[Dan Rentea]

Because I can't afford a 1DXii - the point is that the KP tests just as well, but is a lot cheaper; I have posted images taken at ISO 25600 {at other sites as well as this one} and those who use other cameras are surprised at the Dynamic Range and color fidelity demonstrated at this ISO setting.

Take a look at DXO charts for D5 and D610 - a flagship vs an entry level camera.

D610 score for Sports (Low-Light ISO): 2925 ISO
D5 Sports score for (Low-Light ISO): 2434 ISO

It looks like Nikon D610 blows away the D5 in terms of high ISO. But the reality is totally different.

If I am satisfied with a camera that cost me a mere $700, why would I purchase a whole new system in which the body costs over $5000??

Who said anything about buying other camera from another manufacturer? I just made an observation related to test charts which have nothing to do with reality.

[reh321]

Take a look at DXO charts for D5 and D610 - a flagship vs an entry level camera.

D610 score for Sports (Low-Light ISO): 2925 ISO
D5 Sports score for (Low-Light ISO): 2434 ISO

It looks like Nikon D610 blows away the D5 in terms of high ISO. But the reality is totally different.

Who said anything about buying other camera from another manufacturer? I just made an observation related to test charts which have nothing to do with reality.

The "Photon-to-Photos" charts have totally agreed with my observations. and those of others, in the use of use of higher ISO values.
In the cases you refer to, the lines cross, and the D5 does much better than the D610 in in the "higher" values.

[Serkevan]

Take a look at DXO charts for D5 and D610 - a flagship vs an entry level camera.

D610 score for Sports (Low-Light ISO): 2925 ISO
D5 Sports score for (Low-Light ISO): 2434 ISO

It looks like Nikon D610 blows away the D5 in terms of high ISO. But the reality is totally different.



Who said anything about buying other camera from another manufacturer? I just made an observation related to test charts which have nothing to do with reality.

I'm interested; how is it different in reality? Is the D5 better in low light because of lower noise (doubtful) or because of things like "its AF works better in low light than the one in the D810"? I would be inclined to think that DxO's testing is limited to some technical, relatively easy to measure metrics.

[Class A]

...the point is that the KP tests just as well...

No, it doesn't.

It doesn't "test" as well in terms of measuring DR because the measurements reveal that the KP numbers are inflated. You could easily inflate the DR of the other camera as well by using some denoising process on the raw file. Of course, in an apples to oranges comparison (processing vs no processing), the KP does "as well".

It doesn't "test" as well in visual comparisons either because you again have to create a level playing field by allowing the file from the other camera to be denoised (and subsequently sharpened) as well.

No one is trying to detract from the fact that the KP delivers amazing results right out of the camera, with no post-processing required.
Yet, you simply cannot claim that "it tests just as well" as an FF camera when you deny the latter the same processing out of the camera that the KP received in camera.

[Class A]

I would be inclined to think that DxO's testing is limited to some technical, relatively easy to measure metrics.

The DxOMark "sports" score is simply based on sensor performance. All the scores can basically be derived from fundamental sensor properties (e.g. the full-well capacity). The DxOMark scores reflect sensor performance very well but they don't do anything else.

[SteveinSLC]

The 1Dx and D5 were both released in early 2016. It wouldn't at all be surprising if newer sensors had better performance in some regards. Certainly Canon at that point had the 5DS, which was probably the camera they intended people to get for max image quality. Nikon now has the D850, which I'm pretty sure probably has better image quality than a D5. So it wouldn't necessarily be that surprising if the much newer sensor in the KP(with more megapixels) could outperform the much older D5 or 1Dx in some aspects of image quality.
But that's kind of beside the point. Someone buying for max image quality would likely have bought a 645z, Phase 1, or Hassleblad (or the 5DS or 850). You buy the D5 or 1Dx for speed, ruggedness, battery life, and a whole lot of other features that pro sports and wildlife photographers value beyond just image quality.

