[jsherman999]

What do you mean by "screen tab" ?

Presented up a couple posts, here: https://www.pentaxforums.com/forums/172-pentax-k-3/243897-neocamera-k3-vs-k5i...ml#post3571194

'screen tab' plots a measurement of (basically) per-pixel performance of a sensor, or what you might see at 100%. The 'print tab' is a measurement of the normalized image, which factors out resolution differences (for sensors that have resolution differences) and incorporates total light into the results (which applies to comparisons of different sized sensors.) (and it doesn't mean 'printing', it means 'whole image presented at same size on screen or print' in this context)

It was approached (by me) backwards in this thread, but it wasn't a question of you consulting DxOmark, it was a question of how you would respond to someone else using the 'screen tab' plot to claim the K3 was much more noisy than the K5 at the same display size. Would that be a smart way to make the choice... or would the 'print tab' more accurately reflect the picture?

'screen tab' (pixel-level perf)

vs. 'print tab' (whole image, normalized to 4K output (8MP))

[bdery]

it was a question of how you would respond to someone else using the 'screen tab' plot to claim the K3 was much more noisy than the K5 at the same display size.

These graphs illustrate exactly what I was saying. Thanks.

The print tab as you put it is the only one that makes any sense when comparing varying resolutions. And, lo and behold, the two sensors perform similarly.

I'll still say that DXO's measurements are near meaningless because their sample size is 1. If I presented a sample size of less than 10, ideally 30, at work, I'd get laughed at, with good reason.

[jsherman999]

These graphs illustrate exactly what I was saying. Thanks.

The print tab as you put it is the only one that makes any sense when comparing varying resolutions. And, lo and behold, the two sensors perform similarly.

Thanks.

I'll still say that DXO's measurements are near meaningless because their sample size is 1. If I presented a sample size of less than 10, ideally 30, at work, I'd get laughed at, with good reason.

That's what I've always thought about lenses; Roger Cicala has shown how much deviation the same lens model can show, it's a shame that photozone and slrgear only usually test one lens copy, and then that test can stand for years as judgement on that lens.

I think manufacturers supply the bodies to DXOmark, though, so they can be expected to be in-spec. Even if DXO buys the cameras randomly, If a wafer has imperfections a sensor stamped out of that section is lost in the yield, or in some cases bad pixels can be seen, but I doubt you see many QA issues that translate to significant variations in noise output.

.

[ScooterMaxi Jim]

Important to add the lens/body comparison, as well. Based on the DxO combination numbers indicate for the DA 55:

K-5 - 20 overall quality rating, based on 8 p-mix sharpness rating
K-5iis - 25 overall, 10 sharpness
K-3 - 27 overall, 11 sharpness

Similar, consistent results are computed (probably not actually tested by DxO for all combinations AFAIK) for all lenses. In other words, with the K-3 you have more sharpness available for each lens shot at high ISO in need of some added NR. This is where the "normalization" aspect comes into play. That said, I think you do have to work a bit harder in order to get more out of the denser K-3 sensor. Those who put in the extra work might get a slight edge, which I find to be especially apparent in the finer controls available in Capture One v8-9. Those who have simply carried over the same adjustments from the earlier sensor to the 24 mp sensor appear to be the ones claiming that the K-5 series outputs are clearly better.... They might want to sharpen up and update their raw conversion technique.

[bdery]

it's a shame that photozone and slrgear only usually test one lens copy, and then that test can stand for years as judgement on that lens.

As a reviewer for Pentaxforums, I will say that sadly we all do that. No store or manufacturer will supply 10 copies of a lens for review (and sometimes we supply our own). But with lenses, you have the luxury of building your own sample, by comparing reviews. With bodies it's more complicated, because testing methods vary a lot more. But you can take a sample from one reviewer, assembled from various cameras using the same sensor, and get a good idea of how that sensor performs.

When reading camera bodies reviews, I base my understanding on the assumption that, nowadays, every camera delivers amazing results. I look mostly at how the camera is made, what features it offers, how it performs in the field.

