That's a horrible S/N ratio... even s-video has a s/n ratio of 45db...

I wonder if there was'nt a mistake?...

I couldn't for the life of me, find the info you found at that link....

Suppose an EX3 is recorded (via HD-SDI) to a 10-bit format. Given a 54 dB ratio, how low can for example a face go? Normally a face would be say 70% (e.g. by camcorder zebras) but with 10 bit and a camera having 10 dB less noise than a Z1, could it go down to 7%? Just trying to get a feel for this.

Thinks:

Although the signal format may be 10 bit, there is the question of what is the true levels resolution (from camera's sensor and subsequent in-camera processing) and whether this varies with level (e.g. I wonder if it might diminish at lower levels). In any system, the number of "decimal points" is not necessarily an indication of accuracy.

Results from an initial experiment by others [http://www.dvinfo.net/conf/sony-xdcam-ex-cinealta/135394-ex1-8bit-vrs-10bit-uncompressed-test.html] appear not all that promising. Further tests by other people [http://www.dvinfo.net/conf/sony-xdcam-ex-cinealta/130855-hdsdi-out-420-a-2.html] lead to inconsistent conclusions.

I wonder if anyone here has tried boosting (e.g. by Color Curves) some deliberately underexposed levels from EX1/3 shots recorded at10-bit and found this practical or indeed just whether they have observed any advantage over doing the same thing to the same shots recorded 8-bit.

Just in case there's a misunderstanding.... I was only talking of generally underexposing a stop or 2....

@tumbleweed2: Here's a better link:
http://www.bbc.co.uk/rd/pubs/whp/whp-pdf-files/WHP034_ADD20-Sony-HVRZ1E-FX1E.pdf. I edited it in my earlier post now also. Again, it's towards the end.

The author specifically says the performance is not up to the standards of "proper" HDTV cameras, where 54dB is expected, but the performance is not particularly bad for general production, only a problem for keying and matteing.

I highlight this not to rubbish this classic camera but to know its limitations so as to understand how best to use it.

Why would one underexpose at all?

Exposing so the highlights just kiss 100% isn't under exposing.
If that leaves everything else in the dark then you need to address your lighting.

You need to be very careful with the use of complex gamma curves and uncontrolled lighting. You can find out that any clipping can create very ugly effects such as color banding. I've been there, done that myself and I'm not the only one.

I haven't really tried seriously underexposing the EX1 by 1 or 2 stops, that's quite a lot of dynamic range you're throwing away there. RAW images from the SI-2K can be almost black in post with a linear curve. Sure it's amazing to see the image come up out of the 'negative' but that's a very different outcome to what's being discussed here. Much of what I see is more to do with the capablility of my 6bit LCD than any mojo from the camera. Curiously enough with cameras like the SI-2K and RED lighting becomes more not less critical from my experience. One thing I really don't understand about lighting with these cameras is any shift in CT is much more obvious.

I was on the sidelines of some test shots for 3D with 2 SI-2Ks and the DP was going nuts over a green cast in the shadows under the talent's chin. I eventually found the cause. Light leaking into the test area from uncorrected fluros in the office nearby.There was 2KW of light on the subject. I would have thought that enough to wash out the tiny amount of leakage from the fluros but not so.

Bob.

I think the simple and best answer would be to do a test for yourself. Bracket multiple exposures, grade it, see what you like.

See:
http://www.glennchan.info/video/exposure/exposure.htm

Factors to consider:
-Making the image darker in post lowers noise. Lowering superwhites into the legal range gives you free dynamic range (normally superwhites get clipped off).
- Video knee interaction.
- color correction
- etc. etc.

Simple solution to your overexposed teeshirt, 'bag' it.
Soak garment in tea for a few minutes, don't rinse, dry with hair dryer. Problem solved. Now you know why they're called tea shirts :)
Everyone should have, hair spray, baby powder, tea bags, water and basic makeup kit with them to deal with these problems.
Bees wax is also good for shiny fabrics. Just don't leave it in your kit bag in the car on a hot day, what a mess.

Bob.

"Exposing so the highlights just kiss 100% isn't under exposing."

I'm being misunderstood here....

I'm refering to under exposing whatever your subject is in the frame, not just highlights....

I think we all understand that once you've clipped your whites, you've lost any detail in it, & can't get it back...

Glenn, you've made some fantastic contributions to this and other forums. You have a great deal of technical knowledge. You've probably forgotten more things I'll never know (I'm not "technically" inclined).

In the page of your comparison images (link provided above) you even went so far as to clearly state that what "looks best" is subjective. No one can argue that!

Now, having said all that, I must disagree with the statement: "Clipping is ok if the image looks better." In reality, that's an oxymoron. With certain exceptions (there's always the exception), clipping in the highlights far more often that not indicates "amateur."

