CMOS sensor

[Gunleik Groven]

as to the OP, Graeme stated a number of times - early on - tat silicone has a native response around daylight, and unless you have a filtering of the light, that will be the case for digital cameras.

From the cameras I have had the pleasure to examine, that sounds roundabout spot on...

 

[Graeme Nattress]

Native spectral response of silicon (so this applies to all cameras that use a silicon based sensor) is blue deficient. In as much tungsten light is also blue deficient, you could say that the sensor will produce an even white balance without correction at a colour temperature on the daylight side of things.

Native wb can be altered in a number of ways:

a) optically with a filter (possibly external or part of the OLPF/IR stack)
b) analogue gain
c) digital gain

We currently use digital gain as the optimum balance for our needs, but always re-evaluate this with every sensor design as part of our optimization process. Each sensor is different and behaves differently so it would be stupid to stick to one method dogmatically rather than do the proper evaluation each time.

Graeme

 

[Sidney L. Plaut]

Graeme,

I seem to remember in the Epic manual, that you should be able to see an evenly exposed image in ISO12000 in "false color / video" mode after you have black shaded - depending on
which color temp/ white balance and Gamma, I have not been able to go above iso3200?

 

[Christopher Barrett]

Thank You G!

Man. 80D used to be my go to filter when shooting daylight transparency in florescent. I'm gonna have to dig that thing back out for a test on my Reds.

 

[Blair S. Paulsen]

CTB vs 80

CTB vs 80

FWIW I have had better results with CTB filtration vs 80 series on digital. I believe it is because the mired shift of the 80 filters is designed for shooting daylight film stocks under tungsten which mandates specific photometric conversion properties and commensurate light loss. With a CTB filter you can move the color temp up the scale enough to improve channel balance without losing as much stop.

My typical usage is a 1/4 CTB filter when in a location where practicals play (or shooting available) with a 2700 - 3000K white point. The 1/4 moves me up to +/- 3400K which is back in the wheelhouse of the CMOS sensor/debayering algorithm and I only lose about 1/3 of a stop. You could make a case for correcting all the way to 5000K for "ideal" results but in many situations a filter that eats 1 to 1.5 stops creates overall underexposure issues that may lead to more pain than gain. YMMV.

Cheers - #19

 

[Graeme Nattress]

RED Colorimetry is defined by colour temp. If you use a filter to change the native colour temp, just beware that that is not how the colorimetry at that new colour temp was calibrated.

Graeme

 

[Blair S. Paulsen]

RED Colorimetry is defined by colour temp. If you use a filter to change the native colour temp, just beware that that is not how the colorimetry at that new colour temp was calibrated.

Graeme

Are you suggesting that the decode of my artificially shifted scene referred values via the 1/4 CTB filtration will be less accurate than if I had just shot 2800K clean and let the algorithm do the work? If so, that implies that the processing step adjusts the interpretation of the bayer data to create a more "pleasing" image even if some mathematical accuracy is compromised... If I am off base, I would love to know why.

Great chatting with you at NAB BTW.

Cheers - #19

 

[Graeme Nattress]

Not necessarily - just a "beware".

How it works is that we calibrate the colorimetry of the cameras under a wide variety of light sources, but as I mention above, we don't have any filters in the chain. We pick the appropriate calibration based on the one bit of info we have - the kelvin of the wb, which gives us the indication of tungsten, daylight, somewhere in the middle or "beyond".

In the extreme case of shooting tungsten light, but filtered to a ~6500k wb, you'd be using daylight calibration with what is still essentially a tungsten source, and I could well imagine the colours wouldn't be right. In that case you could set the wb back to 2800k or 3200k and then use RGB gain to adjust the wb and then you'd be using the correct calibration. But that's the most extreme circumstance.

In practical circumstances, you're shooting a warm light and leaving the image looking warm. You may be helping the sensor a bit with some filtering, but keeping the warm look. In that case I don't see an issue either way. Say you put some filtering on a tungsten light and cool it a bit, and wb neutral. You may still be absolutely fine as long as you're not changing the effected correlated colour temp by 1000's of kelvin.

Graeme

 

[paulherrin]

so basically, it's going to be best practice to nail the WB in redcine (or when shooting) and adjust other controls to taste from there. rather, than say - warming up your image by adjusting whitebalance. correct?

 

[Peter Lyons Collister ASC]

Thanks for chiming in Graeme....

Now back to ACES

 

[Graeme Nattress]

Paul, that sounds a reasonable starting point. Always let the image be your guide though - if it looks good it is good, but those recommendations are my starting point.

Graeme

 

[Scott Crawley]

Thanks Graeme.

 

[David Rasberry]

Not necessarily - just a "beware".

How it works is that we calibrate the colorimetry of the cameras under a wide variety of light sources, but as I mention above, we don't have any filters in the chain. We pick the appropriate calibration based on the one bit of info we have - the kelvin of the wb, which gives us the indication of tungsten, daylight, somewhere in the middle or "beyond".

In the extreme case of shooting tungsten light, but filtered to a ~6500k wb, you'd be using daylight calibration with what is still essentially a tungsten source, and I could well imagine the colours wouldn't be right. In that case you could set the wb back to 2800k or 3200k and then use RGB gain to adjust the wb and then you'd be using the correct calibration. But that's the most extreme circumstance.

In practical circumstances, you're shooting a warm light and leaving the image looking warm. You may be helping the sensor a bit with some filtering, but keeping the warm look. In that case I don't see an issue either way. Say you put some filtering on a tungsten light and cool it a bit, and wb neutral. You may still be absolutely fine as long as you're not changing the effected correlated colour temp by 1000's of kelvin.

