Originally posted by taks I have a question about the accelerator chip.
I don't want to sidetrack this thread with a topic that can create heated debates, but since you asked, I'll answer.
Perhaps, if you are still curious, consider creating a dedicated thread about the topic.
Originally posted by taks If I set the camera to limit the ISO to 100-640, will that effectively turn off the accelerator chip?
Yes, that will be the effect on the K-1 II, for example.
Unfortunately, the K-3 III will have the image processing kick in starting from ISO 100 already.
Originally posted by taks And if the image is too dark, boosting exposure in post will not get the manipulation from the accelerator chip?
Yes, of course not.
Originally posted by taks Is that how ISO invariance work?
Yes, that's how an "ISO invariant" sensor will work (not all sensors are like that, for instance, older Canon sensors definitely were not ISO invariant).
Note, however that there are two reasons why you are assuming an idealised picture.
- in practice it can be advantageous to use an appropriate ISO setting. Pushing an ISO 100 exposure will not always give an equivalent result.
- one of the reasons for the above is that some sensors can vary the gain used in A/D conversion. A modern breed of sensors uses a so-called "dual-gain" architecture which reduces the noise floor starting from a certain ISO setting. One can clearly see this behaviour as a discontinuity in ISO->noise graphs and it is this behaviour the Pentax engineers emulated with their software solution which is run on the so-called "accelerator chip".
FYI, the ideas some have expressed here as to why others have reservations about the use of non-optional in-camera image processing do not apply to me. I don't want things to be "more difficult" than they could be or have the mind of a five year old.
Of the top of my head, I can think of three reasons why using mandatory in-camera image processing is not a good idea:
- in the process of "improving" the image, information is lost (smoothed out). From audio research it is well known that weak signals can hide below a strong noise floor. If one takes many exposures (as you do in deep sky astrophotography, for instance), one can average out the noise and retrieve the weak signal. This results in image details you could otherwise not get (one of the reasons PixelShift produces cleaner images is because it uses multiple exposures which help with averaging out noise). Once the image has been manipulated by the "accelerator unit", the sub-noise floor signal is gone. There are respective measurements which show that the "accelerator unit" attenuates certain spatial frequencies (in order to make noisy images look cleaner).
- image processing techniques continue to improve over time. The Pentax engineers worked out an impressive way of treating images, but there is no doubt that in the future better methods will become available. Only if you have the original data from the sensor available, will future methods work optimally. If they have to work on an already manipulated image (cleaned up, saturation increased, partially sharpened, etc.) they won't be able to achieve the best results.
There are two reasons to shoot RAW: a) obtain more manipulation latitude in post-processing, and b) reserve the option to use better RAW developers in the future to squeeze more quality out of the capture. The use of in-camera image processing throws a spanner into the works of the second reason.
- DPReview will always complain about something that isn't a mirrorless camera from their darling companies. However, there is no need to hand them ammunition on a silver plate. If the K-3 III will doctor RAW data starting from ISO 100 with no option to avoid that behaviour, DPReview will make sure its readers will know about it. Forget about any awards, this will be the death sentence for any appreciative review. As much as some people here believe DPReview has no impact on sales, with the amount of exposure DPReview achieves and the influence it has on dealers, etc., there is no doubt in my mind that Pentax will be losing some sales as a result of a damming DPReview "review".