Sigma DP0 Quattro shots

[tagscuderia]

Hi Tom
Are you able to explain how to correct the cyan cast to produce better blues in Lightroom? Also, how do you profile the camera with DNG?
Lee

Hi Lee,
I don't use Lightroom and looking at ACR there doesn't appear to be a global hue adjustment, just per channel and that makes it more complex to correct
Regards camera calibration, this link gives you an idea as to what's required... As an enthusiast only, that's way too much trouble for me when SPP delivers superior results every time. But it will be useful for many.

 

[biglouis]

Louis, as an objective observer I'd say the difference is huge !
The iX3 file is much cleaner: f.i. look around the No Parking text.

Kind regards.

To my tired eyes the difference is so minor I am not sure whether or not my eyes are tricking me. But I'm glad to hear there is a difference (but as you'll see from my comment below I doubt I will use the functionaity in any case).

SFD is auto-bracketing; it's taking 7 shots total, 3 either side of the base exposure. The X3I is a container for the 7 X3F files that can be individually extracted.
Unsure as to why you expect the TIFFs or JPGs to be larger
The SFD image has visibly better IQ in your example thanks to much lower noise. Noise and dynamic range being the raison d'être of SFD.

OK, well I guess I'm a little disappointed that Sigma put their effort into creating SFD when what I would have preferred is for them to broaden the scope of the existing auto-bracketing which I confirmed this morning works with the dng format. For information, the most comprehensive auto-braketing on any camera I own is on the Panasonic GX8 where you can step between 3 to 7 frames in user selectable EV step. The reason why I would prefer to have a richer auto-bracketing function rather than the specialised SFD is that I use HDR Effex in LR6 and it works very well indeed. I'm way too lazy to break the XRIs into their 7 components and export each one as TIF to LR and then to HDR Effex. Too slow and too many steps.

Now I also understand why the files do not change in size. SFD is in effect in camera HDR. I'm not impressed so far with the processing that the X3I files undego when they are put into the editing mode in SPP. I have a powerful PC and the processing is not fast enough to be called 'slow as molasses'. More like 'slow as molasses frozen to below zero...".

But the good news is that I am getting some excellent results from dng, so well done Sigma.

LouisB

 

[tagscuderia]

I agree. I'm unlikely to even bother testing SFD.

N.B, an X3I is like an EXR/HDR file, it requires tone mapping. SPP's defaults nor AUTO will get you anywhere.

 

[furtle]

Hello Louis

Try your test in low light indoors. The X3I file will be much better than any single X3F file. Also, I think the base EV selection is very important. As said above it is easy to deselect a rogue image in the X3I stack before processing but if you deselect too many, the noise comes back.

The EV steps are 1.0 each and I would like to be able to change that in the menu to a smaller number, say, 0.6 or 0.3 EV. Then I believe far more of the individual shots would be useful.

Going back to your test, this was in good light and in this situation, I'm with you; I don't think the advantage of the X3I over a single well exposed X3F is huge. But, once shadows get into the frame the X3I advantage increases.

 

[furtle]

I'm revisiting this post and now the dp0Q had SFD, the interior photos would have been significantly better using SFD. In fact, it was a gloomy day and I think the exterior photo of the house would have been better as well. Just gotta get the base EV right.

Chastleton House, Oxfordshire and near Moreton-in-March but more accurately named Moreton-in-the-Minge. A lovely and interesting Jacobean pile, lived in until the mid 1950s but now owned by the National Trust as the increasingly impoverished owners reluctantly accepted the cash was running out. Yet again a grey day and bravely, I took some indoor, low-light (ooooh noooo!) shots with the ever fickle dp0Q.

 

[biglouis]

Hello Louis

Try your test in low light indoors. The X3I file will be much better than any single X3F file. Also, I think the base EV selection is very important. As said above it is easy to deselect a rogue image in the X3I stack before processing but if you deselect too many, the noise comes back.

The EV steps are 1.0 each and I would like to be able to change that in the menu to a smaller number, say, 0.6 or 0.3 EV. Then I believe far more of the individual shots would be useful.

Going back to your test, this was in good light and in this situation, I'm with you; I don't think the advantage of the X3I over a single well exposed X3F is huge. But, once shadows get into the frame the X3I advantage increases.

Steve, thanks for the information, that would an important benefit. I'll definitely try it as and when I have a need.

