CMOS sensor

[Peter Lyons Collister ASC]

I know that as a long time Red owner and as someone who has access to Graeme and Jim and Jarred,I should understand but people often single our Red sensors as sensitive to 5000 degree kelvin light.
Aren't all CMOS sensors ( Alexa, Canon, etc) balanced to 5000 degree kelvin light????

Sorry Graeme for my stupidity

Peter

 

[Tom.Wong]

most are balanced for daylight but don't have to be, but yes generally it's true which is one of the reasons the blue channel tends to suffer on cmos based cameras. the only exception i know of is the f65 which is actually balanced for 3200 tungsten. but red shouldn't be singled out at all. it's just that cameras from various companies have different ways with dealing with the blue channel better.

and 80d filter usually solves a lot of blue channel issues with any digital camera

 

[Mark Toia]

?... RAW is RAW... What what balance.?. You can shoot what ever kelvin you like and change it later.

I very worried about people when they think what you shoot is what you get. Thats a Arri PRO RES or Canon thinking... RED grading in RAW. You can change your mind what Kelvin you want 5 years later...

 

[Zeb B]

Yes Mark, but if you don't feed the sensor balanced light, you are just gaining up on single channels (thereby producing noise) when doing White Balance in post. RED likes light - lots of it and at 5600K when possible for the best image quality. That's why I gots some of them K5600 Bugs

 

[Andrew Gentle]

Isn't it less that the sensor is intentionally balanced for a particular colour temperature and more that the science behind silicon sensors mean that they are naturally less sensitive to blue light and therefore perform better when there is a greater blue component in your light source?

 

[Gavin Greenwalt]

?... RAW is RAW... What what balance.?. You can shoot what ever kelvin you like and change it later.

I very worried about people when they think what you shoot is what you get. Thats a Arri PRO RES or Canon thinking... RED grading in RAW. You can change your mind what Kelvin you want 5 years later...

Yes but if you have to gain up the blue and down the red you're losing dynamic range when white balancing.

 

[Zeb B]

@Andrew, that's my understanding. Chips, like our eyes, are less sensitive to Blue light - that's why folks can have a hard time pulling keys off a green/blue screen lit with 3200K sources - there's noise in the Blue channel caused by gaining it up through post WB.

 

[Scott Crawley]

I would love to hear something from Graeme on this. I have hypothesized on it some myself. I have read some articles and posts that went partially over my head about the number of blue samples in the matrix being fewer and that is why a sensor is lacking in blue, but is there any reality in this idea that the physical characteristics of silicon play a part?

Isn't it less that the sensor is intentionally balanced for a particular colour temperature and more that the science behind silicon sensors mean that they are naturally less sensitive to blue light and therefore perform better when there is a greater blue component in your light source?

 

[ericyoung]

I know that as a long time Red owner and as someone who has access to Graeme and Jim and Jarred,I should understand but people often single our Red sensors as sensitive to 5000 degree kelvin light.
Aren't all CMOS sensors ( Alexa, Canon, etc) balanced to 5000 degree kelvin light????

Sorry Graeme for my stupidity

Peter

Any photosensor based on silicon semiconductor technology is less sensitive to visible blue light than light towards the green and red end of the spectrum, so yes all silicon based CMOS sensors also "prefer" daylight balanced light.
But silicon CCD sensors are also better with daylight balance. It's the semiconductor substrate material and not the type of phototransistor that determines this.

 

[ericyoung]

I would love to hear something from Graeme on this. I have hypothesized on it some myself. I have read some articles and posts that went partially over my head about the number of blue samples in the matrix being fewer and that is why a sensor is lacking in blue, but is there any reality in this idea that the physical characteristics of silicon play a part?

This is confusing the wavelength sensitivity of silicon (which is poorer towards the blue end), with the choice to have a lower resolution of blue and red pixels compared to green ones in a Bayer pattern sensor, because the human eye is most sensitive to wavelengths in the green part of the visible spectrum, and perceived resolution/sharpness is biased towards the resolution of the green channel in an image. To be more exact, it's the resolution of the luminance (non-colour information) that mostly determines how resolved or sharp an image looks to the human eye, but the green luminance has the larger impact on the perception.

 

[Peter Lyons Collister ASC]

Thanks Eric and others. Mark I think you misunderstood my question. This has nothing to do with RAW and meta data. It has to do with the intrinsic sensitivity of CMOS sensors to Red Blue and Green.

 

[Filip Orlandic]

Aren't all CMOS sensors ( Alexa, Canon, etc) balanced to 5000 degree kelvin light????
Peter

Mysterium 5000K
Mysterium-X 4300K

 

[Gunleik Groven]

Mysterium 5000K
Mysterium-X 4300K

????

Would kind of explain WB oddities I sometimes see bellow 4400k, though, but I'd have thought it was closer to 5000+...

Would've been fun to know.

From where did you pull this?
I'd be surprised if this was really totally accurate.

As tro Mark. It is (in some circumstances) not irrelevant. At some WB the sensor response is "flat". The further beyond that you go either way, the more post-processing is done to ballance the image data. Some of that processing can be good. Some, not so much.

