Originally posted by Peter Zack I really do think there's a connection between pixel density and digital noise. Since I have no engineering background, I can only go by what I see. Look again at Nikon. D3s is the ISO champ (12MP). D3x is the resolution champ (24.5 MP).
The internet is 99% noise. So, by only reading about the topic w/o gaining a true understanding will leave you in the dark. You either have to try out (shoot and print) yourself or listen to engineers or physicists like DxO. Even DPReview wrote an article which was wrong in essence because they did not take into account that Bayer demosaicing reduces noise. The topic is a mine field for those "half in the know". Not unlike DoF ...
Which is why I make this last clarification and won't continue discussing the noise topic (because it would become offtopic).
D3X, D3s, D700:
D3X is the only camera which needs to exist. But because Nikon prefers to maximize profit, they made two crippled versions: D3s in resolution and D700 in resolution
and ergonomics. D3X and D700 have about the same age and almost identical low light performance. The D3X has much better (1.5EV) dynamic range though. The D3s is a newer camera. It still has the lower dynamic range (compared to D3X), but has a slightly improved low light performance (about 0.5EV better) than both D3X and D700. Pixel size has nothing to do with that.
I post this much detail to make you think about it.
Below is another misled "non engineer" let me explain again ...
Originally posted by Winder 1. If size is the most important then MF sensor should be low-light champs.
2. If size is the most important factor then why does the GH1 4/3 sensor outperform most APS-C sensors?
3. If noise is generated at the photosite/pixel level how does having more surface area improve the efficiency of a single photosite?
Larger sensors typically have larger photosites and thus more well capacity.
The GH1 and the 7D have almost the same pixel density, but the 4/3 sensor is only 60% the size of the Canon sensor. The GH1 is with in 1/3 of a stop of the larger APS-C sensor.
Ad 1. They are. Pentax 645D should beat any non MF camera in low light. I already guesstimated (and posted) figures from available data.
Ad 2. Size is the most important factor
for a given technology. No pixel size over 2µm currently is in conflict with existing technology. All dSLRs are over 4µm.
Ad 3. You talk about pixel noise, not image noise. Pixel noise is irrelevant. E.g., what is pixel noise for a film image? The link which ClassA provided for your convenience was there for a reason
Also, I don't know if you know what the full well capacity is. It basically is the number of electrons a sensor can gather before it starts to become "full". It is basically independent of pixel size but dependent on overall sensor area and well depth. Of course, smaller pixels make a pixel's full well capacity smaller. But not that of the sensor simply because it then has more of them. That's just one example where people on the internet really always jump to conclusion too quickly. And this was a trivial consideration to rectify. E.g., the dependence of read-out noise or dark current are more interesting topics really ...
BTW, the sweet spot is nowhere near 12MP. It is about 3µm.
This translates to about 6MP - 8+MP for P&S (1/2" - 1/1.7+"), and led to the 6mpixel.org initiative.
But it also translates to 25MP for FourThird, 40MP for APSC and 100MP for full frame.
Like DoF, I won't participate (beyond this point) in a technical discussion of technical advantages/disadvantages of any particular sensor size or pixel pitch. Threads which do are legion.
I invite more contributions about what Pentax should communicate and do in the next 2 years though.