[Class A]

...so we have no information about how it works - no way of applying Occam’s Razor.

You do not seem to understand the principle. It's bottom line is that one should assume the simplest possible explanation, i.e., do not invent fanciful schemes (like extra signals that are exclusively available to in-camera processing) unless you have proof that they are necessary.

The K-1 II's "accelerator unit" behaves exactly like one would expect from a standard post-processing step.

I'm not making any statements about the K-3 II because I'm lacking comparative images.

What we already know, though, is that it is not the case that Sony sat on their hands and did not improve still photography performance of their sensors since the days of the KP sensor. We now have

• affordable big BSI sensors which objectively have better noise performance (they lose fewer photons to sensor structure),
• dual-gain architectures which improve noise performance (DR) at higher ISO settings, and
• stacked sensors which do not only help with readout speed but also with readout quality.

So, if any of these technologies are used in the K-3 III (and at least the use of a BSI sensor appears to be confirmed) then a claim like "the Accelerator II outperforms a sensor one size up" is simply not tenable.

[Kunzite]

Occam's razor suggests that in the absence of any proof that the "accelerator unit" uses any special input only available during capture, it is more reasonable to assume that it does not.

That's not Occam's Razor, that is prejudice.
US Patent for Method and apparatus for imaging an object Patent (Patent # 10,397,502 issued August 27, 2019) - Justia Patents Search

[repaap]

Yes. I know Fuji has a very closely related sensor, but it can't be exactly the same because it has a different color filter on it and has some pixels devoted to AF, while the Pentax variant doesn't have those. So it will never be a direct comparison. To compare like to like, I suppose you could compare a Fuji 24mp sensor to a Fuji 26mp sensor. The 26mp one was not a dramatic improvement in ISO performance in the fuji world, so I'm guessing that the improvement in the Pentax is mainly due to the accelerator.

Ultimately, I'll be happy if it solves my specific situation where I want to use fast shutter speeds to capture action at festivals in the evening. This image is with a K-01 and the FA 77mm Limited. From what I'm seeing so far, it looks like the K-3 III will enable me to get a much better hit rate with cleaner results, even with a zoom. I'd probably get the 50-135 f.28 for this.

This is what I'm hoping or too. I'm thinking of buying 11-18 and 16-50 PLM for K-3III. as I can use DFA* 70-200 + HD TC with K-3III also(I'v heard good things from this combo on K-p and it should be even better with K-3III). And use Primes with K-1 and it's successors. K-3III should be perfect for kind of events I'v been shooting with K-1. Smaller size, and fast in many ways and on top of tht there is that great ISO performance as we have already seen from examples

---------- Post added 03-03-21 at 10:52 ----------

You do not seem to understand the principle. It's bottom line is that one should assume the simplest possible explanation, i.e., do not invent fanciful schemes (like extra signals that are exclusively available to in-camera processing) unless you have proof that they are necessary.

The K-1 II's "accelerator unit" behaves exactly like one would expect from a standard post-processing step.

I'm not making any statements about the K-3 II because I'm lacking comparative images.

What we already know, though, is that it is not the case that Sony sat on their hands and did not improve still photography performance of their sensors since the days of the KP sensor. We now have

• affordable big BSI sensors which objectively have better noise performance (they lose fewer photons to sensor structure),
• dual-gain architectures which improve noise performance (DR) at higher ISO settings, and
• stacked sensors which do not only help with readout speed but also with readout quality.

So, if any of these technologies are used in the K-3 III (and at least the use of a BSI sensor appears to be confirmed) then a claim like "the Accelerator II outperforms a sensor one size up" is simply not tenable.

Regardless, combination what they have made so far seems to do so. If that senso would be used in Fuji, is that camera beating K-1II in ISO as well as colour precision and DR at higher ISO? If so, then it is mostly the BSI sensor tech what makes the difference. It could be. I suspect that accelerator II does indeed make the difference in making more precise colour with pure black and even at higher ISO, WITH new sensor it is working with.

[Rondec]

I think most photographers will be happy if high iso images look good with less effort. I hope Pentax is careful about where the accelerator kicks in. It really doesn't need to be work at iso 100, for instance, but I think most would agree that it is OK at iso 1600 or above. Even the K5, which has been mentioned fondly in this thread, has been identified as having "smoothing" above iso 1600. Certainly at a certain point APS-C is going to be basically unusable without a little help, I guess the question is where/when.

