K-x Now Has DxO Mark

[miserere_mei]

<p>And it's frickin good! Wow. Below are the Top 4 APS-C cameras in the global DxO mark and <a href="http://www.dxomark.com/index.php/eng/DxOMark-Sensor">3 subcategories.</a> The numbers to the left of each camera indicate their ranking within the whole DxO database, which includes FF and MF digital cameras. You can see the K-x DxO report.<a href="http://www.dxomark.com/index.php/eng/Image-Quality-Database/Pentax/Kx">here.</a></p>

<p> </p><div></div>

[miserere_mei]

<p>You can see a larger version of the list by clicking on the link below.</p><div></div>

[ricardovaste]

<p>To be honest, I've never taken much notice of such sites, they just seem pretty pointless from a photographers point of view. But having heard others talk of it, it does sound like a very good sensor..</p>

[miserere_mei]

<p>Rich, photos are just a side effect of numbers ;-)</p>

[ricardovaste]

<p>Ah, numbers scare me! ;P</p>

[michael_elenko]

<p>Richard, while my artistic side strongly agrees with you, the DxO site has some clever visual aides that help graphically link the numbers to a picture. So if you view the low ISO evaluation graph, you can then manipulate a bar on the right side that shows an example of the impact of the numbers upon an image in the background. That makes it much more real to me.</p>

<p>Miserere, thanks for the update here. One question I have about DxO rankings that perhaps you know: do they state what differences are actually significant between ratings? IOW, for, say, Color Depth, is the 22.7 of the Nikon D5000 significantly lower than the 23.1 of the Samsung GX 20? Would anyone notice in a print or an online image?</p>

<p>ME</p>

[miserere_mei]

<p>M.E., that is a subject of much debate, and I can't give you an answer. I encourage some Googling :-)</p>

[jpo3136b]

<p>Mis, can you explain this to a K1000 operator? What does this mean? Good camera?</p>

[andrewg_ny]

<p>Good for taking pictures. Not as good for driving nails.</p>

<p>Any idea what those little orange exclamation points that are supposed to indicate some sort of additional testing notes on certain cameras mean? Following the links didn't seem to lead to anything specific.</p>

[laur1]

<blockquote>

<p>What does this mean? Good camera?</p>

</blockquote>

<p>It means good sensor. Best used in a Pentax DSLR so far. Interestingly, the previous top global score belonged to the K10D, not to the K20D or K-7.</p>

[miserere_mei]

<p>What Andrew said :-)</p>

<p>Laurentiu, I was quite surprised that the K10D features so high up in many of these standings, especially considering it is a 2006 camera, which is ancient in the digital world. OK, maybe not ancient, but certainly "old" :-D</p>

[mountainvisions]

<p>Looking at the graphs the K-7 and K10D have almost exactly the same SNR18% values through 800 while the K-x is about 2.3db better at most . At 3200 it pulls away from both the K-7.</p>

<p>Putting this into perspective, we are talking about .4 stop difference in noise at 1600 vs the K10D but closer to 3/4 stop vs the K-7. Definitely starts getting significant with the K-7, but the K-x vs. K10D in noise alone is pretty even.</p>

<p>This remains about the same on both print and screen test. Which really is impressive for the K10D being 4 years post release in Fall 2010.</p>

<p>However, while the K-x isn't vastly better in any area, it's cummulatively better in all areas, even if only marginally in most.</p>

<p>Looking at bit depth through the ISO range the K-x takes a slight lead (about .20 bits). But actually the K10D holds it's own and nearly beats young K-x at 800 and 1600. I wonder if this had anything to do with the now non existent 22/16bit DA converter that seems to have disappeared.</p>

<p>Where the K-x really excels without margin is in dynamic range. It is almost 1 stop better than even the K10D at all ISOs except 400 where the K10D drops off (oddly, but then gets better at 800 and 1600).</p>

<p>Actually this was very informative. I am now going to stop using my K10D at 400ISO. It for whatever reason reaches a dead zone at that ISO in all areas of testing. It then becomes competitive again at 800 and 1600. Well, actually I am not going to immediately do this, I want to see what my 400 ISO images look like that I've already shot, and see if I can find a visible drop off. But this gives me something to look at and consider.</p>

<p><B>Anyone care to explain this?</B></p>

<p>What is most interesting is the K10D is actually a better sensor (accounting for age difference) than the K-7 and K20D, unfortunately a lot of the K-7 features are missing from the K10D, and also from the K20D.</p>

<p>However, I have to say, certain things don't show up in this. For instance, the K20D had shadow noise with the DR expansion, while the K-7 eliminated that. The K10D had banding while the K20D eliminated that. The K20D had moire issues (and other inconsistent issues) while the K-7 seemed to eliminate those.</p>

