Adorama Full Frame Shoot-Out

[compuwar]

http://www.adorama.com/catalog.tpl?article=012609&op=academy_new

Looks like almost everyone is almost equal.

Here are the figures that pay (all from DxO):


Body Low Light Color Depth Dynamic Range
5DII 1815 23.7 11.9
1DsIII 1663 24 12
D3 2290 23.5 12.2
D3x 1992 24.7 13.7
D700 2303 23.5 12.2
A900 1431 23.7 12.3

Now, this is a preview, so that may explain why 24 beats 24.7 in color depth.
It also shows the overall difference between raw sensor data and processing the raw data in the case of the D3x and A900. If we look at the graphs for dynamic range, the D3x crosses EV9 at ISO 1600 where the A900 crosses at ISO 800. Fully acceptable dynamic range for an additional stop has value- the signal to noise plot for high-ISO also starts higher and ends better.

The D3/D700 are definitely the high-ISO kings today. The extra pixel density of the 5DII shows up in limiting high-ISO performance enough to make a numerical difference. The numbers between the D3 and D700 are obviously within sample variation.

Color Depth and Dynamic Range at optimal ISOs are so close as to not make a difference. The text and numbers for high-ISO don't seem to match well, since the numbers are pure S/N and the ISOs in the text seem to be more subjective (and don't account for shooting in 1/3 stop increments as most digital cameras allow.)

If we accept the text, the line-up comes out like this:

Pretty-much everything is good up to ISO 1600
D3x and A900 are good up to ISO 800
1DsIII starts to suck before ISO 800

Here's why this part doesn't make a lot of sense to me:

If we say that 24db is the cut-off, then the 5DII comes out between 1600 and 3200- but so does the 1DsIII, D3x, and A900 and the D3 between 3200 and 6400- so if we go by the numbers, I think the text is misleading, but if there's a subjective quality to the noise at a particular point, then the numbers become almost meaningless. In any case, the numbers will at least show the potential to use NR software in post.

 

[fiercetiger224]

I think all the cameras are very much equal in performance. The amazing part is, is that the 5D Mark II is nearly identical to the 1Ds in color depth, using only one DIGIC IV processor vs TWO DIGIC III processors. Now we can only imagine what a 1Ds Mark IV will yield, with TWO DIGIC IV processors!

I'm loving my 5D Mark II! It never seizes to amaze me with every photo that I take with it.

 

[Cliff3]

I think all the cameras are very much equal in performance.

Arguable, but the performance gains seem to be incremental now rather than any fundamental leaps. The effect of higher pixel density on ISO performance is pretty clear. I would imagine further refinement in that area, probably via a-a filters, along with improving data movement through the camera is the next order of business.

Me, the appearance of a spinning beach ball tells me my C2D 2.33 is impressed with 14 bit raw files, so until I upgrade the darkroom, 12mpx is more than enough for my needs.

 

[ChrisA]

Arguable, but the performance gains seem to be incremental now rather than any fundamental leaps.

That is what you'd expect from a silicone sensor. The ratio of electrons of charge per photon of light is fixed by physics. There still are some engineering "tricks" left that could boost this ratio but they are very expensive tricks and in 2009 medium form at far cheaper then using those tricks

Tricks might include thinned back illuminated chips. This makes it so the "wires" don't cast shadows but it's a small increment for a $10,000 cost. Other hyper-expensive tricks used are to evacuate the air from inside the camers then chill the sensor and electronics with liquid gas. This pretty much eliminates thermal noise but at very high cost. But again in 2009 it is way-cheaper to simply buy a Hasselblad if you need that level of performance.

What I'm saying here is that I think they have nearly hit the physical limit of non-exotic silicon sensor designs and "exotic" will never happen because it is cheaper to simply double the size of a conventional sensor.

Years ago film had reached this point, they had pretty much run into the limit is silver and dye based film and only small "tweaks" were left.

All that said, I don't think we are even close to any limit of image processing chips. They might in time get 100 or 1,000 times faster. Same for storage. I can imaging full frame 24 FPS video but not 10X better sensors