Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Tungsten Sensor

Harry Clark

Harry Clark

Well-known member
Joined
Dec 29, 2006
Messages
1,791
Reaction score
0
Points
36
Location
NYC
I know it's never gonna happen but it's snowing outside and I have the day off so I'll start a gratuitous thread here...
In the next version of the camera, it would be great if we had the option of choosing a tungsten-balanced sensor. It's effortless to put on an 85 filter, with minimal light loss (in a daylight situation when presumably there's lots of exposure anyway) It's much harder to make everything daylight. Especially on an indie movie shooting with lots of tungsten practicals.
It's a tungsten world. Just because DSLRs are daylight balanced is not a good reason to make a digital cinema camera daylight balanced. If I only had a choice of one type of film stock on a movie it would be a tungsten medium speed stock like 5217 (although the new Vision 3 stock looks so nice I might carry 5219 instead)
Why do I bring this up?
Because despite the "RAW" nature of Red and the notion that you can make the camera balanced for any Kelvin, I can see quite a bit of noise creep in as I shoot farther away from daylight balance and closer to 3200K.
I'll freely admit that I'm not an engineer and know nothing about designing CMOS sensors, so it may be that I'm incorrect in assuming that it's possible to build a tungsten-native sensor. So apologies if I'm wrong. I'm just saying that's what I would have done, in retrospect.
OK, let the flaming arrows fly! ;)
Cheers,
Harry
 
Its a good idea - tungsten shooting is typically more "light starved" than daylight .

One question - does the camera have a flip switch between tungsten and daylight or do you just shoot the same and correct in post?

Thanks

James
 
I think it would be possible to build one... But you would need to modify the photo sites a bit, like adding a bit of ND onto all the green photo sites, and little less ND on the red photo sites.

Essentially what you need to do is pull back red and green to the same level as blue. I'm not sure how less sensitive the camera would become though, but then again, if you ever hope to have a clean blue channel under tungsten, you need to filter, is just simple physics.

Have you ever seen an unbalanced RAW image? I'm not talking about loading a RAW image into a RAW converter, those never show you the unbalanced image. Just look around for Viper frame grabs shot without optical filtering... The images are always very green, and its just physics, it's easier to capture green light than red or blue. As for red and just blue, it depends a lot on the color temperature you shoot under. Its also pretty hard to capture blue light as well, from what I understand.

I'd root for a Tungsten sensor option.

Hope that helps!
 
It's perfectly possible to trade sensitivity for better tungsten response. The way sensors work, they're always going to be biased away from tugsten as they just don't respond to blue very well - that's the nature of sensors.

But, as sensitivity is an important characteristic also, best that you filter if you want to, rather than us permanently filter to bias the sensor away from daylight and loose sensitivity you can never get back.

Graeme
 
I agree. Nighttime exteriors are hard to get anyway on an indi, even if you CAN correct for your tungsten lights (Although having a 2k or 4k HMI really helps with that). If you have to filter to match the lights then that's only an additional hit against your exposure.
 
Graeme Nattress said:
But, as sensitivity is an important characteristic also, best that you filter if you want to, rather than us permanently filter to bias the sensor away from daylight and loose sensitivity you can never get back.
Click to expand...

Thats a good point, but I also wonder how shooting with a filter would compare to shooting with a tungsten balanced sensor, sensitivity wise. Like how much slower would it be.

I'd love to run some experiments on that, but I can't exactly open up my camera and stick NDs in front of my red and green CCDs :biggrin:
 
I suspect that biasing a sensor for tungsten by driving down the sensitivity would be about the same as putting an 80A filter on the camera; it would be sort of a wash. Or maybe not.

Ultimately the solution may just have to be reducing the noise of the overall system so that one has less noise problems in 3200K lighting. Or rate the camera slower in 3200K, which would be annoying since that tends to be your lowest light level situations.
 
I'm curious about this.

Putting on filters certainly causes more light loss in some channels than others, but it does cause a global loss of light. Its pretty easy to see on my histogram. I can't help but wonder if a tungsten biased sensor (using a design like I mentioned earlier) might be more sensitive when shooting tungsten, with the goal of getting a cleaner image.
 
Why not just balance for daylight and use a blue filter for interiors ?

Would it work ?
 
Antoine Fabi said:
Why not just balance for daylight and use a blue filter for interiors ?

Would it work ?
Click to expand...

Yeah, that can work, but the question I'm asking is this:

With the goal of getting cleaner tungsten footage...