<p>So for those asking, what does this mean? It means objectively the K10D (still) holds it's own, it means objectively the K-x is the best Pentax sensor to date, but in real world use I found the K20D (and K-7) to be better than the test would indicate at lower ISO.</p>

<p>What it also tells me is Samsung sensors are not quite up to par yet, and that the leap from CCD to CMOS isn't quite as amazing as we were led to believe.</p>

[mountainvisions]

<p>ME,</p>

<p>What is interesting about the D5000 is it falls well below it's stated ISO at all ISO, and increasingly so at higher ISO. Whereas Pentax is only off by about 1/3 stop through most of the range.</p>

<p>For instance at 1600 it's ISO only measures 1000.</p>

<p>Assuming this is accurate, then the D5000 is fudging it's ISO by 2/3 of a stop, which would make it look much better at higher ISO (I assume it is sending less voltage to the sensor, keeping signal to noise down).</p>

<p>I'm not sure if PopPhoto list metering accuracy as a objective factor, but if not Nikon has found a creatively sneaky way to make it's cameras look better at higher ISO.</p>

<p>I'm impressed!</p>

[miserere_mei]

<p>Nice post, Justin. But please don't stop using your K10D at ISO 400 just yet :-)</p>

<p>As time has gone by (almost 3 years now) I've become more and more impressed with the K10D, both in terms of build/features, and IQ. The one thing that bothers me, and which will make me upgrade to a new body within the next year, is the lack of ISO 3200. The K2000 used the same sensor as the K10D and did offer 3200, and that's always made me wonder whether the K10D not having it was a hardware or firmware issue.</p>

<p>Pentax did the right thing, of course, because the lack of ISO 3200 means I'll be giving them money for a new body, but that money would have been spent on glass otherwise, so they would probably have earned it anyway. And it's not like the ISO 3200 on the K2000 was useless; I've seen the test images on IR and DPR, and it's as usable as ISO 1600 on the K10D, i.e., expose well and you're golden. Interestingly enough, the K200D (launched 8 months before the K2000) also used the Sony 10MP sensor but didn't have ISO 3200 either.</p>

<p>Justin, I don't think the benefits of CMOS are that great for the user (beyond those of CCD), but as far as I know CMOS makes the manufacturer's life easier in many ways. But a great CCD is better than an average CMOS, and I believe that's what Samsung manufactured: An average CMOS sensor; while Sony have built great CCDs for Pentax. I would have liked to see the K100D scores on DxO. While the handling and speed of the K100D is inferior to the K10D, its IQ still impresses me, and to this day I often use it instead of the K10D when I know I'll be shooting low light.</p>

<p>I'm hoping all new APS-C (and FF?) Pentax DSLRs from now on use Sony sensors. For whatever reason, Pentax really knows how to squeeze the best performance out of them.</p>

[miserere_mei]

<blockquote>

<p>What is interesting about the D5000 is it falls well below it's stated ISO at all ISO, and increasingly so at higher ISO. Whereas Pentax is only off by about 1/3 stop through most of the range.</p>

</blockquote>

<p>To paraphrase Iñigo Montoya... Justin, I don't think this means what you think it means. If I correctly remember how DxO works (don't have time to check it now) it analyses the data coming out of the sensors before it goes through the camera pipeline. So, if the D5000 says it's shooting at a certain ISO, but it's actually shooting 2/3 stops below, then the camera is pushing the images by 2/3 stop to make up the difference.</p>

<p>In short: The D5000 purposely underexposes by 2/3 stop at high ISOs, then increases exposure through internal postprocessing by 2/3 stop to achieve the required exposure. If pics look good, it must have low shadow noise to be able to pull this off.</p>

[jgredline]

<p>In Honor of St Patricks day and the very Good K10D, Here is an old thread and <a href="http://www.pentaxforums.com/forums/post-your-photos/54607-retiring-my-k10d.html">here is my tribute to it. </a><br>

<br /> This thread has pretty much convinced me to have it repaired....</p>

<p>On a second note, much in the same way that the K-X is looked at as the king of noise control over its big brother the K-7, The D90 is looked at in the same way over its big brother the D300S 8-)</p>

[miserere_mei]

<p>Javier, I remember that thread. Or rather, I remember those photographs. Not all of them, just some ;-)</p>

<p>According to <a href="http://www.pentaxforums.com/forums/956896-post3.html">this guy</a> the repair at CRIS should cost you $260.</p>

[mountainvisions]

<blockquote>

<p>If I correctly remember how DxO works (don't have time to check it now) it analyses the data coming out of the sensors before it goes through the camera pipeline.</p>

</blockquote>

<p>This cannot be true. Here is why, if it's purely sensor data than the D200 and K10D would have exactly the same #s. As would the GX20 and K20D. As would every camera that shared a sensor. But as we've seen, all the cameras are not equal. And if that was true, then why measure the camera models at all? Why not just list the sensors, and then put a list of corresponding cameras that use that sensor. Marketing maybe, but I don't think so.</p>

