Web Analytics RiceHigh's Pentax Blog: High K-5 DR of the DxOMark: (Hardware) Mathematical Trick of the New Sony Sensor?

Wednesday, November 10, 2010

High K-5 DR of the DxOMark: (Hardware) Mathematical Trick of the New Sony Sensor?

See what our knowledgeable Russian Engineer and Pentaxian Oleg V analysed and explained:-

http://forums.dpreview.com/forums/readflat.asp?forum=1036&thread=36859818

Oh well, whilst I don't find any flaw with Oleg's arguments, I think a real-world DR test is still required to find out the truth. So, let's wait for DPR's result, as I believe their DR test is closer to real-life and should be more meaningful.

Or, if you are just a K-5 user, you can follow the methodology and test adopted by that D5000 user as quoted by Oleg V, and the *usable* DR of the sensor/DSLR can still be found out briefly. Btw, I still prefer DPR's methodology of shooting a lightness "stair" simultaneously (and of course) as this is just what we want for the widest possible DR practically in daily shooting.

But if the DR of the new Sony sensor (of the K-5/D7000) is just measured to be lower later on, then it is just sad that now even sensor maker also does trick(s) to optimise their hardware for *benchmark*, which is just nothing about real performance, subsequent to what display card chipset/GPU makers have been doing this for years already, say, for the 3DMark.

Comments (23)

Loading... Logging you in...
  • Logged in as
Spotmatic's avatar

Spotmatic · 748 weeks ago

After having seen the K-5's photos and the leeway in the dark shadows... Should we care at all? After all, digital is a trick in itself, from front to back. Mathematical trick or not, I love what the K-5 delivers.
Rice, you don't quite get how it works, do you? It's not gonna show anything bad in DPreview tests (which are garbage BTW), because this trick just brings the shadows up. DPreview DR is going to be lower, because they don't check the total DR of the sensor (just an "usable range"), plus they measure it on JPG files which is totally useless and pathetic. Don't be so cheap - buy the K-5 and check for yourself what can be recovered from the deepest shadows... BTW, why didn't you link that 10EV push ISO experiment? Does it make K-5 look too good?
5 replies · active 748 weeks ago
thanks for mentioning that 10EV experiment, wouldn't even try to up a black shot like that.
I have no money to buy the K-5, which is just too expensive! >:-(
Michael A.'s avatar

Michael A. · 748 weeks ago

Sorry you don't have funds for the magnificent K-5, which is confusing, since you mentioned in past something like money is no problem for you. Perhaps sell some of your less interesting toys like the NEX and get a real tool K-5... :o) Here is the EV 9.5 recovery real live scenario jaad thinks you should comment on as well.. http://forums.dpreview.com/forums/read.asp?forum=...
Imagine now that K-5 in Russia will cost $2300

o_O
Rice, as always you only show what's convienent to your agenda. For example, you made sure to avoid mentioning Gordon B Good findings on this very matter where he actually proves the K5's DR is in fact REAL and not a trick as can be read here:
http://forums.dpreview.com/forums/read.asp?forum=...

And to quote his conclusion:
"In conclusion, no, the raw DR of the K-5 is not a trick or scam, and while it may have been achieved by some sort of innovative noise cancelling/reduction scheme inside the Sony sensor itself, as long as we can not see the effect in reduction of resolution/detail even at very low luminance levels, why would we care?"

Oleg himself was waiting for Gordon to peak in as he knows very well that he's THE authority when it comes to sensors analysis.
3 replies · active 748 weeks ago
What "agenda", boy?

My opinion in my this post is clear: I don't find any flaw in Oleg's arguments but neither do what he suspected have been proven. And afterall, real world exposure tests needed to be carried out to find out the DR, at different EV and lighting levels (which can affect the measured results, against what testers and reviewers have never mentioned!).
You're cherry picking the arguments you want to hear, instead of objectively following the arguments. Gordon pretty much destroyed Oleg's hypothesis, yet you choose to not write about this.

