abrupt highlights clipping

[Michael H. Stevens]

Bob: I don't understand this 150IRE either. I think we must be comparing apples to pears. When you hit 108% on the EX1 it gives you warning telling you to back off?

Anyway I'm out now to do some more shots to test what Bill is saying, but I hear so many different stories here I am actually being the think these cameras might all be different????

[Christopher Barry]

Bob, indeed, how does it record 150%? I can only interepret the graph as a trajectory, in turn that is how narrow the STD REC709 range is? As for other curves being up to 110%, I myself do not understand how there can be detail if 100 IRE is white 255,255,255.

Bob, do you have any idea what the X value on the graph represents, going up to 600 before stop publishing?

Regarding hooking the camera up to an external 10-bit scope, via HD-SDI port, I was considering a good comparison would be do to this, then with the same camera setup, say, locked off tripood, record to SxS card in HQ mode, then play back and compare WFM display from live 10-bit to the 8-bit recorded file, the playing back out of the camera being 8-bit embedded in a 10-bit stream. My HD monitor has a WFM, so I might try this and compare, when I receive my EX1 in due course. This method was what I had in mind to ensure the 8-bit recorded settings are not misleading visually, because of the 10-bit 4:2:2 uncompressed monitoring direct from the HD-SDI port that I will be using, when practical to do so.

[Bill Ravens]

If I had to guess at an answer, and I AM just guessing, 110% is somewhat of an arbitrary cutoff based on the available bandwidth at 8-bit (bit depth)and allowing a bit(no pun intended) for headroom. With 32 bit floating point processing, the bandwidth really isn't greater, just greater resolution within the available bandwidth, certainly not what 10-bit (bit depth)would get you in terms of bandwidth. In conventional NTSC, IRE100 is not 255, it's 235. In fact, IRE108 is RGB255. And the RGB0-255 quanta are derived from the 256 (1.6 million discreet values) values that 8-bit allows. If one were to go to 10-bit, for example, significantly more quanta are possible...1024?

Then again, IRE 0-100 is a non-dimensional number. With greater mathematical precision and bit depth, one could fit a lot more quanta in the IRE scale. Much in the same way that the FCC keeps increasing the number of available frequencies even tho' the bandwidth of available frequencies remains unchanged. Then, a camera that can see IRE 0-150 simply needs to be remapped into 0-100 for present day NL editors, provided the quanta packets are smaller, anyway.

The fundamental problem, I think, is NOT the camera, but, the software we all have to process the recorded data.
But, honestly, I'm just whistling in the dark, here.

Here's an interesting read about Sony CMOS..
http://digitalcontentproducer.com/hd...titude_111106/

[Christopher Barry]

Bill, for a moment, I forgot about the 16-235 and 0-255 differences.

NTSC REC601 and HD REC709, what range applies to the latter?

Questions... I don't have answers.

[Bill Ravens]

I believe REC709 will allow RGB0-255.

[Michael H. Stevens]

Whistling or not, I think Bill is on to something here; makes a lot of sense. Consider all with my opinion that each profile needs its own Zebra number to be valid. That argument was put down but I spent today experimenting.

I locked the camera down on a scene that has a covered wagon, one side in half light, one side in shadow. Hills behind and a big blue sky. I set the Zebras at the #2=100 and shot every scene just when the zebras had all gone, with every possible combination of Matrix (Standard, HiSat, FL, Cinema) and gamma (Std 1,2 &3 and CINE 1, 2 3 & 4). I then took the best grabs and did the same with them adding in Black Master and Black Gamma.

The first thing I noticed is that while most scenes were similarly exposed, many are very dark, despite all being recorded at 100%. I then split the scenes up and without knowing the settings sorted them into the best ten, then five then the best top three. I know this is subjective to some extent, but defining best as closest to the color graduation of film, the winner was CINE4 at HiSat with a little positive black gamma to pull out the black detail.

I'd post some screen catches if I knew how. How do yer do that?

[Christopher Barry]

Michael, not used it myself. When you post a reply, see Manage Attachments, below? Or, top of the screen: Image Gallery >> Upload ?

