[Pentaxor]

Here's to hoping they will implement the Kx sensor in the K7 body soon.

I think I'll probably pass on that one. I'd rather have a full-frame on the next one. too many APS-C camera already and for the sake of only one factor alone which is High-ISO performance, is not worth the upgrade and extra moolah ! I can live without having to shoot at ISO 6400.

[JohnBee]

It's too bad that available samples don't support such claims:

K20D@1600 on imaging-resource
K-7@1600 on imaging resource

I don't see any significant difference in those samples, which isn't that surprising given that these cameras share pretty much the same sensor.

I fully agree and have from the very beginning as I've never had an issue with ISO1600 performances with the K7(in fact I think it's very good), though I've found noise to creep in(see; shadow recovery banding, tinging etc) sooner than on the K20D.

I think most K20D owners who use their camera's for indoor shooting would agree that ISO1600 is often the baseline for low light shooting and thus the reason I mentioned the K20D for higher ISO shooting.

What's even more interesting is that in many of the graphs and tests, the K7 seems to do favorably in the chroma, gray and black channels over that of the K20D. However in RAW sampling, the K20D scores higher than the K7 on all counts.

Here are those scores(respectively):
K-7
12
10.2
8.2

K20D
9.2
7.1
6.9

And the samples:

K-7 ISO3200 NR off


K20D ISO3200 NR off


So though(technically) the K-7 could outperform the K20D where onboard NR in concerned, if you shoot RAW(which we do exclusively) then the K20D outperforms the K-7 in noise performance. In some of our own tests, we've concluded that the K7 began to loose ground against the K20D at/or above ISO1600. At which point the noise performances continued to separate until they were no longer favorable.

And though I will admit were splitting hairs with such differences, the issue came-up(in our case) where we commonly shot between 3200 and 4000(with very good success) whereas we could not do with the K-7.

The oddest thing in all of this is that both camera's apparently share the same sensor.
However... in some strange way, whatever Pentax did with processing(faster FPS?) turned out to have a negative impact on noise performances in retrospect to the K20D.

So... though it may be a worthy trade-off for some people.
In cases such as ours, it was not.

I even emailed Pentax about our findings when we conducted our tests(complete with images) and we were told that this was most likely due to early firmware and that it should be addressed in later revisions. However it doesn't look like it happened.

[BillM]

Would'nt be neat if you could have a camera built to order like a car?
K7's magnesium chasis, stuffed with K20's guts,power options like K200D

[Mohawk]

I was in the market to replace my K10D, and went to the shop with my two workhorses, the DA* 16-50 and 50-135, and an AF540FGZ flash. The K7 just wasn't comfortable in my hands in any combination. So I went with the K20D, and bought the battery grip, which was icing on the cake. Two months of use now and no regrets whatsover.

PS. The battery grip only comes off when I want to go light/play tourist - i.e. with DA 18-250 +/- AF360FGZ.

PPS. Tried out a friend's K7 with small primes and it handled sweeeet. So horses for courses.

[mel]

Jumping in late here and I'm not a guy but I hope I can contribute anyway. I haven't read through all the responses so forgive me if I'm being redundant in any way.

I am currently using both of these cameras. I shot them side by side the other night (a play), one with the DA*16-50 (K-7) and the other with the DA*50-135. I would have to say the one that you would want would depend on your needs. The K-7 is much more responsive than the K20d. I was shooting live action so I couldn't stop to set things up and get it all perfect before firing. The AF on the K-7 would lock much more quickly than the K20D. The shutter being much more quiet than the K20D is also an advantage here. I was shooting under theatre lights, both were set to tungsten for white balance, and the colors came out much more accurate on the k-7 than they did the K20D.

As for image quality, I have been vascillating back and forth. At first I thought the K20D edged out the K-7 but then it would seem to switch. Yes there is noise in both but I think the noise on the K-7 is a bit more pleasing than on the K20D, which seems, I don't know, choppier I guess? Both have outstanding IQ on the best images from the shoot so what it has come down to is the sucky images. I think the sucky ones from the K-7 are just slightly less sucky than the sucky ones from the K20D.

That all being said, they are both outstanding cameras and you should be pleased with either. Don't think about it too much and just buy one

[Class A]

The K20D has long been known(by those who cared) to have one stop over the K7 in terms of noise.

There are others who care who don't arrive at the same "one stop" difference.

Our own falconeye did a very thorough test and found the noise difference to be negligible.

Have you read any of GordonBGood's thread at DPR? IIRC, he also only came to the conclusion that the K20D is slightly better only marginally after you consider the in-camera noise reduction that is higher in the K20D compared to the K-7.

