Sigma Fp review and interview / Cinema DNG RAW

[Waynes]

On 11/10/2019 at 2:27 AM, BTM_Pix said:

The Acsoon Cineeye and a cheap Android phone could give you that sort of thing.

It has a lot of good features like grayscale, RGB, monochrome, false colour, focus peaking, LUT loading etc and the signal can be monitored by up to 4 devices simultaneously.

Decent range and latency as well.

I'm going to do a quick write up about it sometime in the week as whilst my primary interest in it was live view remote with my app for the Pocket4K it has some potential for the FP as well.

With the Cineeye, these 2.5inch £50 Android phones could potentially be combined with a loupe to make a wireless self powered version of the LS300 lash up that I did.

Just need to make sure that the phone has enough grunt to run the Cineeye app, which I doubt will be an issue but the resolution might be.

Yeah, been looking at these to use as a little camera, but only 5mp and who knows what sensor.  :). Bought another phone instead.

[kye]

On 11/10/2019 at 3:21 AM, Llaasseerr said:

Right. It would be great if there was an option to use the "printer lights" feature as a multiply (gain) operation in linear space in Resolve. That's how it is in Nuke, for example.

I haven't checked, but I'm thinking there are probably several ways to get Resolve to do this.  It's a bit like photoshop in this sense - being so powerful that it can do anything but that nothing is immediately obvious.

What context would you want to do this?  I can probably suggest one or two approaches..

[The ghost of squig]

On 11/10/2019 at 1:05 AM, paulinventome said:

Reading between the lines of how they're downsampling 6K to 4K i suspect it will help with moire situations as well. Certainly in the DNGs i've seen, with water and highlights it seems remarkably smooth. C5D had a frame with moire though. One thing i find out the image so far (from samples) is that it's naturally quite soft, in a nice way. Not digitally sharpened or anything. The sort of things i like about Reds for example.

From what I've seen/heard it's line skipping/pixel binning, not downsampling, that plus no OLPF is a recipe for moire/aliasing.

On 11/9/2019 at 2:32 PM, Brian Williams said:

I don’t know, a lot of well received cameras don’t have them, the Pocket 4K being one. I haven’t seen any footage of mine from the FP that concerns me.

The Pocket 4K has a 4K sensor, there's no line skipping or pixel binning so aliasing is minimal. The thing with aliasing is it's always there lurking and you can never be sure when it's going to rear its ugly head. I've had aliasing wreck an important shot with the Pocket 4K, thankfully it's the only one.

[Royalfrakk]

Would anyone upload a 4k 12bit sequence for testing purposes, please? (Maybe some stills as well with a high dynamic range scene) Thank you!

Thank you for this as well! It was nice to test it out.

[paulinventome]

On 11/9/2019 at 3:29 PM, BTM_Pix said:

I've got the same one on my Sigma DP 0 and its very serviceable to be fair.

 

I'm sure that optical one works well. Just that it's not really in line with having a small FF camera. I can imagine for video work it makes more sense because you have a point of contact to help but for day to day stills not so much. 

The 3rd party ones are all too big, i'm looking for something like the Sony EVF, a tiny attachment that wouldn't be out of place on that body. But i cannot see a HDMI based one, maybe the HDMI hardware is too big.

I think once used to an EVF it is difficult otherwise, both for stills and cine shooting - especially in daylight.

I fired off a bunch of tech questions to sigma and they've apologised for taking so long to get back to me, so hoping to hear soon. Then i will decide whether to grab one and see for myself

thanks
Paul

8 hours ago, The ghost of squig said:

From what I've seen/heard it's line skipping/pixel binning, not downsampling, that plus no OLPF is a recipe for moire/aliasing.

 

What have you heard and from where? 

It could be the case but the Jury is out on that. I am assuming the mode they're running the sensor in is 6K FF at 12 bit (from the sensor spec sheet).  If they were binning/skipping i'd expect better frame rates. So i hope that they are reading the whole sensor, then downsampling then bayering and storing - hence the lack of OLPF. I think their bottle neck is data transfer speeds - hence thee 8 and 10 bit DNGs.

Guessing this is the beast

https://www.sony-semicon.co.jp/products/common/pdf/IMX410CQK_Flyer.pdf

You can see the HD and 4K modes but i have no idea what they could actually be doing inside. Also from that spec sheet it isn't clear if Active Pixels mean a crop of the sensor - i sort of assumed it did. 

cheers
Paul

 

[BTM_Pix]

On 11/11/2019 at 9:28 AM, paulinventome said:

I'm sure that optical one works well. Just that it's not really in line with having a small FF camera. I can imagine for video work it makes more sense because you have a point of contact to help but for day to day stills not so much. 

