[Class A]

Thank you for that, and please excuse me if I seemed a little dense.

Well, the gratitude is all mine, because usually this type of exchange of posts ends with my discussion partners going back to square one.

I got a bit tired of talking about the subject because in the vast majority of cases the discussion partners remain unconvinced. I'm glad you found the discussion useful.

I completely agree that the larger sensor should have a better signal to noise ratio if it has correspondingly larger photosites, all other things being equal (which they never are).

Note that the size of the photosites does not matter.

"Everything else being equal" may never apply but there is no better way of discussing the merit of certain differences (say "sensor size").

[Class A]

You forgot to add the qualifier: "(all things being equal, in terms of the sensor technology in use). ..."

Always goes without saying.

[normhead]

The Sigma 70/2.8 EX DG Macro is much better than this. It peaks around f/4, hence at f/8 it already suffers from diffraction.

SInce I use this lens all the time I'm rather surprised they do that. ƒ 5.6 is my preferred setting for this lens for sharpness.

[wahid_satu]

Got some more time on my hands so I decided to look at K3 v K5 IIs at ISO6400



As some here may already have found, at 100% there's more noise on the K-3's. When resized to the same 16MP, K-3's still worse (not shown). When I compare the images at fullscreen (about 2MP) the difference stops being noticable (I didn't compare other resolutions.)

Unfortunately there is no more detail on the K-3's image at ISO6400. So I tried the ISO3200 images and found that at 100% there's slightly more detail on the K-3's, but noise level is still too worse for comfort than the K-5iis. At lower ISOs the noise difference starts to close up and there's noticeably more detail on the K-3's shots.

So it seems, at lower ISOs the K-3 wins, but over ISO1600 the K-5IIS rules. Keep in mind, though that this is compared to the High ISO king of the APS-Cs. Against lesser cameras the K-3 would still win. Now, I feel there will be a market for a K-3S .

[Class A]

SInce I use this lens all the time I'm rather surprised they do that. ƒ 5.6 is my preferred setting for this lens for sharpness.

I'm pretty sure they are not going for a compromise of still nice sharpness and more DOF (as their scene has quite a bit of depth).

[Class A]

As some here may already have found, at 100% there's more noise on the K-3's.

That's a bit of a non-statement.

There is more noise on the K-5 IIs at 16MP than their is noise on the K-5 IIs at 6MP at 100% as well.

It would have been a surprise if Sony had managed to counteract the normal pixel-noise increase that comes with higher MP by better sensor performance.

When resized to the same 16MP, K-3's still worse (not shown).

Too bad you don't show that comparison.
It is the only one worth looking at, AFAIC.

Showing 100% crops of two sensors with a different MP count without normalising the MP for noise comparisons is as useless as choosing different parts of the scene respectively.

Unfortunately there is no more detail on the K-3's image at ISO6400.

In your (non-normalised) comparison, I clearly see more detail in the K-3 image (see for example the small cloth patterns). This isn't surprising as it has higher magnification. I believe some of that extra detail would survive if you downscaled the 24MP image to 16MP and applied a bit of sharpening afterwards to counteract the blurring effect of the downscaling (the need for the latter of course depends on what approach for downscaling you use). However, for detail comparisons, it is probably better to upscale the 16MP image to 24MP.

Now, I feel there will be a market for a K-3S .

Not sure what you mean. The K-3 is already the K-3s version. It has no Bayer-AA-filter.
I wish there were a K-3-non-s version available.

[wahid_satu]

Showing 100% crops of two sensors with a different MP count without normalising the MP for noise comparisons is as useless as choosing different parts of the scene respectively.

Using Lanscoz3 and no sharpening the K-3's is normalized to 16MP. I see more noise on the K-3's (the right bottle cap and rear wall especially noticable) and don't see any detail advantage. In fact, I think the red leaves cloth is better looking on the K-52S. I don't upscale as I don't know why it would be better.

The K-3S as in D3 to D3s .

[Cynog Ap Brychan]

Darn pixel-peepers........!

[Class A]

Using Lanscoz3 and no sharpening the K-3's is normalized to 16MP.

Thanks for sharing.
The red cloth doesn't look that great in the K-3 image, but I think the hue is more accurate than what we see in the K-5 image. Maybe they tweaked the colour filters for better accuracy but lost a bit of sensitivity through that.

It looks like the K-3 image is higher in contrast (which would have a negative impact on what the perceived level of noise is).

The K-5 image looks somewhat smoothed in comparison, but I agree it looks better "out of the box".


Regarding detail: The seams on the black bottle cap look more contrasty in the K-3 image. But that may be just due to a difference of the focus planes.

I don't upscale as I don't know why it would be better.