LouisB

 

[Malina DZ]

Could someone share images with the sun in the frame at various apertures between f/4 and f/11?
What is the max shutter speed at f/4?

 

[TN Args]

Sigma DNG files

Someone on DPR was showing this as well, but also noted that if you increase the color NR to 50 or so in LR, the two look very close.

That would be me who you are quoting.

...Louis, I noticed something interesting--the leaves on the shadow side of the building, near the corner look like the color gets completely lost against the brown bricks in some places. You can see some texture of the leaves, but the color takes on the brown of the building. Do you see that, or am I imagining it? Seems like a Quattro artifact more than X3F vs. DNG though.

Not a Quattro artefact IMHO.

I think biglouis has gotten rightly excited about the DNG, but jumped too early to conclude its superiority to X3F. Lets try to lay down a few factual markers first, before bringing in the subjectives:

1. The DNG is 12-bit and the X3F is 14-bit
2. The DNG is uncompressed and thus approx 105 MB for this camera
3. The X3F is losslessly compressed raw and about 45-55 MB file size
4. The DNG is a linear raw file, and can be considered a true raw file, not like a TIFF for example
5. Adobe products can compress the DNG file losslessly to about 35 MB (assuming Medium JPEG size is embedded)
6. Being 12-bit instead of 14-bit should not limit it in any way except deep shadow noise. Referring to 'fine qualities' being different (other than the 'fine qualities' of deep shadow noise) is certain to be due to processing not source.

I made some preliminary comments on DNG vs X3F when the sdQH first came out, starting from here (warning, link to another site and forum)

In a nutshell, if you process an X3F in the latest SPP and a DNG in Lightroom, using default settings both times, a couple of things will go wrong and dominate early impressions:

• SPP sharpening is a bit high and will create halos (but only minor and at high magnifications of viewing) and image noise (not so minor)
• SPP NR is a little bit high once you turn down sharpening, which in some images will slightly smooth the finest detail at high magnifications
• LR sharpening of the DNG is too high because the default is tuned to Bayer raw files that are naturally soft
• This will give a first impression of wonderful detail in DNG but it is false
• LR noise reduction is too low (again, tuned for Bayer) so again provides misleading impression of high detail
• The DNG will seem blighted with high colour noise, but Colour NR of 50 in LR is a first estimate to fix it

So they might look a lot different due to default software settings, but the differences can be muchly equalized with attention to post processing, and are not actually due to anything inherent in the image files, other than less ultimate dynamic range in the 12-bit DNG which manifests itself as more noise (and less colour retention/accuracy) in the deepest shadows, and would rarely be a big issue. Especially now, with SFD mode for (static) shots with deep shadows.

 

[TN Args]

....in low light indoors....The X3I file will be much better than any single X3F file. Also, I think the base EV selection is very important. As said above it is easy to deselect a rogue image in the X3I stack before processing but if you deselect too many, the noise comes back.

The EV steps are 1.0 each and I would like to be able to change that in the menu to a smaller number, say, 0.6 or 0.3 EV. Then I believe far more of the individual shots would be useful.

Going back to your test, this was in good light and in this situation, I'm with you; I don't think the advantage of the X3I over a single well exposed X3F is huge. But, once shadows get into the frame the X3I advantage increases.

I agree, SFD is for special cases, and will yield obvious benefits in those cases.

I am more the opposite of biglouis in that I am much more likely to use SFD, than work with HDR tools, when I want critically low noise and high dynamic range.

Do you have any tips for how to select the base EV for a SFD exposure?

[edit: quote from the Sigma website about SFD Mode: The new Super-Fine Detail (SFD) exposure mode brings out the full performance of the Foveon X3 Quattro direct image sensor. One push of the shutter generates seven different exposures, creating RAW data in the X3I file format. Using this data with the SIGMA Photo Pro software package, the photographer can create noiseless images with an extensive dynamic range. With this new mode, the more detailed imaging potential of the Foveon X3 Quattro direct image sensor is fully leveraged. From each X3I file, individual X3F files may also be generated. The value of SFD exposure mode is especially apparent in studio photography.]