If it looks good, it's fine.
If you run into issues, it is good to know...

 

[Filip Orlandic]

????

Would kind of explain WB oddities I sometimes see bellow 4400k, though, but I'd have thought it was closer to 5000+...

Would've been fun to know.

From where did you pull this?
I'd be surprised if this was really totally accurate.

As tro Mark. It is (in some circumstances) not irrelevant. At some WB the sensor response is "flat". The further beyond that you go either way, the more post-processing is done to ballance the image data. Some of that processing can be good. Some, not so much.

If it looks good, it's fine.
If you run into issues, it is good to know...

It was said before that it's between 3200 and 5000.

I will try to find the original post.

 

[Gunleik Groven]

It was said before that it's between 3200 and 5000.

Hm... It's NOT close to 3200.
From what I see, I'd say 5000 at its lowest.

Just do a wb test and see where the response is flat and where you introduce gain and clamping (in the red and blue channels respectively).

The M was re-rated from 5600 to 5000 @ B30. But if that is really "clean" response... I am not sure.

Nice way to verify, is to look at plates from the R, G & B channel from the same image at different WB settings.
Expose a couple under tungsten/Daylight and peek at the separate channels to look at exposurelevels and noisegeneration at a given WB rating, look at the RGB waveforms, and I guess you'll get a hunch...

A lot of stuff has been "said" at various points.
Like "RedSpace is the best for everything".

Run tests, and make up a qualified opinion. Then choose, and check that what you've chosen actually works...

That the MX has better lowlight sensibility than the M, makes it less prone to tungsten noiseissues, as you have some extra headroom in the blue channel, and by slightly using that headroom in the lower range, to gain some headroom in the top in the RED by underexposing it 1.25 stops in the top (by shooting @ 800 ISO), thereby avoiding hard clamping in the RED channel @ tungsten is all good.

BUT it is not the same as saying that the flat response is close to 3200.

It's more of a philosophy thing as to what settings that averagely will give the most good results for the most users.
WHICH I BTW think is a perfectly good reasoning and a totally legitimate point for an advice.

Like RC3/RG3 will work "out of the box".
Mostly true for many applications, and probably the most of the userbase most of the time.
So listen.
The advice is good. If not arguably from a purely technical standpoint allways true. It is "true enough".

Edit:
I am a bit wary about publishing tests like these, these days, as they are easily interpreted as "stuff is noisy" or "not right"
I think they are good for a totally different purpose:
To see how a signal chain performs and learn to use and master it.

I am not saying there is anything wrong with it.

But as it is too often interpreted as such, I have leant more and more towards suggesting ways to realitycheck statements, rather than publishing images to state a point...

 

[Peter Lyons Collister ASC]

My original post stems from often reading about "noise" issues shooting Red MX under tungsten lights without any wart ten 80 filtration. I have read about these issues for Alexa or Nikon or Canon DSLRs.
We shot 80 percent of Spiderman under tungsten lights and have run tests to IMAX and it seems pretty clean.

 

[Scott Crawley]

I recently did a series of readings to find and compare the noise threshold of Scarlet under daylight and tungsten at several apertures. I found that under tungsten light i did indeed run into noise sooner that under daylight, though for the most part it wasn't that significant. At stops between f5.8-f8 the difference in light levels ran from 12-25 fc, to rephrase; I needed 12-25 fc more tungsten light to get an image as clean as with a daylight source. So for most shooting it would seem that just walking your tungsten lights in a couple feet would compensate... If I understand my results correctly.

EDIT: corrected Lux to read fc

My original post stems from often reading about "noise" issues shooting Red MX under tungsten lights without any wart ten 80 filtration. I have read about these issues for Alexa or Nikon or Canon DSLRs.
We shot 80 percent of Spiderman under tungsten lights and have run tests to IMAX and it seems pretty clean.

 

[PatC]

I think a lot of tungsten "noise" arguments come from the original M sensor and it has been much less of an issue moving into R1MXs and EPICs.

 

[Gunleik Groven]

Silly me. I don't even have much noise from the M...

 

[Sidney L. Plaut]

My original post stems from often reading about "noise" issues shooting Red MX under tungsten lights without any wart ten 80 filtration. I have read about these issues for Alexa or Nikon or Canon DSLRs.
We shot 80 percent of Spiderman under tungsten lights and have run tests to IMAX and it seems pretty clean.

The reason being you prob. didn't need to push the sensor by underexposing it - you prob gave enough exposure in the shadows and brought them down in post? - only speculating of course

An easy test is to put the body cap on. crank up the ISO to 3200-12800. press "video/false colors" and change the white balance. You will see a difference in the shadow exposure depending on
which White balance you set. - the exposure of the right half or left half of your sensor can be noticeably different.

With build 3x it has much improved and you prob won't see a difference below iso2000 depending on your sensor temp.

- Edit : since redgamma 3 crushes the blacks a bit it is easier too see in RedGamma2.