The human eye is automatically drawn to the brightest area of an image, any image--video, photograph, painting, graphic design, etc.--whatever it may be. If that brightest area is not the focal point, then there is the excellent chance the image fails to accomplish what it was designed to achieve.

The videographer's best, as Bob pointed out above, is to expose for the brightest area in the picture (I have my zebra set at 95%). Let things fall where they may. We should not be afraid of black.

You can see a perfect example of what I’m talking about here in video #2. I have forgotten the name of the cinematographer (maybe Bob or Serena would recognize him—he’s an Aussie), regardless, the gentleman gives very sound, practical advice.

Jay, great to see these again! Got me straightened-up when I saw them for the first time too.

Was it the Sydney Film festival? He and his chum gave some talks and this was a precursor for people to then go out and deliver a a piece of video? Was that it? Something like a 24 hours delivery?

This has got me thinking about when and where . . .

Grazie

@Glenn: Thanks for that link to your experiments, they address exactly what I was wondering about. For Image 5 did you levels-correct using S-curves (Color Curves) or plain (linear) Levels mapping?

Indeed it is worth experimenting further. For the Z1 I understand (from the BBC testing) that use of the (progressively saturating) gamma curves in that camera actually reduces the latitude, so for that camera it is presumably better to apply an S-curve (to compress the superwhites while retaining some of their detail) in post. However it remains to be seen what the (broadly similar) gamma curves do in the EX3.

@Jay: The Videos at Good Dog indeed illustrate clearly how manual exposure is preferable to auto and how richness of colours and shadows can be artfully enhanced. Very nice instructional examples. However, that point is outside this thread since in those scenes the brightest objects (sky, buildings) were also the main subjects.

It's when those reasonably well-lit but relatively (to the brightest things in the scene) shadowy things are the main subjects that the dilemma is in session. In that circumstance it seems (from the discussions here) that the best option is to expose for those subjects and (sadly) let the bright stuff clip, rather than raise the (shadowy) main subjects in post. I notice that is how most TV interviews seem to have been shot.

@farss: I take the point that for controllable scenes (e.g. interviews and studio work) that make-up etc. has a role to remove/reduce the highlights. Possibly I took this a bit too far when applying handfuls of mud to my face and hair, admittedly more for effect than for levels, for a photo sample in a desparate attempt to land myself a background part as a battle-worn soldier. Only served to block the sink...

It depends on what you mean by 'underexposing' and 'well lit'.
You should not expose so that the brightest thing in the frame is at 100% unless it happens to be white or a specular highlight, you'd most likely get some pretty ugly results.
Skin is typically set to expose between 60 to 70%, that's why good cameras have zebras set to or adjustable to that point.
If the subject is dressed in black their skin will be the brightest thing in the frame and should not read 100%. That doesn't mean they're underexposed.
If they happen to be wearing glasses that are producing a small glint off the glass or silver frame you don't decrease exposure to avoid clipping that. In this scenario the shot is not overexposed.
Correct exposure is determined by knowing the sensitivity of whatever is recording the image and measuring the amount of light falling on the subject. The correct exposure for a scene that is all black or all white is the same. Reflective surfaces do complicate things. One could argue that if they're causing problems then the scene is not 'well lit'. You might be able to ignore them if they're just tiny specular highlights (glints). You might have to seriously wrangle them is they're shiny fabrics worn by talent, more so if they're not pure white.

So to answer the original question as best I can: There is good reasons not to under expose relative to correct exposure, a well lit scene, when shooting video.

Bob.

I believe there from a lecture series by this man:
http://www.pieterdevries.com.au/
They may have been presented during a lecture series for videocraft.com.au or Grazie might be on the money, they may have been for a workshop during the Sydney Film Festival.

Bob.

A recent post of mine I think shows where I'm coming from:

http://www.sonycreativesoftware.com/forums/ShowMessage.asp?MessageID=624345

Since I posted that I've changed the CC curve slightly to pull the top node down a bit to keep the highlights (from reflections off shiny skin mostly) below 100%. On some TVs I was seeing ugly color shifts in them.

Bob.

I take your point Bob, it's just that most of my work is on live events which are not so controllable, like sometimes I have to accept the venue's existing lighting arrangement. Typical lecture theatres I have shot in bizarrely have most light falling on the pedestal, much less on the presenter.

Of course I do recce's etc. and try to get the lights (or presenter) repositioned when I can, or supplement the lights by my own ones etc. but quite often this is not possible or acceptable to the venue or presenter (doesn't want to be dazzled, wants to see the audience) stays or progressively during his presentation wanders more and more out of the beam.

For shoots like this I have to accept what I am given and how to make the best of it becomes the issue. Very frustrating to an ex stage lighting man like myself. "Silk purse from a sow's ear" is a phrase that comes to mind, but that;s just it.

