I was in private communication with him about his 14EV DR prediction. He was spot on indeed!
I just thought that there have been some assumptions in his result which made me stay a bit more conservative. However, I concede all his assumptions were legitimate.
Gordon, excellent work!
+1
and stunningGoing by these results
Actually, fourth place is a tie of three. So, the top six are
- P65+
- D3X
- P40+
- 645D, D3s, K-5
However, the smaller sensor of the K-5
must perform worse for noise performance compared to the bigger. It's low light score is just in line with a very good sensor of its size (best APSC ever but the D3s beats it in ISO/surface which would translate to 1400 on APSC compared to 1162 for the K-5). So with an equivalent lens, D3s is the
only camera beating the K-5 in low light performance (by 1/4 stop).
And the medium format cameras excel in color depth and resolution.
But in dynamic range (where the larger sensors used to be better too), the K-5 outclasses everything tested so far. Even the SuperCCD Finepix S3/5 Pro and bigger D3X are beaten by half a stop!
If the D7k for whatever reason stays behind that result then K-5 will wash tons of cash into Pentax which I hope they have good use for
That's correct. But only applies to the overall score which is a bit artificial.
Instead, look at
- resolution
- dynamic range (landscape score)
- color depth (portrait score)
- SNR (or DR) at true ISO 1600 in Print normalization (some use the sports score instead which is poorer though)
and you have four reliable figures which are much more trustworthy than the overall score. Of course, K-5 doesn't rank high in resolution or ISO 1600 SNR because of its relatively small sensor. But its very good elsewhere and the best APSC sensor ever, with only the 7D sensor having higher resolution.
See above. Best testing lab with a broad service. Run by physicists
My verdict on DxO is rather constant independently from their last test. Great in individual results. A bit artificial when it comes to overall presentation.
BTW,
Anybody note how they tested the K-5 prior to D7000 and how they quote an indicative price of 1374 USD (which is French street price w/o VAT)?
I'm sure the K-5 is a big hit at DxO Labs in Paris
One last note...
Everybody wondering how DR may increase in the future beyond what we have seen now ...
According to Gordon's analysis and this DxO result, the full well capacity is ~40,000 electrons and read-out noise is ~3.2 electrons (Gordon guessed 4) where luminosity reaches 3.2 electrons too (SNR 1:1). However, at 3.2 electrons, the photon shot noise is 1.8 electrons already and the pure electronic read-out noise is only 2.6 electrons. That's an amazingly good result, chapeau to Sony engineering.
On the other hand, it only leaves maybe one more stop DR which is feasible with a sensor of that size, i.e., DR 15.0-15.5 is kind of a barrier for APSC sensors without offering native lower iso (an
ideal sensor of the same full well capacity would have 15.7 EV DR).