Sigma DP Quattro

[MrSmith27]

If I remember correctly about a year ago the newest Merrill, DP3M, became available. Since then, nothing. Having used and compared the cameras I'm confident that I get better image quality than I get from my mid format setups. Especially when printing, and even moreso when working in black and white. At the same time I am effectively using a sensor that is at least 2-3 years old. And I'm using cameras that dropped 50+% in price and seem to be at the end of their life-cycle.

So where is my 2014 Merrill?

I would happily spend a lot of money on an updated DP2M, say, with a full-frame sensor and some kind of optical stabilisation. I'm not even asking for better low-light performance or less insane battery drain.

Edit: And there it is.

http://www.sigmaphoto.com/article/sigma-dp-quattro

[NancyP]

Sigma's got a lot on its plate, and the series of lenses is the priority. I think that Sigma may have something in the R and D phase. Designing not only a new sensor but a new processor is not easy. The APS-C files are 55 to 65 Mb apiece, considerably larger than their Bayer CFA counterparts, and just think how big the full frame files would be - you'd need Massive Throughput Improvement to make the camera usable. I think that they could make a SLR with decent live view first.

[Quentin]

The Quattro models may be the answer:  http://sigma-rumors.com

[RobertJ]

That's awesome but why is the focusing ring so much smaller?

Edit: I think it's either fake or a concept for the upcoming show.  Car makers do the same thing.

If it's real, I'm hoping for an improved/upgraded sensor.

[RobertJ]

Wow, never mind.  Check it out http://www.thephoblographer.com/2014/02/09/sigmas-new-dp-quatro-cameras-feature-medium-format-quality-ugly-looks/

[capital]

Well, these cameras appear to be real as imaging resource posted a press release on them. (see their info-graphic about the Quattro layout below)

Though I am not sure how I feel about the "Quattro" layout of the blue pixels vs the green and red sub-layers. Typical Bayer arrays rely on the green channel for luminance while this Foveon Quattro will rely on a blue channel. Now I know why they scaled down the sub layers to 1/4th the top layer for boosting color sensitivity, Sigma it seems has diverged from the "true" color at every pixel mantra. Not sure how this will play out with people who enjoy that 3 color concept, even if it is or might be pyschological effect more than a real measurable phenomenon. Perhaps this "Quattro" layout is adequate, will be nice to see the head to heads on this one.



One more thing, I guess Sigma thinks they could only extract so much signal from the "old" layout and this was the alternative, not sure if this is a real noise limit or a practical R&D budget limit. Actually reflecting on it a bit more, I kind of wonder if Sigma might have been doing binning of color data in their raw processor with the Merrill series.

Sigma North America's article about the new Quattro:

http://www.sigmaphoto.com/article/sigma-dp-quattro

[BernardLanguillier]

Interesting development for sure!

Cheers,
Bernard

[hjulenissen]

I kind of wonder if Sigma might have been doing binning of color data in their raw processor with the Merrill series.

For binning to make sense, I believe that you have to do it in the analog domain (prior to ADC).

Once you have digitized the signal, you stand to gain nothing, but loose something compared to a well-implemented noise reduction algo.

-h

[capital]

hjulenissen, it may stand to reason then why did not Sigma opt for a user selectable option to bin sub-layer data as opposed to a permanent 1:1:4 layout. I guess a simple answer again is lower R&D cost, though not that satisfying. :)

[hjulenissen]

hjulenissen, it may stand to reason then why did not Sigma opt for a user selectable option to bin sub-layer data as opposed to a permanent 1:1:4 layout. I guess a simple answer again is lower R&D cost, though not that satisfying.

This design _may_ mean lower cost, lower defect rates, higher throughput, smaller file sizes,...

All of which are positive for the end-user.

-h

[LKaven]

Can't wait to see the samples.  I've been enjoying the images from these cameras in the DP1/2/3 permathreads.  Sight unseen, I already want a full-frame version of the Quattro...and a medium-format version. 

[Hulyss]

Point is the major problems of the current foveon technology is fixed with this new chip.

- The files weight less.

- The processing time in camera is far quicker, New chip AND new sensor. It is designed to be quick and efficient.

- Battery autonomy.

- New SPP.

- 14 bits raw vs 12 bit.

- Better dynamic range, by far.

It is still a foveon chip, extremely optimized. The 4.5 mp RED and GREEN layers are what SIGMA thought would be essential for color capturing.

And for the design, you'll get used to it.

Now, this architecture might be very interesting because it merge speed and foveon advantages. It is probably now possible to go full frame.