[Piotr Wozniacki]

Bob, Bill at all,

I knew my gut feelings were not deceiving me. Have just browsed through the very first recordings made with the EX1 on day one (right of-the-box); unfortunately don't remember what settings exactly I was using (it might be cine gamma, but almost most certainly not Hisat matrix). Anyway, the sky behind foliages is treated correctly; it beats me why anybody should think the ugliness I showed before was a normal thing!

As soon as we have more sun here, I'm going to try and re-create. If it's the Hisat matrix at fault, it probably is badly programmed or for special purposes only (like lowlight/ indoors - without highlights, anyway). If I can't reproduce the clean look below, at least I'll know the camera needs servicing, after all...Or does it have any factory reset switch / key combination?

[Alexander Ibrahim]

Originally Posted by Piotr Wozniacki

I knew my gut feelings were not deceiveing me. Have just browsed through the very first recordings made with the EX1 on day one (right of-the-box); unfortunately don't remember what settings exactly I was using ...

Piotr, trust me- you are seeing mostly a normal photographic effect. It was made more obvious (and a bit ugly as you say) by incorrect exposure. The codec had some effect, but you would have seen it in 35mm too.

I know you are seeing different things now than with the earlier shots, but there are lots of factors, including the lighting, that have changed.

Take your light meter out there. Take an incident reading at the camera, then take some spot readings at the trees, the sky behind the branches and then at the unobstructed sky. I suppose you may want to take readings at the fence- look just go crazy and meter everything.

(Don't forget to calibrate the light meter and the camera together!)

Take a few bracketed exposures for a few subjects, writing down all your settings... not just the picture profile but the aperture, zoom and focus settings- and your meter readings too. Then go back in and evaluate the images with that information at hand.

If you have a DSLR or 35mm camera go out there with that and do the same thing. Heck, try it with another video camera too.

Let me know if you still think you have a problem when you are done with all that.

[Bob Grant]

Originally Posted by Piotr Wozniacki

Bob, Bill at all,

I knew my gut feelings were not deceiveing me. Have just browsed through the very first recordings made with the EX1 on day one (right of-the-box); unfortunately don't remember what settings exactly I was using (it might be cine gamma, but almost most certainly not Hisat matrix). Anyway, the sky behind foliages is treated correctly; it beats me why anybody should think the ugliness I showed before was a normal thing!

As soon as we have more sun here, I'm going to try and re-create. If it's the Hisat matrix at fault, it probably is badly programmed or for special purposes only (like lowlight/ indoors - without highlights, anyway). If I can't reproduce the clean look below, at least I'll know the camera needs servicing, after all...Or does it have any factory reset switch / key combination?

The camera does have a reset in the menus.
All the other recent Sony cameras have a hardware reset function, the EX1 is the only camera I've come across with a software reset. Just be careful as I don't know how much it resets. The reset in the Z1 does not reset the picture profiles. It does reset one thing though, if you buy a "PAL" variant it resets to 50Hz and the opposite if you bought a "NTSC" variant. The PPs each have their own reset in the Z1, haven't investigated the EX1s reser results as yet.

[Piotr Wozniacki]

Yes you're right Bob, there is the Reset option in the menu - not being certain what it does, I won't do it as for now. I have reset all the PP's and will try to recreate the path that lead me from correct to bad results, that made me start this thread in the first place.

[Piotr Wozniacki]

OK, so I reset my EX1, and shot the same scenes (sorry they are boring) with many different settings, but steering away from the Hisat matrix. The two grabs below are just the factory settings. The scopes in Vegas still reach above 100, but the artefacting is mostly gone - just the tree branches against the sky!

I guess two important conclusions (not scientific at all) may be drawn:

1. The Hisat matrix may be good for making dull scenere more punchy (like low light, indor scenes), but not where white cliping may occure

2. While other cameras (well, at least the V1, HC1, Canon A1) tend to spill the backlight into and over contrasty details (making the twiggies even thinner), on the EX1 we can see something quite opposite: the dark objects against highlighted backgroud tend to "overshadow" it, resulting in retaining the original colour of the backgroud which otherwise would already bee blown to white.

