File :-(, x, )
Anonymous
o lawd, is dat sum EOS 5D MkII ISO 25600?

Suck it up, Nikon! Most cameras can't do this smooth at 3200, and this is three stops faster.
>> Anonymous
From what I've seen, the D3/D700 sensor is cleaner than that at the same speed.
>> Anonymous
>>258632
False.
>> Project !dashI8UpO.
>>259127
stop bumpan old'd threads
>> Martin !!ve2Q1ETWmJH
to bad the colours are incorrect, ty swatch chart
>> Anonymous
Ugh, I wouldn't touch 25,600 on that thing. BANDING!
12, 800 looks usable, though. There's something I didn't think I'd hear myself saying this year O_O
>> Anonymous
     File :-(, x)
yeesh at 1600

i could use this all day long with no worries

goddamn
>> ac !!VPzQAxYPAMA
>>260086
Agreed. The bump in noiseless speed is akin to the difference between a P&S and an SLR. The ISO2000 samples posted on dpreview's page look as clean as the ISO400 that my SLRs put out.

It's like my 50/1.8 has magically become a Noctilux.

(Or it will if I can ever actually save enough money to get one of these, anyway)
>> Anonymous
>>260095

i'm the one who made that should i upgrade to a 40D from a xti thread recently, asking if there was a huge difference

this is what i'm looking for, very usable 1600 and up compared to okay but not great 1600
>> Anonymous
hm, 25600 doesn't seem _that_ impressive, but given that my 300D has about equal quality to that at 1600, I am definitely looking forward to this. Yes please extra four stops.

And yeah, looking at the other samples on dpreview, 12800 is probably usable for news shots, and 3200 and less are essentially noiseless. It's pretty goddamn impressive.
>> ac !!VPzQAxYPAMA
>>260166
Well, since I didn't respond to that thread...

Yeah, difference in ISO performance between the XTi and 40D is negligible. ISO800 and 1600 look about the same on both of 'em. 40D's worth the upgrade for the extra controls (knob to pick your AF point, dual control wheels, C1/C2/C3 modes), better AF system (cross-type on all points), and live view, but not for better noise performance.
>> Project !dashI8UpO.
Imagine how much less noise ISO 25600 would have if its sensor had only 12 megapixels.
>> ac !!VPzQAxYPAMA
>>260184
There's gotta be a reason neither Nikon nor Canon has done this beyond just "People look at the resolution numbers." Maybe for the low-end, but the higher-end buys are a lot more impressed by ISOs than by megapixels. So I wonder why Canon didn't just release a 5DII with a 15MP sensor that could do a noiseless ISO6400 and usable ISO512000...
>> Project !dashI8UpO.
>>260190
Canon has severe mismanagement issues.
http://uk.news.yahoo.com/techdigest/20080917/ttc-exclusive-canon-engineers-held-back-e870a33.html
>> Anonymous
>>260184
>>260190

Warning: long read with sensor physics involved

You're all assuming that "bigger photosites = more photons = better SNR = less noise". That may be generally true, but there's more going on than that.

ISO sensitivity (equivalent) is a physical constant. A sensor set to ISO 400 should theoretically render a medium gray exactly the same as a piece of tri-X, with exactly the same amount of light exposed to each sample. At some point, both sensor and film will blow out because they have too many photons hitting them. This can be cut down through use of aperture, shutter, ND filters, so on -- but the crucial thing to notice is this:

- for a given sensitivity or ASA equivalent, the number of photons required for a medium gray is a constant across all equipment -

Okay? So say you have a piece of tri-X and a piece of tech pan. Each one can be shot in any light; the difference is that you need a long exposure with the tech pan and short with the tri-x. They each take the amount of photons that they need to make the exposure. The only difference is in the amount of grain.
>> Anonymous
>>260207
cont'd

The fundamental difference between film and a digital sensor is that, with a sensor, a single device has to work at all sensitivities. The photon wells are a set size, and the amount of light coming in is a constant for a given ISO equivalent (remember above). The next crucial thing to notice is

- no matter the size of the photon wells, a camera at a given ASA equivalent must always react to the same amount of light -

So your camera is sitting around set to ISO 25000. You take a shot, and a tiny number of photons squirt in, equivalent to the number needed for a proper exposure at that ISO. They go into these massive photon wells designed for shooting at ISO 50 and sit there at the bottom. The only way that having larger wells improves noise characteristics is by integrating the photons over a larger area -- the other key factor in noise production is, obviously, the lowest ISO the camera was designed for!

Basically, because these photon wells are so underfilled, there's always going to be a terrible signal-to-noise ratio. If your camera had a base ISO of 1600, then the wells would only need to be deep enough for that level of light, and ISO 25600 would be that much closer to a "full" exposure of the wells. When your camera sensor has to react to an enormous range of photon levels, increasing the frontal area of the sensors can only do so much to improve noise characteristics.

In essence, Canon has recognized this and is doing something else to improve noise (and they really are, just look at these photos). Incidentally, the idea that you can shoot ISO 3200 basically noiselessly is stunning to me, since Delta 3200 is pretty much a mess of grain.

TL;DR: Increasing the size of the sensor sites will only work to improve noise to a certain extent, and if you really want high ISO performance, you're going to have to give up the low end in exchange.
>> Anonymous
>>260210
We know, but because it's generally true, it's true enough to talk about it in this way.
>> Anonymous
>>260210

first informative /p/ post ever?
>> Anonymous
>>260211
Generally true means that it is a factor. But you can't say something like "well just drop it to 15 megapixlulz instead of 21 and you can shoot with a pinhole camera in a lead box". They might be able to get some astonishing noise performance if they went to 6 or 4 megapixels on the same sensor, but the ~15% to 25% increase in size that would come with a slight drop is not going to make a huge difference.
>> ac !!VPzQAxYPAMA
>>260207
>>260210
Useful and informative post is useful and informative. Thanks!
>> Anonymous
Welcome to the camera industry. Nikon will release something that does it even better next year, and then Canon will take the lead back the year after, and so on.