[MarkJerling]

Can someone please explain to me, why the K-5 and K-5IIs scores higher than the new K-3?
Much appreciated!
See Pentax K-3 versus Pentax K-5 IIs versus Pentax K5 - Side by side camera comparison - DxOMark
This makes no sense to me.
Yes, I know, the sensor score is just that. But, I would still have expected the K-3 to score higher.

DXOMark says this about their sensor score:
What does Sensor Overall Score show?
Sensor Overall Score shows a camera’s:
* Sensor quality in terms of noise.
* Ability to render high contrast.
* Formation of colored noise.
* Ability to shoot in low light.
Sensor Overall Score does not show a camera’s:
*Resolution, i.e., its ability to render fine details.
*Lens quality.
*Optical aberrations.

[adpo]

The K-5 and K-5II both have ISO80, whereas the K-3 only hits ISO100. That, along with larger photo sites, ends up giving that half-stop or so advantage in dynamic range to the older chip. DxO heavily weighs their score on DR response, hence the 2 point difference. If you look at the measurement charts produced by DxO, you'll find that they're essentially on-par in performance

[dane.dawg]

Had the K-3 offered an ISO80 setting like its predecessors then the score would almost certainly have been on a par with them.

[Rondec]

I think it is interesting to see what Falk Lumo says about it.

https://www.pentaxforums.com/forums/172-pentax-k-3/243728-how-much-do-you-tru...ml#post2594468

I will let you look at it, but basically, the K5 and K3 test the same (a difference in score of 2 is not going to be perceptible), but the K3 will tend to have higher resolution at low iso due to higher megapixels.

[goubejp]

Had the K-3 offered an ISO80 setting like its predecessors then the score would almost certainly have been on a par with them.

I completely agree with this sentence. In the score the dynamic range is taken in account whatever iso is. You may notice on the dynamic range curve that the K3 is very good in this aspect compared to the K5, but has less MAXIMUM dynamic range simply because it can't go down to 80 iso.

[MarkJerling]

Thanks everyone. It seems to me that DXOmark places too high a score on the 20 iso difference.

[dosdan]

Thanks everyone. It seems to me that DXOmark places too high a score on the 20 iso difference.

A difference of 2 points in a DXOMark is insignificant in the real world. Furthermore, achieving high DR in real world situations is problematic, with veiling glare in lenses (particularly zooms) in high contrast situations likely to place a much lower limit on an image's DR.

BTW, a formula that was empirically determined by Peter Van Den Hamer (Peter.vdHamer.com | Reusable thoughts on pixels and bits.), and based on the DxOMark of the Leica M8's relationship between its individual scores and the DxOMark score, is:

59 + 4.15x(PORTRAIT - 21.1) + 3.6x(LANDSCAPE - 11.3) + 4.6xLOG2(SPORTS/663) + 0.036

The co-efficients give you an idea of the relative weighting given to each score in the overall DxOMark.

This formula gives calculated scores of 83.6 vs 82 (DxOMark) for the K5, and 81.4 vs 80 (DxOMark).

You can see how well the formula performs by downloading my spreadsheet that compares a range of other Canon, Nikon & Pentax scores:

https://www.dropbox.com/s/yjrzcn7vk26phn7/DxOMark%20Scores.xls

Dan.

[MarkJerling]

A difference of 2 points in a DXOMark is insignificant in the real world. Furthermore, achieving high DR in real world situations is problematic, with veiling glare in lenses (particularly zooms) in high contrast situations likely to place a much lower limit on an image's DR.

BTW, a formula that was empirically determined by Peter Van Den Hamer (Peter.vdHamer.com | Reusable thoughts on pixels and bits.), and based on the DxOMark of the Leica M7's relationship between its individual scores and the DxOMark score, is:

59 + 4.15x(PORTRAIT - 21.1) + 3.6x(LANDSCAPE - 11.3) + 4.6xLOG2(SPORTS/663) + 0.036

The co-efficients give you an idea of the relative weighting given to each score in the overall DxOMark.

This formula gives calculated scores of 83.6 vs 82 (DxOMark) for the K5, and 81.4 vs 80 (DxOMark).

You can see how well the formula performs by downloading my spreadsheet that compares a range of other Canon, Nikon & Pentax scores:

https://www.dropbox.com/s/yjrzcn7vk26phn7/DxOMark%20Scores.xls

Dan.

Thanks Dan. Very interesting!

[Rondec]

Thanks everyone. It seems to me that DXOmark places too high a score on the 20 iso difference.

Well, two points difference is unlikely to be visible in real life -- 15 points means one stop of difference. If you look at the graphs, the SNR curve and the dynamic range curve for both cameras lie almost on top of each other. For whatever reason, the base iso of the K5 was actually 70 (as measured by DXO Mark), the base iso of the K3 is 90. I'm sure that if Pentax could figure out a way to get a little more dynamic range out of the K3, they would have, but I guess the sensor just wouldn't do iso 80.

[dosdan]

I'm not interested in the DxOMark overall score per se, as the weighting of each individual score is DxO's idea, and others may place a difference emphasis of what performance parameter is more important to them. However two of the sensor scores I find very useful:

Portrait Score, based on optimum Color Depth. I can't relate this to real world situations.

