Problem... solution.

[Roberto Lequeux]

As an additional aside, it is also interesting to note that noise often enhances visual processing, and can substantially improve the efficiency with which our brains can encode visual information. So, some noise may therefore be advantageous for our perception of moving images!

Do you mean to say that a little noise makes people "look harder"?

 

[Daniel Browning]

This would definitely increase the chance of photon detection and reduce read noise

Actually it's just the latter.

While current production sensors may well be approaching 90% efficiency in some wavelengths, they are only producing one electron pair. Very difficult to detect.

The photon is detected either way, the only difference is the read noise:

When 1 photon causes 1 electron, and read noise is 6 electrons, the ratio is 1/(6/1), which is poor.
When 1 photon causes 200 electrons, and read noise is still 6 electrons, the ratio is 1/(6/200), which is very good.

In either case, the number of photons detected is the same, and so luminance and photon shot noise, too, is the same. The difference is less read noise. That is, black silicon does not affect sensitivity (QE) at all; nor could it, since we're near the theoretical limits already; however, it it does improve something that is very important: read noise.

Higher QE, if it were possible, would improve photon shot noise in all exposure zones of the image, including highlights, midtones, and shadows, and, more importantly, pushes the read-noise dominated zones down even further into the shadows.

Read noise, on the other hand, affects only the parts of the image that are not already dominated by photon shot noise, which depends on the circumstances of the camera and scene. Typically, read noise only dominates shadows, so that is where something like black silicon would help.

For a practical example, try shooting a Canon 5D2 DSLR at ISO 100 pushed 4 stops and compare it with ISO 1600. The photon shot noise is the same, but read noise is about three stops better (23.5 electrons -> 2.5 e-). The ISO 1600 mode does not have higher sensitivity or higher photon detection, just less read noise.

Kind regards,

 

[JD Holloway]

lol be a screamer in the sack!

 

[JD Holloway]

I think you would agree that we're quibbling about semantics here, as I've already said, the word sensitivity is used indiscriminately. I can't/won't argue your point: I agree with you. I only wish I had a 5D2 to try your test...Or that new FF 700D nikon.

That being said, if Black Silicon (B.S. funny, no?) can help that's great. Call it system-sensitivity, because it has the potential to accurately deliver low light photons to my eyes...reguardless how it happens. :thumbsup:

 

[Daniel Browning]

...the word sensitivity is used indiscriminately.

That's true. It's yet another word in a long line of illustrious words that have lost their meaning thanks to years of misuse. At least in this case we have other words to replace it, such as QE. Other words were not so lucky, and the concept formerly represented by these words has now been lost to all but a few; e.g. aperture, prime lens, ironically, literally.

 

[Mike S. M.]

It seems that the easy way to understand this, for the proverbial lay man, is that if you want/have to use tungsten on a subject, try to use daylight balanced lighting for all your fill light needs. This will provide blues on the less exposed areas, the ones that need it most, and also increase the contrast between cool and warm tones from the subject to the background, which should be a good thing. This is what I've done with the F900 and other cameras, from the Ex1 to the minuscule Canon HG20, and works well, though then again, I've never been afraid to crush those blacks to eliminate noise if needed.

But then where do you keep your white balance? For tungsten?
The EX1 has pretty much the same blue channel problem as the Red One.

 

[Ian Andolina]

Do you mean to say that a little noise makes people "look harder"?

Well, noise seems to make very small contrast differences more visible, so one can make discriminations not possible without the noise. How this would translate into perception of actors[1], I'll leave to the imagination :beer:

Our brains have long been thought to be in a constant battle against internally generated noise - cells in the brain never respond identically to identical stimuli, however it appears that the stochastic (to us) responses enhance complex processing. Perhaps the noise in the blue channel can be used to the advantage of the system capturing the image (be it human brain or Red One [which is a neo-lithic EPIC brain after all!])?

----
[1] I'm interested in what you mean by harder? I think noise in an image can "mean" things outside of its purely perceptual effect, and this will combine together. Perhaps noise in a sequence will remind the viewer of a high contrast grain aesthetic like Raging Bull, and at the same time it will change the detectability of edges and movement at a perceptual level.

 

[Roberto Lequeux]

Oh. And how is it that noise brings out more contrast?

[1] In the same way that the slight stutter of using 24fps likely forces the viewer to engage by having to smooth out the motion by -sort of- creating frames in between. IMO this excites parts of the brain that may not be engaged with faster framerates or real life viewing. I guess I believe that images can to some extent be de-noised in our brains, and by doing this you once again ask the viewer to take more of an active role instead of sinking into their chair and going along for whatever the ride may be.

I also see that there are other values of having noise. I am no DP, but I do think that grain helps tie everything in the frame together by giving it all an underlying common texture. Also, I think that noise must have the same effect, only we haven't seen enough phenomenal pieces of art with high levels of noise, so it may not carry that value of association.

