Sigma Fp review and interview / Cinema DNG RAW

[Lars Steenhoff]

Smallrig cage is planned to ship 23 December 

[Brian Williams]

1 hour ago, paulinventome said:

I don't see why alternate lines at a different ISO would reduce resolution but maybe it's not what i imagine, i don't know how many sensors let you set iso on a per line readout basis. Is it the same line being read out with two ISOs?

Its reducing it because it would, for example, shoot the even lines from the sensor at IOS 100, then take a second exposure immediately after of just the odd lines at ISO 1600, then merge them into one frame, blending the two exposures. And the sensor isn't setting the iso per line, rather Magic Lantern is taking two separate images, each using half the resolution (just even lines or just odd line).

From ML- " its a trick that samples half of the sensor at ISO 100 and the other half at ISO 1600, for example. If you mix these two, you can get almost the entire dynamic range the sensor is capable of (around 14 stops)."

[kye]

Noam Kroll did a review, and some test footage...

https://noamkroll.com/full-sigma-fp-review-4k-raw-video-samples/

 

[Brian Williams]

17 hours ago, MikhailA said:

"after powering off the camera, or after power save sleep, the camera loses basic settings such as video resolution, codec, audio recording level"  - is that true?

17 hours ago, rawshooter said:

Unfortunately yes.

So I just now got around to testing this- it seems that turning off the camera vs letting the auto power-save put the camera to sleep are basically the same thing, have the same effect on settings. If you in one of the Custom modes (C1, C2, C3) and you change a setting and let the camera turn off, then yes, when you turn it back on the settings will have changed to their default settings; which seems right to me, that’s how all programmable custom modes work on cameras, the whole point of them is to remember a specific setup of settings. The difference here is that the “sleep” mode is actually turning off the camera, or doing all the same things that turning off the camera does, instead of just pausing the camera as most cameras do when going to sleep.

But in any of the other modes, PSAM, anything you set stays set that way the next time the camera turns on, whether you turned the camera off or if it went to “sleep”.

12 hours ago, rawshooter said:

The problem is that you need custom modes for zebras to be visible.

This isn’t true, you can set any of the display modes to have zebras on, and this applies to both custom modes as well as PSAM modes. (I only now just discovered this in the menus). Again, it sucks, but you can’t have zebras and focus peaking on at the same time.

[The ghost of squig]

I was hoping the FP would be a 4K replacement for my 5D MK3 with Magic Lantern, but so far I haven't really seen anything that suggests it's much of an upgrade. Here's some sunrise beach shots taken with Magic Lantern 14bit lossless raw upscaled to 4K for comparison. The FP is definitely an improvement shooting high ISO, but it doesn't look like there's any dynamic range improvement, which is disappointing. The thing that concerns me the most is the aliasing and moire, which hasn't been a problem with the 5D.

 

[Brian Williams]

11 minutes ago, The ghost of squig said:

I was hoping the FP would be a 4K replacement for my 5D MK3 with Magic Lantern, but so far I haven't really seen anything that suggests it's much of an upgrade. Here's some sunrise beach shots taken with Magic Lantern 14bit lossless raw upscaled to 4K for comparison. The FP is definitely an improvement shooting high ISO, but it doesn't look like there's any dynamic range improvement, which is disappointing. The thing that concerns me the most is the aliasing and moire, which hasn't been a problem with the 5D.

 

Looks very nice, though I have to say that I don’t think this video shows particularly good dynamic range, though that could be the grade on it. But yes I know, you’re saying the FP doesn’t seem to show much more than this, if any improvement, I agree. Canon’s sensors aren’t typically rated the best, at least DXO-wise.

Well going from upscaled 4K to actual 4K is an upgrade. (And I know a lot of people say 4K doesn’t matter to them, but even if you only care about HD, 4K downscaled to HD will certainly have somewhat better dynamic range than native HD, if the sensors are close).

The last time I used magic lantern was maybe a year ago with the EOSM, and many years before that with the 50D, and I’m sure it’s a much much better experience with the 5D than that little EOSM camera, but the thing that drove me crazy and made it kind of unusable for me was the fact that the real-time preview during shooting was either a significant center crop or, if you wanted actual framing, the preview dropped to a low-framerate, B&W preview; both versions made it significantly hard to both compose and stay in focus. Is the 5D able to give a good real-time preview with proper composing?

[mercer]

31 minutes ago, Brian Williams said:

Is the 5D able to give a good real-time preview with proper composing?

ML Raw on the 5D3 works like any camera would. In fact, with the exposure and focus aids, it works better than most cameras. The only issue I have found is that playback is wonky. But liveview is in real time with the correct full frame image.

Here's a frame from a clip I shot a while back on the Canon 24-70mm f/4...

[The ghost of squig]

32 minutes ago, Brian Williams said:

Looks very nice, though I have to say that I don’t think this video shows particularly good dynamic range, though that could be the grade on it. But yes I know, you’re saying the FP doesn’t seem to show much more than this, if any improvement, I agree. Canon’s sensors aren’t typically rated the best, at least DXO-wise.

