Sigma Fp review and interview / Cinema DNG RAW

[eleison]

1 hour ago, rawshooter said:

the camera downscales the 6K raw sensor data into simulated raw sensor data of a virtual, software-made 4K sensor. Which creates all kinds of weird artifacts, as demonstrated above in the visual comparisons of the camera's 6K stills with the camera's 4K video.

Oh, so basically a firmware update is the only option right now.  Well, that cured my GAS for a bit.

[rawshooter]

1 hour ago, eleison said:

Oh, so basically a firmware update is the only option right now.  Well, that cured my GAS for a bit.

IMHO not even a firmware update can fix this, because there may be no clean way of scaling undebayered 6K to undebayered 4K. 

The elephant in the room is RED's patent on compressed raw video recording. Recording the full 12bit 6K raw sensor data uncompressed would require 864 MB/s (=7 Gbit/s), hit the limits of USB 3.1 and exceed the speed of most external SSD drives. So Sigma went for downscaling instead of compression as a way to reduce data throughput, almost like in the old days of interlaced video...

The camera would be able to record 6K compressed RAW (since 1:2 compression is rather trivial - simple zip/lzw compression would do the job) if RED didn't have the patent stranglehold on this option.

[Lars Steenhoff]

It not about red, its about money.

They could have made a 6k camera with an nvme ssd. 

But I guess that would require a higher fee to sony to enable the 6k readout . and also a faster bus.

[Lars Steenhoff]

Question, if you put the camera in M mode and you set the shutter speed to 50 why does my fp make the shot at 100?

anyone have the same problem?

( I use manual focus lenses )

[Andrew Reid]

Which frame rate and rec. mode are you in?

Mine stays at 50 in 4K 25p

[Lars Steenhoff]

in stills mode only, video is fine

[cpc]

On 1/17/2020 at 1:52 PM, paulinventome said:

Beg to differ on this one 😁

If you have a high contrast edge and use various scaling algorithms, Lanczon, Sinc, etc,. Often the scale results in negative lobes (like a curve overshoot). It's quite easy to see this. Happens in compositing all the time. It's exacerbated by working Linear as we always do (i suspect because of the dynamic range). So it's a well established trick to cover to log, scale, and convert back to linear. I suspect some apps do this automatically (maybe resolve does, i don't know)

https://www.sciencedirect.com/topics/computer-science/ringing-artifact

It's a bit geeky but there's a diagram of a square pulse and lobes that extend beyond it.

Sony had this issue in camera on the FS700, very bright edges would give black pixels as the scale went negative.

My gut tells me in the case of a bayer sensor it could be even worse - partly because of the negative signal but also the tendancy to have stronger edges because of missing bits of image?

I will be surprised if Resolves does rescale in anything different than the image native gamma, that is in whatever gamma the values are at the point of the rescale operation. But if anything, some apps convert from sRGB or power gamma to linear for scaling, and then back.

You can do various transforms to shrink the ratio between extremes, and this will generally reduce ringing artifacts. I know people deliberately gamma/log transform linear rendered images for rescale. But it is mathematically and physically incorrect. There are examples and lengthy write-ups on the web with what might go wrong if you scale in non-linear gamma, but perhaps most intuitively you can think about it in an "energy conserving" manner. If you don't do it in linear, you are altering the (locally) average brightness of the scene. You may not see this easily in real life images, because it will often be masked by detail, but do a thought experiment about, say, a greyscale synthetic 2x1 image scaled down to a 1x1 image and see what happens.

I have a strong dislike for ringing artifacts myself, but I believe the correct approach to reduce these would be to pre-blur to band limit the signal and/or use a different filter: for example, Lanczos with less lobes, or Lanczos with pre-weighted samples; or go to splines/cubic; and sometimes bilinear is fine for downscale between 1x and 2x, since it has only positive weights. On the other hand, as we all very well know, theory and practice can diverge, so whatever produces good looking results is fine.

Rescaling Bayer data is certainly more artifact prone, because of the missing samples, and the unknown of the subsequent deBayer algorithm. This is also the main reason SlimRAW only downscales precisely 2x for DNG proxies.