[Dan Rentea]

I'm interested; how is it different in reality? Is the D5 better in low light because of lower noise (doubtful) or because of things like "its AF works better in low light than the one in the D810"? I would be inclined to think that DxO's testing is limited to some technical, relatively easy to measure metrics.

It does have better noise than D610 and a lot better colors from ISO 3200 and above. But DXO can't test this in a proper way because they use cameras in controlled environment where the light is very good and the differences can't be seen as you can see them when you shoot in real life situations. I've seen differences in the quality of files even between my 5D Mark IV and the old 1Dx. The 1Dx files were better.

I've never seen a crop camera from the same generation of a full frame camera or a generation newer that can produce better images at high ISO than a full frame. And I've played with quite a lot cameras... I would really want a crop camera that can take smaller lenses than the one used on full frame and which can match the image quality of a full frame. At ISO between 100 and 400 (800 on new crop cameras) you need eagle eyes to see the differences, but from ISO 800 things are starting to change and full frame cameras have a slight advantage in terms of noise up until ISO 3200. From 3200 ISO the difference become more evident. Again, the quality of light have a strong impact in terms of noise at high ISO. Often you see images from spectacles taken with crop cameras at ISO 10.000 that look very good due to the quality of light.

[house]

Of course you can artificially increase measured DR.

DR is defined by the distance between full saturation and noise floor. By lowering the noise floor -- achieved by smoothing and hence the removal of some detail -- you can create a file that exhibits more DR. More precisely, if that file had been produced by a camera without further manipulation then said camera would have more DR than the one producing the original file.

Thanks for that explanation. I knew that PP could increase DR but I hadn't found the logic that makes it true.

[reh321]

The DxOMark "sports" score is simply based on sensor performance. All the scores can basically be derived from fundamental sensor properties (e.g. the full-well capacity). The DxOMark scores reflect sensor performance very well but they don't do anything else.

We're not talking about DxO scores here - DxO, as far I know, doesn't even test the KP, because the KP performs denoising on data before it ever reaches the 'raw' files - but tests by other independents have shown that the KP does do very well in reducing noise and maintaining color despite very high ISO settings .... 'tests' using actual photographs show that the KP does well even at 'insanely high" ISO values.

Pentax KP First Shots: Going head-to-head against the Nikon D500 at ISO 819,200!

[Belnan]

from some KOL in China
Sensor : 26 mp IMX571
This source is very reliable.

A image about autofocus system maybe 153 point （similar to Nikon D500）
new system

Nikon D500

New APSC different from old pentax cameras
KP

Any info about a hybrid viewfinder over there?

[reh321]

No, it doesn't.

It doesn't "test" as well in terms of measuring DR because the measurements reveal that the KP numbers are inflated. You could easily inflate the DR of the other camera as well by using some denoising process on the raw file. Of course, in an apples to oranges comparison (processing vs no processing), the KP does "as well".

It doesn't "test" as well in visual comparisons either because you again have to create a level playing field by allowing the file from the other camera to be denoised (and subsequently sharpened) as well.

No one is trying to detract from the fact that the KP delivers amazing results right out of the camera, with no post-processing required.
Yet, you simply cannot claim that "it tests just as well" as an FF camera when you deny the latter the same processing out of the camera that the KP received in camera.

Pentax claims that their adjustments to higher ISO values are somehow connected to the sensor in a way that must be performed at the time, rather than being standard denoising that can be performed at any time; somehow they improve both noise and Dynamic Range {and they appear to improve the resilience of color data} as ISO value increases, but do not say how they do it. Testing of 'raw' files from the KP shows that the KP files have essentially as much DR (*) as a particular camera, which happens to be "FF". Either file could receive further "PP" processing, but we do not know exactly how that "PP" would work out because it would depend upon skill of the individual involved.

(*) DR relates to the number of meaningful bits of data used to store the image; it is an inherent characteristic of the data, and it not dependent upon how they came into being.