I think manufacturers supply the bodies to DXOmark, though, so they can be expected to be in-spec.

That's a risky assumption. I remember reading a photoaone review where the reviewer found strong decentering, returned the lens to the manufacturer, and was told that everything was in spec.

The other hypothesis, that manufacturers send "perfect" or '"tweaked" samples, is even more worrisome, and would require us to discard every such test as non representative from the real product.

If a wafer has imperfections a sensor stamped out of that section is lost in the yield, or in some cases bad pixels can be seen, but I doubt you see many QA issues that translate to significant variations in noise output.

The sensor itself is one element. In particular for noise, it will be influenced by the electronics, battery current, heat flow, etc. Those will vary from sample to sample.

[panonski]

Yes, and if you really must compare on-screen with JPEGs, set the K-3 to M (14 MP) and compare to the K-5 at 16 MP. The K-3 will still come out ahead.

I just did that , a few days ago, and results were fantastic. At 4500 Iso, I was getting very beautifull shots almost in imposible dark condition where I was able to shoot moving instrumental players, and record them with 1/60 and above.

I noticed this 14 mpix pics were 300 dpi, and large one are 240 dpi.... Don't know what that means, but I'm sure that in dark you need creamier pic, and 14 vs 23 mpix is not so different, specially for screens...

[jsherman999]

I suppose it's possible that non-sensor variations in the ADC or some of the other components could result in a more-noisy camera within the same model line. I just have not run across anyone complaining about that possibility, either in the fora or on review sites, which suggests to me that it's not likely, or at least not as likely as lens variation, which you hear about constantly. It would be a good question to pose to the DXO engineers in their feedback section: "Are you concerned about solid state QA issues that can lead to noise variations between cameras of the same model? Why or why not?"

[bdery]

I noticed this 14 mpix pics were 300 dpi, and large one are 240 dpi.... Don't know what that means, but I'm sure that in dark you need creamier pic, and 14 vs 23 mpix is not so different, specially for screens...

Indeed 14 and 24 MP are not that different for most people, except when you crop. I have a 3x2 FEET print on canvas in my living room which was shot with a 10 MP camera.

would be a good question to pose to the DXO engineers in their feedback section: "Are you concerned about solid state QA issues that can lead to noise variations between cameras of the same model? Why or why not?"

No need to ask these guys. And given how they test equipment, and how little they appear to understand (or take in two account) statistical variations, I wouldn't put much weight on their reply.

There are variations in every manufactured product. A DSLr is so complex that it's impossible to have identical units. It starts with the AF, microlenses, sensor (shot noise, dark current, etc), pre-treatment, ADC, and you haven't factored in the electrical connexions yet

[jsherman999]

Indeed 14 and 24 MP are not that different for most people, except when you crop. I have a 3x2 FEET print on canvas in my living room which was shot with a 10 MP camera.



No need to ask these guys. And given how they test equipment, and how little they appear to understand (or take in two account) statistical variations, I wouldn't put much weight on their reply.

There are variations in every manufactured product. A DSLr is so complex that it's impossible to have identical units. It starts with the AF, microlenses, sensor (shot noise, dark current, etc), pre-treatment, ADC, and you haven't factored in the electrical connexions yet

I suppose. I would be very surprised to find a significant variance in noise/DR between examples of the same model, though. AF, metering, a few other things possibly, but noise introduced into the processing chain from partially failing ASICS? Doesn't seem highly likely... The fact that we haven't heard about this as an issue or a worry from any of the review sites or from users/forums suggests that's the case too. Or at least I haven't heard about that issue, perhaps I'm out of the loop on that.

[bdery]

The fact that we haven't heard about this as an issue or a worry from any of the review sites or from users/forums suggests that's the case too. Or at least I haven't heard about that issue, perhaps I'm out of the loop on that.

The fact that we don't read about it doesn't mean it doesn't exist. It CAN mean that, or it can mean that testing techniques are flawed and that many reviewers, being photographers and not scientists/engineers, don't really understand the fundamentals of products qualifications and testing.

