[Leumas]

Also here is a comparison

The X3 sensor was incredibly detailed

Attached Images

 

[Leumas]

and another

---------- Post added 11-09-19 at 04:48 PM ----------

Those being a couple modern FF sensors vs an old crop X3 Foveon.... that is pretty impressive in my book

[stevebrot]

Also here is a comparison

The X3 sensor was incredibly detailed

You understand that the other shots are downsampled, right?


Steve

[BigMackCam]

Also here is a comparison

The X3 sensor was incredibly detailed

You have to be very careful looking at DPR comparisons.

See below two screen captures of comparisons between the K-1, KP, K-3 and DP1 Merrill...

The first is of the raw files, but of course they still have to be rendered somehow which means whatever software is used (probably Lightroom for the Pentax shots) converts the raw data to something it can display using default settings. I imagine different software was used for the Sigma, since it doesn't use regular Bayer demosaicing... and the tone curve / local contrast here is considerably harsher, which - along with the greater noise (even at base ISO) - contributes significantly to the sharper / more detailed appearance.

The second capture shows what each camera has achieved using in-camera JPEG processing. Now, the Sigma looks awful here - but let's assume that's just a very poor JPEG engine... However, see how much better the Pentax camera images look compared to the raw renderings. The in-camera JPEG processing is similar (though probably not as good) to what most folks would carry out in Lightroom or other raw processing software.

My point here is, with suitable raw processing, the gap between Bayer sensor Pentax cameras and Foveon sensor Sigma cameras isn't as pronounced as you think. And it's certainly not as pronounced as those comparisons in the linked article would have you believe (again, I think there is a manual focusing issue on the K-5 shots in that article - not to mention a lack of appropriate processing)...

Attached Images

   

---

Last edited by BigMackCam; 11-09-2019 at 04:47 PM.

[BigMackCam]

Just for grins, I thought I'd show what happens at a rather modest ISO 800 too...

Or, as I like to say, "No such thing as a free lunch"

[PJ1]

You have to be very careful looking at DPR comparisons.

'nuff said!

[Leumas]

You understand that the other shots are downsampled, right?


Steve

.....If a higher res photo doesn't look better down sampled than a native res shot, than it is not a higher fidelity image to begin with. Super sampling in video games is done all the time to this effect. (Render at 4k, downsample to 1080p for superior image than native 1080p)

You have to be very careful looking at DPR comparisons.

See below two screen captures of comparisons between the K-1, KP, K-3 and DP1 Merrill...

The first is of the raw files, but of course they still have to be rendered somehow which means whatever software is used (probably Lightroom for the Pentax shots) converts the raw data to something it can display using default settings. I imagine different software was used for the Sigma, since it doesn't use regular Bayer demosaicing... and the tone curve / local contrast here is considerably harsher, which - along with the greater noise (even at base ISO) contributes significantly to the sharper / more detailed appearance.

The second capture shows what each camera has achieved using in-camera JPEG processing. Now, the Sigma looks awful here - but let's assume that's just a very poor JPEG engine... However, see how much better the Pentax camera images look compared to the raw renderings. The in-camera JPEG processing is similar (though probably not as good) to what most folks would carry out in Lightroom or other raw processing software.

My point here is, with suitable raw processing, the gap between Bayer sensor Pentax cameras and Foveon sensor Sigma cameras isn't as pronounced as you think. And it's certainly not as pronounced as those comparisons in the linked article would have you believe (again, I think there is a manual focusing issue on the K-5 shots in that article - not to mention a lack of appropriate processing)...

The Jpeg processing will add sharpness, but it still falls short of the crispness of the Foveon.

Just for grins, I thought I'd show what happens at a rather modest ISO 800 too...

Or, as I like to say, "No such thing as a free lunch"

For sure. ISO performance one of the big hurdles for Foveon. It's a iso 100-200 camera. Much like my K10D. Unless you shoot B&W than it can be a pleasant grain effect.

[BigMackCam]

The Jpeg processing will add sharpness, but it still falls short of the crispness of the Foveon.

See my previous comments on tone curve and local contrast. Crisp is easy.

