[reh321]

Is that documented somewhere?

It seems very odd that DPReview would try and influence what a third-party had to say about the K-1 II, doesn't it?

If that is true it gives credence to the notion that someone at DPReview has an axe to grind with Pentax. Maybe because Ricoh didn't gift them a review copy, or something like that.

Were you not following the saga of the K-1ii in this section when it first came out?
At first the review appeared as it does now; then there were major changes to the section on "noise", repeating the claims of DPReview and mentioning being contacted by them (*).
After further review, the only remnant are the words

A previous version of this review made some hasty conclusions about how the K-1 Mark II’s noise performance affected stacked astrophotos. I’ve removed those changes and am investigating the issue further before deciding to change my original review, if at all.

.

Most important for this discussion is his final conclusion

Conclusion and Verdict

Pentax K-1 Mark II is a great camera. All things considered, it’s excellent for astrophotography and I was genuinely impressed with the function of both Astrotracer (on the night sky) and Pixel Shift resolution mode (on my landscape foregrounds and daytime landscapes). Pixel Shift worked wonders on fine detail in the right conditions.

(*) It is natural there would be some contact between the two publications, since "Lonely Speck" is also supported by Amazon.

Disclosure

We are a participant in the Amazon Services LLC Associates Program, an affiliate advertising program designed to provide a means for us to earn fees by linking to amazon.com and affiliated sites. We are also a participant in the B&H Affiliate Program which also allows us to earn fees by linking to bhphotovideo.com.

---

Last edited by reh321; 09-26-2019 at 06:50 AM. Reason: got rid of extra tag

[dafbp]

Two things that it's apparently missing:
- metering area selector, a la K5. I really miss that on the KP. Have to use the third wheel or the menu :P
-3rd wheel and mode selection by lever, a la KP. Using a big wheel just for mode selection is just plain dumb. KP had it right
-grey body (that's actually a certainty)

One improvement over the KP (back to Kn roots) are the protruding buttons. Those flush buttons in the KP are a pain in the dark, or when you're in a hurry to find them.

[BROO]

1. Hybrid viewfinder patent and LCD (flocs) size used there.
(The paper on k-1 is the lcd screen size).

2. Af line on mirror. (5x13 ??)
(Af sensor entrance and guidelines are wider than before)

Links to some interesting patents.

Thanks lhsms2

Would you be able to summarize some of this and explain how it relates to the new ASP-C flagship?

[Kunzite]

Is that documented somewhere?

It seems very odd that DPReview would try and influence what a third-party had to say about the K-1 II, doesn't it?

If that is true it gives credence to the notion that someone at DPReview has an axe to grind with Pentax. Maybe because Ricoh didn't gift them a review copy, or something like that.

Ian was asked by our friend Rishi to look into this (they knew each other and were talking regularly). It's not as sneaky as looking for favorable reviews to change them - but they did influence an independent review.

A positive point - after I suggested to look into demosaicing algorithms - which I've found to influence cross-hatching artifacts - Rishi contacted Ian and:
"A previous version of this review made some hasty conclusions about how the K-1 Mark II’s noise performance affected stacked astrophotos. I’ve removed those changes and am investigating the issue further before deciding to change my original review, if at all."
I guess Ian couldn't find an issue after all.

[FozzFoster]

Hey, sorry if this has been addressed before - not really caught up on all 76 pages haha..

I haven't noticed a wi-fi button or indicator light on the new camera... and details provided yet?

Thanks!

[Class A]

just because some high-frequencies have been affected does not imply that all high-frequencies were.

Correct, but judging by the competence you display in a lot of your contributions, I'm assuming that you know full well that it isn't possible to create the perfect oracle that always correctly decides when smoothing should occur and when not.

Thus, the statistical evidence does prove that the RAW is "cooked" but that does not mean it makes mush.

Correct.

There is the fact, though, that one can see the mush in some of the images.
I'd never say that all high-ISO images of the K-1 II are unusable or that they all lack detail. However, with my own eyes -- the same that I'm using to read back what I'm typing right now -- I saw the mush. I'm sorry.

I'd bet that Pentax was surprised by the pushback on the K-1ii's accelerator given that they'd put the same technology in the K-70 and KP without complaint.

Yes, that's what I wrote a few posts earlier myself, adding that Ricoh may not care at all. After all, they are not too bothered with severe problems in their distributor network either. Possibly Ricoh were surprised, possibly they couldn't care less what DPReview writes. I don't know.

You'd think that if the accelerator ate stars or made mush, it would have been noticed but it was not.

The KP review actually contains a passage where they wonder whether the unusual lack of noise could be down to post processing.