Some detail in the 24MP image that exists as pixel-to-pixel contrast becomes sub-pixel detail in the downscaled 16MP version. In other words, if there is contrast at the pixel-level, it will be compromised in downscaling. Hence, for detail comparisons, it seems more appropriate to me to upscale the 16MP image to 24MP.

[Barry Pearson]

So it seems, at lower ISOs the K-3 wins, but over ISO1600 the K-5IIS rules. Keep in mind, though that this is compared to the High ISO king of the APS-Cs. Against lesser cameras the K-3 would still win. Now, I feel there will be a market for a K-3S .

I concur (although I think perhaps the K-3 holds up a bit higher than ISO 1600).

Printing from DNGs at A3+, at ISO 51200 the K-3 is noisier than the K-5IIs for prints that show exactly the same scene. In other words, all debate about pixels has become irrelevant because the prints show the same scene taken with presumably the same lens at the same subject distance.

I hope I'm wrong, but a friend checked my prints and agreed with this observation.

Important edit: I have now done my own side-by-side tests with my own cameras, and come to significantly different conclusions from those I came to from processing the Imaging Resource DNGs. See another post here.

[azimuthal]

The shutter speed difference is equivalent to -0.64 stops.

BTW, the comparison to the full-frame sensor is not fair, AFAIC, as f/8 on FF is ~f/5.3 on APS-C.

physics isn't fair.
this is for dof equivalence, not particularly meaningful in this situation.

That gives the full-frame sensor a bit more than a stop of an advantage (which it pays with shallower DOF).

one stop of advantage in what?

So bear in mind that we are not only comparing different sensor technologies, but also different lens speeds.

you seem to think that taking a full-frame lens like the 70mm sigma off of an aps-c and putting it on a full-frame magically causes more light to go through it. this is not the case, it is true that more of the light is used in the case of the full-frame, this is what you might call an advantage of full-frame.

if we continue with the sigma 70mm and take pictures with full-frame, aps-c, m4/3
the metering(iso, aperture, shutter) of a scene whether it is aps-c, full-frame, m4/3 is going to be the same assuming the actual ISO of the cameras is the same and assuming the metering of the cameras agrees.

what will change with the increasing/decreasing light used by the sensor is the field of view.

Well, the gratitude is all mine, because usually this type of exchange of posts ends with my discussion partners going back to square one.

because what you are saying is at best: ambiguous, at worst: nonsensical.

I got a bit tired of talking about the subject because in the vast majority of cases the discussion partners remain unconvinced.

promising news.

Thank you for that, and please excuse me if I seemed a little dense. I completely agree that the larger sensor should have a better signal to noise ratio if it has correspondingly larger photosites, all other things being equal (which they never are).

Note that the size of the photosites does not matter.

you either
a) didn't read this
b) didn't understand this

so i'll try to help. this article says that if you keep the sensor size the same and increase the resolution while lowering the size of the photosites, you can still find an advantage in SNR, despite the smaller photosites.

it does not say that photosite size is meaningless.

the post by Cynog that you were responding to was keeping everything the same (megapixel, sensor technology) the full-frame will have better snr due to larger photosites. (it was a true statement)

the d600(24MP) has better snr than the k3(24MP) because although they have the same megapixels, the d600 has larger photosites. (assuming similar technology)
the d800(36MP) has better (normalized) snr than the k5(16MP) because although they have the same pixel size, the D800 has higher resolution.

[wahid_satu]

Printing from DNGs at A3+,

You have the resources to print A3+ just for noise evaluation? Wow, as a hobbyist I find that amazing. Kudos to you! And thank you for the information. It will be useful someday.

[Barry Pearson]

You have the resources to print A3+ just for noise evaluation? Wow, as a hobbyist I find that amazing. Kudos to you! And thank you for the information. It will be useful someday.

Yup. I get through A3+ paper at an alarming rate. In effect, I not only use it for competitions and exhibitions, but also for test prints! (I use A4 for hand-around prints). I love printing.

[jsherman999]

because what you are saying is at best: ambiguous, at worst: nonsensical.

Actually, everything Class A said was correct.

you either
a) didn't read this
b) didn't understand this

so i'll try to help.

this article says that if you keep the sensor size the same and increase the resolution while lowering the size of the photosites, you can still find an advantage in SNR, despite the smaller photosites.

it does not say that photosite size is meaningless.

It's in fact telling you that for purposes of image SNR, the photosite size is not what matters because the SNR can be averaged over several photosites.

The 'larger photosite = better' myth has been debunked for a few years now.

.

[SteveB]

This is the best I can get out of ACR8.2 at iso 6400 on the IR K3 file....with a bit of help from a Noiseware plugin so it doesn't really count.