My experience with the K-1 II was that it gave files that were a little cleaner at high iso. I didn't find it night and day difference, but I was more able to use iso 6400 and up without noise reduction. Below that, I didn't feel that there was a particular need as the K-1 is really nice already.

[MMVIII]

What type of brain would care about 300 pixels lost when you have 67.987 pixels of glowing stars in your background or "astrophotography"?
And this clientele never applies and postprocessing noise reduction?

I consider this a topic purely for poorly skilled borderline brickwall photo fetishists who create so poor images that they have to care about a handful of pixels.

I think you nailed it. The major point is: there is no such thing as the perfect depiction of a scene. There are so many variables involved and with any image you choose a set of them to get a resulting image. In best case you are skilled and knowledgeable and can choose the parameters you know will give you the best results for your purpose. From this perspective some of these extreme side by side pixel peeping tests at 200% make sense. To know the behaviour of distinct parts of your equipment and the effect changes have. But as you can see on the comparison of the review of the KP with the K-1II: the one got praised for the high ISO charactar of the images. The other one got caught up in a theoretical debate, that got fueled by false claims ("star eater on steorids") and until now failed to support the "theory" with real examples. It might be so, that stacking of these images will be more problematic, but I have not seen an example of that. It would be easy to produce, as for the K-1 there are these two versions.

But in reality no one would have something like a reference picture of every scene that is photographed. The resulting image is influenced by so many factors that a (theoretical) loss might never be realized. Choose another lens (sample variation), choose another focal lenght, use the lens at another aperture, choose another shutter time, the clouds moved and the light changed, the clouds moved and the light changed, use another RAW converter, change one slider at the RAW conversion, use another monitor to look at the image, print it...

[texandrews]

This thread is starting to get snippy, personal, and will get shut down if it continues. Rules of thumb: don't make unnecessary comments about someone else's intentions, broad statements about groups of people or individuals, don't rise to the bait if you feel something's been directed at you---leave it alone.

[gatorguy]

I think most photographers will be happy if high iso images look good with less effort. I hope Pentax is careful about where the accelerator kicks in. It really doesn't need to be work at iso 100, for instance, but I think most would agree that it is OK at iso 1600 or above. Even the K5, which has been mentioned fondly in this thread, has been identified as having "smoothing" above iso 1600. Certainly at a certain point APS-C is going to be basically unusable without a little help, I guess the question is where/when.

My experience with the K-1 II was that it gave files that were a little cleaner at high iso. I didn't find it night and day difference, but I was more able to use iso 6400 and up without noise reduction. Below that, I didn't feel that there was a particular need as the K-1 is really nice already.

On the K3III it will be active at 100ppi. Canon's newest camera's "raw baking", if you want to call it that, kicks in early too. While I saw it noted in a review someplace, perhaps DPR, I don't recall seeing any mentions of negative effects from it.

[Serkevan]

On the K3III it will be active at 100ppi. Canon's newest camera's "raw baking", if you want to call it that, kicks in early too. While I saw it noted in a review someplace, perhaps DPR, I don't recall seeing any mentions of negative effects from it.

There's not a lot that you need to do to an ISO 100 image - I sincerely doubt the strength of the denoising is gonna be nearly as strong as the one applied at ISO 1600, for example.

[reh321]

You do not seem to understand the principle. It's bottom line is that one should assume the simplest possible explanation, i.e., do not invent fanciful schemes (like extra signals that are exclusively available to in-camera processing) unless you have proof that they are necessary.

I am a trained scientist / experienced engineer. I most certainly do understand the principal. You are misapplying it.
We have heard from Pentax’s designers several times. Your words do not agree with their knowledge. Nothing more needs to be said.

By making the 'accelerator' act "full time" on the K-3iii, the designers are now saying that they have complete confidence in it despite our continued discussions - talking about it is as useful as talking about a "flippy" LCD on the K-3iii.

[Rondec]

There's not a lot that you need to do to an ISO 100 image - I sincerely doubt the strength of the denoising is gonna be nearly as strong as the one applied at ISO 1600, for example.

Not sure about it. To me, it probably is a bit like pixel shift. That is to say, you actually see the most benefit of pixel shift in terms of noise reduction, color improvement, and dynamic range at high iso, but you definitely can get cleaner at shadows at low iso too.

I use pixel shift a lot at low iso simply because it allows me an extra half stop that I can push shadowed areas without having noise start to come in to the overall photo. I suppose this will give the appearance of having more dynamic range than competitors due to the cleaner shadows.

