Pentax K-3 wins over the expensive boys with the richest color rendition

[beholder3]

Hi,

for many photographers the level of color details you can capture with your camera are quite some topic for them. This is often referred to as "tonality" or "color depth".

How many nuances of blue does your picture contain? If many, the gradients will be smooth and nice.

Other than rely on theory created by third parties I decided to have a look myself and propose you all do. Only trust what you can personally verify.

The logic here is:

1. It all depends on the raw data. If it's not in there, it's faked or just a matter of postprocessing.
2. All color is created in demosaicing the bayer array in the raw converter so the foundation are the 4 color subpixels red, green1, green2 and blue.
3. As color values are coded in raw as linear values between 0 and 16384 (2^14 for a 14bit raw file) the question to answer is: how many different values for a RGGB color subpixel does the sensor differentiate? The more the better. The more, the more nuances of color. The more, the more nuances of color after debayering.
   If it's not there captured in the RGGB channel in the raw right at the start you can only fake it in there in postprocessing.

What we need is raw files of the same scene. Here I took the files from the new studio test scene on DPR all at ISO 100, daylight simulation. I guess most here will accept DPR's skill in setting up an acceptable test scene under controlled conditions.

Loading…

www.dpreview.com

Loading…

www.dpreview.com

Loading…

www.dpreview.com

Loading…

www.dpreview.com

Loading…

www.dpreview.com

Loading…

www.dpreview.com

Then we need a good tool for raw file analysis. Here RawDigger comes into play, as it nicely counts the pixels for each RGB layer.

RawDigger

Everything You Always Wanted to Know About Raw(But Were Afraid to Ask)

www.rawdigger.com

In its histogram window you can save the full raw data for each RGGB color channel and as extra nicety it all by itself counts the number of differentiated values.

You just need to read and compare.

Here is the result:


Pentax K-3 offers more color depth than many expensive competitors

Yes, the little K-3 beats them all hands down in this controlled test environment. All the Nikon D810 and Nikon D7100 are no match. The Canons 5D3 and 1DX are far behind and the biggest loser is the Nikon D4s with the crudest color depth (more than +25% step up needed to reach the Pentax). It seems, a lot of expensive imagined fairy dust and esoteric wishful thinking does not hold up to the test. ;-)

The Pentax engineers really seem to have done a good job there.

Feel free to prove it different or add more data. Or enjoy your great little Pentax camera.

 

[altis]

I'm sorry but that really is a very poor measure of ADC accuracy.

Is it monotonic? Is it linear? How much noise is there?

 

[Tom Lusk]

Wow!

Once all the top photographers in the world see this, they will be dumping their 1D and D4 cameras.

They can then use the funds realized to buy a whole whack of K-3s and still have money left for lenses.

 

[OPTMEKX7]

+1, seems Pentax engineers dialed the color rendition better than the big boys.

cheers,

Rene

 

[MightyMike]

Something doesn't seem right.. how about adding a couple more Pentax cameras to the mix to see if this conclusion holds true across different models... also throw in a few more from the other brands

EDIT ADD: the thing is these values could be from more impressive CFA technology or simply a better cooking job of the RAW data

--
Mike from Canada
"I am not a great photographer! God is a great creator! All I do is capture His creation with the tools He has provided me."
'I like to think so far outside the box that it would require a telephoto lens just to see the box!' ~ 'My Quote '
http://www.michaelfastphotography.com/galleries/VP-BDI_3a.jpg
http://www.airliners.net/search/[email protected]&thumbnails=

 

[Antal I Kozma]

Hi,

for many photographers the level of color details you can capture with your camera are quite some topic for them. This is often referred to as "tonality" or "color depth".

How many nuances of blue does your picture contain? If many, the gradients will be smooth and nice.

..................

Hi,

Thanks for the info. I have always had high opinion of the K-3.

That said, I am so much satisfied with the D810 that I could not ask for anything more. I have just retired form being behind the camera for some 45 years. During these years I used everything from 35 mm to 8x10 sheet film and anything in between.