Well going from upscaled 4K to actual 4K is an upgrade. (And I know a lot of people say 4K doesn’t matter to them, but even if you only care about HD, 4K downscaled to HD will certainly have somewhat better dynamic range than native HD, if the sensors are close).

The last time I used magic lantern was maybe a year ago with the EOSM, and many years before that with the 50D, and I’m sure it’s a much much better experience with the 5D than that little EOSM camera, but the thing that drove me crazy and made it kind of unusable for me was the fact that the real-time preview during shooting was either a significant center crop or, if you wanted actual framing, the preview dropped to a low-framerate, B&W preview; both versions made it significantly hard to both compose and stay in focus. Is the 5D able to give a good real-time preview with proper composing?

Yeah the 5D MK3 is way better than the EOSM and 50D in every respect, IQ, DR, high ISO, and yes preview is real-time with proper framing, just like regular liveview only with a lot more display options. Magic Lantern has the same dynamic range as 14bit raw photo mode: 11.7 stops @ 100 ISO. It's a filmic grade, so the blacks are somewhat crushed. I'm not a big fan of low contrast or HDR type looks unless it's a creative decision driven by a particular narrative.

I have a Pocket 4K too, but I still like the look of the 5D MK3, the color science, cadence, and highlight rolloff are very good. About the only things I don't like about the 5D are its high ISO performance, fixed pattern noise, and limited dynamic range. It looks great upscaled to 4K. I always use a promist filter on the Pocket 4K, it looks too digital without out.

My gut feeling is: I should wait for a 3rd party OLPF before I take the plunge with the FP. I do really like the color science and the form factor.

15 minutes ago, mercer said:

ML Raw on the 5D3 works like any camera would. In fact, with the exposure and focus aids, it works better than most cameras. The only issue I have found is that playback is wonky. But liveview is in real time with the correct full frame image.

Here's a frame from a clip I shot a while back on the Canon 24-70mm f/4...

Nice shot.

[Brian Williams]

8 minutes ago, mercer said:

But liveview is in real time with the correct full frame image.

What resolution/crop are we talking about here? Because even now as I google it, everywhere I look I’m seeing it written that’s it’s as I remember it, either a b&w 7fps liveview, or a vertically stretched one. Is this something that was implemented recently? Are we talking 1:1 crop or a binning mode?
I’m not trying to question anyone’s truth here, if what you say is correct, I’d be interested in giving ML another go.

 

4 minutes ago, The ghost of squig said:

Yeah the 5D MK3 is way better than the EOSM and 50D in every respect, IQ, DR, high ISO, and yes preview is real-time with proper framing

Yes, I realize the EOSM and 50D pale in comparison to the 5D, I only mention them to explain my usage of Magic Lantern, which is pretty similar across all compatible Canons, just better resolution and framerates (and apparently Liveview too) for some than for others. But they all can do dual iso, 14bit, etc, just to different effect.

[The ghost of squig]

8 minutes ago, Brian Williams said:

What resolution/crop are we talking about here? Because even now as I google it, everywhere I look I’m seeing it written that’s it’s as I remember it, either a b&w 7fps liveview, or a vertically stretched one. Is this something that was implemented recently? Are we talking 1:1 crop or a binning mode?
I’m not trying to question anyone’s truth here, if what you say is correct, I’d be interested in giving ML another go.

It's 1080p lineskipped up to 60fps, just like regular liveview. If you wanna shoot higher res than 1080p, then it's lower frame rate and not always centred, and can be greyscale. I don't usually shoot higher res than 1080p because the rolling shutter is worse at higher resolutions, and I'm content with the IQ of the 4K upscaling and a touch of sharpening.

[rawshooter]

5 hours ago, Brian Williams said:

So I just now got around to testing this- it seems that turning off the camera vs letting the auto power-save put the camera to sleep are basically the same thing, have the same effect on settings. If you in one of the Custom modes (C1, C2, C3) and you change a setting and let the camera turn off, then yes, when you turn it back on the settings will have changed to their default settings; which seems right to me, that’s how all programmable custom modes work on cameras, the whole point of them is to remember a specific setup of settings. The difference here is that the “sleep” mode is actually turning off the camera, or doing all the same things that turning off the camera does, instead of just pausing the camera as most cameras do when going to sleep.

But in any of the other modes, PSAM, anything you set stays set that way the next time the camera turns on, whether you turned the camera off or if it went to “sleep”.

This isn’t true, you can set any of the display modes to have zebras on, and this applies to both custom modes as well as PSAM modes. (I only now just discovered this in the menus). Again, it sucks, but you can’t have zebras and focus peaking on at the same time.

Thank you so much for your help and clarification! Turns out that, reading the camera manual, I got completely confused about the difference between Custom modes (C1, C2, C3) and "Custom Mode Settings" (which only affect the display) - I had thought that "Custom Mode Settings" were about configuring the Custom modes... (Sigma's language is not exactly helpful here...) Maybe it would be good to write a practical guide for the camera, because it has its own quirks and learning curve.

- I just found out that the display for video recording can be much improved for judging correct exposure, and seeing more details in shadows and highlights, when setting Color to "Portrait" and "Tone" to "Manual" with +5 in the Shadows and -5 in the Highlights.

