Sigma DP Quattro

[Quentin]

I have been "testing" the DP2Q over the last few days, coming to the camera expecting a lot based on my experiences with its predecessor, the DP2M.

I really believe you need to test cameras yourself, as each users tolerances to particular faults is likely to be different.  This seems particular evident with initial reports from users of the DP2Q.

It has been a bit of a roller coaster ride.  It started favorably, then fell away a bit, but has come back strongly

City of London, from the "Shard".  DP2Q, 1/500 sec, F/5.6, ISO100.



1.  I quite like the odd form factor (looks like it was designed to appeal to mobile phone users), but the lens seems huge in comparison with the small front element.  I would have thought they could make the lens at least an F/2, instead of the meagre F/2.8 without increasing the circumference of the lens .  Still, the lens is as sharp as the old Merrill so nothing lowst there.  Have to buy 58mm filters, though.

2.  The SD card cover is a dodgy rubber push-on bodge.  The rest of the camera seems well made and slick.  Pity about the SD card door.

3.  Menu system is simplified and better.

4.  Batteries are larger and last longer.  You still get two batteries in the box with the camera.

5. Sigma Photo Pro software is utter rubbish (in my opinion).  Fortunately you can do most post processing in Photoshop. Turn down sharpening and NR and export as a 14 bit TIFF file, Adobe 1998 profile.

6.  What matters is the image quality.  Here, it is mainly good news.  The colour balance is more neutral - no green hue, as there is with the DP2M.  Pixel level accutance is as good as the Merrill, and the higher pixel count means a larger finished image.  But it really gets interesting if you (a) turn down sharpness and noise reduction as much as possible and (b) save out a 16 bit TIFF, using the S-HI setting, which upscales the image to around 39mp.  Then apply moderate sharpening in Photoshop.  Because the Quattro image is "cleaner" than a Merrill image, what you get, if you are careful, is a near medium format quality image, a 40mp or thereabouts alternative to a bulky entry-level medium format camera.  If the DP1Q is as good then things might get interesting for landscape photographers, particularly as I understand a cable release may be available. Of course, this is my assessment after just a week of shooting in good light, so your mileage may very well vary.  Also out of camera jpegs are pretty decent, not that I intend to shoot jpeg much, preferring the masochistic experience of using SPP :-)

On the other hand, a shot from the Q will not have the same immediate "punch" as a DP2M image.  I'd say in this respect alone, it has 30-40% of the punch.  A tad "bayeristic", one might say, but still little or no Moire and more accurate colour.  Loads of sharpness too, of course.

I have now added a 58mm UV filter and a 58mm adapter for my Lee 75 filter adapter.

Here is a link to a jpeg of an image with plenty of detail taken today

https://dl.dropboxusercontent.com/u/9806585/London_From_Shard_2.jpg

[Quentin]

From the 72nd floor of the Shard, with the DP2Q.

[capital]

Try before you buy Quattro arrived and testing against the Merrill.

There seems to be some variation in sharpness, perhaps down to variation in the AF point.

Here's some A/B tests so far. Hopefully a pattern emerges.


All shot on tripod, ISO 100, 2 second timer, 1/200 second, F/8, zero sharpening, zero noise reduction.

[David Anderson]

Thanks for the sample image Quentin - the detail is great.  

Thanks Alan for the write up as well.

[capital]

Same images as above, though comparing in camera Quattro JPG (middle) versus SPP exports of Merrill on left and Quattro right.

Seems these Quattro images are softer than their Merrill counter parts, will need to explore what sharpening parameters produce better than all settings at minimum.

[Quentin]

I have found the best output solution for me is export from SPP as a S-HI (39mp), TIFF-16, Sharpening and NR turned "off", sharpening to taste in Photoshop using the tool of your choice. 

[pflower]

I am sitting on the fence somewhat regarding the Quattro (especially since I have just bought a backup DP2M).  However I am puzzled by your comments about exporting as an S-Hi 16 bit Tiff and saying that this represents 39mp).  My 16 bit tiffs from the DP2M are all around 88MB.  What exactly are you doing?  Can you elaborate a bit?

