Sigma Fp review and interview / Cinema DNG RAW

[imagesfromobjects]

5 hours ago, paulinventome said:

which version of the Elmarit - the newest one? Can you do me a favour and let me know what the corners are like at infinity on the fp?

cheers!
Paul

When my new fp arrives, sure. I was using a rental, but the corners showed some cyan - about the same as on a7S. Its 11606, 1st Asph version, so not optimized for digital, though still great by most metrics. Are you asking about field curvature at infinity? I normally shoot at <3m but there's minor FC wide open, which is masked around f/5.6. No "smearing", though, but you one could easily conflate FC with the corners being soft. They're indeed sharp from wide open, assuming one is *focusing* on the corners.

 

[paulinventome]

20 minutes ago, Scott_Warren said:

Thanks for being a fountain of information, Paul! The obsession for seeing the ground-truth image is very real, haha. I've dabbled in baby steps of color management with my work in games, so it's great to be made aware of how deep the rabbit hole goes.

That's really nice of you to say so, i appreciate it. I think in these times of quarantine we can all obsess a bit.

Looked at the DNG and ACES. My observations (as i don't generally use ACES)

- For DNGs the input transform does nothing and you have no choice over colourspace or gamma in the RAW settings which makes sense. But the question is what is Resolve doing with the data - is it limiting it to a colourspace? Is it P3? Is it 709? Is it native camera?

- This ACES version uses a better timeline colourspace. But Resolve really works best (in terms of the secondaries and grading controls) when using something like 709. If you qualify something when you have the timeline set to a log style colourspace you'll find it really difficult. I don't know enough about this set up to know whether Resolves works at its best. I would hope that being a basic setting Resolve would know about it but i made a huge mistake once setting my timeline to RedWideGamut/LogG10 and couldn't work out why everything just didn't work very well!

- The outputs are plentiful, and you should match your monitor with what output matches the best. If i switch it on my P3 display between P3 and P3 (709 Limited) then the image doesn't change - therefore in this scenario i believe that resolve is debayering into 709 space from the sigma.

- The sigma controls let you boost saturation, the white point of those images is around 9000 as you can push saturation and balance til it's mostly grey.

I thought that ACES include a RRT which is a rendering intent - like film emulation - maybe i can't find this in Resolve at the moment.

So what i would be interested in is comparing CIE diagrams for this vs other routes to see if the DNGs hold beyond 709 colours.

Clearly we are nitpicking at a level that is frankly embarrassing but i'm bored...

cheers
Paul

[paulinventome]

2 minutes ago, imagesfromobjects said:

When my new fp arrives, sure. I was using a rental, but the corners showed some cyan - about the same as on a7S. Its 11606, 1st Asph version, so not optimized for digital, though still great by most metrics. Are you asking about field curvature at infinity? I normally shoot at <3m but there's minor FC wide open, which is masked around f/8.

As far as i understand the 1st version of both the Elmarit and the Summicron had huge issues on non M cameras because the filter stacks are too thick and the rear of the lens produced very shallow ray angles. Jono Slack did a comparison with the V1 and V2 summicron on a Sony a7 and the results were very different. The performance of it on the a7 was abysmal before. Again, infinity focused.

But the fp has no OLPF and there's no data as to how thick that stack is. I've been after a 28 cron for a while, but the V2 is expensive and elusive whereas there are plenty of V1 around - but there are no tests on an fp.

The 50 and 90 are incredible on the fp but i need a wide to go with them. 

So would be super interested to see how that fairs when you do get your fp.

This is less FC and more smearing. More an issue with the camera than the lens.

thanks
Paul

[Scott_Warren]

@paulinventome

It turns out the stock X-Rite software doesn't do a super accurate job of calibrating the toe of the display curve for sRGB. It makes it too crunchy. I calibrated with DisplayCAL and can now see into the toe more accurately, so I don't need to compensate in Resolve with using the flatter P3 ODT to offset what was looking too crunchy on both my P3 and sRGB monitors. I can only imagine what a dedicated LUT box would offer!

After leaving the fp DNG as-is and then white balancing and pushing saturation right up to the vectorscope target boxes (Saturation set to 80-90 on a node, not the RAW saturation controls which behave weirdly), the AP1 primaries look pretty true to life. I took a low tech approach to adjusting the saturation by eyeballing the corner of the living room and comparing it to the monitor and it looks pretty dead-on at this point. 