It could be a plus for Pentax, assuming the flaming from DP Review over the accelerator's application to RAW images isn't too great.

[Class A]

That's not Occam's Razor,...

In what sense am I not applying Occam's razor?
I specifically only talked about the K-1 II and said that the latter's results can be explained with simple post-processing. No need to evoke any notion of in-camera processing that goes beyond that. Applying Occam's razor in this case means that fancy theories should not be adopted, because they are not needed.

...that is prejudice.

How so? I was only talking about cameras I know, not any cameras I don't know yet.
I specifically said "I'm not making any statements about the K-3 II because I'm lacking comparative images."
Please justify your allegation of "prejudice" or take it back.

US Patent for Method and apparatus for imaging an object Patent (Patent # 10,397,502 issued August 27, 2019) - Justia Patents Search

Did you read the patent?

If you did, you will have noticed the following passage:

"...a charge-reading processor that reads out accumulated charges in a given pixel via a pixel circuit of the charge accumulated pixel; and a noise-reading processor that reads out noise signals from a pixel circuit for a given pixel. The noise-reading processor reads noise signals from a pixel circuit for a pixel that is an object of noise acquisition, in parallel with the reading of accumulated charges by the charge-reading processor."

(emphasis is mine).

The Sony sensor used in the K-1 II does not read out noise signals in parallel to reading out accumulated charges (signal). As a result, the K-1 II cannot apply the technique described in the patent. Note that only the sensor itself -- not an added chip like an "accelerator unit" -- can perform the necessary per-pixel noise recording. The patent specifically (and appropriately) states that a dark frame taken at any other time does not yield the benefits the patent is after, so there is no way they could have emulated their idea using the existing Sony sensor.

As far as the K-3 III is concerned, it is unlikely that it will make use of the patent because in that case it seems very likely that we would have heard of a cooperation between Ricoh and a sensor manufacturer already. After all, the sensor would be employing a Ricoh patent. Furthermore, Ricoh themselves stated that they are using -- I quote -- "image processing". In the respective interview, the Ricoh engineer did not claim to have improved upon the Sony sensor technology used in the K-1, he instead specifically said that they are performing improved "image processing".

If I don't respond to further quotes or claims, it won't imply a tacit approval on my behalf or lack of counterarguments. It will just mean that I don't see a point in keeping engaged anymore.

[BarneyL]

In what sense am I not applying Occam's razor?
I specifically only talked about the K-1 II and said that the latter's results can be explained with simple post-processing. No need to evoke any notion of in-camera processing that goes beyond that. Applying Occam's razor in this case means that fancy theories should not be adopted, because they are not needed.

Well, if they can be explained with simple post processing, I assume it's pretty simple to give an example that illustrates the point. Like, K-1 image shot at high iso and post-processed compared to K-1 II shot without such processing applied.

Otherwise, the premise does not hold.

[davidsladek]

Not sure if this was discussed, but the Accelerator unit doesn't only handle noise reduction but also the colour fidelity and maybe other variables too (dynamic range...). so for it to kick in at 100 it may not be bad news depending on what it is programmed to do at this sensitivity...

[Serkevan]

Not sure about it. To me, it probably is a bit like pixel shift. That is to say, you actually see the most benefit of pixel shift in terms of noise reduction, color improvement, and dynamic range at high iso, but you definitely can get cleaner at shadows at low iso too.

I use pixel shift a lot at low iso simply because it allows me an extra half stop that I can push shadowed areas without having noise start to come in to the overall photo. I suppose this will give the appearance of having more dynamic range than competitors due to the cleaner shadows.

It could be a plus for Pentax, assuming the flaming from DP Review over the accelerator's application to RAW images isn't too great.

Oh, I meant only in terms of denoising for single shots (like the AU works now) - pixel shift is an entirely different matter.

[Kunzite]

Did you read the patent?

To avoid a thread closure (I feel we're very near that point), I will only respond to this specific thing then stop.

Yes, I read the patent. It's an example of information not available when post-processing your images. A warning against over-simplifying a very complex problem.
We know next to nothing about what the Image Accelerator Unit does. I guess also about integrating imaging sensors in a camera. Let's not make convenient assumptions under the cover of an unrelated principle.
You're saying it's all software and can be done better in post? Prove it.

This thread is about the K-3 III, by the way. So... can we get back to being amazed at the results?