Do you have more sensitivity using a daylight sensor with a conversion filter (32K to 55K).... or is it more sensitive to only hold back green and red light via a specially designed sensor?

Global filtering VS selective filtering of channels.

I've been messing around with this for months, I'd lean towards the purpose built sensor being more sensitive. I'm using magenta filters, and while it helps pull back the green, it also pulls back all the other channels as well, so I'm assuming selective would work better.
 
Unlike film which is balanced to a particular white point by default (say Tungsten), typical digital sensor are not necessarily balanced to any particular white point at the sensor level. However, quantum efficiency may be important in the design of a sensor to aim for certain wavelengths for maximum SNR. It is important to realize that we are talking about metamers (colors that have the same visual appearance) here, as any full spectrum chromacity such as that of Tungsten (close to standard Illuminant A) would be matched by a metamer-based respose, whether it is our tri-chromatic human eye, or camera RGB.

During many stages of internal processing, the individual R, G, B, signals are adjusted in proportion to match any intended white point -- it could be result of picking a preset such as outdoor/indoor/etc., or auto-white balance, or manually by using a gray card (say 18% gray).

Hence, I would concur with Graeme Nattress that it is better to do such operations in post-processing than adjusting the sensor for that, and might loose some sensor sensitivity.
 
The 80A blue filter that converts 3200K light to 5500K balance loses two stops of exposure, so your 320 ASA RED camera becomes an effective 80 ASA, which is pretty slow for a tungsten-lit night interior or exterior.

A compromise may be a half-blue correction with only a one-stop light loss, but considering most interior tungsten film work is done with 500 ASA film, most people are going to want to shoot digital at similar low levels.

Of course you could rate the camera faster to compensate for the blue filter and get more noise, but since you were using the blue filter to reduce noise, it's sort of pointless and you might as well not use the filter and rate normally under 3200K lighting.
 
David Mullen ASC said:
but since you were using the blue filter to reduce noise, it's sort of pointless and you might as well not use the filter and rate normally under 3200K lighting.
Click to expand...

If you wanted to reduce noise in RED in low light interior shots what would you try? Would using daylight kinos and blue cc filters around practical bulbs make any sense? Is there any other way to get practicals looking 4-5000k ish? Or is this train of thought delusional?
 
Given that silicon is poor on sensing blue and nothing can be done about that:

putting a filter on the front to reduce green and red will do just that, but, will always reduce some blue as well - hard to avoid, really. Even if the glass was clear it would attenuate a little.

So, you get the most blue light hitting the sensor with nothing else infront of the lens. That means least noise on blue.

The main reason for balancing ahead of the sensor with a filter is to deal with mismatched clipped highlights, which DRX does a fine job of in software.

If blue noise is a bother, the new noise reduction does a very fine job on chroma noise.

Graeme
 
The F900 has been "tweaked" to make the zero gain level in each channel balanced for 3200K, but of course, it's also rather noisy in the blue channel as a result.

I don't think there's a perfect solution -- you could go for a halfway balance towards blue using camera filters or lighting gels (or daylight lamps too). But perhaps the most practical solution is to just not rate it too fast when shooting tungsten-lit scenes.
 
I have yet to receive my RED (we're expecting our big bundle of joy next week, actually!), but I'm curious how much noise you folks are experiencing when balancing in post to 3200K. Are we talking tiny increments, or a lot of added noise? Since I'll be shooting quite a bit in dark interiors, this is a special concern of mine.

Thanks much.

Stephen
 
I suggested tentatively elsewhere in this forum that Zeiss glass might be particularly suitable for Red because of this very issue..

Michael Lindsay said:
I suspect that since Zeiss glass transmits more blue light they will work perfectly for Red (even if just psychologically) in low light level, low colour temp, situations.. (Remember Red's slightly starved blue channel in low colour temp light)...
Click to expand...

what do people think? Enough of a difference to make a difference.. David?

Michael
 
Michael Lindsay said:
I suggested tentatively elsewhere in this forum that Zeiss glass might be particularly suitable for Red because of this very issue..

what do people think? Enough of a difference to make a difference.. David?

Michael
Click to expand...

Sorry, not enough to make a difference.


A sensor in between tungsten and daylight would still cost you about one stop. It's a good compromise however it does mean that if you using HMI's then you're losing a stop unnecessarily.
 
You must log in or register to reply here.
Back
Top