It was the Nikon D3 that first made me feel that finally digital is "the tool". Since then I kept hopping up the street of upgrading as Nikon introduced new cameras. When we got to the D800 I thought that "this is it", this is my last camera. That lasted till the D810 came out with a few features that I was missing from the D800.

Now I can say for sure that this camera is as good as I shall ever need. Plenty of detail for prints up to 36x48 images, very nice colour rendition, reasonably fast shooting speed with good buffer. I am not saying that something better will not come. All I am saying is that as a retired professional I do not need anything better. I shall be "upgrading" when this D810 will end its life cycle.

Back to the K-3, it is a very fine instrument and I keep an eye on Pentax gear. I am emotionally attached to the brand name, I started out on Spotmatics. I am looking forward to seeing how the FF Pentax will turn out.

Best to all, AIK

 

[Gary Martin]

Now I can say for sure that this camera is as good as I shall ever need.

Famous last words

 

[Antal I Kozma]

Now I can say for sure that this camera is as good as I shall ever need.

Famous last words

 

[beholder3]

Here with some more models:


more models compared with their color nuances

I added the Phase One IQ180 (using 16 bit raw) and the Olympus E-M1 (only using 12 bit raw) to show where the upper and lower edges of the market are.

The Samsung is in there for the higher pixel count and different sensor tech.

 

[Holger Bargen]

Thank you, beholder3, for your work and the nice analysis. At least it is nothing assessed from an impression which was transferred to a pseudo-measurement. It's a true measurement with an exact description of the method and the material used as well as numeric values as results from the measurements.

The result is impressive - but do we know the reason for these differences?

A camera is not just a sensor and the processing of the data from this sensor.

You can't use a camera without a lens.

Was the same lens used for all photos that were analysed - it would have been possible - I think that the Sigma 70 mm macro lens should be available for all camera systems.

But if different lenses were used this could be another reason for the results we found in this analysis.

Many of us here in the forum use Pentax not just because of great camera bodys but because Pentax lenses are very special in a positive way. There is a certain colour rendering in Pentax lenses - we all know it, we all like it - but we never were able to tell in detail what is so special about these lenses - maybe the results of this little study could give us the answer - maybe the picture Pentax lenses display on the sensor is much richer in colour than lenses of any other brand!

We all know the lens tests at DxO - maybe they should add another item they should add to their analysis - maybe simply in the way it was done here - and they would find that their ranking so far was much to simple and that they were not able to detect the true winners.

If there were different lenses used for the analysis in this thread I strongly believe that the main reason for the different results is not the camera body but the lens.

Best regards

Holger

 

[Holger Bargen]

Thank you, beholder for your work and the nice analysis.

I would like to see an analysis of a Sigma DSLR with the foveon sensor - many people say that it is a very special sensor with nice and very natural colour rendering. So I am curious if we could find this impression with the results in the analysis you did.

It would also be interesting if the old CCD sensors - let's say of K10D would do as good or even better than the more actual models. There was much improvment of sensor technology regarding high ISO performance (and it is improtant for me and I like it a lot) - but I still have my K10D and I know that if it comes to colours it would be a camera body I may should have in my bag if the advantages of modern sensors do not count for the work.

Best regards

Holger

 

[beholder3]

I would like to see an analysis of a Sigma DSLR with the foveon sensor - many people say that it is a very special sensor with nice and very natural colour rendering. So I am curious if we could find this impression with the results in the analysis you did.

It would also be interesting if the old CCD sensors - let's say of K10D would do as good or even better than the more actual models. There was much improvment of sensor technology regarding high ISO performance (and it is improtant for me and I like it a lot) - but I still have my K10D and I know that if it comes to colours it would be a camera body I may should have in my bag if the advantages of modern sensors do not count for the work.

I agree this would be interesting but the scope of my work is limited to the given set of really comparable raw images. Since I do not own a camera shop inventory I have to rely on the new studio test scene raw images here for now. Sadly no old models and no foveons available there.

 

[beholder3]

I now added Sony RX100III, A7s, A7R, Fuji X-T1 and Panasonic GH4 as well.


added 5 more cameras

Having seen Sony's raw data I have to say I am seriously underwhelmed. That really doesn't look promising. Very reduced color data with little difference between the point and shoot and multi thousand dollar models. If I were shooting color critical subjects Sony was dead for me.