Thanks

I have found the best output solution for me is export from SPP as a S-HI (39mp), TIFF-16, Sharpening and NR turned "off", sharpening to taste in Photoshop using the tool of your choice. 

[Quentin]

I am sitting on the fence somewhat regarding the Quattro (especially since I have just bought a backup DP2M).  However I am puzzled by your comments about exporting as an S-Hi 16 bit Tiff and saying that this represents 39mp).  My 16 bit tiffs from the DP2M are all around 88MB.  What exactly are you doing?  Can you elaborate a bit?

Thanks

This refers to megapixel equivalence, not file size.

So, a S-HI file out of SPP is 7680 x 5120 pixels = 39,321,600 pixels.  Round that down and its equivalent to approx a 39mp file.

By way of comparison with a standard merrill 16bit image, a 16 bit S-HI Quattro file is about 225mb (not mp).

[capital]

Two unrelated things regarding Quattro:

First, I am wondering aloud here, why is that we are experiencing this pixel level blurring if the top layer is supposed to act as the luminance layer? Is this an artifact of the code or a fundamental idiosyncrasy of this new layout? *If* the image were extracted from the topmost layer only, I think I should expect the same pixel level acuity of the x3 designs. I also remember reading something to the effect that human vision is such that perceived color resolution is lower than luminosity resolution. Which has me wondering what is going on at the pixel level of these Quattro renderings.

Second point, I am noticing if I increase sharpening for even the ISO100 DP2Q test shots that this also is sharpening noise. So a standard Photoshop "smart sharpen" is not cutting it, instead a round trip into Lightroom to add some noise reduction + sharpening appears to do better. Is this the experience of others?

[Alan Smallbone]

I believe from what I understand is the luminance layer is also based on the blue layer and not a true luminance image. By that I mean all visible colors forming the luminance  For example for my cooled astronomy camera I have a luminance filter which is clear and lets in the visible spectrum for the most part, then I have separate filters for R,G, B. I can shoot the color at a lower resolution, by binning, and then use the Luminance for the detail.

For the Quattro it looks like they may be generating a synthetic luminance from the blue and the essentially binned red and green layers and generating the luminance that way might lead to some loss of resolution or blurring. Just a guess.

Like every camera and image sharpening always seems to be different, at least that is what I find. So yes it does take some different methods to prevent artifacts.

Alan

[RobertJ]

Quentin or someone, can you do a test of RAW vs. JPEG out of the Quattro?  It seems every comparison I've seen shows the Quattro JPEGs being way too superior to the converted RAWs.  Now, this might be because they had NR turned up in SPP, but I don't know.

So, in-camera, turn off any noise reduction (or set it to lowest setting), set sharpening to zero, take a JPEG + RAW, and process the RAW with the same settings and compare it to the JPEG.  This would be very helpful...

[capital]

Ok, so I have been doing some more digging on the A/B Merrill/Quattro image series I took.

I have two versions, one at 100%, it shows the Merrill pulling ahead of the Quattro, but then when you zoom in to 200%, your perception changes, and you can actually begin to see why the Merrill looks it is resolving more... except it is not, it actually not resolving the small yellow fruits you thought it was resolving at 100%. Instead, the Quattro actually is showing the roundness and discerning them better.

Note: These were taken from SPP 16 bit TIFFs all NR and sharpening at the lowest positions.

[capital]

So, in-camera, turn off any noise reduction (or set it to lowest setting), set sharpening to zero, take a JPEG + RAW, and process the RAW with the same settings and compare it to the JPEG.  This would be very helpful...

Hi Robert,

I am uploading something now for you to obtain.

By downloading these files you agree to use them for private non-commercial viewing only, please do not redistribute the files.

SDIM0027.JPG is out of camera, SDIM0027-2.JPG has been exported from SPP and further processed in LR.