Is the need for pushing saturation dramatically due to the nature of how big the AP1 gamut is compared to sRGB/ 709? No normal monitor can display AP1 so I imagine there has to be some base pushing of saturation in order to get a good starting point within a smaller gamut, unless I have that totally backward. 

[imagesfromobjects]

12 hours ago, paulinventome said:

As far as i understand the 1st version of both the Elmarit and the Summicron had huge issues on non M cameras because the filter stacks are too thick and the rear of the lens produced very shallow ray angles. Jono Slack did a comparison with the V1 and V2 summicron on a Sony a7 and the results were very different. The performance of it on the a7 was abysmal before. Again, infinity focused.

But the fp has no OLPF and there's no data as to how thick that stack is. I've been after a 28 cron for a while, but the V2 is expensive and elusive whereas there are plenty of V1 around - but there are no tests on an fp.

The 50 and 90 are incredible on the fp but i need a wide to go with them. 

So would be super interested to see how that fairs when you do get your fp.

This is less FC and more smearing. More an issue with the camera than the lens.

thanks
Paul

 

No problem at all. There is a bit of conflicting info out there re: wide angle rangefinder lenses on mirrorless, with a few factors. You already know about the sensor glass (and obviously a lot else) so I don't need to go all "101" with this BUT, some of the Sony cameras seem to handle them better than others. The a7S is supposed to be one of the best, whether due to pixel pitch or just by virtue of being lower res, who knows. The 1st-gen a7R was supposed to be the worst, and the a7 somewhere in between. It also very much comes down to one's specific use case, whether the potential tradeoffs are worth it. I ended up with the 11606 after using a quite few other 28's - one of my favorite focal lengths - because of its size, handling, color and tone rendition. I used it happily on the a7S for a year or so, even after reading how "terrrible" it was supposed to be, even on the a7S. So, basically, YMMV, but here are a couple uncropped stills I shot, with no color shade corrections - click for larger versions:

The corners show some cyan, no doubt. Whether its' enough to bother you depends on you. It was pretty much the same on Sigma fp. As for more info, there's some really good stuff in this thread on Fredmiranda.com, relating to specific M lenses on the fp. so maybe peruse at your leisure:

https://www.fredmiranda.com/forum/topic/1619170/0

 

 

 

 

 

 

[Chris Whitten]

All the test clips I recently uploaded used the Leica M 50mm f1.4 (1980's version).

I can take a shot using my Elmarit 28mm f2.8 (pre-asph). I think it was vignetting wide open.

 

I also have a Colour Checker chart I can shoot if need be.

 

[BenLemmid]

Hi,

did you notice, that Sigma silently released the v1.02 firmware yesterday?
https://www.sigma-global.com/en/download/cameras/firmware/

There aren't any official news / announcements towards the delayed update - I had to search for it actively.
I really hope, that it will solve the video fragment problems.

B

[Chris Whitten]

Its was released on March 18th and most of us jumped on it.

[paulinventome]

10 hours ago, Scott_Warren said:

@paulinventome

It turns out the stock X-Rite software doesn't do a super accurate job of calibrating the toe of the display curve for sRGB. It makes it too crunchy. I calibrated with DisplayCAL and can now see into the toe more accurately, so I don't need to compensate in Resolve with using the flatter P3 ODT to offset what was looking too crunchy on both my P3 and sRGB monitors. I can only imagine what a dedicated LUT box would offer!

After leaving the fp DNG as-is and then white balancing and pushing saturation right up to the vectorscope target boxes (Saturation set to 80-90 on a node, not the RAW saturation controls which behave weirdly), the AP1 primaries look pretty true to life. I took a low tech approach to adjusting the saturation by eyeballing the corner of the living room and comparing it to the monitor and it looks pretty dead-on at this point. 

Is the need for pushing saturation dramatically due to the nature of how big the AP1 gamut is compared to sRGB/ 709? No normal monitor can display AP1 so I imagine there has to be some base pushing of saturation in order to get a good starting point within a smaller gamut, unless I have that totally backward. 