As I said, looking at RAW files is not looking at the sensor since Pentax applies NR to RAW files and not all models receive the same NR. NR also varies with the ISO setting and the K-7 is known to use less NR then the K20D at the higher settings.

Also, you'd really need to look at the spatial resolution plotted against ISO as well. What if the K20D achieves better out-of camera noise values but at the expense of resolution? Did you try to use extra NR on K-7 files to see if you could bring them to the level of the K20D?

BTW, according to GordonBGood's finding the K-x's oh-so-fantastic high ISO performance is only due to some part because of better sensor performance. The other part comes from clever noise reduction which is so clever that it is even hard to show that is being done.

In summary, I think the whole issue is a lot more complicated than just stating that the K-7 has a sensor that is one stop noisier than the K20D.

[Phaser]

lool at this k7 AlunFoto - PESO/Snowproofed

[JohnBee]

There are others who care who don't arrive at the same "one stop" difference.

Our own falconeye did a very thorough test and found the noise difference to be negligible.

That test(if we can call it that) is hardly conclusive.
There are so many issues with his method that it makes one wonder what he ever expected to obtain from this other than more questions.
Personally I'd like to see it done right(same lens, same file format and most of all with a real world subject(color)).

Have you read any of GordonBGood's thread at DPR? IIRC, he also only came to the conclusion that the K20D is slightly better only marginally after you consider the in-camera noise reduction that is higher in the K20D compared to the K-7.

Yes, I've followed GB's work and make much use of his utility(which is a godsent btw).
Unfortunately... since noise is exponential in nature, terms such as "slightly better" and "marginal" have no place in performance evaluations.
ie. slightly better becomes substantially worst under Ev adjustments and so on and so forth.

As I said, looking at RAW files is not looking at the sensor since Pentax applies NR to RAW files and not all models receive the same NR. NR also varies with the ISO setting and the K-7 is known to use less NR then the K20D at the higher settings.

You could go there.
Or... your could look at the real world data and accept it for what it is.
In our case, practicality trumped theory as one unit(s) clearly outperformed the other, no matter what.
So, I simply can't agree with such a conclusion. Having said that... I'm not doubting NR really is the issue here either.
Therefore if Pentax applied some fancy processing to even the score between both units, I for one wouldn't be bothered by this at all.

Also, you'd really need to look at the spatial resolution plotted against ISO as well. What if the K20D achieves better out-of camera noise values but at the expense of resolution? Did you try to use extra NR on K-7 files to see if you could bring them to the level of the K20D?

Yes, absolutely. In fact, we spent an entire week shooting and applying various NR methods in our own attempts to match the K7 to the K20.
Having said that, no matter what we did, we were never able to match the K7 on any of the K20D files at higher sensitivities.
And though the K7 did produce sharper files with higher curves by default, the same could be done with the K20 in PP also.

Additionally... adding NR(which was also suggested by Pentax) on the K7 did not resolve the situation as we would have hoped.
The main reason was that the added NR came at the expense of detail loss in the final output(moreso than that of the K20 files).
So the short answer was no... adding onboard did not level the score between the two units.

BTW, according to GordonBGood's finding the K-x's oh-so-fantastic high ISO performance is only due to some part because of better sensor performance. The other part comes from clever noise reduction which is so clever that it is even hard to show that is being done.

Great!
All we need now is for Pentax to update the K-7 with the Kx's clever NR processing and were golden. Though something tells that's just not going to happen

Without disrespecting others comments and observations, I'm going to go with the notion that the Kx performance is just as reliant on the sensor as it is on the clever NR algorithms that accompanies, it.
IOW. if it were simply an processing issue, don't you think others would follow suit? And more importantly, wouldn't we be seeing this on the soft side of things?

Having said "all this"
I am a firm believer in a practice trumps noise solutions too(pun intended).
And so I would be more than happy to take images with you on a color chart with both unprocessed and processed output(from RAW) in hopes of achieving something tangible in the process.

Because who knows...
I might end-up eating my words and picking-up another K7
But as it stands, for me... the K7 is inferior to the K20D where noise in concerned.
And I for one, would love to be wrong about that.

[mel]

So is noise the only issue at hand? I think the K-7 outperforms the K20D in so many other areas it's a trade off. And the noise differences aren't enormous really.

To me, shooting the two side by side was like driving a full size SUV down a winding highway versus driving the same in a sports car.