[Quentin]

The more I dig in to Quattro files, the more astonishing micro detail  becomes apparent.   Upscaling using photozoom pro  and Lanczos rivals S-HI direct out of SPP.  Increased crunchiness can be achieved using local contrast enhancement for a more Merrill type look, but if that is what you want, better off using a Merrill.

My biggest issue is low level purple blotchiness.  This seems to affect areas of water and some shadow areas, particularly even greens.  We saw this also with the Merrills.  I hope Sigma can mitigate it.

[RobertJ]

Thanks capital.

[capital]

Imaging Resource updated their online gallery with more images, also including some raw files too.

http://www.imaging-resource.com/PRODS/sigma-dp2-quattro/sigma-dp2-quattroGALLERY.HTM

[capital]

Is there any difference in bleeding between a Merrill & Quattro?

Ratio is 8:1.

Note: I raised the shadows on both to +100 to expose into the shadows, an area thought to be bad for dark reds.

[capital]

Tried to do a macro comparison.

I have to give the nod to the DP2M, it has better user ease of use when zooming to 10x. For some reason Sigma Corp have regressed here and only allow 8x magnification on screen for the DP2Q. The difference in nailing focus is tremendous when dealing with minute details. I tried this setup on two different days taking a number of shots on each camera, this was the better result for the DP2Q. I tried both 8x manual focus and auto focus, there might still be improvement focus-wise, but the DP2Q seems harder to manage getting a good hit ratio versus the DP2M. I think perhaps the only thing going for the Q is color rendering of this scene.

So some detail about processing, exported from SPP with all NR sharpness settings at minimum values. Took both Q & M files into Photoshop and applied two rounds of sharpening first at higher radius then a lower radius of about half the first pass. Then on the Q file to match the "grit" I did a low level % unsharp mask w. a radius of about 9 pixels.

I have another scene waiting to be processed that has a lot of fine red details in highlights and shadow that I hope to post soon, perhaps someone out there wants to give the x3f raws a try to make sure I am doing this processing as fairly as possible.

[Quentin]

I slightly prefer the Quattro shot, but honestly, its a matter of taste.

I have a close-up lens with my DP2M

Tried to do a macro comparison.

I have to give the nod to the DP2M, it has better user ease of use when zooming to 10x. For some reason Sigma Corp have regressed here and only allow 8x magnification on screen for the DP2Q. The difference in nailing focus is tremendous when dealing with minute details. I tried this setup on two different days taking a number of shots on each camera, this was the better result for the DP2Q. I tried both 8x manual focus and auto focus, there might still be improvement focus-wise, but the DP2Q seems harder to manage getting a good hit ratio versus the DP2M. I think perhaps the only thing going for the Q is color rendering of this scene.

So some detail about processing, exported from SPP with all NR sharpness settings at minimum values. Took both Q & M files into Photoshop and applied two rounds of sharpening first at higher radius then a lower radius of about half the first pass. Then on the Q file to match the "grit" I did a low level % unsharp mask w. a radius of about 9 pixels.

I have another scene waiting to be processed that has a lot of fine red details in highlights and shadow that I hope to post soon, perhaps someone out there wants to give the x3f raws a try to make sure I am doing this processing as fairly as possible.

[joofa]

I agree that the top layer doesn't strictly record Luminance, ...,  still makes a nice, not accurate (because overweighted in Blue and Red) but nice, substitute for Luminance from which a useful contribution to the other channels can be calculated. Blue wavelengths are also relatively the least affected by diffraction, and the top layer exhibits the least diffusion/scatter.

Hi,

I used a linear model that is a variant of polynomial regression to obtain an approximation of CIE Y (luminance) from the Quattro curves. The Quatro sensitivity data was graciously supplied by Ted (XpatUSA). It seems that one can do a reasonable approximation to Y as shown below:

For more info please see the following threads:

http://www.dpreview.com/forums/post/54094414
http://www.dpreview.com/forums/post/54095986

Joofa