 

[biglouis]

^^^^ TN, that is really useful information, thanks for sharing ^^^^

 

[Sapphie]

^^^^ TN, that is really useful information, thanks for sharing ^^^^

Agreed, great summary from TN. Processing Quattro files has been a long and hard 'relationship' for me. I liked the JPEG output from the outset but shadow colour was often desaturated. SPP was originally very slow and the output way too crunchy, before they introduced that Detail slider. I have also since learned to reduce sharpening in SPP. More recent versions seem very heavy-handed with luminance NR, blurring detail in distance on landscape shots, so I now routinely move that to the left. SPP is also now much faster, especially with the GPU acceleration.

I was seriously thinking (again) of selling my Quattros. Then the magical DNG option, which I had hoped for, arrived sooner than expected. Not had much time to experiment yet but initial impressions are good. Also, worth noting that LR has some of the Quattro colour modes available in the Profile drop-down of camera calibration section. I seem to be starting with Portrait as a good base. As TN says Adobe can compress the DNGs to half the size, either by just running them though Adobe DNG Converter or in LR by selecting a DNG in Library view and then selecting Menu option Metadata/Update DNG Preview and Metadata - I am not sure exactly what else the latter does though.

What I am not sure about is WB - I assume those in the drop-down in LR are Adobe's presets and not grabbed from the DNG?

Final question (not specifically related to DNG) - can anyone explain why the Sigma Monochrome mode always renders the image 0.5-1.0 stop underexposed compared to colour? You can see it happen with this Profile selected in LR but it happens in SPP too.

One day, maybe Sigma will give us the option to save DNG and X3F at the same time!

Lee

 

[biglouis]

Some additional thoughts from me on the new dng option. But first, what it is all about, the photograph.

Ironic, or iconic? John Gay was a British photograph (actually born a German but emigrated to the UK in the 1920s) who lived and worked where I live in north London and shot many superb photographs around the King's Cross area. The elevated photograph taken from the height of the top of an old gas street lamp on Midland Way looking towards the gasholders on Good's Way taken in the early 1950s is classic in terms of the black and white MF photography and the time it captures. Since the early 'naughties' of this century this entire area has been redeveloped. The iconic Victorian gasholders - some of the best examples of their type - have been moved to a location about half a mile away on the canal and now adorn the exterior of circular up-market appartments - called Gasholder Park (worth a visit if you are in London).

Seeing Gay's work on a billboard is wonderful but I am not sure if it is appreciated by the casual passer-by. I wanted to record the coincidence of his work being displayed on the very road of his subject, Good's Way, as it now in 2017.

This is taken from the temporary canal path beside the construction works on the canal, you can see the tops of the new 'Gasholder' appartment blocks in the distance.

The tranquility of the Regent's Canal early on a sunday morning when no one else is about! The main lines into St Pancras station go over the canal at this point.

Bottom line, I am impressed with what I can get out a dng file. Working in Lightroom is so obviously superior to SPP it doesn't really need to be stated and being able to do so has rewarded my perservering with the DP0Q. Personally, I think it helps to support the argument that the 14mm lens on the DP0Q is a very good optic indeed. I've owned many 21mm lenses and most the same or sometimes several times more than a DP0Q, so as someone who is inherently cheap(!) I am satisfied (at last) with my purchase. I'm going to keep shooting dng until I discover whatever bugs or problems reveal themselves. You never know, maybe Sigma have finally got it right.

Just my two cents

LouisB

 

[octagone]

Re: My new Sigma DP0Q

I feel like I have been given a brand new camera. A different camera. One with a great 21mm lens and superb sensor. Why, oh why did Sigma not do this sooner????

Thankyou, thankyou, thankyou, Sigma for finally giving us dng!

My feeling is dynamic range is improved, sharpness at 100% is maintained, less appareance of flattening of subjects, better texture handling, better colour in shadows.

A few files for you to play with. Tell me if I am fooling myself, please. There are links to full size jpegs exported from LR and the original DNG files to play with (but be aware the DNGs are 100MB files). Right click and 'save as' to download the files

NOTE: please do not share the files - direct any interested parties to this post.

Full sized jpeg | dng

Full sized jpeg | dng

Full sized jpeg | dng

I have done some treatment trials with RawTherapee, with the first and second photo. I get almost the same results as the examples you posted.