 

Yeah, it was just a general comment on the quality of it with a typical Sigma LCD. 

The DP0 is predominantly a landscape camera so its a slower pace where it doesn't get in the way in the same respect. Although with the Foveon's achilles heel of being a low ISO camera the point of contact is welcome when you're fighting shutter speeds all the time.

Video use on the DP0 is moot

On 11/11/2019 at 9:28 AM, paulinventome said:

Yhe 3rd party ones are all too big, i'm looking for something like the Sony EVF, a tiny attachment that wouldn't be out of place on that body. But i cannot see a HDMI based one, maybe the HDMI hardware is too big.

I think once used to an EVF it is difficult otherwise, both for stills and cine shooting - especially in daylight.

Yeah its a perennial topic on here, usually punctuated by me moaning about why I can't have small tilting EVF like my Leica ones.

The Leica T/TL is in the same vein as the FP and whilst its perfectly usable as is, the addition of the external hotshoe mounted EVF is transformative.

When you look at the PCBs for these cameras (and those of Olympus who provided Leica with the early version for the X Vario etc ) it is tapped from the same place as the rear LCD. 

Keeping it in that MIPI fornat gives a lot more options in terms of physical screen size but also in overall form factor as you don't need the converter board or additional powering source as you do with an HDMI one, hence why the other cameras such as Leica, Olympus, Canon and Sony can have such neat little packages.

When your only outlet from the camera is HDMI then you are a bit stuck, although you can get HDMI>MIPI converter boards so you can then roll your own solution but the drawback is the board is roughly half the size if not more than the FP itself so it becomes a bit self defeating.

It would have made perfect sense for Sigma to have included an EVF option and its a bit of a headscratcher why they didn't. There is no standard for the multipin connector that carries the signal and power but they still had the option of making their own (they have EVFs in their SD Q/H range so its not alien technology for them) or licensing the one from Olympus.

The reasoning could be that they have something USB-C based, which is plausible as it could be compact but of course rules out one of its primary features, or that, more, likely they have a different model coming with an inbuilt EVF.

Pretty much like the Leica CL is to the Leica T/TL.

[Lars Steenhoff]

 

 

[paulinventome]

20 hours ago, BTM_Pix said:

The Leica T/TL is in the same vein as the FP and whilst its perfectly usable as is, the addition of the external hotshoe mounted EVF is transformative.

 

I can imagine. I find it very difficult and unnatural to shoot via a screen these days. I agree and find it an odd move to not even make a nod to an EVF connection at a future point. I think that will be a negative for stills photographers and it's a bit expensive for the point and shoot crowd. But once i hear back from sigma about the nitty gritty of what they may alter in firmware with the recording formats i will decide whether to grab one as a B/C Cam.

How bright is the screen?

@Lars Steenhoff That's a nice looking video for once, sigma should really push that one. Although i see flicker in some of the longer shots and the grade is a bit all over IMHO but certainly one of the best ones so far.

cheers
Paul

[BTM_Pix]

4 hours ago, paulinventome said:

How bright is the screen?

Sigma don't publish figures so its all a bit subjective but as way of a quick crude illustration, its a pretty bright day here today so here is the DP0 alongside a Panasonic GX80 which is roughly representative of a lot of average LCDs.

The DP0 is significantly brighter (which it needs to be as unlike the GX80 it doesn't articulate at all) and whilst I prefer using it with the loupe, I've got no issues with the brightness using it in these sort of conditions.

You'll have to get someone to do something similar on an actual FP for a more definitive illustration but bear in mind that the LCD on the FP is two generations on from the DP0 (2.1million dots versus 900K dots) so I would expect a brightness bump to have accompanied the resolution increase.

[mechanicalEYE]

Still emphasis with a few video samples but a good look at the camera.

 

[Llaasseerr]

On 11/10/2019 at 5:46 PM, kye said:

I haven't checked, but I'm thinking there are probably several ways to get Resolve to do this.  It's a bit like photoshop in this sense - being so powerful that it can do anything but that nothing is immediately obvious.

What context would you want to do this?  I can probably suggest one or two approaches..

Yes there are - I know of a few approaches already, we are really talking OOB though. Yes Resolve is powerful for colour, but it does not make that power explicit and precise like Nuke. It's dumbed down in some ways.