Landscape Score, based on maximum Dynamic Range. A high technical DR is indicative of either a high FWC or a low total read noise figure (the RMS combination of 3 non-correlated noise sources: Sensor RN; PGA noise; ADC noise), or both. Since the max DR (used for this score) is at the true base ISO, a high DR score usually indicates a very low-noise ADC implementation. (Sensor RN is more significant at mid-to-high ISOs.) A very low-noise ADC is important when boosting underexposed images (e.g. flash failures) and when boosting deep shadows. So a high Landscape score (e.g. 13-15 EV/stops) indicates good useful performance in these areas.

Sports Score, based on Low-Light ISO. It is the highest ISO at which the imaging system meets all 3 of the following criteria:

SNR of at least 30dB
DR of at least 9 EV/stops
Colour depth of at least 18bits.

Meeting all 3 criteria should produce a good quality image (technically, at least). Usually it's the SNR performance that is the dominant factor in determining the score. A bigger sensor will have a higher score because it captures more photons and has less shot/photonic noise in comparison to the signal. Shot noise is an intrinsic property of light (due to the quantatised nature of photons) and the signal-to-shot-noise ratio is proportional to the square root of the number of photons captured. A higher score in the same-sized sensor comparison (i.e. the same format e.g. APS-C) indicates a better fill-factor/better micro-lens performance/better Quantum Efficiency). As fill-factors have improved, due to better micro-lenses, QE has started to exceed 50% in some cameras. (In old micro-lens-less sensors, QE was less than 25%.) It will be technically impossible to get more than another stop of high ISO performance improvement, (can't exceed QE=100%), but a high Sports score indicates that a sensor is efficiently capturing and converting more incident photons hitting it and will have have a bettersignal-to-shot/photonic noise ratio. Practically, it means you can use a higher ISO for the same amount of visible noise. And, for similar QEs, but when comparing across formats, the ratio of Sport/Low Light scores should be reasonably close to the ratio of sensor areas:

Camera	Format	Sensel Pitch	ISO/Score	Score Ratio	Sensor Area Ratio
D800	FF	4.7µm	2853	2.34	2.34
K-3	APS-C	3.9µm	1216	1	1
OM-DE-M5	MFT	3.7µm	826	0.61	0.68

Dan.

[MarkJerling]

Well, two points difference is unlikely to be visible in real life -- 15 points means one stop of difference. If you look at the graphs, the SNR curve and the dynamic range curve for both cameras lie almost on top of each other. For whatever reason, the base iso of the K5 was actually 70 (as measured by DXO Mark), the base iso of the K3 is 90. I'm sure that if Pentax could figure out a way to get a little more dynamic range out of the K3, they would have, but I guess the sensor just wouldn't do iso 80.

Thanks Rondec. I suppose, I've always felt my K-5 could do better in low light situations. It would be interesting to compare, in that sense, with the K-3.

[MarkJerling]

I'm not interested in the DxOMark overall score per se, as the weighting of each individual score is DxO's idea, and others may place a difference emphasis of what performance parameter is more important to them. However two of the sensor scores I find very useful:

Portrait Score, based on optimum Color Depth. I can't relate this to real world situations.

Landscape Score, based on maximum Dynamic Range. A high technical DR is indicative of either a high FWC or a low total read noise figure (the RMS combination of 3 non-correlated noise sources: Sensor RN; PGA noise; ADC noise), or both. Since the max DR (used for this score) is at the true base ISO, a high DR score usually indicates a very low-noise ADC implementation. (Sensor RN is more significant at mid-to-high ISOs.) A very low-noise ADC is important when boosting underexposed images (e.g. flash failures) and when boosting deep shadows. So a high Landscape score (e.g. 13-15 EV/stops) indicates good useful performance in these areas.

Sports Score, based on Low-Light ISO. It is the highest ISO at which the imaging system meets all 3 of the following criteria:

SNR of at least 30dB
DR of at least 9 EV/stops
Colour depth of at least 18bits.

Meeting all 3 criteria should produce a good quality image (technically, at least). Usually it's the SNR performance that is the dominant factor in determining the score. A bigger sensor will have a higher score because it captures more photons and has less shot/photonic noise. Shot noise is an intrinsic property of light (due to the quantatised nature of photons) and the signal-to-shot-noise ratio is proportional to the square root of the number of photons captured. A higher score in the same-sized sensor comparison (i.e. the same format e.g. APS-C) indicates a better fill-factor/better micro-lens performance/better Quantum Efficiency). As fill-factors have improved, due to better micro-lenses, QE has started to exceed 50% in some cameras. (In old micro-lens-less sensors, QE was less than 25%.) It will be technically impossible to get more than another stop of high ISO performance improvement, (can't exceed QE=100%), but a high Sports score indicates that a sensor is efficiently capturing and converting more incident photons hitting it and will have less shot/photonic noise. Practically, it means you can use a higher ISO for the same amount of visible noise. And, for similar QEs, but when comparing across formats, the ratio of Sport/Low Light scores should be reasonably close to the ratio of sensor areas:

Camera	Sensel Pitch	ISO/Score	Score Ratio	Sensor Area Ratio
D800	4.7µm	2853	2.34	2.34
K-3	3.9µm	1216	1	1
OM-DE-M5	3.7µm	826	0.61	0.68

Dan.

Thank you Dan, most informative.