Even calman isn't perfect. The guys at Light Illusion really, really know their stuff and some of the articles on their site can be really useful. I tend to profile a monitor with at 17 cube and then use that as a basis to generate a LUT for different colourspaces. So even on a factory calibrated 709 display (Postium reference monitor) i found that i still needed a LUT to get the calibration the best i could. I even used two different probes. And sanity checked by calibrating an iPad Pro (remarkably good display) matching that and then taking that to the cinema and projecting the test footage with the iPad so i can eyeball. Once i know i have devices where i can be pretty calm about things then that's a good place to start. I work from home and i'm not a dedicated facility so do the best with what i have.

One issue with the iDevice displays though is that OLED is an odd creature. You will find that only OLED can create full saturated colours in shadows. So if you watch a profile on, say, 100% Red then as the brightness decreases the red stays in the same chroma coords whereas on pretty much all other display technologies the chroma will desaturate.

I mentioned the saturation as a way to eyeball the white balance. push it up and you see colour casts very easily. 

I found your DNG looked perfectly natural with default settings.

As to the question of what is Resolve doing with the fp colours i still don't 100% know. I think it does the same in ACES as DNG. In your case i think you were using an AP1 space and if you saturate in that then you can push the colours to the edge of AP1 which is expected? Especially via a node because the sigma colours start off as a small portion of AP1 but the Resolve node can push those within that colourspace.

Working in too big a colourspace is also problematic. The original ACES (AP0?) was too big as a working space as the grading controls would be too heavy handed as they had a huge space to have to work within. The original AP0 was considered an archiving space, not a working one. Math also can work differently. There were some great examples of math failing in unbound RGB and even P3. So going with the biggest is not always the best!

I think to see what the camera is doing would be a case of shooting some super saturated colours, beyond pointers gamut (which is the gamut of natural surface colours) and then taking that DNG into Resolve. Doing it as 709, then P3 and comparing whether a) it looks the same save for some parts and b) do the CIE diagrams clearly show colours beyond 709 without any tweaks.

I wonder if i shoot an RGB Chart off my P3 reference monitor and see how that fairs?

If this is a Sony sensor then i would pretty much guarantee the Red is beyond 709 - this was an issue i had with sony sensors way back with the FS100 and 700 and the a7 series.

cheers
Paul

 

[paulinventome]

6 hours ago, imagesfromobjects said:

No problem at all. There is a bit of conflicting info out there re: wide angle rangefinder lenses on mirrorless, with a few factors. You already know about the sensor glass (and obviously a lot else) so I don't need to go all "101" with this BUT, some of the Sony cameras seem to handle them better than others. The a7S is supposed to be one of the best, whether due to pixel pitch or just by virtue of being lower res, who knows. The 1st-gen a7R was supposed to be the worst, and the a7 somewhere in between. It also very much comes down to one's specific use case, whether the potential tradeoffs are worth it. I ended up with the 11606 after using a quite few other 28's - one of my favorite focal lengths - because of its size, handling, color and tone rendition. I used it happily on the a7S for a year or so, even after reading how "terrrible" it was supposed to be, even on the a7S. So, basically, YMMV, but here are a couple uncropped stills I shot, with no color shade corrections - click for larger versions:

Thank you!

Yes, i had most of the voightlanders, from 12mm upwards on various Sonys. They were pretty good, even the 12. The sigma is a bit better with the 12 as well.

I was understanding that the 28s in particular were problematic because of the exit pupil distance in the design. It was after viewing the slack article that made me obsess a bit more about it all.

I was never happy with the voightlander 1.5 and finally decided to try the summicron. Fell in love with the look and the render. Found them matched my APO cine lenses better. So decided to sell most of the voigtlanders and just focus on 3 Leica lenses, 28, 50 and 90. 

There's no where i've found to rent in the UK to test so hence asking people!

Love the photos.

The only issue is that most of my photography may be like that, but i do also need f8 landscapes/architecture and as flat a field as possible. So looking for samples like that too!

cheers
Paul

[Chris Whitten]

Here's two DNG (not FP jpegs), Elmarit 28mm f2.8 lens. First image shot at f8 ISO100, no processing or lens correction, no colour correction in Capture One except tiny exposure balancing. Second is f4 at ISO100, again no correction or processing. Both times focussing at infinity. The f4 looks soft in the corners.