[BernardLanguillier]

Now, this architecture might be very interesting because it merge speed and foveon advantages. It is probably now possible to go full frame.

That last comment must both be generating amounts of saliva dwarfing the throughput of Niagara falls while sending shrivers of concerns in more than a few camera company HQs... 

Cheers,
Bernard

[Bart_van_der_Wolf]

That last comment must both be generating amounts of saliva dwarfing the throughput of Niagara falls while sending shrivers of concerns in more than a few camera company HQs...  

Unfortunately, the Foveon concept (color discrimination from Silicon penetration distance) doesn't lend itself very well to oblique rays of light (as in large sensor dimensions with shorter focal lengths to create a large enough image circle). There is a reason why the physical sensor dimensions have remained pretty constant over time ...

There also seems to be some misinterpretation of "the "Quattro" layout of the blue pixels vs the green and red sub-layers". There are no Red/Green/Blue sub-layers, and Sigma calls them Top, Middle, and Bottom, for a reason. The top layer records R+G+B (essentially Luminance, and therefore clever to use that for boosting resolution), the middle layer records mostly R+G, and the bottom layer records mostly Red. The subtraction of the layers will reconstruct R/G/B, but noise is always increased by subtracting random noise signals. Hence the limitations on High ISO.

Cheers,
Bart

[MrSmith27]

Could somebody with technical knowledge please shed some light on why they don't do exchangeable lenses?

The way I understand it is that the only difference between the Merrills and Quattros are the lenses with completely identical sensors in each series. Wouldn't it make sense to sell a DP body plus three or so exchangeable lenses? Now as I said many times I'm using the DP2M and DP3M and I love the image quality. I just don't understand the technical reasons for having bolted on lenses.

[Hulyss]

That last comment must both be generating amounts of saliva dwarfing the throughput of Niagara falls while sending shrivers of concerns in more than a few camera company HQs... 

Cheers,
Bernard

For sure :p

Over the world many ppl think that the new sensor is not a "true" Foveon anymore, and they are wrong.

It is still a full 3 layers sensor. Just look at the picture below :



The signal is captured on the 3 layers, as usual.

The thing is that the GREEN and RED layers have much bigger pixels, 4 pixels binned in one. That is done for the CAPTION part.

The signal after processing is pretty simple, without any interpolations as some said: This is division of the GREEN and RED layers in a certain %age, after the caption.

The whole trick is in the binning technology.

The RED and GREEN layers have the same surface as the Blue layer : 23.5×15.7mm

On top of that, we go from 15 MP To 20 MP, 30% more resolution.

[LKaven]

That last comment must both be generating amounts of saliva dwarfing the throughput of Niagara falls while sending shrivers of concerns in more than a few camera company HQs... 

So they say it's a 39MP equivalent (bogopixels)?  39 x 2.25 = 87.75MP bogopixels full frame.  It'd take a little work to make the most of the sub-diffraction layer, but it's promising.

[The Ute]

I'm content to wait and let the results speak for themselves.

As the old saying goes, "one picture is worthy a thousand words".

 

[MrSmith27]

So they say it's a 39MP equivalent (bogopixels)?  39 x 2.25 = 87.75MP bogopixels full frame.  It'd take a little work to make the most of the sub-diffraction layer, but it's promising.

The whole Megapixel-metric is silly. You can crop out portions of Foveon-images and use them. When I look at any other pictures that's simply not possible. I always end up scaling them down, then sharpening them in Photoshop. So the pixel-size of an image without additional info on the quality of the file is useless. That being said I would personally describe that 4x1x1-sensor layout as having 20 Megapixel, because that's the pixel in the raw ouput.

[LKaven]

The whole Megapixel-metric is silly. You can crop out portions of Foveon-images and use them. When I look at any other pictures that's simply not possible. I always end up scaling them down, then sharpening them in Photoshop. So the pixel-size of an image without additional info on the quality of the file is useless. That being said I would personally describe that 4x1x1-sensor layout as having 20 Megapixel, because that's the pixel in the raw ouput.

From the press release:
"Newly developed Foveon X3 Quattro direct image sensor / Now featuring 39 megapixel-equivalent ultrahigh resolution"

They call it a "Super High" mode.  It can deliver 39MP JPGs out of the camera.  I suppose their raw platform can deliver a 39MP TIF as well.

I agree with you though.  There are significant qualitative differences in the pixel level data between the Quattro and a bayer sensor, and sensors beyond.

[The use of the name Quattro probably irritates Audi just a little bit, though what can they do about it.  It's not an all-wheel drive camera.]