Originally posted by Class A That test(if we can call it that) is hardly conclusive.
There are so many issues with his method that it makes one wonder what he ever expected to obtain from this other than more questions.
Personally I'd like to see it done right(same lens, same file format and most of all with a real world subject(color)).
Quote: Have you read any of GordonBGood's thread at DPR? IIRC, he also only came to the conclusion that the K20D is slightly better only marginally after you consider the in-camera noise reduction that is higher in the K20D compared to the K-7.
Yes, I've followed GB's work and make much use of his utility(which is a godsent btw).
Unfortunately... since noise is exponential in nature, terms such as "slightly better" and "marginal" have no place in performance evaluations.
ie. slightly better becomes substantially worst under Ev adjustments and so on and so forth.
Quote: As I said, looking at RAW files is not looking at the sensor since Pentax applies NR to RAW files and not all models receive the same NR. NR also varies with the ISO setting and the K-7 is known to use less NR then the K20D at the higher settings.
You could go there.
Or... your could look at the real world data and accept it for what it is.
In our case, practicality trumped theory as one unit(s) clearly outperformed the other, no matter what.
So, I simply can't agree with such a conclusion. Having said that... I'm not doubting NR really is the issue here either.
Therefore if Pentax applied some fancy processing to even the score between both units, I for one wouldn't be bothered by this at all.
Quote: Also, you'd really need to look at the spatial resolution plotted against ISO as well. What if the K20D achieves better out-of camera noise values but at the expense of resolution? Did you try to use extra NR on K-7 files to see if you could bring them to the level of the K20D?
Yes, absolutely. In fact, we spent an entire week shooting and applying various NR methods in our own attempts to match the K7 to the K20.
Having said that, no matter what we did, we were never able to match the K7 on any of the K20D files at higher sensitivities.
And though the K7 did produce sharper files with higher curves by default, the same could be done with the K20 in PP also.
Additionally... adding NR(which was also suggested by Pentax) on the K7 did not resolve the situation as we would have hoped.
The main reason was that the added NR came at the expense of detail loss in the final output(moreso than that of the K20 files).
So the short answer was no... adding onboard did not level the score between the two units.
Quote: BTW, according to GordonBGood's finding the K-x's oh-so-fantastic high ISO performance is only due to some part because of better sensor performance. The other part comes from clever noise reduction which is so clever that it is even hard to show that is being done.
Great!
All we need now is for Pentax to update the K-7 with the Kx's clever NR processing and were golden. Though something tells that's just not going to happen
Without disrespecting others comments and observations, I'm going to go with the notion that the Kx performance is just as reliant on the sensor as it is on the clever NR algorithms that accompanies, it.
IOW. if it were simply an processing issue, don't you think others would follow suit? And more importantly, wouldn't we be seeing this on the soft side of things?
Having said "all this"
I am a firm believer in a practice trumps noise solutions too(pun intended).
And so I would be more than happy to take images with you on a color chart with both unprocessed and processed output(from RAW) in hopes of achieving something tangible in the process.
Because who knows...
I might end-up eating my words and picking-up another K7
But as it stands, for me... the K7 is inferior to the K20D where noise in concerned.
And I for one, would love to be wrong about that.