Please note that I didn't have to drastically reduce exposure, and the foreground objects are bright enough and NOT oversaturated, like they had to be with the Hisat matrix if I wanted to prevent the trees-against-sky artefacting.

PS. The more I work with the EX1, the more I hate the ND filter switch! Unfortunately, this is not going to be improved with a firmare update:(

[Bill Ravens]

From Adam Wilt's most extensive review of the EX1:

"I’ve found that the EX1’s knee does a fine job except when highlights are strongly colored. Saturated highlights show more hue shift and harsh clipping than I’d like. I’m exploring this further out of curiosity, but even if the knees were perfect I would still shoot with cine gammas, because I prefer the progressive compression to the look of a traditional knee."

"• Standard – the camera’s default color matrix.
• High SAT – boosts all colors by about 15%-20%.
• FL Light – boosts the red-cyan axis by 15%.
• Cinema – shrinks the red-cyan axis by 20%."

[Bob Grant]

Looking at Image83, the whitest cloud above the end of the top of the roof is clipped and has lost detail. It's not visually objectionable because clouds can be all white (no pun intended) but if that was a patch of blue sky you'd be likely back to your original problem.

Which I think goes back to something I said many pages previously. Clipped specular highlights are generally visually acceptable because they're almost invariably white anyway. Your clipped cloud is another example of where only those who go looking notice. It's predictable and visually acceptable. Your previous problem with the sky is predictable but visually unacceptable.

I posted a link to an article on Luminous Lanscapes about histograms and what they mean and how to use them. The author made a very applicable point. A histogram inside a view of the scene is very useful, you need to look at the scene AND the histogram to guage what is going to be the result.

For certain, Hi Sat could bring you undone. If not, a logical question would follow. Why make it selectable. If it's always such a desirable thing why not just have it like that on every camera. Again a good point from the Luminous Landscapes article about a very similar scene to yours. We're trying to fit a scene with 11 stops dynamic range into a medium with 5 stops and it will not fit (well it wouldn't on his DSC or film camera). All the image adjustments these Cinealta cameras have gives us a better shot at squashing it in but it's a risky business. Some of the possible adjustments are as you've noted designed for different kinds of scenes or use under controlled lighting. Many of them are features from high end cameras and many users here and elsewhere have for years be asking why Sony make us pay large sums of money to get a camera with them. Well Sony have delivered but it is a case of buyer be aware.

[Piotr Wozniacki]

Bob, Bill, Leonard at all,

Thank you very much for your input. I don't need to tell you how much relief the output of all that is to me; I was really very frustrated over the last two days!

I feel flattered by the fact that again, my observations coincide with (by far the most comprehensive so far) EX1 review from Adam Wilt; the statement that Bill has quoted:

"I’ve found that the EX1’s knee does a fine job except when highlights are strongly colored. Saturated highlights show more hue shift and harsh clipping than I’d like. I’m exploring this further out of curiosity, but even if the knees were perfect I would still shoot with cine gammas, because I prefer the progressive compression to the look of a traditional knee."

- has been practically illustrated in this thread, I hope. Note the terms "hue shift and harsh clipping"; add the Hisat matrix and you will get frost on bare trees! Or patches in the sky that almost look like ink stains.

Also, I couldn't put it better myself, Bob - your statement:

"Clipped specular highlights are generally visually acceptable because they're almost invariably white anyway. Your clipped cloud is another example of where only those who go looking notice. It's predictable and visually acceptable. Your previous problem with the sky is predictable but visually unacceptable."

- may well be considered the final conclussion of this thread.

PS. As a side note, Bob, the software EX1 reset takes EVERYTHING back to where it was when still in the box, including PP's. But also, of course, any button assignment, zebra settings, and the recording format (with the latter depending on the PAL or NTSC area one is in).

PS.PS. A note to Leonard Levy: thanks Leonard for consistently supporting my opinion on the effect unacceptability throughout the whole thread!