Graeme

I have experienced some funky red/blue color balance issues using blue filtered GE reveal light bulbs for practicals with my video camera. Canon's white balance and magenta/green tint controls don't compensate properly. I'm sure its much easier to compensate with raw adjustments.

 

[Stephen Williams]

so basically, it's going to be best practice to nail the WB in redcine (or when shooting) and adjust other controls to taste from there. rather, than say - warming up your image by adjusting whitebalance. correct?

It makes no difference, best practice would be to add some blue filtering to either the light or the camera, you don't need the full 2 stops, however the more the better if your shooting a greenscreen & want fairly high compression. (8:1)

 

[Scott Crawley]

While I would agree, One could argue that it doesn't matter if you CTB a tungsten light and then effectively alter exposure by +1.5 EV, or just light tungsten +1.5 EV and adjust Kelvin setting to -gain in the red channel. So long as you are not clipping the Red channel in the first instance the net effect is the same; +1.5EV (or approx 7lux) worth of energy in the blue spectrum. Makes sense, no?

I think I am making a mess of my conversions, but hope the underlying idea is clear.

It makes no difference, best practice would be to add some blue filtering to either the light or the camera, you don't need the full 2 stops, however the more the better if your shooting a greenscreen & want fairly high compression. (8:1)

I'm sorry. It just occurred to me that I am rephrasing what Graeme already said, only he said it better.

In practical circumstances, you're shooting a warm light and leaving the image looking warm. You may be helping the sensor a bit with some filtering, but keeping the warm look. In that case I don't see an issue either way. Say you put some filtering on a tungsten light and cool it a bit, and wb neutral. You may still be absolutely fine as long as you're not changing the effected correlated colour temp by 1000's of kelvin.

Graeme

 

[Gavin Greenwalt]

While I would agree, One could argue that it doesn't matter if you CTB a tungsten light and then effectively alter exposure by +1.5 EV, or just light tungsten +1.5 EV and adjust Kelvin setting to -gain in the red channel. So long as you are not clipping the Red channel in the first instance the net effect is the same; +1.5EV (or approx 7lux) worth of energy in the blue spectrum. Makes sense, no?

I think I am making a mess of my conversions, but hope the underlying idea is clear.

Ideally you would balance your lights so that the RAW red green and blue histograms all clipped to white 'simultaneously' on a white card. That way you can push the entire image up into the cleanest range of sensitivity without any clipping channels. The problem with just lighting tungsten and pushing the white balance is that the blues in your skintones and grays will start to clip before the rest of the scene so you'll have to give the entire exposure an extra stop of headroom which you could normally move more usable information into.

 

[Scott Crawley]

The problem with just lighting tungsten and pushing the white balance is that... you'll have to give the entire exposure an extra stop of headroom which you could normally move more usable information into.

But isn't that exactly what you do when you rate the camera at 800 ISO? I don't see a problem with that. :wink5:

I don't suppose that it makes any difference if you invoke -red gain via Kelvin and -green via tint adjustments as opposed to pulling them down anywhere else in the look panel. Graeme, is it the same operation?

 

[Wayne Morellini]

Wow, a thread with people that actually talk sense and understand things around here

 

[ericyoung]

Not necessarily - just a "beware".

How it works is that we calibrate the colorimetry of the cameras under a wide variety of light sources, but as I mention above, we don't have any filters in the chain. We pick the appropriate calibration based on the one bit of info we have - the kelvin of the wb, which gives us the indication of tungsten, daylight, somewhere in the middle or "beyond".

In the extreme case of shooting tungsten light, but filtered to a ~6500k wb, you'd be using daylight calibration with what is still essentially a tungsten source, and I could well imagine the colours wouldn't be right. In that case you could set the wb back to 2800k or 3200k and then use RGB gain to adjust the wb and then you'd be using the correct calibration. But that's the most extreme circumstance.

In practical circumstances, you're shooting a warm light and leaving the image looking warm. You may be helping the sensor a bit with some filtering, but keeping the warm look. In that case I don't see an issue either way. Say you put some filtering on a tungsten light and cool it a bit, and wb neutral. You may still be absolutely fine as long as you're not changing the effected correlated colour temp by 1000's of kelvin.

Graeme

Surely putting various strengths of CTB gel on tungsten lamps is reducing the level of the tungsten source green and red closer to the level of the blue. Ie REALLY making it more daylight balanced - it's not a "bodge". It's effectively NOT still a tungsten source then.

OK, CTB gel light transmission curves are not completely smooth, so the transmitted light will not be a perfect "natural" black body daylight spectrum, although it's certainly good enough for film.
So filtering with full CTB, makes the light source 5600K (or as close as CTB gel manufacturers can make it), and I would expect putting the camera colour temp to 5600K should look right, as it usually does when I've used it.

Or are you just saying that pushing tungsten towards even more extreme blue ~6500K, the camera is expecting a certain smooth colour spectrum for white, and the actual spectrum due to gel imperfections at that correction strength may well be too distorted and "peaky", so it'd be better to adjust RGB gain in the camera settings while lighting with an ungelled tungsten source, at least as far as colorimetry goes, if not noise performance.

Of course, it'd be better to start with a daylight source and only modify that a little if we want WB at 6500K, rather than a tungsten source modified a lot. Not to mention the extreme light level loss!

 

[Scott Crawley]

Graeme, if I understood you correctly you were saying that the MX sensor and RAW 16 bit could handle digital gain to the tune of a couple thousand degrees kelvin without any apparent loss of fidelity, and therefore you do not advocate analogue gain over digital gain. Is that correct?

So are you saying that analogue gain is not inherently better than digital with the MX, or simply that within reasonable ranges of manipulation that there is no appreciable difference between them? ... Because the system is robust enough for that?

Sorry again if it seems I am just re-phrasing all of your remarks. I'm just trying to get my head around this.