I'm assuming it mattered that a certain individual, who thinks less of Pentax than a dog thinks of their fleas, was not a co-author of the KP review. Incidentally, the same author responsible for the KP review also wrote the K-70 review.

Now don't you think this could be more than a coincidence?

In any event, the fact that DPReview didn't pick up the issue immediately has no bearing at all on the question whether the issue matters or not. Or are you suggesting that they are such good journalists that there is no way they could have missed an issue the first time around, meaning that when they eventually wrote about it, they were making blowing things up against their better knowledge?

I find the theory that it matters who is in the author team much more plausible.

It seems that what Sony calls lossy raw files are effectively JPGs

That's not fair at all. Even the lossy raw files have much, much more bit depth than JPEGs do and the data they lost due to the lossy compression is indeed not that crucial. One can see artefacts in certain situations but for most images and viewing conditions the differences will be pretty inconsequential (very much like the K-1 II downsides of the processing).

Physicist falconeye actually wrote about the lossy compression back then stating, IIRC, that it was a rather clever approach and that you can recreate pretty much the same image by adding artificial noise after decompression. The reasoning is that it does not matter whether you faithfully store true noise or not store it (saving bandwidth) and then add the same kind of noise later. Again, the question is how do you avoid removing anything that isn't true noise but part of the signal and whether anyone would be able to tell if there was tampering. In both cases, lossy Sony format and K-1 II processing, there are cases where the tampering is observable but as you heard from BigMackCam the downsides of Sony's approach rarely made a difference and as you can hear from many K-1 II users, the downsides made no difference to them and they only enjoyed the positive effects.

I think the whole "accelerator" unit discussion was not completely off-topic because we are discussing future Pentax products and it could very well matter which approach Ricoh will take with future cameras but I think the debate has pretty much run its course. I don't want to repeat the same arguments over and over again and will very much try to not engage in this discussion anymore unless someone directly asks me something.

[MMVIII]

Two things that it's apparently missing:
- metering area selector, a la K5. I really miss that on the KP. Have to use the third wheel or the menu :P
-3rd wheel and mode selection by lever, a la KP. Using a big wheel just for mode selection is just plain dumb. KP had it right
-grey body (that's actually a certainty)

One improvement over the KP (back to Kn roots) are the protruding buttons. Those flush buttons in the KP are a pain in the dark, or when you're in a hurry to find them.

The metering area selector is now a button, left to the viewfinder. While I always like selectors where you can also see the chosen mode while the camera is turned off, a button might offer more than three options, if they would make sense.

Maybe I'm fetching to far, but the larger top display might show the selected mode, as it would for the smart functions assigned to the upper right selection wheel. And in this I could imagine that an e-paper display instead of the (more probable) ("calculator")LCD could be an energy efficient and the settings would be readable without having to turn the camera on. (not going to happen, but still...)

[Mistral75]

Thanks lhsms2

Would you be able to summarize some of this and explain how it relates to the new ASP-C flagship?

For instance the entrance to the AF sensor is wider than before (it goes closer to the borders), meaning that (i) the coverage of the AF points is wider (it further extends to the left and right) and/or (ii) the central AF points are more sensitive than those of the K-3 / K-3 II / K-1 / K-1 II (the rangefinder base is larger).

See the comparison below. The angle of shooting is not exactly the same (it's higher for the K-3 II) but the outcome is unambiguous.

[Class A]

Were you not following the saga of the K-1ii in this section when it first came out?

I witnessed the DPReview saga but not the developments around Ian Norman's review.

...and mentioning being contacted by them

Ah, OK. Good on him to disclose that.

It is natural there would be some contact between the two publications, since "Lonely Speck" is also supported by Amazon.

I don't find it natural at all. Amazon is huge and I don't think there is anyone caring whether or not there is consistency between what independent sites state.

It seems more plausible to me that DPReview's "science editor" Saurarshi Jyoti ("Rishi") Sanyal has some axe to grind with Pentax and since he apparently knew Ian, he thought he could further his cause by trying to "correct" an independent review.

It's not as sneaky as looking for favorable reviews to change them - but they did influence an independent review.

I think it was downright shady. If Ian didn't see an issue on his own, why go and try to influence him?
The noble interpretation is that Saurarshi Jyoti ("Rishi") Sanyal only had the customer in mind and wanted to protect innocent customers from really bad Pentax cameras but having been in his line of fire myself and having seen the mud he threw at Pentax in his comments, I have difficulty seeing him as a saint figure.

I guess Ian couldn't find an issue after all.

Yes, that seems to be the case. However, Ian wrote that he wasn't able to make comparisons with a K-1, so we don't know whether he would have been even happier with the K-1.

[stevebrot]

The source images, however (IIRC), at least include lens cap shots. In other words, you are only looking at what the sensor-processor pipeline produces.

