DX0 k-5 scores up

Class A · 11-08-2010, 01:29 AM

Originally posted by falconeye

How exactly do you think they can measure such dark luminosities?

So you're saying "measurement error"?
Do you have an idea why it is an "extended" setting?

Class A · 11-08-2010, 01:34 AM

Originally posted by garyk

New camera for next year and lens array is simply un real.
Does anyone know the standing of sigma in the dslr market.

Which Sigma lens you would like to have that is not available in K-mount?

The Foveon sensors in Sigma cameras has advantages (no colour-moiré, no demosaicing required) and disadvantages (noise problems and odd colours in very low light). They have a new version of the sensor (many more MPs) which sounds promising.

rawr · 11-08-2010, 02:16 AM

Originally posted by Class A

They have a new version of the sensor (many more MPs) which sounds promising.

Yes, the SD1 (46MP or 15MP however you look at it) looks interesting. I wonder how much the body will cost though.

falconeye · 11-08-2010, 02:37 AM

Originally posted by Class A

So you're saying "measurement error"?

That's pretty obvious, isn't it?

The 4-digit precise DR figure 14.12 EV at ISO 70 and 13.70 EV at ISO 91 are derived from the 0dB luminosities. You can compute it for yourself (the SNR numbers appear when hoovering with the mouse):

14.12 vs. 14.127 = -log_2(0.008%) + log_2(4992 x 3284/8M)/2 (ISO 80)
13.70 vs. 13.668 = -log_2(0.011%) + log_2(4992 x 3284/8M)/2 (ISO 100)

So, the difference in EV is log_2(0.011%)-log_2(0.008%) = log_2(11/8) = 0.46.

BTW, the published EV difference is 0.42 = log_2(111/83) giving an indication of possible luminosity measures prior to rounding (0.0083% and 0.0111%). However, all luminosities are probably rounded to 0.001% for a reason.

Therefore, the EV differences are within log_2(106/84)=0.34 and log_2(114/76)=0.58 from the measured data.

The expected EV difference between ISO 100 and 79 (1/3 stop less) is 0.33, the expected EV difference between ISO 70 and 91 is log_2(91/70)=0.38, the published EV difference is 0.42, the EV difference from published SNR data is 0.46. The published difference (0.04) from expectation is small meaning that they measure a bit better than the rounding to 0.001% would suggest.

But there is absolutely no reason to believe that the deviation of DR vs. ISO from expected behaviour (at ISO 80) is from anything but measurement uncertainties.

Originally posted by Class A

So you're saying "measurement error"?
Do you have an idea why it is an "extended" setting?

Because "80" is an ISO setting Pentax has never used before?

I'm sure, 80 was a compromise between two schools of thought: "real photographers don't care about the minimal difference between 80 and 100 and it is confusing" vs. "the minimal difference is what will make the K-5 shine".

So, they've put 80 where it belongs too: into the geek section aka extended setting.

Last edited by falconeye; 11-08-2010 at 02:45 AM.

gnd · 11-08-2010, 05:30 AM

DxOMark Sensor Scores

Overall Score 80
Portrait (Color depth) 23.5 bits
Landscape (Dynamic range) 13.9 Evs
Sports (Low-Light ISO) 1167 ISO

kittykat46 · 11-08-2010, 05:53 AM

Originally posted by garyk

When this thread started i started looking around. I ran across the Sigma web site. And was blown away...... I am a pentax guy that is going to buy the pentax k-5 but the line up for sigma is off the charts... New camera for next year and lens array is simply un real.
Does anyone know the standing of sigma in the dslr market.

Sigma is even more of a niche player than Pentax.
Previous implementations of the Faveon sensor have very, very good resolution, colour and dynamic range at base ISO, but everything falls apart rather quickly at moderate ISOs which even the K-7 wouldn't break a sweat....

Class A · 11-08-2010, 05:54 AM

Falk, thanks a lot for the calculations. First I wondered why the sensor normalisation contained a "/2" but I figure that this due to the fact that SNR only increases with the square root of an area factor.

Originally posted by falconeye

Therefore, the EV differences are within log_2(106/84)=0.34 and log_2(114/76)=0.58 from the measured data.

