Serena...

Thanx for having done this. And thanx for providing your test method.

I'm using cine2 all the time, because I don't want to have to deal with above 100IRE information. Also cine1/2 looks most linear/neutral.
Cine4 has heavy highlights-compression (too much in most situations). Cine3 also has a lot of highlights-compression, but additionally the darks get pushed down a bit relatively to cine1/2 leading to contrasty low-mids. Therefore cine3 actually is the most complex curve.

Btw, I guess that the dynamic range/latitude gets maximized with cine1/2, because the noise of the ex1 is too much that black-stretching could actually retrieve extra-information in the darks which else would get lost because of 8bit-quantization. But heavy highlights-compression of cine4 together with 8bit quantization will definitely crush highlight-information.

Thank you Serena, Bill, Alister and all who have shed more light on the Cine Gamma settings.

Thankfully we have this forum to express the need for this thread, as well as the good folks who are willing to generously give of their time-talent-experience for the benefit of all. Cheers!

i second that. i wouldn't be where i'm at today without the contributions from all of you. i've learned more in the last year since i bought my EX1 then just about all the previous years combined. I don't really consider myself a Newbie anymore. maybe a veteran Newbie.

This is the Alan Roberts review of the Cines in the EX1.
I have copied and pasted this so I hope that this is ok?

Cine2 is the only curve suited to production without grading, since it clips at 100%.
Cine1 is similar but copes with overexposure by extending beyond 100% video level.
Cine3 and 4 differently share the contrast range, use these to taste.

If using Cine1, 3 or 4, make sure that video will not be clipped in post-production. Or that grading can cope with the over-voltages.

Here is a plot of the four cine gammas plus standard 1 gamma. Given in two forms: the usual log of subject brightness, plus with a linear subject brightness which is the form in Sony's sketches.
The data was taken with the following conditions:
subject surface brightness: 16000 lux; 5600K; white; broad; diffuse.
Effective subject brightness at 1.0 on curves 900 lux approx.
WB: 5600K
camera gain: 0dB
shutter speed:1/50
HQ mode, 25fps
PP properties: default with knee "off" and changing only the gamma selected.
When iris was closed the brightness reading was 3%. This could have been adjusted using black level, but I left defaults as is.

The LCD shows image % brightness to a whole number, so accuracy decreases with decreasing reading. The readings are repeatable within 1 digit (generally get the same reading).

Thanks serena.
Looking at the graph's I really like the curve of Cine 2.
I'm really suprised that Cine 1 peaks at 110 as I thought the top end droped off a bit in brightness.

Thanks agina for your effort.

Here is a plot of the four cine gammas plus standard 1 gamma. Given in two forms: the usual log of subject brightness, plus with a linear subject brightness which is the form in Sony's sketches.
Fantastic info, Serena. Thank you very much.

I'm happy to see that Cine2 uses a knee rather than a hard clip to be asymptotic to 100%. The drawback is the "somewhat" flattening of the midrange(contrast reducing), relative to C1, to get the knee that I wonder about.

Sometime back when I first tried to evaluate the gamma curves the SAW curve generation facility made me suspect that the EX camera is natively balanced to 3200K. Still I don't know the answer to that, but the SAW facility did suggest that the gamma curves vary with WB. Out of curiosity, using my more direct method, I've looked again at the effect of changing WB and the results are attached.
I used the same subject source and added a full orange gel, which cut the source by a stop. The camera white balanced at 2800K, so I accepted that.
The effects are greater than I expected. The camera showed greater sensitivity and the results are plotted to retain relative subject brightness of 1.0 matching the previous plots. STND 1 gamma exhibits a steeper roll off (or knee) prior to clipping.

In the 5600K plots the plotted dynamic range is 10 stops, with some unreliable data going down another half stop with black half a stop further down. So you could say 11 stops, with a bit of a stretch. However at 2800K the range (with the same considerations) is 12 stops, so the BBC statement of 11 stops dynamic range certainly holds under tungsten.

I'm interested in any comments on these observations.

Very interesting work. For what it's worth, if you're going for accuracy I would think an offset of 400k from 32k is starting to push it a bit, assuming you're correct about the native 32k design. I'm sure you know temperature change is detectable at 200k intervals to the eye, becoming very obvious as when reach 500k. If you can't hit 32k +/-100k personally I would draw the line at 2900k for your test. I also think it'd be interesting to see how great the effect of this variance is on the results.

Fair enough comment, Max. In this instance I was more interested in observing any change and had not expected the extent of it. Now it might be worth redoing for 3200K.