 

[Ian Andolina]

http://www.ncbi.nlm.nih.gov/pubmed/18003873

We found that the addition of random channel and synaptic noise yielded a significant improvement in the accuracy of the network's encoding of light intensity across eight light levels spanning 3.5 log units (P<0.001, modified Levene test). The benefits of noise remained after controlling for several consequences of randomness in the model. Additionally, improvements were not confined to perithreshold stimulus intensities. Finally, the effects of noise are not present in individual neurons, but rather are an emergent property of the synaptically connected network that is independent of stochastic resonance. These results suggest that noise plays a constructive role in neural information processing

Noise can push a subthreshold stimulus over threshold which is why I suggest edge contrast may be enhanced.

Actually, the 24fps flash is highly correlated, and thus I believe negatively interferes more significantly with visual processing than the more uncorrelated sensor noise. Neurons can detect flicker up to 80-100hz and use temporal correlations themselves to encode information, thus flicker responses interfere with their temporal encoding. It is a testament to the brain that we can even see such badly degraded input. I do think higher frame rates will be a great advance for particular types of film making, just as hyper-focus vs. shallow focus is. We obviously integrate patterns of light in spite of the work the brain is forced to do, but then why not shoot at 12fps to invole our audiences even more (chungking express or in the mood for love?), or use Chris Marker's still images of La Jetee to really tax the viewer into extrating narrative from images . I think however a large part of the allure of 24fps is more environmental - we've learned to associate that (degraded) aesthetic to the cinematic experience!

 

[brainburst]

If you're looking for a very warm tone, it's okay if the blues are rather low, right? The problem arises if you try to raise those underexposed blues to make the image more neutral, because the information that used to be all close together at the bottom end of the blue channel gets spread out unnaturally, resulting in noise. Is that correct?

Never throw away information in acquisition when the look you want is achievable in post.
Under exposure of any channel just leaves you with noise. You can always take blue out later, but it will be more difficult to add it later if you never had info to start with. The discussion here is about light not the look.

 

[David Mullen ASC]

I don't really agree - I think whatever gets the original as close to the final look without requiring pushing the RGB channels around drastically, the less noise there will be. If you want a deep orange look, then recording an image with a deep orange look will be less noisy than pushing a white image to deep orange.

You just have to take into account that what is really a neutral image on the RED is around 5000K when deciding what you need to do to record a warm or cold image.

Now with minor tints, then you should be able to record a neutral image and shift one way or the other without much problem. But when you're talking about an extreme correction, it would be better to get that look on the set so that you aren't pushing the chroma or channels around heavily in one single direction.

But this is assuming you know what you want and you don't want to be able to change your mind in post. If you want maximum flexibility to change your mind, then yes, record a neutral image so that you have as much information to work with as possible.

But otherwise, I tend to believe (in an ideal world) that you should know what you want in advance and design the photography to get the look, whether in camera or in post or a combination of both. Not only is that more artistically satisfying, but it tends to get you a better image in post.

However, I also believing in leaving yourself a little wiggle room, so practically speaking, I tend to do a combination of in camera and in post, sort of go halfway there and finish the look in post. Like a dark blue moonlight scene -- I'll light it blue and underexpose it, but not as much as the final image will be. But I don't believe in always shooting a neutral image, not if I know what I want to achieve.

 

[shashbugu]

the blue channel issue is true with most digital cameras

 

[Rich Schaefer]

Since we are on the subject,

How about a tungsten based sensor as one of the Epic options? Jim is it possible to build such a sensor? Perhaps it could be the "Monstro T"

 

[Jannard]

Since we are on the subject,

How about a tungsten based sensor as one of the Epic options? Jim is it possible to build such a sensor? Perhaps it could be the "Monstro T"

Nope.

Jim

 

[bobaandy]

The native daylight white balance is a function of CMOS sensors themselves, not an "option" RED has when manufacturing them.

 

[Evin Grant]

Silicon is just inherently more red sensitive, hence the over sensitivity to the IR spectrum. The best course is just to strive for better sensitivity and S/N ratio for the chip overall.

 

[Graeme Nattress]

Correct Evin.

 

[Antoine Fabi]

Evin,

From which ND filter strenght you think we have to use IR filter then ?

Antoine

 

[albert rudnicki]

I am not Evin, but I'll say that 1.2 is where you should be gearing up with some IR consideration.
What makes things dificult is a sandwiching the filters.
Usually in front of ND you stuck a polar or grad, that will cut more light.
I am personally feeling safe up to .9 with all kinds of other filters.
Recently I've tested ND1.5 IR Pancro with .9 ND Lee (no IR) and shot all kinds of tests with sun directly in the lens. No problems.
Here is an example.

 

[Antoine Fabi]

Hey thanks Albert !!

Very much appreciated.

Antoine