Originally posted by Class A The burden of proof is on those who claim that the "accelerator unit" can not be separated from the image pipeline.
Wrong.
It's the malcontents who should prove that the effort is not significant, and worth it - after considering all the effects.
Originally posted by Class A The "accelerator unit" is a chip which processes digital signals. Instead of processing the digital data coming from the Sony sensor, it could instead directly be fed to the PRIME processor. It is obvious to anyone with an engineering background that while at most a redesign of the "accelerator unit" chip might be necessary, there is zero hurdle whatsoever in principle to bypass it or operate it in neutral mode.
Is it?
"At most a redesign" and "there is zero hurdle" points rather to a management position, than engineering.
Originally posted by Class A The claim that no one would have complained about the K-1 II, if the K-1 had not been available for comparisons, has no leg to stand on as the image manipulation can be detected by spatial frequency analysis without any reference model. DxOMark has been detecting smoothing of RAW data from Pentax cameras long before the "accelerator unit" was introduced. The difference between this type of denoising and the "accelerator unit" behaviour is that previously RAW data manipulation did not start before ISO 1600. In other words, it was in line with other manufacturers did who also manipulate their RAW data but not in the lower ISO regions.
The claim is plausible at least; at it was the introduction of the K-1 II which triggered people into endless attacks on the accelerator. They stopped looking at the images when they realized that the only difference is the mainboard.
Originally posted by Class A If that is a too technical view for some, I clearly remember a forum member (MJKoski) presenting images of some dead animal, strongly complaining how much the fur detail had been smeared by the K-1 II. Visual comparisons to cameras using the same sensor (e.g., Nikon D800) are also capable of revealing some loss of detail in some circumstances (the dust on the PCBs on images published by pentaxforums.com springs to mind; the K-1 II clearly smeared that a bit).
I remember that blunder. More noise reduction applied by default by the RAW app on the K-1 II image, which invalidated any comparison - after fixing that, the difference was debatable; people were studying OOF area trying to find something.
It was this comparison which made me decide that you guys are nitpicking, and to go forward with the upgrade.
Originally posted by Class A The tip to underexpose and then push in post, exploiting the ISO-invariance of Sony sensors, is appreciated but there are not only the disadvantages that have been pointed out by @Breakfastographer but the K-1 III will throw a spanner in the works. It will engage the processing at ISO 100 level already.
You have no idea what it does, what the images look like, yet you complain.
Originally posted by Class A As for the topic of the "accelerator unit" being a perennial bone of contention, it would help tremendously if its supporters would just acknowledge that unadulterated RAW data has advantages for some and then moved on, simply stating that they personally don't mind or even prefer the baked RAW data, instead of accusing people with a different view as seeing ghosts or being (I quote) "disingenuous".
Or maybe it's you being unable to drop the subject. This is bashing.
Originally posted by Class A Personally, I'm happy for everyone who purely enjoys the "accelerator unit" and I don't contest that they see no loss of detail and only see benefits. That's 100% fine by me. What is not OK for anyone is to make unsustainable claims about how the "accelerator unit" is an "integral part of the image pipeline" to the extent that it could not be removed without causing image deterioration (or similar) and that anyone who sees potential downsides is only imagining the latter.
The accelerator unit is an integral part of the image pipeline; that's obvious for anyone with an engineering background.
Removing the accelerator unit means removing essential processing steps, besides the NR stage; my guess is that you wouldn't get an usable image. Well, my guess is actually that the camera will crash but let's say you worked to avoid that.
You would want to say "disabling the NR stage on the accelerator unit" instead. Even for that, the effects on the image and the interaction with all the later processing stages would have to be very carefully considered; and I assume that's what Mr. Iwasaki was talking about.
It's all about effort.