I'll recheck anyway, because that apparent difference makes me a little uneasy that there may be an error in data reduction (compensating for the change in source luminance).

whoa....this whole test has set me kind of on end. I just never considered the gamma change when resetting white balance. I shoot exclusively at 5800K. If I want to dial out too much yellow, I do it in post. So, as long as the gammas I'm studying are good for 5800K, I should be OK.

whoa....this whole test has set me kind of on end. I just never considered the gamma change when resetting white balance.

Bill, it may be that I'll be embarrassed by being too quick in publishing that result. Have to find some time to check, so make no changes to your approach in the meantime!

Just for the ones that are having a hard time understanding these charts (probably only me on this entire forum) your units ascending on the left are IRE levels right? The subject brightness levels in 0.5 increments is what is confusing me the most though. If someone could just help me understand how that bottom line works that would be most appreciated.

Sorry for the dumb question in a technical thread guys...

There are no dumb questions! Sometimes dumb answers!
The vertical axis is IRE, although I understand that term as really belonging in the analogue world. So 100 is broadcast max and generally our digital cameras capture up to 109 before clipping. The horizontal axis is subject brightness relative to a chosen test set up. As the subject waxes or wanes (in brightness) so you move along the horizontal axis, each halving or doubling representing 1 stop change in brightness. The test point brightness was chosen to get 109 using cine gamma curve 4. You'll see that the standard gamma curves (1,2,3,4) reach 109 at much lower than that subject brightness.

Measuring the data is easy if a little tedious. In reducing the data allowances have to be made for changes in the test source, such as adding gels to adjust colour. I suspect my plot of the effects of WB show that I adjusted in the wrong direction, so greatly exaggerating any effects. I'll let you know! In the meantime, don't worry about the WB effects plot.

The horizontal axis is subject brightness relative to a chosen test set up. As the subject waxes or wanes (in brightness) so you move along the horizontal axis, each halving or doubling representing 1 stop change in brightness. The test point brightness was chosen to get 109 using cine gamma curve 4.


Ahhh I get it now! Thanks for explaining Serena

I've looked at the effect of changing WB and the results are attached.
I used the same subject source and added a full orange gel, which cut the source by a stop. The camera white balanced at 2800K, so I accepted that.
The effects are greater than I expected. The camera showed greater sensitivity and the results are plotted to retain relative subject brightness of 1.0 matching the previous plots. STND 1 gamma exhibits a steeper roll off (or knee) prior to clipping.



Unfortunately, for me, these interesting conclusions are wrong. I did indeed muck up reducing the results of that test, compensating in the wrong direction for the effects of the added gel. I redid the measurements for 3200K and differences in the curves between daylight and tungsten are small (measurement error band). Apologies for the trip up the garden path.

whew...

thanx for setting things straight. It was a mind boggler, for sure.

I have been shooting weddings with C1 (Bill's settings), detail off and have had good results. Yesterday I used C4 instead (all other settings unchanged). I found the image substantially "softer". Is there any reason that this may occur?

I calibrated my lens after the last firmware upgrade but that was 4 weddings ago. Is it possible that the calibration needs to be repeated?

Thanks,
Mervin

I shot some footage going the other way Mervin by going from Cine 4 to Cine 3 and found Cine 3 to be softer. Oh I forgot that I shoot 720/50p with Cine 4 but on this shoot I shot Cine 3 1080/50i.
I have found that Cine 1 combined with Cine 3 yeilds the best combo in my opinion.

I calibrated my lens after the last firmware upgrade but that was 4 weddings ago. Is it possible that the calibration needs to be repeated?

Thanks,
Mervin

Hi Mervin,

When you say calibrated your lens I guess you're talking back focus. I can't say this is the cause for softness but for interchangeable lens cameras the best routine is to calibrate that not only when a lens is mounted but also after the camera is moved/transported or there is an environmental temperature shift of a few degrees. Not always practical but recommended. If you have an EX1 perhaps this rule can be relaxed (fixed lens) but I would check it before every shoot really as preflight.

Here is a plot of the four cine gammas plus standard 1 gamma. Given in two forms: the usual log of subject brightness, plus with a linear subject brightness which is the form in Sony's sketches.
The data was taken with the following conditions:
subject surface brightness: 16000 lux; 5600K; white; broad; diffuse.
Effective subject brightness at 1.0 on curves 900 lux approx.
WB: 5600K
camera gain: 0dB
shutter speed:1/50
HQ mode, 25fps
PP properties: default with knee "off" and changing only the gamma selected.
When iris was closed the brightness reading was 3%. This could have been adjusted using black level, but I left defaults as is.