[bobmaxja]

Independantly of any test, if the user like the photos he take, that what count.
I have a KP and It’s perform better than my K5. Seem also in reading many post on KP, I have not read one comments that user are not happy despite peoples who dont own it , complaining about the accelerator. For me and others , this trick do the job.

[Class A]

Pentax claims that their adjustments to higher ISO values are somehow connected to the sensor in a way that must be performed at the time, rather than being standard denoising that can be performed at any time;

You seem to be referring to a single vague statement that was once made as part of an interview and that could easily be just marketing speak.

Please let me know if you can point us to something that clarifies that a Pentax engineer unambiguously makes the claim.

The reason why one ought to be very sceptical about such claims is that the Sony sensor is designed to be a black box. You cannot interfere with its inner workings. You may obtain dark frame images and use them to improve noise, but

1. such a technique wouldn't leave the denoising fingerprint that is detectable in the KP files. Dark frame substraction does not attenuate high spatial frequencies.
2. this technique is available to users of cameras without an "accelerator" unit as well.

Testing of 'raw' files from the KP shows that the KP files have essentially as much DR (*) as a particular camera, which happens to be "FF".

You can repeat that as many times you want, it doesn't become true when "DR" is straightforwardly interpreted as the "sensor DR".

You are referring to a notion of "data DR" that does not make sense if one allows any kind of data manipulation.

DR relates to the number of meaningful bits of data used to store the image; it is an inherent characteristic of the data, and it not dependent upon how they came into being.

Of course it matters how the data "comes into being".

If the KP analysed the raw data and every time it determined it to be a recording of no light at all and then replaced all bits with "zeros" and, likewise, replaced all the bits of frames it determines to be complete over exposures to "ones" then it would be measured to have maximum DR (e.g., 14-bit) independently of the ISO value.

It, obviously, matters what happens to the bits after they were delivered by the sensor.

Either file could receive further "PP" processing, but we do not know exactly how that "PP" would work out because it would depend upon skill of the individual involved.

Feel free to treat both cameras with any amount of PP you see fit to squeeze out the last bit of performance. Then compare.

This may not be entirely fair in that both cameras may not have received the optimal post-processing, but it is vastly more fair than allowing one output to be massaged before testing while the other is tested as is.

Your "fair" test is the equivalent of comparing JPEGs from two different cameras and allowing one of them to have its settings tweaked, including contrast and noise reduction and while the other is evaluated based on its default settings.

Independantly of any test, if the user like the photos he take, that what count.

Agreed.

I said before, that I'm not arguing that the KP isn't a great camera that delivers great results.

All I'm doing here, is to counter the spread of misinformation that the KP manages to match an FF camera in terms of DR.
This debate should never have lasted this long, because the graphs used by other people (not by me) explicitly communicate that you cannot compare the data points (the KP data points above ISO 640 are shown with downward pointing triangles, which are only used when there has been denoising).

[Mistral75]

From what I heard I would guess that the accelerator unit is performing the same kind of de-noising process as DxO PRIME; at least it resorts to the same kind of algorithms.

[Kunzite]

The reason why one ought to be very sceptical about such claims is that the Sony sensor is designed to be a black box. You cannot interfere with its inner workings.

You are saying that:
- noise is generated exclusively on the sensor
- no external factors could possibly "inject" noise into the image
- there is no way to detect anything about the noise, except by analyzing the final "RAW" image
I believe this is blatantly false - an oversimplification from the same species as the "pure RAW". Just a "noisier" power supply could affect the image you're getting. Or the SR system, which - as we know - it's using electromagnets. Then there's the operating conditions - temperature, exposure duration, etc.

Here's some random patent which talks about such issues:
US Patent for Method and apparatus for imaging an object Patent (Patent # 10,397,502 issued August 27, 2019) - Justia Patents Search

The fact is, we do not know which types of noise the accelerator attempts to identify and correct (AFAIK there's also a higher level noise processing done by the PRIME). All I can say for sure is that it results in better images.