The next Foveon camera?

Scottelly wrote:

Doppler9000 wrote:

abera wrote:

tarmov wrote:

Jozef M wrote:

If all you can tell me is that you don't trust me and that I'm telling lies, ... then I give up.

That is far from all I can tell you. I have already told you much more.

I have told you that Foveon cameras (via SPP) produce much lower horizontal pixel level jpeg compression ratios than Bayer cameras - and that is a clear evidence of more information.

If I may give a couple of comments and then I'll run away as I don't think the conversation would end up going towards a fruitful direction.

JPG compression results are hardly a good measuring metric of anything. Not only it's not meant for that, but also there are too many unknown variables. Not all encodings are created equal.

More importantly since image processing is involved the results will tell even less about the image sensor itself, more about differences in processing the data.

I have also already noted that one can take shots with lens cap on and then convert raw to jpeg via SPP and see the resulting file sizes to get an estimate of the relative impact of noise + postprocessing. And I can tell you with certainty that extra noise + postprocessing is unable to describe the jpeg compression ratio difference between Foveon and Bayer images. Which means the Foveon extra information is real, not imaginary.

Image sensor performance metrics can be measured properly and many of them can be done at home. Been there, done that. Google for "photon transfer curve" and work from there. You'll actually get results in terms of SNR and photoelectrons.

Foveon has reasonable FWC and QE, unreasonable read noise (and for color imaging significant noise from color conversion). SNR can be reasonable when the exposure is large enough, but low exposure areas are uncompetetive, thus DR is also somewhat low. Resolution is fine and apart from B&W cameras it has the least false color artifacts. Color accuracy is relatively poor, especially in less exposed areas or difficult light.

It's an interesting sensor, but for most photographers the drawbacks aren't worth the only real advantage (false color artifacts, e.g. moiré). Unfortunately the disadvantages can't really be fixed as they're inherent to the technology.

The multiple layer idea on the other hand may well make a comeback with different materials (e.g. perovskites or maybe organics), but not in this decade.

Thank you for laying this out in such a clear way.

Two points to amplify.

The problematic limitations of the Foveon are inherent. There is no amount of development that will make the sensors materially more efficient.

Second, the insistence on analyzing sensors at the pixel level makes no sense. Bayer sensor designers are using smaller pixels

No, they're not . . . at least not for the most part. Until the need for more pixels to make 8K video the APS-C sensors have been stuck around 24 MP for more than ten years! The only one that has actually made a significant step up in APS-C sensor resolution is Fuji. And when was the last time a micro-4/3 sensor was made with more megapixels? It's been years since they stepped up from 16 MP to 20 MP. That extra 25% more pixels won't do much to change resolution or aliasing. In fact I don't know why they bothered. I think it was most likely a marketing decision, because the 20 MP cameras with 1" sensors were taking some of their customers. Even in the full-frame world they just aren't interested in making cameras with more megapixels. Canon's latest, the R8, is just 24 MP, and all of Panasonic's recent full-frame cameras these days have 24 MP sensors (i.e. the S5 and their new S5 II). If they're stepping up, then why are there so many new full-frame cameras on the market with just 24 MP sensors?

The 24 MP cameras are built for speed, which is important for certain photographers and for hybrid video users.  These are not markets where a FFF is likely to be attractive.

We’ll know more when the Covid supply chain issues are truly behind us regarding higher resolution cameras employing the 3 um sensor tech used in the X-H2.

The fact remains that using performance metrics based on pixel- and not sensor-level will arbitrarily handicap smaller pixel cameras.

both to increase resolution and to reduce aliasing. These advances are ignored in the “compression ratio” analysis, hopelessly skewing the results.