Panasonic also seems to be doing weird things here. While they do go beyond 12 bit their linear encoding ends at the value 5320. I would still call that "12bit with nominal extra".

Due to the Fuji's non Bayer pattern I created the virtual "total" value by adding the (highest) green value (12199) a second time to make it somehow comparable. Still you can see the nuances captured are on the lower end of what the other 14 bit APSCs capture. Suprising given all the tales of Fuji colors. Seems the raw files dont really hold up to a K-3 here.

Here is the maximum number of values a certain number of bit depth can hold:


bits as numbers

 

[Rosember]

Thanks for the very interesting analysis, beholder, appreciated. What is obvious for me from the data is the red "weakness" of most of the sensors. The NX1's 28 MP sensor looks a little bit better in this respect than the Sony sensors: although it does not top the K-3 in the overall value the color resolution seems to be distributed a tad more evenly over the spectrum. Samsung seems to have done their job well with this sensor.

 

[MightyMike]

How about providing the numbers in percentage form as an efficiency level giving the bits available provided in their RAW format.

Obviously if the value is low but still over the next step that may have been a reason to go up to the next bit level. so a system that offers 14 bit RAW might have a value of 4500, that would be low efficiency but the next step down in industry standards is 12 bit as they skip 13 bit and had the system been 12 bit it would have been throwing out values as it would top off at 4096...

The funny bit is the average human can't even distinguish between 3 adjacent shades in 8 bits per channel colour and most(all) display devices can't actually display all those shades (12, 14, 16 bits) so if viewing the photo digitally all those extra bits are wasted even if the monitor says it can do 32 bit colour. Finally no printer can print that many different shades accurately or at all so if printing you aren't getting all those values. So we can squabble over tonality all we want but at the end of the day its a false value as the only place you can find the difference is in the file data, its not seen, nor printable, nor displayable graphically.

 

[Chris Mak]

I now added Sony RX100III, A7s, A7R, Fuji X-T1 and Panasonic GH4 as well.


added 5 more cameras

Having seen Sony's raw data I have to say I am seriously underwhelmed. That really doesn't look promising. Very reduced color data with little difference between the point and shoot and multi thousand dollar models. If I were shooting color critical subjects Sony was dead for me.

Panasonic also seems to be doing weird things here. While they do go beyond 12 bit their linear encoding ends at the value 5320. I would still call that "12bit with nominal extra".

Due to the Fuji's non Bayer pattern I created the virtual "total" value by adding the (highest) green value (12199) a second time to make it somehow comparable. Still you can see the nuances captured are on the lower end of what the other 14 bit APSCs capture. Suprising given all the tales of Fuji colors. Seems the raw files dont really hold up to a K-3 here.

Here is the maximum number of values a certain number of bit depth can hold:


bits as numbers

Interesting! The values for the Sony A7 cameras are food for thought though: both the A7r and the A7s received quite a bit of praise regarding their color richness. Yet their values in your test put them on compact camera level quality.
So what does this say? Are these numbers largely irrelevant in real world use, meaning that other qualities are crucial in taking good images? Or are the hardware devices we currently have (monitors, printers) obscuring the benefits of these (when higher) values? Or a third option: is the test flawed?
I have both the Sony A7r and the Pentax K3, and view my images (from raw) on a calibrated Eizo monitor, and print to an Epson SC P-600 printer. Where the colors in my Pentax K3 images should downright outclass the colors in my Sony A7r images when taken with the same Zeiss lenses (according to the values in your test), the opposite is true: the Sony images benefit from a significantly wider DR (in a single unedited exposure), have richer colors and more color detail.
Any ideas on how to put this in line with these test values? I have good eyes (or like to think so) so I'm intruiged with the relevance of this test.

Also, how does this correlate with the DxO values for "color depth", where the Sony A7r makes it to a value of 25,6 while the Pentax K3 trails at 23,7? Somehow the DxO values are more in line with real life use.