I think my 35mm f2 Summicron is a better lens, and smaller.

[paulinventome]

1 hour ago, Chris Whitten said:

Here's two DNG (not FP jpegs), Elmarit 28mm f2.8 lens. First image shot at f8 ISO100, no processing or lens correction, no colour correction in Capture One except tiny exposure balancing. Second is f4 at ISO100, again no correction or processing. Both times focussing at infinity. The f4 looks soft in the corners.

 

Thanks for doing that! It confirms that the fp is probably similar to the a7s. The f4 is clearly smeared as expected. I'd hoped with no OLPF then it would be closer to an M camera.

So it seems that it's the latest versions that i will be looking for... The 28 cron seems spectacular in most ways. 35 is too close to the 50 and i find the bokeh of the 35 a little harsher than the others (of course there are so many versions so difficult to tell sometimes!)

Thanks again
Paul

[imagesfromobjects]

On 3/25/2020 at 11:34 AM, paulinventome said:

May be a while before I can provide any infinity shots, because we're all basically on house-arrest here in Philadelphia for a while. Without any other info, I'd say go for the newest (11677) version of the Elmarit 28/2.8, because it was allegedly tailored to perform better on [Leica] digital than the 11606. It may be tricky to find a good deal on a used one, because most Leica people baby the s*** out of their gear, and try to sell it LNIB, but if you're really particular about edge-to-edge sharpness at f/8 + infinity, it's probably your best bet. The 11606 does well enough for me, but if you need edge-to-edge, it's probably best at f/8-11, and one may need to pull back just a hair from the infinity hard stop to mask FC.

 

My fp arrived today (sweeeet!!) so I'll have some time to do testing, but I can't promise infinity.

 

 

[paulinventome]

11 hours ago, imagesfromobjects said:

My fp arrived today (sweeeet!!) so I'll have some time to do testing, but I can't promise infinity.

Some small compensation for the lock down. I've been taking mine out on our daily exercise outside. It's pretty quiet around here outside as expected. Dogs are getting more walks then they've ever had in their life and must be wondering what's going on...

I am probably heading towards the 11672, the latest summicron version. I like the bokeh of it. As you say, getting a used one with a lens this new is difficult.

Have fun though - avoid checking for flicker in the shadows and enjoy the cam first!!

cheers
Paul

 

[Chris Whitten]

FWIW, three M lens shots (DNG) at f4. 

28mm, 35mm and 50mm. There are older 1987-92 era lenses, pre-asph. Focus is at infinity. Both the 28 and 35 show smear in the corners at f4.

Even trying to expose correctly using the FP light meter and histogram, I can comfortably drop my shots 1 stop in Capture One.

Without any FP profile in CO, and no colour correction myself, the default is a bit bluish/purple in the shadows imo.

[Scott_Warren]

@paulinventome 

I tried a test using your IPP2 workflow, and I have to say I'm now a huge fan  The only thing different is I'm using Hard highlight rolloff since I like the extra snap it gives to light sources, but the color behavior out of the gate + the OneShot chart is just fantastic. Skin is maybe a hair red compared to reality, and a Medium rolloff makes it look more correct, but at the expense of the hotness of the lamp. To my eye, this is what the corner of my TV console looks like in reality. It's actually a bit scary how quickly this approach gets to a solid image, haha. I feel like I'm cheating!

IPP2 kind of feels like RED's own baby ACES in that you can control global aspects as you wish. It's super powerful. Huge fan of global operators in place of knob parties!

 

[Chris Whitten]

I will try that at some point (soonish).

[Scott_Warren]

Definitely play around with it, Chris! The highlight roll off control alone allows for range compression depending on what's going on. Kind of seems like a local highlight tonemap operator.

For this shot, I used a Soft rolloff to help the roof lights transition more smoothly into the fog. It's acting like an extra diffusion layer on top of the natural fog scatter. 

 

[Scott_Warren]

More stills from the progression of the pan. I'd upload a video, but the motion was nothing to write home about. Really just sweeping around finding interesting compositions at random. 

There's something appealing about the behavior of blues and teals using IPP2. It has a certain iciness to it that's hard to describe. 

[redepicguy]

Gosh I love this cam! Shot on vintage Nikon F Mount glass. 😉 Gosh it compresses the heck out of the image on this site. haha