The LCD shows image % brightness to a whole number, so accuracy decreases with decreasing reading. The readings are repeatable within 1 digit (generally get the same reading).
Serena Thank you so much for sharing this! ;) and everyones inciteful help.

I don't know how you guys derived at the "conclusion" that Cine gamma 1 is for bright/outdoor scenes, and Cine gamma 3/4 are for dark/indoor.

My tests have shown the following.

If you expose for skin tones at 70-75% in Cine 1, you will overexpose the white background/bright sky much faster, than you would in Cine 3 and 4.

If you set zebra 1 to 75 and enable zebra 2 which is at 100, you will expose the color chart "Caucasian" square at the same time. So 75% skin = 100% white at same exposure in Cine 1

Set the gamma to Cine 3, the white square will be sitting at 97% with the skin square is at 75%

Go down to Cine 4, and the white square will sit comfortable at 95%

This all means that you will have a better chance of maintaining your bright background and deep blue sky in Cine gamma 3 and 4.

I would also argue that the same logic can be applied when the situation is turned around.
In Cine 1 you will have more exposure in the poor lit shadows when shooting indoor, while Cine 3 and 4 will be more prone to crush the blacks.
The only reason for using Cine 3 and 4 indoor must be that the curve allows for less light and therefore less gain.

This all depends on what your subject is and therefore where your main exposure lies. But in most situations we are aiming the camera at a person.
75% exposure being your target, you want to make sure that the next most important thing, either the sky or the dark room is as close as possible.

Finally you can use Cine 2 for scenes which could introduce banding in the 8bit colorspace. Since Cine gamma 2 compresses the image to 100% white clipping (which btw can be changed back to 109 if you set the gamma level to -5), You will effectively have a better "dynamic range resolution" pr stop.

Same goes for all of you who use -3 DB gain.. If you set any of the Cine Gammas to -4 (cine gamma 2 to -9) You can maintain the dynamic range at 109%

Cinegammas are designed to be graded. They are not designed to be used in the same way as you would with a standard gamma and you need to adjust where you put your skin tones accordingly. Whenever you take a large range of something (in this case brightness range) and squeeze it into a narrower range (by displaying it on a TV or Monitor with a standard gamma) you are compressing the image.

Cinegammas will look flat when exposed correctly on a standard TV or monitor because the monitors gamma curve does not have the range to correctly display what the cinegamma has captured.

To restore your cinegamma footage to something that look best on a standard TV you need to grade it. Given that the main aim of the CG's is to capture more highlight information, and some of the mid-hi range is used for this extra highlight data, then you need to be prepared to move your CG midtones and skin tones upwards to traditional levels in post.

If you shoot using a CG and you have your skin tones at the traditional 70 ire (which is rubbish anyway as everyones face is different) then those skin tones are in the compressed part of the gamma curve and no amount of grading will ever make them look as good as they should. You must keep skin tones and natural textures at a lower level when using the CG's to avoid this compression. When using CG's you need to modify the way you expose to get the most out of them.

...If you shoot using a CG and you have your skin tones at the traditional 70 ire (which is rubbish anyway as everyones face is different) then those skin tones are in the compressed part of the gamma curve and no amount of grading will ever make them look as good as they should. You must keep skin tones and natural textures at a lower level when using the CG's to avoid this compression. When using CG's you need to modify the way you expose to get the most out of them.

Rats. Just when I thought I understood what to use. I've been using Cine3 and 4 setting the skin highlights to 60-70 using false colors. Alister, for caucasian skin in a studio with black background, what ire would you use for skin highlights?

I'm exposing faces up to 65, and I'm good with it.

But that's just me,,, Alister?

Here is a good article by Adam Wilt on the effect of Cinegamma on images.

ProVideo Coalition.com: Camera Log by Adam Wilt | Founder | Pro Cameras, HDV Camera, HD Camera, Sony, Panasonic, JVC, RED, Video Camera Reviews (http://provideocoalition.com/index.php/awilt/story/ex1_gammas/)

Cinegammas are designed to be graded. They are not designed to be used in the same way as you would with a standard gamma and you need to adjust where you put your skin tones accordingly.


Yes I realize that, but it does not change the fact that you OVEREXPOSE your bright background quicker (If exposing for Caucasian skin tones at 75%) when using cinegamma 1.. Even beyond 109%, then you cant grade yourself out of that.

I'm exposing faces up to 65, and I'm good with it.