Using Gain is the obvious method if you can't use the Raw tab (Prores footage?). As I mentioned, I think using an ACEScc log curve will make the printer lights behave the same as a Gain - I need to check, it's been a while. I personally dont't know though what is a 1 stop increment wiht printer lights. Using Gain is easier, you double it for each stop up or halve it for each stop down. But of course, there should just be a linear Exposure mode in stops that can be toggled right there on the panel as a fully fledged part of the UI. It would require that you tell it what the input transfer function is, then it would bracket that behind the scenes.

[KnightsFan]

@Llaasseerr making linear gain adjustments is easy in resolve, just use a CST node to go to linear. Its very explicit. Or am i mistaking what you are trying to do?

[kye]

2 hours ago, Llaasseerr said:

Yes there are - I know of a few approaches already, we are really talking OOB though. Yes Resolve is powerful for colour, but it does not make that power explicit and precise like Nuke. It's dumbed down in some ways.

Using Gain is the obvious method if you can't use the Raw tab (Prores footage?). As I mentioned, I think using an ACEScc log curve will make the printer lights behave the same as a Gain - I need to check, it's been a while. I personally dont't know though what is a 1 stop increment wiht printer lights. Using Gain is easier, you double it for each stop up or halve it for each stop down. But of course, there should just be a linear Exposure mode in stops that can be toggled right there on the panel as a fully fledged part of the UI. It would require that you tell it what the input transfer function is, then it would bracket that behind the scenes.

Like @KnightsFan says, a node operating in Linear might do it via Gain.  The Offset control is also worth examining although I suspect it operates in Log mode.

In terms of calibrating it, take two test shots and expose one a stop up, then do the above and match the levels on the +1 shot.

In terms of being obvious, there are only so many things you can make obvious, and once a piece of software gets more features than screen real estate you need to start hiding them.  Saying that Resolve is like PS is quite true in the sense that "it can do almost anything" is true for both due to the flexibility in structure they both offer, rather than it being a 'feature' that was programmed deliberately by someone.  Resolve is also the F1 of colouring packages and the learning curve is somewhat similar - most people stall an F1 car before they've managed to make it go further than its length, and most people probably wouldn't be able to get footage into Resolve and edited and coloured and exported their first time either  

[paulinventome]

3 hours ago, Llaasseerr said:

As I mentioned, I think using an ACEScc log curve will make the printer lights behave the same as a Gain - I need to check, it's been a while. I personally dont't know though what is a 1 stop increment wiht printer lights

This all largely depends how you have colour management set up. What colourspace is your timeline set to? For example most of the grading controls, especially secondaries, are really designed to work in gamma space (like 709 or similar). If your timeline is set to a log space then pulling keys is very difficult. What you can do though is right click on each node and choose a custom space and gamma on a per node basis, this is a godsend if you're mixing approaches in a single tree.

Gaining in log space is different to gaining in Linear. Printer lights as a concept AFAIK are really a linear gain operation. In log space Add is the same as gain in Linear but log curves usually have a toe in which case all bets are off.

Whilst linear space (32 bit float) can bring it's own issues, if you are doing real world photographic operations, exposure, blurs and so on then linear is the only space you should be in.

cheers
Paul

[mechanicalEYE]

 

[Lars Steenhoff]

Nice!

[majoraxis]

That does look good - kind of like a cross between Sony sensor but with Blackmagic CNG style recording format - wait, that's what it is right? JK ; )

 Anyways, that's a very nice image for the money.

[Brian Williams]

On 11/9/2019 at 7:29 AM, BTM_Pix said:

I've got the same one on my Sigma DP 0 and its very serviceable to be fair.

I actually also used it to make a DIY low cost EVF with a £30 raspberry pi panel and a thin USB power bank with some success.

Having some experience with raspberry pi's, out of curiosity, how exactly were you able to use one to create an HDMI-in enabled EVF?

[BTM_Pix]

7 hours ago, Brian Williams said:

Having some experience with raspberry pi's, out of curiosity, how exactly were you able to use one to create an HDMI-in enabled EVF?

It was using just the LCD panel that I had for a Pi rather than the Pi itself.

Like this one

 

If you actually did want to incorporate a Pi into the equation though, you could use an HDMI to CSI-2 bridge such as this one to get the signal into it in combination with OpenCV to make a solution with monitoring functionality such as false colour, focus peaking etc.

[mechanicalEYE]