Also your statement above about the need for "real world exposure tests" before known the DR is a sign that you don't appreciate the scientific method and proper test results that do not suit your agenda of hating current Pentax products ;) (Regarding the scientific method, I do want to see more proper tests, and not "real life shots" where a wild number of unrepeatable and uncontrollable variables come into play and make objective analysis difficult or impossible.)

Also, it seems - judging from above - that you do not understand dynamic range, what it is. Unless we're talking about long exposures (multisecond), and the light levels associated with those shots, the light levels do not influence DR basicly at all. The size of the well remains the same regardless of the light level, and the read noise also remains basicly the same. If the exposure is long, noise levels do tend to go up somewhat, but as long as we're talking about regular hand held photography, that is totally irrelevant.

If you do not see flaws in Oleg's arguments, please read Gordon's arguments in the same thread (multiple posts).
RH: What "agenda", boy?

LMAO!
Take two steps back and look at your blog.
Huston... we have a problem!
RuffRidahHK's avatar

RuffRidahHK · 748 weeks ago

Posters on the DPReview website are telling each other that you are banned from that site. Why you bann on the forum?
1 reply · active 748 weeks ago
I don't know why, you should ask Phil Askey, not me.
That Pentax forum is nasty, full of insecure types who shoot messengers carrying bad news and bring the worst out of anyone who tries to get to the truth.
new most stupid blog entry ever
RH, you may not have the knowledge to find a flaw in Oleg's argument but GordonBGood has the brains, you seem to be lacking: At http://forums.dpreview.com/forums/readflat.asp?fo... he writes "Oleg, while perhaps something like what your describe could be done to fool some Dynamic Range (DR) scores, it is unlikely to fool DxOMark's method since they do not depend on an actual noise measurement at 0 db and in fact can't take one for cameras such as the Nikon D5000 where the black and near black noise levels have negative noise excursions clipped to the zero level. For the Nikon cameras it has been speculated that they may actually shift the zero point slightly positive and actually clip slightly too high. Pentax does not zero clip the black data for the K-x, K-r and K-5, so it can be directly measured from the black border photosites, but DxOMark definitely compensate for this normal black offset (Canon do this too)."

So all is good. But let me guess, you will not pull back your useless blog entry, will you?
GordonBGood further writes "In conclusion, no, the raw DR of the K-5 is not a trick or scam, and while it may have been achieved by some sort of innovative noise cancelling/reduction scheme inside the Sony sensor itself, as long as we can not see the effect in reduction of resolution/detail even at very low luminance levels, why would we care?" (http://forums.dpreview.com/forums/readflat.asp?forum=1036&message=36869348&changemode=1)

Get it now, RH?
rice you call everyone "boy" who has a different opinion than you, how should we call you mr sony?
1 reply · active 748 weeks ago
Not a "different opinion", but simply a personal attack. See it?
RH, last I checked Oleg himself is holding off to do testing as he is no longer convinced of the examples he used being valid, so why are you?
Wow just shows how biased you are against Pentax. GordonBGood overturned Oleg's arguments, but of course you ignore that.

Aren't you ashamed of yourself?
Toomas Kadarpik's avatar

Toomas Kadarpik · 748 weeks ago

Rice, save some money and buy the camera, the DR is there and it is real thing, even at iso 2500 the camera operates quite perfectly from noise and DR point of view, 60D,7D and other APS-C except D7K can not get even close to that. I have restored already dozens of shots with missed flash exposure and they are usable.

You compare K5 with 5D1 even, I think it is not feasible, this new sony sensor beats the 5D1 in every aspect, except FF lens performance and diffraction limits - but this all depends what you are trying to do.

Sorry but you are not in the right track.
The math is wrong, if any black offset S=S+S0, the DR should be Smax/Smin approx= Smax/S0, not Lmax/Lmin=inf. If the dxomark does the latter, it is dxomark's implementation bug (I don't think they do).

Post a new comment

Comments by