<p>DXO has to be analyzing the data as it comes through the image processor. There is no other explanation as to why the K10D outscores all the other Sony 10MP sensor based cameras if only the sensor is tested. It has to have something to do with the 22 bit D/A converter, while Nikon obviously falls flat because of the 4 channel converter.</p>

<p>The other thing is why doesn't DXO just normalize the image data, that is if the camera states it's working at 6400, extrapolate (or force the camera during test to actually output) to go to 6400. Instead they list the actual ISO that it was tested at.</p>

<p>Regardless, the image processor has to be tested for the results to differ as much as they do, these aren't purely sensor test.</p>

[mountainvisions]

<p>BTW, I agree with the your post above.</p>

<p>Basically what you address is my complaint that firmware isn't used to prolong the life of the camera.</p>

<p>These cameras can have as little as a 12-18 month life span based on features, obviously many of us have decided features alone won't get our money. But for me things like auto bracket, live view MLU (true mirror lockup with multishot mirror up), IR based bulb and so on really make the K-7 super appealing. Some (if not all) of this stuff could be added with firmware, but unfortunately Pentax needs to sell cameras to stay relevant, firmware doesn't get the front page of magazines. </p>

<p>I really don't need the K-7 for my uses for IQ (obviously, since it is at best an even push with the K10D), but a lot of the features, the size, and the build are all what I was asking for. Puts me in an interesting spot. The K20D was a good call on everyones part that passed up on it, I know Javier noted it felt like a finished K10D to him, but to me it was like a K20D beta, or K10D Semi Super. Just too many unfinished features, and the sensor while arguably better because it offered about the same IQ with more pixels was not really leaps better overall. </p>

<p>The K-7 on the other hand really was a nice step up, but we are still getting overall (although again, I really feel in real world use the IQ is better than the DXO #s would indicate at ISO 100-400) September 2006 IQ. That isn't to say this is bad, before the D90 the K10D was the only sub $1000 or even sub $1500 camera in the top 10.</p>

<p>But like you said, an average CMOS is not better than a really good, and really well integrated (into the image processor) CCD. I wonder if we blame Pentax for not being able to get the most out of the Samsung sensor, or Samsung just isn't that good and Pentax is getting the most they can out of it.</p>

<p>It's really a shame Pentax scrapped the K10D so quickly. The K10D will forever be the PZ-1P of digital, the best camera no one really appreciated. Obviously it made a splash, but Pentax seemed like they wanted to mothball the K10D as quickly as possible, even putting out an inferior K200D with a higher price to feature ratio to replace it.</p>

[miserere_mei]

<blockquote>

<p>DXO has to be analyzing the data as it comes through the image processor.</p>

</blockquote>

<p>Maybe the truth is somewhere in between. It's the D5000 ISO numbers that have me thinking. If the camera says ISO 1600 but DxO measure 1000, does the resulting image look like the K-x at ISO 1600, or at ISO 1000? It it's the former, then Nikon are pushing the image by 2/3 stop and DxO are measuring data before this push occurs; if it's the latter, then the D5000 image will be darker than the K-x image at a given ISO, and DxO are measuring the data coming out at the end of the pipeline.</p>

<p>Bear in mind that how data is read out and interpreted <em>before</em> it goes into the imaging pipeline is brand dependant. Because of this DxO could be looking at the data before it hits the imaging pipeline and see different results from two different camera brands using the same chip. However, I have not had the pleasure of spending countless hours reading up on RAW data manipulation, so I may very well be wrong. If anybody here knows for sure, I would be happy to listen (or read). Some enlightening links would also be nice :-)</p>

<p>Maybe tomorrow I'll read through the DxO tech files to figure this out. I sure know how to have fun... :-s</p>

[deaner66]

<p>Here, Here, Justin.</p>

<p>Long live the K10D.</p>

[mountainvisions]

<p>I guess the flip side of what I originally said would be if it is the way your are explaining it, and seeing as how I haven't done the legwork either to figure it out, Nikon has figured out a way to do something that generally doesn't work all that well.</p>

<p>They are essentially using DR expansion all the way through the ISO range and/or pushing the image.</p>

<p>From my experience with the ist D, it was better to shoot at 1600 than shoot at 800 and push it a stop in post processing. I did this over hundreds if not thousands of shots before I came to the conclusion that pushing wasn't effective. The camera was much better at processing at the correct ISO.</p>

<p>If Nikon solved this problem that is a hell of a feat, looking forward to the real explanation.</p>

[jemal.yarbrough]

<p>I guess I'll spend my extra money on lenses and flashes and continue to use and love my K10D (although at the moment its been replaced by the Sony A350 cause I love the Minolta 50mm f1.4 I just purchased, that and it, unlike the K10D had to go in for repair). Long live the K10D.</p>