Yes and no. The data in the DNG/PEF is always the product of the pipeline. The dark frames would be the product of the sensor output after data shaping/trimming and whatever other pipeline processing might be done and would be an indication of intrinsic system noise, including any pipeline NR. The assumption that dark frame sensor output prior to pipeline processes is zero mV from all sensels is not true.

If you then see smoothing, it comes from the processor (as the sensor is a standard part that doesn't apply such frequency-dependent noise reduction).

The problem is "seeing" the NR. Doing so is dependent on knowledge of the pre-processed state and cannot be done be examination of pipeline output alone. An automobile analogy might be attempts to describe ignition/fuel injection algorithms and final drive ratios based on 1/4 mile times and measurement of rear wheel diameter.

There are two different things being discussed here. The one is the methodology at Photons to Photos, the other is the apparent softening of results above some ISOs observed by some users. I was talking about the former. P2P does not evaluate detail capture, they are only concerned with guessing reasons for observed noise, so their analysis is a red herring in regards to your concerns.

If Pentax is applying in some sort of smoothing to mask NR artifact in pipeline processing, it would be readily visible in the results of controlled shots of a standardized, high detail subject. That being said, there are concerns associated with the RAW processing. ACR does a lot of remapping of pixels as part of its standard approach to low-end values (shadow detail) rendering. This is done to overcome the intrinsic paucity of tones at the low end inherent in digital capture. Pulling low values only makes it worse.* As such, it is important that any testing involve details in the mid-range, value-wise.


Steve

* Yes, I tested this out, but am not in a position yet to provide an article.

[Class A]

While I always like selectors where you can also see the chosen mode while the camera is turned off, a button might offer more than three options, if they would make sense.

Another advantage of a button-based manipulation style as opposed to a selector with discrete physical states is that the respective choice (e.g., the metering mode) can become part of a stored user style or may change when modes are changed to re-assume the value the option had when the camera was last in that mode.

And in this I could imagine that an e-paper display instead of the (more probable) ("calculator")LCD could be an energy efficient and the settings would be readable without having to turn the camera on. (not going to happen, but still...)

Yeah, I don't see an e-paper display happening either, but I agree that a matrix display would be nice so that choices made (say which CH/CL option was chosen) can be fed back via the top LCD rather than requiring the user to operate blindly or having to turn on the back LCD for that purpose only.

[stevebrot]

Thanks a lot for including the following part when quoting Ian Norman:

"I wish that manufacturers would stop trying to apply noise reducing techniques to RAW files. These operations are much better saved for JPEGs or for post processing when the photographer can make the choice of how much or how little noise reduction to apply."

Does Ian's wish make him a "raw purist" who is needlessly concerned about principles instead of just wanting to get the best images possible? I personally don't think so.

My perspective as well. I am picky about NR on my images and apply it as sparingly as possible. If I wanted the camera to do it for me, I would shoot JPEG.


Steve

[stevebrot]

P.S. I'd bet that Pentax was surprised by the pushback on the K-1ii's accelerator given that they'd put the same technology in the K-70 and KP without complaint. You'd think that if the accelerator ate stars or made mush, it would have been noticed but it was not.

Different market segment. One of the selling points of FF is intrinsic low noise.


Steve

[Class A]

The assumption that dark frame sensor output prior to pipeline processes is zero mV from all sensels is not true.

I don't see where anyone made that assumption.

The problem is "seeing" the NR. Doing so is dependent on knowledge of the pre-processed state and cannot be done be examination of pipeline output alone.

No, there is no need for knowledge about the pre-processed state (in the sense you seem to be implying).

It is known how the noise distribution (the amplitudes over spatial frequencies) should look like in the absence of any processing. A Gaussian distribution is typically a very good approximation of what to expect; for some components a Poisson distribution is more appropriate. In any event, deviations from the expected distribution unequivocally prove that the data wasn't left alone but has been tampered with.

FWIW, DxOMark also published when they detected smoothing of raw data. BTW, many manufacturers apply some noise reduction to raw files. The (not insignificant) difference to the K-1 II case is that most do it much, much later, i.e., starting with much, much higher ISO settings.

[stevebrot]

1. Hybrid viewfinder patent and LCD (flocs) size used there.
(The paper on k-1 is the lcd screen size).

2. Af line on mirror. (5x13 ??)
(Af sensor entrance and guidelines are wider than before)

Links to some interesting patents.

??2018-151588 | ???????IP Force?

??6515452 | ???????IP Force?

??2017-138544 | ???????IP Force?

??2017-138508 | ???????IP Force?

??2017-194526 | ???????IP Force?

??2017-191295 | ???????IP Force?

j-platpat

Yay! Change of subject! Pretty pictures too!


Steve