Given that they appear to measure with more precision than they publish, wouldn't it be fairer to say the EV differences are within log_2((111-0.4)/(83+0.4))=0.41 and log_2((111+0.4)/(83-0.4))=0.43?
The mid-point of that range is the published 0.42, whereas the mid-point of your range is your calculated 0.46.

The expected 0.38 (log_2(91/70)) are definitely outside the range one obtains if they measure a magnitude better than you were assuming.

The fact that all luminosities are rounded to 0.001% may be just for data presentation reasons?

I'm probably too optimistic regarding the precision with which they can measure but I don't understand how you can assume a measurement precision just from what is probably a presentation related rounding and proximity of the measured result to the expected result.

Originally posted by falconeye

I'm sure, 80 was a compromise between two schools of thought: "real photographers don't care about the minimal difference between 80 and 100 and it is confusing" vs. "the minimal difference is what will make the K-5 shine".

Quite possible.

Class A · 11-08-2010, 06:07 AM

Originally posted by gnd

DxOMark Sensor Scores

Seems like enabling ISO 80 allowed the K-5 to beat the D7000 in the overall DxO score. The minute amount of better DR (+ a tiny bit of help from the colour sensitivity) beats the (insignificant) "Low-Light ISO" advantage of the D7000.

BTW, I don't know why DxO don't show rounded crop factors (focal multipliers). I get 1.53 (1.528) for the K-5 and 1.54 (1.536) for the D7000.

johnmflores · 11-08-2010, 06:34 AM

Don't know if anyone has seen this, but here's an example of a guy pulling a usable photo from a severely underexposed shot:

Flickr: Discussing My best illustration of dynamic range in the K-5 in Pentax K-5 Fan Club

It's pretty amazing - the original shot was so underexposed that it was almost pure black!

mcjm78 · 11-08-2010, 06:54 AM

From DxO, the only reason D7000 has a higher "Low-Light ISO" score is because they "cheated" with their ISO values.

Quotes from DXO on how they evaluate that "Low-Light ISO is then the highest ISO setting for the camera such that the SNR reaches this 30dB value while keeping a good dynamic range of 9 EVs and a color depth of 18bits."

If you go by the graphs, K5 and D7000 virtually are the same until ISO 1600, where Pentax applied some "magic" K to pull ahead.

The only real significant advantage Pentax against D7000 is actually the dynamic range from ISO400 onwards, even before the "smoothing algorithm" applied from ISO1600. Thats the most interesting part of the result, which I think Pentax's Engineers might have done something right in that department.

Mike

mcjm78 · 11-08-2010, 06:56 AM

Originally posted by johnmflores

Don't know if anyone has seen this, but here's an example of a guy pulling a usable photo from a severely underexposed shot:

Flickr: Discussing My best illustration of dynamic range in the K-5 in Pentax K-5 Fan Club

It's pretty amazing - the original shot was so underexposed that it was almost pure black!

Thats amazing man! Still must resist.... :ugh:

AngryCorgi · 11-08-2010, 07:14 AM

Originally posted by mcjm78

From DxO, the only reason D7000 has a higher "Low-Light ISO" score is because they "cheated" with their ISO values.

Quotes from DXO on how they evaluate that "Low-Light ISO is then the highest ISO setting for the camera such that the SNR reaches this 30dB value while keeping a good dynamic range of 9 EVs and a color depth of 18bits."

If you go by the graphs, K5 and D7000 virtually are the same until ISO 1600, where Pentax applied some "magic" K to pull ahead.

Cheated? Seriously?? I've seen everything from "conspiracy theories at DXO" (Prior to K-5 results being released, mind you) to now "they're cheating with their ISO values". And in-camera-NR forced on RAW files is considered "engineering magic".

Thank goodness for the zany interwebz...there's always someone here willing to give me a laugh and a smile on a dull Monday.

falconeye · 11-08-2010, 07:17 AM

Originally posted by Class A

I'm probably too optimistic regarding the precision with which they can measure

No problem. Try to do a similiar measurement and your optimism turns into pessimism

falconeye · 11-08-2010, 07:34 AM

Originally posted by mcjm78

From DxO, the only reason D7000 has a higher "Low-Light ISO" score

Has nothing to do with iso and their scores are identical. Does anybody really think we are talking 4-digit precision here? The two results are 99.6% identical which shows how well DxO measures and how small the sample variation is.