Chris

 

[Toccata47]

But really, this is entirely academic, isn't it? Or let's look at this a different way. How many professional landscape photographers shoot pentax? How many studio photographers shoot pentax?

I'll wait...

How many of the same use 1dx or d4?

Color depth and accuracy is great, particularly if you don't post process. I haven't seen time in pp decrease at all from the d3 to the d800e generation though depth and accuracy has improved.

With that said, the native pentax color palette (which does vary by sensor generation) is the reason I bought a K5 and later a K01 (which is actually my favorite pentax camera to date!).

I've been waiting for ricoh/pentax to throw the knock out punch for years now...I hope the FF is it.

 

[Holger Bargen]

But really, this is entirely academic, isn't it? Or let's look at this a different way. How many professional landscape photographers shoot pentax? How many studio photographers shoot pentax?

But this is no reason why Pentax could be second best. Why do people make their decision for or against certain brands? Often they go with the crow, use the material they knwo from the school where they learns their craft. I believe that a decision for a brand is most of the time not based on intensive testing of different systems.

One reason why professionals may not decide for Pentax is the availability of a replacement if anything is damaged. A new professional Pentax lens within a few hours? Good luck!

I'll wait...

How many of the same use 1dx or d4?

Color depth and accuracy is great, particularly if you don't post process. I haven't seen time in pp decrease at all from the d3 to the d800e generation though depth and accuracy has improved.

With that said, the native pentax color palette (which does vary by sensor generation) is the reason I bought a K5 and later a K01 (which is actually my favorite pentax camera to date!).

I've been waiting for ricoh/pentax to throw the knock out punch for years now...I hope the FF is it.

 

[steelski]

The reason it is flawed is because most people have no idea about this data. It is only expressing quantity, and not quality.
As you rightly pointed out, the A7 series has a great rep and clearly one of better color reproduction.
I have been hesitant in picking up a K3 simply because I do not seem to enjoy the results from it. On paper, it would seem that it is far and away the best APS-c camera ever made for some like myself (read into that what you will). yet there is something not ticking the color box for me. Maybe if I invested more time into getting a good profile from it, it would suit me more? To illustrate what I mean, just use the comparometer in the DPR tools to compare the 645z and the k3, A7r and Nikon d810. Focus on each of the faces in the picture to see the difference in rendition.
This could just be ACR and the profile could be wrong, but the image looks wrong to me(k3) and the 645z looks incredibly balanced. If it is just the RAW then I can work around that. If it is the actual file characteristic of the k3 then it seems disappointing despite all the "data" to the contrary.
On thinking more and more about it I would be interested in making a comparison with different RAW conversion software.

 

[miles green]

But really, this is entirely academic, isn't it? Or let's look at this a different way. How many professional landscape photographers shoot pentax? How many studio photographers shoot pentax?

But this is no reason why Pentax could be second best. Why do people make their decision for or against certain brands? Often they go with the crow, use the material they knwo from the school where they learns their craft. I believe that a decision for a brand is most of the time not based on intensive testing of different systems.

One reason why professionals may not decide for Pentax is the availability of a replacement if anything is damaged. A new professional Pentax lens within a few hours? Good luck!

I hear this argument over and over again. imho it doesn't apply to everyone.

I guess if you live in a major city, it is a valid argument. But in many cases, if the replacement / alternate solution is not in your camera bag from the beginning, you're out of luck.... I'm mainly thinking of wedding photographers here in Corfu, if something brakes during a wedding shoot, the next plane from camera headquarters is the next morning, and no camera brand is about to change that.

Service turn-around time is more important i would think. Unfotunately Pentax lacks here too.

I'll wait...

How many of the same use 1dx or d4?

Color depth and accuracy is great, particularly if you don't post process. I haven't seen time in pp decrease at all from the d3 to the d800e generation though depth and accuracy has improved.

With that said, the native pentax color palette (which does vary by sensor generation) is the reason I bought a K5 and later a K01 (which is actually my favorite pentax camera to date!).

I've been waiting for ricoh/pentax to throw the knock out punch for years now...I hope the FF is it.