up to 25k). I would actually prefer, if it could be done with less than 36MP, so I don't have to buy a new computer yet Other than that I really don't care how much MP it has, as long as it doesn't go below 12. (Of course there is some other criteria like pricerange, size+weight, AF and so on.)

So perhaps my conclusion, Canon could do it if only they wouldn't squeze so many pixels on their sensors, is not scientifically correct. But I'm afraid it might be true in the reality of the cameras to come next.

Sorry for the long post...

Post (hide subjects) Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow