Which cameras have the most pleasing grain structure?

[PannySVHS]

With the Lumix S1 I had the first consumer camera with nice grain in lowlight instead of mush. Until that camera all DSLMs produced mushyness due to codec compression. C300ii does that, ugliest noise reduction in a 10bit cine camera body. By far the best mode, possibly even good, was the Cine 4K mode in that regard, 1080p 444 12 bit is terrible once the noise reduction kicks in. It then is as bad as a 8bit 420 codec of the same generation.

Just sayin, that until some time you either had to spend big bucks or shoot with an external recorder to have any grain at all in lowlight.:)

I recognized fixed pattenr noise, fpn, with my bmmcc but C300ii in the same scenarion gave me nasty mushines and the weirdest artefact I have witness, background shining through the face of a person. Pure wizzardry so to say😄

So how you all judge the S1 S1H noise? For me it is awesome when it has a chance to show with the great low light sensitivity of this cam. S5 must be the noise beauty bargain champ these days!

[webrunner5]

As great as the output is on it Red needs to come up with some kind of fix. Then they will have a Very desirable product. You can't expect clients to be perfect all the time.

[webrunner5]

1 minute ago, PannySVHS said:

With the Lumix S1 I had the first consumer camera with nice grain in lowlight instead of mush. Until that camera all DSLMs produced mushyness due to codec compression. C300ii does that, ugliest noise reduction in a 10bit cine camera body. The best mode was the Cine 4K mode in that reagard, 1080p 444 12 bit is terrible once the noise reduction kicks in as bad a 8bit 420 codec of the same generation.

Just sayin, that until some time you either had to spend big buck or shoot with an external recorder to have any grain at all in lowlight.:)

I recognized fixed pattenr noise, fpn, with my bmmcc but C300ii in the same scenarion gave me nasty mushines and the weirdest artefact I have witness, background shining through the face of a person. Pure wizzardry.

So how you all judge the S1 S1H noise? For me it is awesome when it has a chance to show with the great low light sensitivity of this cam. S5 must be the noise beauty bargain champ these days!

I think the original Canon C100 has about as clean of output there was. Sure 1080p but..

[PannySVHS]

C100, really? I would imagine mushyness kicking in due to the slim codec, which is 28mbit or something like that.

FPN in the important areas of the image are a no go for sure and one cannot always pull shadows down, if the wanted look does not allow. So lighting and knowing your camera noise are indeed important subjects. Lumix S1 is my bargain cinematic noise champ for now.

[webrunner5]

Yeah pretty wimpy codec but it was down sampled from 4K. Canon proved you don't need 400mbit a second to make stuff happen if you can somehow make magic happen. Their Color Science at the time helped a lot also.

You need to look at some of its footage. Even the original Sony A7s is no slouch at 1080p.

[projectwoofer]

31 minutes ago, PannySVHS said:

With the Lumix S1 I had the first consumer camera with nice grain in lowlight instead of mush. Until that camera all DSLMs produced mushyness due to codec compression. C300ii does that, ugliest noise reduction in a 10bit cine camera body. By far the best mode, possibly even good, was the Cine 4K mode in that regard, 1080p 444 12 bit is terrible once the noise reduction kicks in. It then is as bad as a 8bit 420 codec of the same generation.

Just sayin, that until some time you either had to spend big bucks or shoot with an external recorder to have any grain at all in lowlight.:)

I recognized fixed pattenr noise, fpn, with my bmmcc but C300ii in the same scenarion gave me nasty mushines and the weirdest artefact I have witness, background shining through the face of a person. Pure wizzardry so to say😄

So how you all judge the S1 S1H noise? For me it is awesome when it has a chance to show with the great low light sensitivity of this cam. S5 must be the noise beauty bargain champ these days!

It's not that I find the noise on my S5 ugly or something but I'm currently a bit paranoid so I denoise even the slightest amount of noise and then add film grain.

[independent]

1 hour ago, webrunner5 said:

Sure in a studio environment on a tripod, run n gun, moving from scene to scene, changing lens quickly. Try doing sports, running around shooting your 3 year old son outdoors not so easy.

That is the reason you Should shoot raw so you Can recover mistakes, but if you don't have the ability to recover than not the best tool to be using. The Komodo does not sound like the best overall option then. Who wants to worry about having to do perfect exposures all the time.

Perfect exposure is very different from overexposing or underexposing by two stops. If you just look at your monitor, viewfinder, or LCD, you wouldn't make such gross exposure errors. For example, if you were shooting your kid outdoors, two stops overexposed, the image would look obviously blown out.

Anyways, CineD has made some mistakes with their tests, especially the Komodo and not understanding how the Komodo retains and recovers highlights. Also for their latitude test, they also chose to expose hot for the Komodo (a mistake), which led to those skewed results. CVP exposed more sensibly and got 3 stops over and 3 stops under, a more balanced result for latitude. 

[Django]

55 minutes ago, independent said:

Protecting the highlights is something you should do for every digital video camera. What kind of extreme situations or incompetent DP would make such gross exposure errors? Latitude matters much less than dynamic range and highlight roll-off, where the Komodo leads significantly. 

Actually it really depends on your camera, log/codec & DR... if shooting RAW or S-Log2 advice is usually to expose to the right 2 stops above... but with DGO sensors you'd probably expose more to the left... and in run & gun situations you sometimes just have to make hard choices, protect highlights or blow them out to preserve some shadow info. 