[creampuff]

...we get full size prints up to ISO4000 with our K20D's which is something the K7 doesn't do either with or without NR and PP(trust me, we tried).

...We own several K20 and then saw and shot the entry level Kx only to find it ran circles around them.

In some of our own tests, we've concluded that the K7 began to loose ground against the K20D at/or above ISO1600.

In fact, we spent an entire week shooting and applying various NR methods in our own attempts to match the K7 to the K20.
Having said that, no matter what we did, we were never able to match the K7 on any of the K20D files at higher sensitivities.

Well you've merely regurgitated and included dpreview's images and results, but I'm wondering if you can share some of the comparative images of your own considering you or others with you, have conducted tests to support your assertion. With your extensive tests, I'm sure you got a couple of high ISO images lying around that you can post with the exif intact to enlighten all of us. Also I find it strange that you use "we" rather than the first person "I" in your posts... are you a company or a testing facility rather than an individual?

[JohnBee]

So is noise the only issue at hand? I think the K-7 outperforms the K20D in so many other areas it's a trade off. And the noise differences aren't enormous really.

To me, shooting the two side by side was like driving a full size SUV down a winding highway versus driving the same in a sports car.

I totally agree with you on that.
I actually loved the K7 for a number of reasons(including the ergonomics).
Unfortunately, since my wife and I had an investment in low light(indoor) shooting, we found ourselves under a very specific set of conditions. I think this played a heavy hand in our tests and conclusions as the effects of noise seemed far more relevant than most people seemed to care about. Perhaps our experience was a little out of bounds in terms of specifics, but... I also think we uncovered an interesting discrepancy between the two units in the process also.

[Pentaxor]

I believe it comes down to whether the real world ISO 3200 image results would be enough to keep the owner happy. if the results are acceptable or looks nice, I don't think the marginal difference would matter.

just employ the ETTR/ETTL method (with or without NR) or use post-processing NR afterwards, and you'd be happy.

since the ISO past 3200 is an extended software-based ISO, I believe it is possible for Pentax to update it thru firmware, whether they would apply a stronger NR, but retain details as well.

[JohnBee]

Well you've merely regurgitated and included dpreview's images and results, but I'm wondering if you can share some of the comparative images of your own considering you or others with you, have conducted tests to support your assertion. With your extensive tests, I'm sure you got a couple of high ISO images lying around that you can post with the exif intact to enlighten all of us. Also I find it strange that you use "we" rather than the first person "I" in your posts... are you a company or a testing facility rather than an individual?

First off, dismissing valid references as regurgitation does nothing for the credibility of your intentions in this discussion. Secondly the tone you set in your comment completely undermines all possibilities of keeping this a beneficial conversation. If I wanted to argue, I would have posted a thread on dpreview and invited someone like Rice High to play along(think about it).

Having said that, I don't have any K7 samples from our tests off hand.
Though I may be able to dig something up I honestly don't know if we even kept them and it's been nearly a year since we conducted our tests let alone referenced anything from them since.

For what it's worth, here are two ISO4000 samples from our worst K20D. I say worst because one camera has more noise and banding than the other and this these were taken with that particular unit. So I for what it's worth, think we could get a little better in terms of detail or final IQ.





FTR. both images were taken in very low light and PP for print(naturally).

[Javaslinger]

Those are terrific images for ISO4000... It would be nice to see contrasting images from K7...

[Laurentiu Cristofor]

I am a bit suspicious of the dpreview shots because they seem to have either positioned cameras at different distances or they mixed up their test shots. The K-7 and K20D crops should cover the same area, but they don't.

I fully agree and have from the very beginning as I've never had an issue with ISO1600 performances with the K7(in fact I think it's very good), though I've found noise to creep in(see; shadow recovery banding, tinging etc) sooner than on the K20D.

But then, how can the K20D have an advantage of one stop as you claimed? Given that performance is similar at 1600 (without splitting hairs), the K20D cannot be one stop better at 3200 (which would mean it's still as good as it was at 1600). And 3200 is the top ISO available by default on both cameras. Extended range is disabled by default for a reason.

And though I will admit were splitting hairs with such differences, the issue came-up(in our case) where we commonly shot between 3200 and 4000(with very good success) whereas we could not do with the K-7.

In this case, calling an alleged superiority of the K20D in the 3200-4000 ISO range as a "one stop advantage" is misleading. A one stop advantage without any additional qualifications would imply that a camera is as good at 200 as another at 100, and so on. It's particularly important to be clear in this case, because most people rarely shoot at 3200 and above, which makes this alleged advantage irrelevant for most, even if it were true.