BTW, the color response tab shows a neglegibly stronger Bayer filter for the K-5 which won't effect the result but is enough to inverse a -0 into a +0 (so to speak) when looking at the color sensitivity curve at ISO 800, rather than the 18% gray curve.

ogl · 11-08-2010, 09:36 AM

Originally posted by falconeye

Has nothing to do with iso and their scores are identical. Does anybody really think we are talking 4-digit precision here? The two results are 99.6% identical which shows how well DxO measures and how small the sample variation is.

BTW, the color response tab shows a neglegibly stronger Bayer filter for the K-5 which won't effect the result but is enough to inverse a -0 into a +0 (so to speak) when looking at the color sensitivity curve at ISO 800, rather than the 18% gray curve.

https://www.pentaxforums.com/forums/pentax-news-rumors/121512-d7000-cant-catch-k-5-dxo-mark.html

what do you think about D7000 sensor? is it another one? or the same?

11-08-2010, 05:30 AM	#290
gnd New Member Join Date: Feb 2009 Posts: 6	Nikon D7000 scores DxOMark Sensor Scores Overall Score 80 Portrait (Color depth) 23.5 bits Landscape (Dynamic range) 13.9 Evs Sports (Low-Light ISO) 1167 ISO

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
DXO Scores	Adam Lucas	Pentax DSLR Discussion	6	09-27-2010 11:36 AM
DA*60-250mm F4 ED SDm scores high ...	jpzk	Pentax SLR Lens Discussion	7	01-02-2010 01:19 PM
NFL Scores Predictions thread	LeDave	General Talk	9	11-18-2009 05:00 AM
Dx0 for Pentax	Rick	Pentax DSLR Discussion	2	10-27-2007 01:57 PM
DX0 now supports K10D	Rick	Pentax DSLR Discussion	6	06-01-2007 11:03 AM

11-08-2010, 01:29 AM	#286
Class A Pentaxian Join Date: Aug 2008 Location: Wellington, New Zealand Posts: 11,251	Originally posted by falconeye How exactly do you think they can measure such dark luminosities? So you're saying "measurement error"? Do you have an idea why it is an "extended" setting?

11-08-2010, 01:34 AM	#287
Class A Pentaxian Join Date: Aug 2008 Location: Wellington, New Zealand Posts: 11,251	Originally posted by garyk New camera for next year and lens array is simply un real. Does anyone know the standing of sigma in the dslr market. Which Sigma lens you would like to have that is not available in K-mount? The Foveon sensors in Sigma cameras has advantages (no colour-moiré, no demosaicing required) and disadvantages (noise problems and odd colours in very low light). They have a new version of the sensor (many more MPs) which sounds promising.

11-08-2010, 02:16 AM	#288
rawr Veteran Member Join Date: Jun 2009 Posts: 11,913	Originally posted by Class A They have a new version of the sensor (many more MPs) which sounds promising. Yes, the SD1 (46MP or 15MP however you look at it) looks interesting. I wonder how much the body will cost though.