On 1/17/2020 at 1:52 PM, paulinventome said:

But if that scale is done properly - is it not going to improve the image OR is the fact that scaling those 4 layers, which are already missing pixels in-between actually makes it worse?

It is actually possible to do Bayer aware interpolation and scale 3 layers instead of 4. This way the green channel will benefit from double the information compared to the others. You can think of this as interpolating "in place", rather than scaling with subsequent Bayer rearrangement.

Similar to how you can scale a full color image in dozens of ways, you can do the same with a Bayer mosaic, and I don't think there is a "proper" way to do this. It is all a matter of managing trade offs, with the added complexity that you have no control over exactly how the image will be then debayered in post. It is in this sense that rescaling Bayer is worse -- you are creating an intermediate image, which will need to endure some serious additional reconstruction. Ideally, you should resize after debayering, because an advanced debayer method will try to use all channels simultaneously (also, see below).

On 1/17/2020 at 1:52 PM, paulinventome said:

The Ideal is take each RGGB layer then do an interpolation, so that instead of going from 3K (one channel) you interpolate that up to the full 6K by interpolating the missing pixels and *then* scale it down to to 1920 and then the resulting bayer image might actually look really good....?

This is possible, and you can definitely extract more information and get better results by using neighboring pixels of different color because channels correlate somewhat. Exploiting this correlation is at the heart of many debayer algorithms, and, in some sense, memorizing many patterns of correlating samples is how recent NN based debayering models work. But if you go this way, you may just as well compress and record the debayered image with enough additional metadata to allow WB tweaks and exposure compensation in post, or simply go the partially debayered route similar to BRAW or Canon Raw Light.

 

In any case, we should also have in mind that the higher the resolution, the less noticeable the artifacts. And 4K is quite a lot of pixels. In real life images I don't think it is very likely that there will be noticeable problems, other than the occasional no-OLPF aliasing issues.

[rawshooter]

2 hours ago, Lars Steenhoff said:

Question, if you put the camera in M mode and you set the shutter speed to 50 why does my fp make the shot at 100?

anyone have the same problem?

( I use manual focus lenses )

I just checked, with an adapted manual focus lens as well. 1/50 in M mode yields 1/50 in the picture, both on the camera review display and in the DNG meta data. Are you sure that you were actually in M (and not, for example, in A or P) mode?

[Lars Steenhoff]

yes I'm in M mode, here a video that shows how my cam does this.

I will try a factory reset, see of that helps

I'm talking about stills not video, just to be sure we are talking about the same.

Also look at the histogram before taking the shot and in playback, they don't match, the playback shot is underexposed

[rawshooter]

On 1/17/2020 at 12:52 PM, paulinventome said:

The Ideal is take each RGGB layer then do an interpolation, so that instead of going from 3K (one channel) you interpolate that up to the full 6K by interpolating the missing pixels and *then* scale it down to to 1920 and then the resulting bayer image might actually look really good....?

But the problem still is, as far as I can wrap my head around it, that any edge of two different colors in the image, we can't really properly interpolate and scale the single red, green and blue matrices, because their correct scaling would imply knowledge of the full color spectrum they are part of.

To give an example, we might have an edge of a red object against a grey object. In the Bayer pixels representing the red object, red values would be high, green and blue values would be low; in the Bayer pixels representing the grey object, red, green and blue values would be about the same. But if we isolate the red matrix and only take the edge, then we'd see a high red value for the red object against a lower red value for the grey object. If we interpolate and scale, we'd average the two red values, which would be visually wrong - because we're (a) blurry the edge and (b) in the mix with the other blue and green values and their processing, create false colors.

This is exactly what's happening in the test sample images I uploaded where the silver synthetic fabric with its fine, pixel-level white-vs.-black pattern provoked (a) blurry edges and (b) false (violet) color in the downscaled 4K CinemaDNG image. 

In order to avoid such artifacts, the scaling algorithm would actually need to know context - that is, which full-spectrum colors the single color matrices are part of. And that would require debayering the image...