The important point to note here is, the JPEG can add whatever the processing engine is set for. Just as, in Sigma's JPEG engine, the end result is worse. My point is, the way the Sigma raw is being rendered by whatever software was used is quite different from the way the Pentax raw is being rendered. The latter can be made to look extremely similar through raw processing (try it - I just did and you can get remarkably close).

Again, I'm not saying the Foveon sensor isn't capturing more detail. I believe it is. But that original article you linked to, and the DPR comparisons, have to be very carefully considered. Look at them wanting to be impressed by Foveon, and you'll be very impressed. But beware of confirmation bias. Start looking for the reasons why the Foveon images appear to be better in certain respects, and means of balancing / equalising rendering aspects between the two, and you may still be impressed, only less so.

[stevebrot]

.....If a higher res photo doesn't look better down sampled than a native res shot, than it is not a higher fidelity image to begin with. Super sampling in video games is done all the time to this effect. (Render at 4k, downsample to 1080p for superior image than native 1080p)

I think, perhaps, you may not understand what a down-sample does. It is a selective discard and remapping of pixels, a winnowing and removal of data, and always results in less detail than the original. Alternatively, an up-sample of the Foveon image to 24Mpx will introduce synthetic data with unavoidable dilution of detail. That is why is is always necessary to compare like-with-like on both comparison crops and MTF resolution testing. Never mix formats or pixel resolution in the comparison; doing so is never valid.

(Super-sampling is a form of computational photography and fine for games, I suppose...)

BTW...you are aware that the DPR images are the famous green fuzz shots


Steve

[Digitalis]

Foveon's technoplogy was quite seductive but it was doomed by the physics of silicon.

This phrase is by far the best nutshell explanation of Foveons shortcomings I have seen. Another important point is the physiology of human trichromatic vision doesn't work by absorption depth, making the name Foveon rather deceptive. I have also seen some rather rare and unusual artifacts from Foveon equipped cameras in the studio that simply do not occur with other cameras, texture artifacts, bizarre colour shifts when used with unusually slow or extremely fast duration flash pulses, I have seen luminance moire show up - nothing can be done about that in post.

Foveon's dynamic range and high ISO performance was competitive with early CCD and CMOS sensors but not with the more modern ones.

At High ISO's dynamic range is hardly the biggest issue with foveon..Delta E slides off into oblivion five stops above base ISO with foveon cameras, in contrast DSLRS contemporary with foveons introduction as well as current models remain quite accurate at ISOs 8 stops above Base ISO and beyond. The colour inaccuracies at high ISO stem from the inherent absorption characteristics of silicon.

[aaacb]

Sigma cameras, and especially the merrill series, cost quite a bit on the used market (about $450 for a dp2 merrill, which was introduced for $1000 in 2012). That's surprising to me, at least for a niche camera. Looks like others are also excited for them too

[stevebrot]

I have also seen some rather rare and unusual artifacts from Foveon equipped cameras in the studio that simply do not occur with other cameras, texture artifacts, bizarre colour shifts when used with unusually slow or extremely fast duration flash pulses, I have seen luminance moire show up - nothing can be done about that in post.

Thanks for adding your observations from use. These support the list of reported issues I mentioned in a comment above, though without the benefit of actual usage.


Steve

[Rondec]

The thing about Foveon images is that I believe a lot of the benefits come from fairly aggressive algorithms that are built into Sigma's processing software. I believe that you can get similar effects if you push your post-processing settings with Bayer style sensors as well, but I don't think most of us want to go there. It's a bit like why people show up and say that their cell phone takes better photos than their SLR. It doesn't really, but the sharpening and saturation are bumped up in comparison to the default settings on your Pentax SLR and so the SLR images look kind of bland in comparison.

The other thing to mention about SA mount cameras is that there is no third party support for lenses (significantly worse than the K mount)!

[BigMackCam]

The other thing to mention about SA mount cameras is that there is no third party support for lenses (significantly worse than the K mount)!

On a more positive note, though, Sigma hasn't withdrawn support for SA mount (yet)

[Rondec]

On a more positive note, though, Sigma hasn't withdrawn support for SA mount (yet)

My comment was mostly tongue in cheek...

It does seem like with them investing in the L mount alliance the odds of more SA mount gear coming out in the future is pretty small. But I suppose they will continue to have lenses for SA mount cameras for awhile anyway.