My opinion is that the pictures in DNG format has less noise, and color drifts than the X3F, and the details are extremely fine.
But after doing my tests of treatment with RawTherapee on your examples and those of Dpreview, it seems easier to achieve good results from an X3F. When opening with SPP, portrait mode, NR luminance, default value SPP. NR chrominance 0. Sharpness 0, see -4. And details completely left.
And then by processing the TIFF with rawTherapee.
There are more noises and sometimes (not often) there is still a drift of red and green colors in the TIFF, but it corrects very well with RawTherapee. And it's easier to have a crunchy rendering.

The DNG requires more processing, to have the right colors, as well as sharpness and micro contrasts if you want a crispy rendering (it is more difficult to bring out the micro details). In addition there are color drifts in some details (more present with dpq, at least on your examples, than with SdqH).

An example with the Dp2q, X3F - SPP portrait mode, RawTherapee.
Full res on Flickr
SDIM2827-NEW-Bis by Noël Billy, sur Flickr

29Mpix:
SDIM2827-28Mpix-1-F-1 by Noël Billy, sur Flickr

 

[tagscuderia]

What I am not sure about is WB - I assume those in the drop-down in LR are Adobe's presets and not grabbed from the DNG?
Final question (not specifically related to DNG) - can anyone explain why the Sigma Monochrome mode always renders the image 0.5-1.0 stop underexposed compared to colour?

The presets in ACR (so I'll presume Lr as well) are generic. And judging from the "As Shot" WB in the OOC DNGs and X3F metadata, SIGMA don't record Tint — that's likely the reason behind the shift to cyan so the correction might be easier still.

The difference in EV is down to the gain applied to each layer for colour conversion and/or the matrices to go from camera space to RGB. Ted/xpatUSA is the man to ask regards that though!
In SPP with Merrill you can also observe the change in EV for each layer, the bottom i.e, 100% red on the mixer always being less exposed.

Cheers,
Tom.

 

[Sapphie]

My first 'fail' with DNG (using dp2 Quattro) whilst doing some test shots - buttercups amongst some greenery, half of the petals had overexposed to white. I don't know why the Quattro sometimes does this, especially with flowers in the sun, but I do know that with X3F I could probably have recovered with the overexposure compensation tick box.

Lee

 

[ggibson]

My first 'fail' with DNG (using dp2 Quattro) whilst doing some test shots - buttercups amongst some greenery, half of the petals had overexposed to white. I don't know why the Quattro sometimes does this, especially with flowers in the sun, but I do know that with X3F I could probably have recovered with the overexposure compensation tick box.

Lee

Isn't that overexposure compensation just an intelligent fill option?

 

[Stoneage]

Isn't that overexposure compensation just an intelligent fill option?

Kind of. The dpQ sensor has "overexposure pixels" on the sensor that are less sensitive. So SPP tries to fill blown out areas with the color information it gets from those pixels.
It can work on even surfaces, like skies. But mostly i don't find it satisfying.

 

[biglouis]

My first 'fail' with DNG (using dp2 Quattro) whilst doing some test shots - buttercups amongst some greenery, half of the petals had overexposed to white. I don't know why the Quattro sometimes does this, especially with flowers in the sun, but I do know that with X3F I could probably have recovered with the overexposure compensation tick box.

Lee

I should have made it clear in my initial enthusiastic meanderings that the problem with the Quattro blowing highlights has not been 'solved' by introducing dng. Nor would one expect it to be as it is a function of the sensor and not the software. I'm only hoping that Sigma are not in denial about this problem with their sensors as this is something which really does need to be addressed in future products.

I don't know if the propensity to blow highlights has been reduced with the new SD Quattro cameras - especially the H.

LouisB

 

[tagscuderia]

The highlight "blowing" issue is a direct consequence of the Analog Front End, not the sensor per se; see SD14 vs SD15 for example, so SIGMA are definitely aware of it! Given the tick-tock nature of SIGMA's past, Quattro mkII will probably be AFE-less and thus won't suffer the same issue.

P.S, Kalpanika doesn't clip as severely as SPP but does require tint correction. So SIGMA potentially clip at the point at which the data is unreliable/inaccurate. If they clip the DNG similarly then that's a shame seeing as other software feature powerful highlight recovery options.

 

[furtle]

Collioure, Languedoc-Roussillon region of France. Pyrénées-Orientales and close to the Spanish border easter Pyrenese. Some exceptional wines being made there now in spite of Languedoc's reputation for filling up the European wine lake with industrial grade plonk.