So, in other words, if we don't debayer, we'll never get a clean downscale. In theory, we could debayer and then algorithmically reconstruct a 4K Bayer matrix - but then, IMHO, it makes much more sense to encode debayered 12bit Log video right away.

4 minutes ago, Lars Steenhoff said:

I'm talking about stills not video, just to be sure we are talking about the same.

I'm talking about stills, too, and can't reproduce the behavior of your camera on mine...

[Lars Steenhoff]

1 hour ago, rawshooter said:

I'm talking about stills, too, and can't reproduce the behavior of your camera on mine...

Is the histogram the same before and after the shot? ( its darker on mine ) 

[rawshooter]

6 minutes ago, Lars Steenhoff said:

Is the histogram the same before and after the shot? ( its darker on mine ) 

You're right, it's darker.

Well, this is a buggy piece of a camera...

[Lars Steenhoff]

Yes it is, I'm still trying to love it, but for stills this makes me crazy.

It all depends in what fixes the next firmware will bring.

and the shutter lag, that also makes me crazy, because when I think I have the shot, it takes it a fraction later.

[Andrew Reid]

3 hours ago, Lars Steenhoff said:

yes I'm in M mode, here a video that shows how my cam does this.

I will try a factory reset, see of that helps

I'm talking about stills not video, just to be sure we are talking about the same.

Also look at the histogram before taking the shot and in playback, they don't match, the playback shot is underexposed

 

Can confirm I have same issue! In stills mode at least, not in video or cine mode.

It only happens with a passive manual lens adapter where the aperture can't be read by the camera (F--)

Sigma or Pana L-mount lens it is fine

I also get a jumpy flickering of brightness in live-view, like the camera is trying to work out the metering

I think it's a LV metering bug.

Picture comes out same exposure as LV in cine mode... But switch back to Stills LV and exposure is brighter.

Very weird.

1 hour ago, Lars Steenhoff said:

Yes it is, I'm still trying to love it, but for stills this makes me crazy.

It all depends in what fixes the next firmware will bring.

and the shutter lag, that also makes me crazy, because when I think I have the shot, it takes it a fraction later.

Can confirm the slight shutter lag as well.

It's a lot longer than average.

[rawshooter]

What is also funny with the shutter lag: 
The picture shown first on the viewfinder after pressing the shutter button is taken with a different (shorter) shutter speed than the eventual picture recorded. You can test that when shooting fast-moving subjects with a long shutter speed. 
 

(Again, an indication that most camera reviews are worthless today, because no reviewer has spotted/described this issue yet.)

[Andrew Reid]

Now the good news... That Contax Zeiss 28-70mm F3.5 is a stunner, a real gem. So small and light. Almost as small as the Zeiss 55mm F1.8 FE

It does not seem parfocal but it has character. Amazing macro mode too and the Zeiss look.

I have quite enjoyed putting it on the Leica M AF adapter on my Sony a9 too... Autofocus it

Even the Sigma 45mm F2.8 is a fatter lens.

[Lars Steenhoff]

Yes the zeiss lens is a keeper, I received mine today, and it's not perfectly parfocal with the adapter I use, I read somewhere that parfocal depends also on the distance to the sensor, its needs to be perfect. 

My copy is focussing very smooth, really happy with it.

[rawshooter]

I can also warmly recommend the Tokina 28-70mm/2.8 which is a bought-up and continued Angenieux design:

https://cameragx.com/2018/04/11/the-truth-about-the-angenieux-28-70-af-zoom/

[Andrew Reid]

Have you checked out the ISO 6 mode yet? Would be good if they got that working in 4K... Goodbye NDs.

I am working-around the M mode bug by using Auto ISO and Expo.comp, not ideal... They really need to fix that, it's fundemental stuff! Also could not get Exposure bracketing to work at all. I will just shoot stills 99% in A priority and lock exposure during a shot.

[Lars Steenhoff]

Iso 6 only works for stills ... and yes it would have been nice if it could be used for video.