On Komodo from what I understand there is no base or native ISO which is kinda unique. I guess you expose in the middle and push/pull in post right?

[independent]

100% agree, cameras do behave differently and people should know their cameras. I'm just addressing erroneous ideas about latitude. Based on cineD's test, Gunther could/should have exposed lower for a more balanced over/under to help avoid confusion; some people are interpreting that as the Komodo only having 1 stop overexposure latitude...period? No. He just used 60% EV as an arbitrary reference to show that the camera had 7 stops latitude. If he had exposed it as CVP (or anybody else) would, his results would "look" more balanced (+3/-3) to uninformed readers. 

Yeah, RED cameras including the Komodo are quite different in terms of Native ISOs and how data is allocated off the sensor. ISO 800 is not a native ISO, it's just a starting point for exposure strategy.  ISO is more like a LUT to help the shooter expose properly. For example, shoot at lower ISOs in low light (to protect shadows) and higher ISOs in brighter scenes (to protect highlights)

For example, if something is just under clipping at ISO 400, you can bump your ISO to 3200 and the image still won't clip. The image will LOOK brighter, but you can check your exposure using false color, goal posts and stop lights to see how data is being distributed and what's happening in the highlights and shadows. This is different from other cameras where you raise ISO to increase gain. 

All this can seem counterintuitive when comparing w/ other cameras, which I guess is why there's so much confusion. 

 

 

 

[QuickHitRecord]

2 hours ago, webrunner5 said:

I think the original Canon C100 has about as clean of output there was. Sure 1080p but..

Here are some C100 (i) tests I did that give an idea of the grain (download for a less compressed experience):

 

[webrunner5]

That is not the experience I had after they did the firmware update to raise the top ISO to 80,000. But what you show is pretty ugly lol. Could be, things progress.

[M_Williams]

4 hours ago, independent said:

Protecting the highlights is something you should do for every digital video camera. What kind of extreme situations or incompetent DP would make such gross exposure errors? Latitude matters much less than dynamic range and highlight roll-off, where the Komodo leads significantly. 

It's not just about recoverability or exposure errors or whatever.

HOW dynamic range is allocated is intrinsically part of that "film" look, as well as highlight roll-off. You don't get smooth highlight roll-off if most of your dynamic range is bunched up in the shadows.

[M_Williams]

triple post

[M_Williams]

triple post sorry

[TomTheDP]

On 2/16/2022 at 8:28 PM, M_Williams said:

The Komodo has pretty terrible highlight latitude. It's definitely a camera that's better to underexpose than overexpose. I think CineD found it was good with 7 stops of underexposure but only 1 stop over. It's quite extreme and the main reason I don't care for the camera. Exceptional highlight latitude is a hallmark of film (as in, actual film).

The tests I've seen indicate 4 and 1/2 stops over and about the same under. Some of the highlight retention seems to be software based like davinci highlight recovery. So less color information but still retaining detail. 

Pretty much similar dynamic range to other RED cameras. 

[Nikkor]

In stills cameras CCD cameras have much nicer noise. A shame they stopped developing CCD sensors.

[Django]

Agreed, my Leica M9 is by far the closest to film camera I own. Not just noise but also image thickness and colors. It just nails that film look. Digital Bolex (arguably the camera with most lo-fi mojo) also used the Kodak CCD.

Another great benefit from CCD was global shutter.

But sadly the DR & ISO spec race killed CCD in favor of CMOS.

[Rhood]

2 hours ago, Nikkor said:

In stills cameras CCD cameras have much nicer noise. A shame they stopped developing CCD sensors.

Well it’s not only in stills camera’s.
The Sony F35 I mentioned earlier uses a CCD sensor as well.

It’s digitally as close as you can get to film / print.

The Alexa is of course pretty close as well, but the grain is nicer on the Sony F35.

2 hours ago, Django said:

Agreed, my Leica M9 is by far the closest to film camera I own. Not just noise but also image thickness and colors. It just nails that film look. Digital Bolex (arguably the camera with most lo-fi mojo) also used the Kodak CCD.

Another great benefit from CCD was global shutter.

But sadly the DR & ISO spec race killed CCD in favor of CMOS.

I think it’s actually due to production costs. CCD sensors are more difficult to make, and cost more to produce if I’m not mistaken. 

[Nikkor]

I have a phase one p45+. This back is CCD but very old, it has a Kodak sensor and it's not that great in terms of DR, BUT compared to CMOS, I enjoy the shadows even if they are noisy. With modern CMOS there is a lot to recover from shadows, but it looks horrible, wrong, desaturated colors, with the p45+ it goes noisy but the color still look right. I guess if you pixel peep or do some measurments the DR will be better on the CMOS, but watched from a distance, the image as a whole looks better on the CCD to me. Not to speak about colors, but I believe this is because of the bayer filter strength.

I think the reason to go for cmos was because it's cheaper and consumes less energy.

[M_Williams]

19 hours ago, TomTheDP said:

The tests I've seen indicate 4 and 1/2 stops over and about the same under. Some of the highlight retention seems to be software based like davinci highlight recovery. So less color information but still retaining detail. 

Pretty much similar dynamic range to other RED cameras. 

Yes the Komodo does have built-in highlight recovery that uses other color channels to reconstruct clipped data, exactly like Da Vinci.