11-08-2010, 02:37 AM	#289
falconeye Veteran Member Join Date: Jan 2008 Location: Munich, Alps, Germany Photos: Gallery Posts: 6,871	Originally posted by Class A So you're saying "measurement error"? That's pretty obvious, isn't it? The 4-digit precise DR figure 14.12 EV at ISO 70 and 13.70 EV at ISO 91 are derived from the 0dB luminosities. You can compute it for yourself (the SNR numbers appear when hoovering with the mouse): 14.12 vs. 14.127 = -log_2(0.008%) + log_2(4992 x 3284/8M)/2 (ISO 80) 13.70 vs. 13.668 = -log_2(0.011%) + log_2(4992 x 3284/8M)/2 (ISO 100) So, the difference in EV is log_2(0.011%)-log_2(0.008%) = log_2(11/8) = 0.46. BTW, the published EV difference is 0.42 = log_2(111/83) giving an indication of possible luminosity measures prior to rounding (0.0083% and 0.0111%). However, all luminosities are probably rounded to 0.001% for a reason. Therefore, the EV differences are within log_2(106/84)=0.34 and log_2(114/76)=0.58 from the measured data. The expected EV difference between ISO 100 and 79 (1/3 stop less) is 0.33, the expected EV difference between ISO 70 and 91 is log_2(91/70)=0.38, the published EV difference is 0.42, the EV difference from published SNR data is 0.46. The published difference (0.04) from expectation is small meaning that they measure a bit better than the rounding to 0.001% would suggest. But there is absolutely no reason to believe that the deviation of DR vs. ISO from expected behaviour (at ISO 80) is from anything but measurement uncertainties. Originally posted by Class A So you're saying "measurement error"? Do you have an idea why it is an "extended" setting? Because "80" is an ISO setting Pentax has never used before? I'm sure, 80 was a compromise between two schools of thought: "real photographers don't care about the minimal difference between 80 and 100 and it is confusing" vs. "the minimal difference is what will make the K-5 shine". So, they've put 80 where it belongs too: into the geek section aka extended setting. Last edited by falconeye; 11-08-2010 at 02:45 AM.

11-08-2010, 05:53 AM	#291
kittykat46 Veteran Member Join Date: Dec 2007 Posts: 923	Originally posted by garyk When this thread started i started looking around. I ran across the Sigma web site. And was blown away...... I am a pentax guy that is going to buy the pentax k-5 but the line up for sigma is off the charts... New camera for next year and lens array is simply un real. Does anyone know the standing of sigma in the dslr market. Sigma is even more of a niche player than Pentax. Previous implementations of the Faveon sensor have very, very good resolution, colour and dynamic range at base ISO, but everything falls apart rather quickly at moderate ISOs which even the K-7 wouldn't break a sweat....

11-08-2010, 05:54 AM	#292
Class A Pentaxian Join Date: Aug 2008 Location: Wellington, New Zealand Posts: 11,251	Falk, thanks a lot for the calculations. First I wondered why the sensor normalisation contained a "/2" but I figure that this due to the fact that SNR only increases with the square root of an area factor. Originally posted by falconeye Therefore, the EV differences are within log_2(106/84)=0.34 and log_2(114/76)=0.58 from the measured data. Given that they appear to measure with more precision than they publish, wouldn't it be fairer to say the EV differences are within log_2((111-0.4)/(83+0.4))=0.41 and log_2((111+0.4)/(83-0.4))=0.43? The mid-point of that range is the published 0.42, whereas the mid-point of your range is your calculated 0.46. The expected 0.38 (log_2(91/70)) are definitely outside the range one obtains if they measure a magnitude better than you were assuming. The fact that all luminosities are rounded to 0.001% may be just for data presentation reasons? I'm probably too optimistic regarding the precision with which they can measure but I don't understand how you can assume a measurement precision just from what is probably a presentation related rounding and proximity of the measured result to the expected result. Originally posted by falconeye I'm sure, 80 was a compromise between two schools of thought: "real photographers don't care about the minimal difference between 80 and 100 and it is confusing" vs. "the minimal difference is what will make the K-5 shine". Quite possible.

11-08-2010, 06:07 AM	#293
Class A Pentaxian Join Date: Aug 2008 Location: Wellington, New Zealand Posts: 11,251	Originally posted by gnd DxOMark Sensor Scores Seems like enabling ISO 80 allowed the K-5 to beat the D7000 in the overall DxO score. The minute amount of better DR (+ a tiny bit of help from the colour sensitivity) beats the (insignificant) "Low-Light ISO" advantage of the D7000. BTW, I don't know why DxO don't show rounded crop factors (focal multipliers). I get 1.53 (1.528) for the K-5 and 1.54 (1.536) for the D7000.

11-08-2010, 06:34 AM	#294
johnmflores Veteran Member Join Date: Apr 2009 Location: Somerville, NJ Photos: Gallery \| Albums Posts: 5,361	Don't know if anyone has seen this, but here's an example of a guy pulling a usable photo from a severely underexposed shot: Flickr: Discussing My best illustration of dynamic range in the K-5 in Pentax K-5 Fan Club It's pretty amazing - the original shot was so underexposed that it was almost pure black!

