Exposure question - How do you explain this paradox?

[nightpiper]

From my understanding, silicon sensors have linear response. The companies specially tweak them to have a gentler roll off at the 2 ends. Which i believe is done via the processing algo & not tied hard wired to the sensors.

Edited: Non linear response on semi conductor has very undesireable efx on image acquisition. No engineer wants that can of worms but it has its specific use.

 

[Dream Merchant]

Hmm, I haven't touched my LF gear in literally years and it wouldn't surprise me if my lenses are now fungus ridden... I would go look but it's buried in the back of my store room!

I miss those days, I miss 5x4 trannies, I don't miss lugging the thing around!

I did in the early days briefly use a DSLR to preview the scene exposure wise, but I never found it particularly good because I always used the screen on the back of the camera and that isn't calibrated even to the camera's own images. Between competent use of an accurate light meter and a DSLR I would use the light meter anytime, if for nothing else because the screen on a DSLR is horrible I usually can't recognise the images on my screen later either!

The other problem of course is different cameras are calibrated differently in terms of exposure and exposure response, and then have the aforementioned different tone curves. So for example my D300 has a tendency to overexpose that my D3 doesn't have, while overall the D300 has brighter shadows - and tbh I've been meaning to tweak the curve for ages now.

So it's going to be a little bit difficult to compare a random DSLR with film (and what film? ) And also, digital has more DR particularly in the shadow areas, so I always find film blocks up in the shadows far more quickly, but I'm not sure if that's down to differences in tone curves or just poorer DR (my inclination being the latter).

Remember also if you've used any lens movements on the LF, you'll be needing to compensate for that.

I wasn't aware that some manufacturers calibrate to 12% grey... do you know which ones?

Can't answer your light meter question I'm afraid, the incident ones I've used are all pretty accurate and if they're accurate to even 1/3 of a stop then their performance is linear enough for me

So far from what I could gather, Nikon and Sekonic go the 12% route.

 

[cabbySHE]

Waitng patiently for the final verdict/conclusion of this topic of discussion.
If it is...what shall we do, if not...then how ?

One illusion which many of those who had been learning art or design will know is...when an small element of grey centre within a black backgrd, it will appears lighter, but the same grey on a white backgrd and it will turn out to be darker.

 

[Dream Merchant]

I don't think there is any one conclusion, but several sets of applicable physics at play.

Anyone know if a reference list of DSLR sensor response Gamma curves exist, and where to find it? I did a search but numerable articles of how sensors work come out.

 

[Dream Merchant]

From my understanding, silicon sensors have linear response. The companies specially tweak them to have a gentler roll off at the 2 ends. Which i believe is done via the processing algo & not tied hard wired to the sensors.

Edited: Non linear response on semi conductor has very undesireable efx on image acquisition. No engineer wants that can of worms but it has its specific use.

That's the thing. I keep on reading that digital sensors have absolutely linear response, so I have to assume that any 'S' shaped curve that exists are actually algo tweaks by the engineers.

 

[fotoju]

Ok let me see if I can explain this.

We all know a camera decides on the correct exposure based on "middle grey". And that it uses reflective metering as opposed to incident metering which should be the case for spot on exposure for all scenarios.

- ok first of all unless the camera makers have changed things dramatically, all cameras and light meters 'sees' exposure as 18% grey. In other words the camera references all exposure based on a shade of grey that is 18% grey. You used to be able to buy a calibrated grey card with this reference 18% grey color from Kodak. I think it is still available from places like Cathay Photo or Ruby Photo but you gotta ask them if they have it in stock these days.

Now, I'm sure all of us have experienced "hot spots" or "flashing light" on our camera's LCD when we try to take pictures of something white or bright colored in the sun. That means that part of the scene has been overexposed.

- this flashing light on the camera lcd shows the highlight areas of the image, some of which may or may not be over exposed. If you really want to check if your exposure is correct and whether you have captured all the tones in your picture from black to grey to white check your in camera histogram. That will give you a better representation of what your camera has captured.