By "well-lit" I meant not requiring anything other than zero gain in the camera, to avoid any confusion with the low-light issue discussed on another thread. I didn't mean it in the sense of a job well done. An ambiguity of the language I used.

,,,and thanks for the link.

Now, having said all that, I must disagree with the statement: "Clipping is ok if the image looks better."

To some degree, all shots will have clipping in them since there's usually some highlights (especially specular highlights) that are just way brighter than everything else. And then on the bottom end you have noise obscuring shadow detail (and glare hitting the viewer's monitor). So ultimately you have to decide how much shadow and highlight detail you want to throw away. And usually it's right to throw away detail in specular highlights and such (e.g. glints off metallic objects or reflective surfaces).

I think I've said this before but anyway
For events we expose the for the brightest lighting and leave it at that. Sometimes that can fail spectacularly. White satin dresses in fog lit by deep blue light and strobes, YUCK.

You do what you can. We used to shoot operas. The only good opera footage I've seen the local crew spent a week relighting the production for the TV show.

Bob.

Fausse,
I used color curves and not levels.

Some information on my method:
http://www.glennchan.info/articles/vegas/color-correction/tutorial.htm

The curves:
http://www.glennchan.info/articles/vegas/color-correction/curves-and-secondary-presets2.veg

farss wrote For events we expose the for the brightest lighting and leave it at that

Exactly what I do several times a year as well.
My thinking is that, if the stage gets dark, it's dark for the audience so why should the TV viewer be any different.
If it gets really dark, I may open my iris one stop but that's about it.

BTW, I agree with the statement "Clipping is ok if the image looks better."
I do a lot of talking heads and always expose for the face, no matter what the background is.
Sometimes I get lucky and have total control and other times I'm shooting someone in an office with a lot of windows which, depending on the time of day and the weather, can mean an overexposed background.
On a waist shot of the interviewee, I don't care if the background gets blown out as long as the exposure on him/her is OK.

Mike

usually it's right to throw away detail in specular highlights and such (e.g. glints off metallic objects or reflective surfaces).

I can agree with that. Specular highlights, depending on size in the frame, would be acceptable. However, I think something as large as someone's clothing, i.e., a white shirt, is way too much--hard to ignore.

As I've said too often, I like the look of film. In the context of this thread the particular characteristic of film is that S-curve which flattens response in both shadows and highlights. The basic video response is much more linear with significantly restricted dynamic range, so over exposing highlights (ie clipping) leaves "holes" in the image. In the real world it isn't possible to avoid scenes of excessive dynamic range and clipping is often unavoidable. You see it frequently in broadcast TV (e.g. white clouds, bright lights) so I wouldn't say this screams amateur video; it does say 'video'. I agree with Glen (and undoubtedly Grazie) that what looks right is right, and I expect to achieve that through the combination of camera-work and post processing (as also in using film).
One thing to watch carefully ( a matter alluded to by farss) is that not all the RGB sensors clip at the same time. This was shown very nicely in a thread on another forum concerning unrealistic image artefacts when trees with sky were over exposed. We see this much less dramatically where skin tone (Caucasian) highlights take on an apricot plastic appearance; much more objectionable than clipped clouds or Glen's t-shirt.highlights.
What do do then? I believe that if people are in the scene then you must expose for skin tones and build everything around that. No people (or people in LS) then expose for best fit to scene dynamic range -- if shadow detail doesn't matter, let it go. Clouds, let them go. If you're shooting sail racing with lowish sun on white sails it's hard to let them go but with a Z1 the mid range will be under-exposed badly and have noisy low-tones when corrected. But hey, film has grain so I'm tolerant of a bit of noise (looks OK to me!).
Given that mostly we're talking about cameras recording to an 8 bit dynamic range, we need to do as much in-camera as possible (plus lighting if that luxury is available). Graded filters are valuable for making it easier for the camera, but often not applicable. Generally the range of the sensors is a lot more than 8 bits, so if the camera allows then this is the area to be mastered. Setting knee (where the slope of the gamma curve is reduced for highlights) is the 'standard' fix for extending highlight range, allowing good mid-range exposure without clipping important highlights; however does not prevent hard clipping. A much better approach (IMO) is that of the hyper-gamma curves in Sony higher end cameras (cine-gamma 4 in the EX series) -- these roll off highlights so over-exposure doesn't (unless severe) look like a hole in the image (loss of detail is gradual). Not forgetting controls at the toe of the gamma curve (black gamma, etc) that enable separate adjustments in recording shadow detail. However images shot with cine-gammas are not intended for direct viewing, just for recording a wide dynamic for post processing.
That plastic skin look? Keep skin at about 60% on an EX. Now you've got me going I'm going to do an experiment using using a Wratten #85 filter shooting with a WB of 3200K, because I think the EX is natively set up for 3200K. Why on earth?? Excellent question, and I'll let you know any result. Comments are welcome, naturally.