 

 

[Lars Steenhoff]

It also helps to put the sharpening to the max when shooting raw. it helps with focussing.

And its not baked in the raw

[paulinventome]

5 hours ago, The ghost of squig said:

Yeah the 5D MK3 is way better than the EOSM and 50D in every respect, IQ, DR, high ISO, and yes preview is real-time with proper framing, just like regular liveview only with a lot more display options. Magic Lantern has the same dynamic range as 14bit raw photo mode: 11.7 stops @ 100 ISO. It's a filmic grade, so the blacks are somewhat crushed. I'm not a big fan of low contrast or HDR type looks unless it's a creative decision driven by a particular narrative.

 

Well on paper DXO says that 5D3 is as you say, the A7sII is down as 13.3 and side by side with the fp the range is the same or actually a bit greater with the fp. So i'd infer that the fp is going to be tested (in stills) around that ballpark. 

In fact don't we understand the Panasonic S1 having the same sensor (not sure about S1H) - in which case the tested range is 14.1 (but in a 14 bit container? Hmm)

Then the cDNG mode of the fp is a stop/stop half less. So that would put the native range around the same as the 5D3 in cinema mode. Add some highlight reconstruction and you'd be getting 12/13 stops in a typical scene if you expose right. This matches my testing - and compared to the Red (down as 15.2 range) i think it all matches up in real world.

So i would say that the fp ought to be capable of producing a better image than the 5D3 technically - at UHD and better noise handling.

Of course there's a colour gamut and general feel to contend with, but there has only been a few examples i've seen that i think look good, but that's not the cameras fault!

It's all a bit IMHO. I've a shoot in the next month of so, and will be using this as a riggable B cam so i'm looking at it from my perspective. I *wish* it had more than 12 bit cine mode (and if it is that sony sensor that *could* do a 4K crop at 14 bit but i also have to step back sometimes and just look at how much you get out of something so small and cost effective...

cheers
Paul

[Lars Steenhoff]

https://www.dropbox.com/sh/fwd55ee51c6h0jv/AADa_V1Qg1c-vc95wo0E5dk3a?dl=0

Here a test file, 8 bit, 10 bit, 12 bit

Both 8 and 10 bit shadows break up the same way as Paul said, It a bit a shame because I really need 25 fps and 10 bit does not do the job.

[paulinventome]

4 hours ago, Lars Steenhoff said:

Both 8 and 10 bit shadows break up the same way as Paul said, It a bit a shame because I really need 25 fps and 10 bit does not do the job.

 

But 10 does a better job in the extreme highlights - so expose up a bit? 

What is it you need to achieve?

I'd also consider shooting 24p and playing it back at 25p. An awful lot of film transfers are just sped up in PAL land and very few people notice. Check the running times of UK vs US DVDs...

But at the same time Sigma have always been very polite about replying to me and whilst they're unable to comment, i do believe the information is actually going back to the right place.

EDIT : I can see you underexposed those samples by around 4 stops to test. But in reality what would you be doing? I mean testing is all well and good but i guess this camera is all about exposing normally - that's where the tonality is. If the shadows are important then double the ISO to grab an extra stop at the bottom?

cheers
Paul

[KC Kelly]

Has anyone used the fp with an external monitor?  If so, is there any lag between the camera and the monitor display?

[Lars Steenhoff]

Thanks for the tip @paulinventome to expose to the right more with 10 bit.

And yes I could speed up the 23.97 to 25

I also wish there was true 24 fps

In the 8 and 10 bit I'm just a bit concerned about the green in the last bit of the shadows, it seems that perhaps the black level is interpreted wrong, I have had green issues on the canon 5d with magic lantern raw before when the black level wat set wrong.

 

@KC Kelly

I use a SmallHD Focus Oled and there is no noticeable lag.

Other monitors may differ.

[Lars Steenhoff]

By the way 25 frames per second is also important if you want to use the 180 shutter rule and you have lights that operate at 50 hz. then you can match the shutter to the lights frequency.

And the lights where I live mostly still do 50 hz.

[rawshooter]

7 minutes ago, Lars Steenhoff said:

By the way 25 frames per second is also important if you want to use the 180 shutter rule and you have lights that operate at 50 hz. then you can match the shutter to the lights frequency.

And the lights where I live mostly still do 50 hz.

You can also do that by choosing 172.8 degree shutter with 24p/23.98p.

[paulinventome]

Here's a quick comp of a test i did a while back

The lamp clips, 2nd row down 12/10 and 8 bit. Clipped waveform under that. So this is sensor clipping.

The top row and the bottom row show highlight reconstruction. The DNGs have the non clipped highlights rebuilding clipped (Red)

The 12 bit and 10 bit are basically identical, both linear format and that final stop of light in the DNG has tonnes of tonality before clip. The 8 bit with it's linearisation has much less latitude in that final stop (because it has a better distribution in the mids). The knock on effect is that, like the shadows which go green, the highlights go magenta. As @Lars Steenhoff mentions i do wonder if this is a bug/issue or just a side effect of the lack of tonality.

But the take away is that 10 bit is the same a 8 in the shadows but does have a better highlight.

cheers
Paul