Say you are taking a picture of 2 friends. Friend A wears a dark-colored shirt, while friend B wears a white one. After taking a shot, you review your LCD screen and see that Friend B's shirt has flashing highlights to denote over-exposure. But Friend A's shirt is properly exposed.

What do you do from here?

Suppose you only take a picture of Friend A. He looks well-exposed and you can say you've gotten the exposure right. Now let Friend A step out and take only a picture of Friend B. Using the same exposure as that for Friend A, you now find Friend B to be over-exposed.

- what has happened is that the camera is basically a dumb tool. Since your friend is wearing a dark shirt your camera will produce a exposure reading that will render the shirt in terms of 18% grey thus giving you an exposure that will reproduce the dark shirt in a lighter shade from what it actually is. That is why you can see detail in the shirt and you think that it is correctly exposed when viewed on the lcd screen. In other words the camera has over exposed the dark shirt. Now using the same exposure obviously the white shirt being lighter than the dark shirt is going to be over exposed and you will see the white shirt showing up as flashing highlights on your lcd screen.

Normal layman's logic will be that it's expected. White reflects more light while darker colors absorb more light and hence don't radiate as much light as white. But in photography, as I mentioned earlier, it's not the reflected light that determines what the exposure should be (that could work if what we photograph is close to middle grey) but the light that is INCIDENT on the subject.

- of course using an incident light meter to do the exposure metering will give you a more accurate representation of the scene but you will still need to understand how to compensate your exposure reading to best represent how you want to present the scene to your viewers. In other words you will need to decide if you want to expose more for the shadows to get more detail there or to emphasize the highlight areas and get more detail there instead. In many cases if you do not use an incident meter correctly you can end up with lots of underexposed images rather than correctly exposed ones. An incident meter can tell you how much light is falling on your subject but it cannot tell you whether you should deliberately overexpose for shadow detail or underexpose to hold highlight details and sacrifice shadow details. That is for the photographer to decide.

For my example, both Friends A and B are standing under identical lighting. So why should Friend B's white shirt be overexposed while Friend A's darker shirt is properly exposed?

- One way to decide exposure is to put a 18% grey card in the same lighting as your subject then use your camera meter or a reflected meter to take an exposure of the grey card. For this to be effective you need to move close until the grey card fills the entire viewfinder. Then you can use the exposure the camera gives you. Of course if your camera has a built in spot meter you can use that to spot the grey card.

Anyone care to comment on my analysis?

 

[cabbySHE]

This seems to be the conventional approach of which I'd learned. But in order to get details out of the dark and details out of the light area, the contemporary approach would be using the HDR processing method.
If my guess is correct, some camera manufacturer are already working on it.

 

[engrmariano]

use gray card. problem solved.:sweat:

 

[sabee]

Wow some people don't even read what the TS is asking :<

 

[nightpiper]

That's the thing. I keep on reading that digital sensors have absolutely linear response, so I have to assume that any 'S' shaped curve that exists are actually algo tweaks by the engineers.

Definitely linear response for sensors. The roll off at the ends is like what we do in tone curves, software manipulated. Noticed u dun want to make it really close to "S" shape like? Juz a little is all it takes to make the image more contrasty.

 

[diediealsomustdive]

Wow some people don't even read what the TS is asking :<

TS does not appear to be participating anymore.

 

[Michael]

Most camera tone curves are S shaped as I explained somewhere in one of my super-ridiculously-long posts. So not straight certainly. Easiest reference is to have a look at one of the more recent DPR reviews. Also, some cameras with the right software will allow you to set your own tone curve.

I was thinking about the film comparison but not sure I can give you one off the top of my head really; there's probably an article somewhere. From my own experiences the issues are usually very muddied by different DR between film and digital which makes it difficult to compare precisely how the two stack up. Sorry if that's a bit of a copout answer!

i believe the sensor response is linear but the S-curve is a software issue applied later in in-camera post processing

 

[cabbySHE]

According to Ts, she said ...no books has give an explanation on this, then how did Jed and Dream Merchant know so much ?

 

[daredevil123]

According to Ts, she said ...no books has give an explanation on this, then how did Jed and Dream Merchant know so much ?

Must be either engineers by training and profession, or, have friends who are engineers.