[jpzk]

Those who have simply carried over the same adjustments from the earlier sensor to the 24 mp sensor appear to be the ones claiming that the K-5 series outputs are clearly better.... They might want to sharpen up and update their raw conversion technique.

That is an interesting point you made here.
I've been using a K5 for years with great results with my DA*300/4 and now that same lens is mounted on a K3.
So perhaps it is my technique that is faulty - both with using the camera and the RAW conversion "method" I use - but I find that the "apparent" noise of the K3 to be more pronounced than with the K5.
What exactly do you mean by sharpening up and updating the RAW conversion technique?
I am truly interested to find out.
Cheers!

JP

---------- Post added 04-07-16 at 09:11 PM ----------

Originally posted by bdery Yes, and if you really must compare on-screen with JPEGs, set the K-3 to M (14 MP) and compare to the K-5 at 16 MP. The K-3 will still come out ahead. I just did that , a few days ago, and results were fantastic. At 4500 Iso, I was getting very beautifull shots almost in imposible dark condition where I was able to shoot moving instrumental players, and record them with 1/60 and above.
I noticed this 14 mpix pics were 300 dpi, and large one are 240 dpi.... Don't know what that means, but I'm sure that in dark you need creamier pic, and 14 vs 23 mpix is not so different, specially for screens...

Well, I am going to try that too!
Why didn't I think about it anyway?
Thanks!!

[ScooterMaxi Jim]

That is an interesting point you made here.
I've been using a K5 for years with great results with my DA*300/4 and now that same lens is mounted on a K3.
So perhaps it is my technique that is faulty - both with using the camera and the RAW conversion "method" I use - but I find that the "apparent" noise of the K3 to be more pronounced than with the K5.
What exactly do you mean by sharpening up and updating the RAW conversion technique?
I am truly interested to find out.
Cheers!

Many of us got into a specific routine over several years with sensors in the 14-16 mp range. It is true that these sensors (especially 16 mp) have especially low noise levels unless pushed to extremely high ISO. So, sharpen them a bit (depending on AA strength of the sensor), allow NR to stay at default, or close to it. The amount of work is less. The 24 mp sensor on the K-3 not only inherently yields a sharper result (when compared at similar screen magnifications), but there is far less smoothing done at high ISOs, and a bit more pixel noise (but not much more at equal screen magnifications).

Now, I'm not going to say what is "right" in terms of specific settings because - first of all I use multiple platforms (Capture One primarily now, as well as some Lightroom) - and the "right" amount is situational and based on personal preferences. Bottom line, the K-3 sensor will yield its best results with added sharpening at an increased radius (based on pixel pitch), clarity, vibrance, structure, etc., coupled with higher noise reduction settings accordingly. If left to default settings (which are in the ballpark for the 16 mp sensors), you won't be getting the most out of that sensor. I will say that you can go a bit further with Capture One than Lightroom due to added features (Structure in particular) and the more-refined noise reduction, but Lightroom also produces a higher quality image with the K-3. It just requires more effort - both on the sharpening/contrast side and the NR side of it.

[jpzk]

Many of us got into a specific routine over several years with sensors in the 14-16 mp range. It is true that these sensors (especially 16 mp) have especially low noise levels unless pushed to extremely high ISO. So, sharpen them a bit (depending on AA strength of the sensor), allow NR to stay at default, or close to it. The amount of work is less. The 24 mp sensor on the K-3 not only inherently yields a sharper result (when compared at similar screen magnifications), but there is far less smoothing done at high ISOs, and a bit more pixel noise (but not much more at equal screen magnifications).

