p.2 #1 · 5D3 and DR (nothing new here, just hoping :D )
Here is a video from Sony's Hugo Gaggioni discussing discussing Sony's sensor technology. Starting at 22:14 he talks about CDS (Correlated Double Sampling) and how it's used to normalize reset noise. At 24:30 he talks about Sony's digital CDS implementation and how it eliminates the banding seen on competing analog CDS designs. The context is for video broadcast technology but is applicable to both their Video and Still products. Sony calls this their EXMOR technology and I believe the first DSLR sensor to utilize it was their 12MP sensor in the A700 and derivatives for the Nikon D90/D5000, followed by the 16MP variant in the D7000/D5100/K5 and 16MP alphas.
p.2 #2 · 5D3 and DR (nothing new here, just hoping :D )
fotosculptor wrote:
In 2009 I attended a talk by one of the engineers who designed the sensor for the Sigma cameras. He mentioned that at the time sensor tech captured 40% of all available light. Or in other words, there was only 2.5 stops of available advancement room 4 years ago. While electronics and algorithms can improve, reducing sensor, pattern, and non-uniformity noise, we're already at the end-game for DR on the 35mm format.
In my opinion, 3-stops is unrealistic. I'd be thoroughly impressed if RAW performance at the same MP level increased even 1 stop.
Yes but low ISO DR is a different issue, they can fix up pattern banding and lower read noise, Canon can at least, 3 stops better for that isn't crazy at all considering that the best Sony sensor already had Canon sensor beat there by something like 2 stops. I think the 1D4 sensor itself grabs a lot more low ISO DR than they actually end up getting from it.
Normalized to an 8MP print the Canon 5D2 gets not even 12 stops measured engineering DR (and usable due to banding is even lower) while the D3x and D700 get just about 14 stops measured engineering DR (and most of it is usable). The 5D2 is a full 2 stops worse max DR measured and getting closer to 3 stops worse usable DR so it is not at all crazy to hope that the 5D2 could provide nearly 3 stops more usable DR and at least 2 stops better measured DR.
Now talking about say improving mid tone SNR by 3 stops is not going to happen, since as you say, they are already too close to the limits for that to even be possible.
But Canon has lots of room to get better low ISO DR still. The DR curve bends way over at low ISO.
Also notice that the Nikon D3s has much worse maximum DR than the much higher MP D3x so max DR is not all about keeping MP low. The tech used matters far more.
p.2 #3 · 5D3 and DR (nothing new here, just hoping :D )
snapsy wrote:
Here is a video from Sony's Hugo Gaggioni discussing discussing Sony's sensor technology. Starting at 22:14 he talks about CDS (Correlated Double Sampling) and how it's used to normalize reset noise. At 24:30 he talks about Sony's digital CDS implementation and how it eliminates the banding seen on competing analog CDS designs. The context is for video broadcast technology but is applicable to both their Video and Still products. Sony calls this their EXMOR technology and I believe the first DSLR sensor to utilize it was their 12MP sensor in the A700 and derivatives for the Nikon D90/D5000, followed by the 16MP variant in the D7000/D5100/K5 and 16MP alphas.
One thing I wonder is if all that stuff is under Sony Patent and if there is any other way to match their low ISO DR, if not, we may be stuck for a long while with much worse DR. I hope not.
p.2 #4 · 5D3 and DR (nothing new here, just hoping :D )
skibum5 wrote:
One thing I wonder is if all that stuff is under Sony Patent and if there is any other way to match their low ISO DR, if not, we may be stuck for a long while with much worse DR. I hope not.
Very interesting video, thanks for sharing, snapsy. You both are right, "if all that stuff is under Sony Patent" then it will be a very long long long time for Canon to get rid of the pattern banding noise.... i wish we have 1dx raw black frame to check the latest Canon read-out processing.
p.2 #5 · 5D3 and DR (nothing new here, just hoping :D )
skibum5 wrote:
One thing I wonder is if all that stuff is under Sony Patent and if there is any other way to match their low ISO DR, if not, we may be stuck for a long while with much worse DR. I hope not.
Based on what I've read about Canon's process technology and the captive investment in their fab lines, the issue may be as much about fab economics as it is about technology. All just speculation though.
p.2 #6 · 5D3 and DR (nothing new here, just hoping :D )
snapsy wrote:
Based on what I've read about Canon's process technology and the captive investment in their fab lines, the issue may be as much about fab economics as it is about technology. All just speculation though.
Is it 25% modification requirement of someone certified patent to become yours and that will cost alot from one baseline to another in term of implementation.
p.2 #7 · 5D3 and DR (nothing new here, just hoping :D )
While it would be nice to have Sony EXMOR style clean readout, realistically I would be satisfied with the DR properties of the 1D IV.
Priority 1 is to avoid pattern noise, especially in a high MP camera where the DR at the image level at base ISO is improved if the DR at the pixel level is constant and the noise is random.
To be honest I am more concerned about the type of insufficient column gain calibration seen in the 7D. Vertical banding in the higlights, at the -46 dB level, is going to show up in smooth well exposed areas after modest contrast enhancement and sharpening. For my style of shooting it has been a more frequent problem than the shadow banding in the 5DII.
Technology and fab issues aside, Canon needs to recognize the demands and provide sufficient calibration and slow readout as an option.
Every canon shooter deserves better treament from Canon from now on, do you think. And high resolution cam don't have to be extremely hi-iso performance. We can live with 12 stops uniform DR from 25-3200 iso, can't we
and that will cost alot from one baseline to another in term of implementation.