11-08-2010, 06:54 AM	#295
mcjm78 Senior Member Join Date: Aug 2007 Location: Perth, Australia Posts: 268	From DxO, the only reason D7000 has a higher "Low-Light ISO" score is because they "cheated" with their ISO values. Quotes from DXO on how they evaluate that "Low-Light ISO is then the highest ISO setting for the camera such that the SNR reaches this 30dB value while keeping a good dynamic range of 9 EVs and a color depth of 18bits." If you go by the graphs, K5 and D7000 virtually are the same until ISO 1600, where Pentax applied some "magic" K to pull ahead. The only real significant advantage Pentax against D7000 is actually the dynamic range from ISO400 onwards, even before the "smoothing algorithm" applied from ISO1600. Thats the most interesting part of the result, which I think Pentax's Engineers might have done something right in that department. Mike

11-08-2010, 06:56 AM	#296
mcjm78 Senior Member Join Date: Aug 2007 Location: Perth, Australia Posts: 268	Originally posted by johnmflores Don't know if anyone has seen this, but here's an example of a guy pulling a usable photo from a severely underexposed shot: Flickr: Discussing My best illustration of dynamic range in the K-5 in Pentax K-5 Fan Club It's pretty amazing - the original shot was so underexposed that it was almost pure black! Thats amazing man! Still must resist.... :ugh:

Thread Tools	Search this Thread
	Search this Thread: Advanced Search

11-08-2010, 07:14 AM	#297
AngryCorgi Forum Member Join Date: Sep 2010 Location: Northern VA Posts: 73	Originally posted by mcjm78 From DxO, the only reason D7000 has a higher "Low-Light ISO" score is because they "cheated" with their ISO values. Quotes from DXO on how they evaluate that "Low-Light ISO is then the highest ISO setting for the camera such that the SNR reaches this 30dB value while keeping a good dynamic range of 9 EVs and a color depth of 18bits." If you go by the graphs, K5 and D7000 virtually are the same until ISO 1600, where Pentax applied some "magic" K to pull ahead. Cheated? Seriously?? I've seen everything from "conspiracy theories at DXO" (Prior to K-5 results being released, mind you) to now "they're cheating with their ISO values". And in-camera-NR forced on RAW files is considered "engineering magic". Thank goodness for the zany interwebz...there's always someone here willing to give me a laugh and a smile on a dull Monday.

11-08-2010, 07:17 AM	#298
falconeye Veteran Member Join Date: Jan 2008 Location: Munich, Alps, Germany Photos: Gallery Posts: 6,871	Originally posted by Class A I'm probably too optimistic regarding the precision with which they can measure No problem. Try to do a similiar measurement and your optimism turns into pessimism

11-08-2010, 07:34 AM	#299
falconeye Veteran Member Join Date: Jan 2008 Location: Munich, Alps, Germany Photos: Gallery Posts: 6,871	Originally posted by mcjm78 From DxO, the only reason D7000 has a higher "Low-Light ISO" score Has nothing to do with iso and their scores are identical. Does anybody really think we are talking 4-digit precision here? The two results are 99.6% identical which shows how well DxO measures and how small the sample variation is. BTW, the color response tab shows a neglegibly stronger Bayer filter for the K-5 which won't effect the result but is enough to inverse a -0 into a +0 (so to speak) when looking at the color sensitivity curve at ISO 800, rather than the 18% gray curve.

11-08-2010, 09:36 AM	#300
ogl Banned Join Date: Dec 2007 Location: Sankt Peterburg Photos: Gallery \| Albums Posts: 8,382	Originally posted by falconeye Has nothing to do with iso and their scores are identical. Does anybody really think we are talking 4-digit precision here? The two results are 99.6% identical which shows how well DxO measures and how small the sample variation is. BTW, the color response tab shows a neglegibly stronger Bayer filter for the K-5 which won't effect the result but is enough to inverse a -0 into a +0 (so to speak) when looking at the color sensitivity curve at ISO 800, rather than the 18% gray curve. https://www.pentaxforums.com/forums/pentax-news-rumors/121512-d7000-cant-catch-k-5-dxo-mark.html what do you think about D7000 sensor? is it another one? or the same?