Now, I'm not going to say what is "right" in terms of specific settings because - first of all I use multiple platforms (Capture One primarily now, as well as some Lightroom) - and the "right" amount is situational and based on personal preferences. Bottom line, the K-3 sensor will yield its best results with added sharpening at an increased radius (based on pixel pitch), clarity, vibrance, structure, etc., coupled with higher noise reduction settings accordingly. If left to default settings (which are in the ballpark for the 16 mp sensors), you won't be getting the most out of that sensor. I will say that you can go a bit further with Capture One than Lightroom due to added features (Structure in particular) and the more-refined noise reduction, but Lightroom also produces a higher quality image with the K-3. It just requires more effort - both on the sharpening/contrast side and the NR side of it.

Thanks for the reply.

What I did during the past few days was testing the K3 with the DA*300/4 on backyard birds.
I did set the K3 with no NR whatsoever.
Shooting RAW DNG.
TAv and letting the ISO float (trying not to go higher than ISO1000)
Then, used "Natural" colours with no other tweaking.
Opened the RAW files in Photoshop CS5 (via Camera RAW).
Found that noise was present at ISO 800 and up, and often even at ISO 400 if I inadvertently underexposed.
Clean that "minimally" with the NiK plug-in DFine. That helped some when the image is viewed un-cropped. Much worse of course if cropped at 50% for instance.
Any use of the NR would reduce details in my images (when seen at 50% magnification).

You say that the K3 sensor will render better results with added sharpening coupled with (higher) noise reduction - - I always thought that sharpening would increase the apparent noise ??
Do you sharpen first and then use NR ??

Obviously, I need to continue trying all sorts of setting combinations, both in-camera and RAW converter, until I can reach some satisfactory (for me) results ... I mean "for me": no noise or the least possible).

Thanks again for the time -- much appreciated.

JP

[ScooterMaxi Jim]

I haven't used the NiK plug-ins, but I'm sure they are helpful (based on reputation).

Your shooting process sounds fine. Just as a note, the color setting in camera has no effect on the actual RAW file (but provides direction for a starting point in the processing) - and I think the Natural setting is just that (which is good in my book).

Pretty much all RAW images need some form of sharpening, and how that can be done well relates to radius and micro-sharpening that yields contrast. Your RAW converter is set to a specific amount of sharpening. It isn't a flat truth that all sharpening results in greater noise. How you mix sharpening and noise reduction will yield the best result - and even then that's based on what will be a normal viewing distance (screen size, print size, etc.).

As for my personal workflow, I did use some plug-ins when using Lightroom extensively. Now, I use Capture One most extensively, and the tools available there are applied simultaneously prior to processing. In that regard, they simply work out better than any post-processing changes I've used. I'm not sure that it really matters whether you are better off applying sharpening or noise reduction in particular order on a processed TIF (either way is a compromise compared to the non-destructive tweaking available at the converter stage of processing). As a matter of workflow needs, when Lightroom output has required tweaking, it has been in the noise reduction area for me - as that has been a weak point in the program (I'm not a huge fan of color sliders either, frankly).

[Canada_Rockies]

Originally posted by panonski
Originally posted by bdery Yes, and if you really must compare on-screen with JPEGs, set the K-3 to M (14 MP) and compare to the K-5 at 16 MP. The K-3 will still come out ahead. I just did that , a few days ago, and results were fantastic. At 4500 Iso, I was getting very beautifull shots almost in imposible dark condition where I was able to shoot moving instrumental players, and record them with 1/60 and above.
I noticed this 14 mpix pics were 300 dpi, and large one are 240 dpi.... Don't know what that means, but I'm sure that in dark you need creamier pic, and 14 vs 23 mpix is not so different, specially for screens...
Well, I am going to try that too!
Why didn't I think about it anyway?
Thanks!!

I used 10 Mpx for nearly 8 years, and just naturally set the camera to best quality JPEG when I replaced the K10 with the K3 -- but this is something I am going to try, probably tomorrow, if I have the energy after a recent "vacation" in the local health centre -- nothing permanent, but nothing by mouth for 3 days tends to lower my old phart's energy levels.

My back yard is on the North side of a hill, and forested with Lodgepole Pine, so there is little or no bright light back there. It's great for birds, but not for photographers. Thank you for the tip!