Wow, I didn't realize everyone in Poland had the same shade of skin. In my country, we have whites, blacks, latinos, asians, and all different kinds of skintones to deal with. 65% on one guy might be fine, but totally wrong for the next guy. No matter how you slice it, using zebra on skintones just comes down to guessing because every face is different. Sure, an experienced shooter can still guess right most of the time, but it's still just guessing. I can do that without zebra. Zebra is one of the most important functions on a pro video camera, but it is only correct if you know the value of WHAT you are measuring -- and usually that should be white card, gray card, or chip chart. Anything else is just guessing.

Wow, I didn't realize everyone in Poland had the same shade of skin. .

Ha, ha :) Actually yes, Doug - 99.9% of this nation is of the same race, and similar skin color.

I just wanted to stress that - in spite of the common "recipe" for the Caucasian skin tone to be exposed with zebra at 70-75% - after some experimenting, I settled with 65% as being safer from decolorization due to highlight compression.

Piotr,
Well, one of these days I'm going to get over there to Poland and see the faces for myself! I've traveled all over the rest of Europe better never made it to Poland. It's on my list.

Either I don't understand you Dominik or I believe you are wrong.

The standard gammas (without knee) will give you about 2.5 stops of head room above 70IRE while the cinegammas give you around 5.5, stops of headroom. Total DR for standard gamma with no knee is about 7.5 stops and cinegammas is 12. A well set knee will extend the standard gammas out to around 10-10.5 stops giving you about 5 above 75IRE. If you take a look at the plots prepared by Serena earlier in this thread, or the published curves you can clearly see the extra headroom the cinegammas give you. But the key principle when using the cinegammas is that you should lower your midtone levels to keep them in the more linear part of the curve so that you can make use of the gradually compressed highlights. Taking something that with a standard gamma you would have exposed at say 75ire and now exposing that at 60ire gives you around an extra 1.5 stops of useable headroom. Then in the grade you sort out you contrast range to give the tonal range that you want.
You don't want to do this with a standard gamma plus knee because the knee area is either compressed or not, there is nothing in between and due to the sharp onset of the knee you don't want the knee point set too too low, so to get good dynamic range you end up using just the range from 85ire to 100 (109)ire for the knee (15-19% of recorded data) for 4 to 4.5 stops which is a lot of compression. As a result if you try to grade this it will look bad as you have not really used sufficient data to enable you to extract useable picture information.
Cinegammas on the other hand start the compression at around 30ire with it steadily increasing especially above 65ire. As a result the same highlight range that with the knee is squashed into just 15 to 19% of the available data is now spread over 35%-39% of the recorded data so you have much more data to work with so the information recorded can be manipulated in post without so much degradation, thus making this useful information that you can do something with. In addition the gentle increase in compression means that more data is allocated to the areas that are most important, gradually decreasing as you approach overexposure and clipping.

Alister

Apparently you don't want to understand me because I never mentioned Standard gammas.. I'm sure you are a experienced guy, but try to read my post for once before you begin taking it apart.

I will repeat for you once more. CINE gamma 3 and 4, will give you more headroom to the bright sky, if you are exposing for a caucasian skin tone.

Even when grading, you would appreciate that the second most important subject in your shot (the sky in this example) laying closely exposed to your first (the person in front of it). Otherwise you are just "fixing" in post what you did wrong in the field.

If thats what you like to do. ok. It still doesn't make it optimal.

And I repeat. No mention of Standard gammas in this post.


----
Im sorry for sounding arrogant, but it annoys me endlessly when people pick apart my arguments, and making me sound stupid, with fancy "I know it all" answers that don't even relate to what I wrote in the first place.

Ok Dominik, I've re-read your post and I apologise for bringing standard gammas into my reply, but I still don't believe you understand the way the Cinegammas are supposed to work.

I don't know how you guys derived at the "conclusion" that Cine gamma 1 is for bright/outdoor scenes, and Cine gamma 3/4 are for dark/indoor.

My tests have shown the following.

If you expose for skin tones at 70-75% in Cine 1, you will overexpose the white background/bright sky much faster, than you would in Cine 3 and 4.

If you set zebra 1 to 75 and enable zebra 2 which is at 100, you will expose the color chart "Caucasian" square at the same time. So 75% skin = 100% white at same exposure in Cine 1

Set the gamma to Cine 3, the white square will be sitting at 97% with the skin square is at 75%

Go down to Cine 4, and the white square will sit comfortable at 95%

Your treating the CG's as you would standard gammas in your tests and thats not how they are supposed to be used. For starters straight shooting with CG's without grading will lead to a flat, un-natural image on a monitor due to the large miss-match between the cameras cinegamma curve and the monitors gamma curve (rec-709 for an HD monitor).
The reason why CG's 3 and 4 work better for darker scenes is not to do with the top end of the curve but the middle and bottom end. CG3 and CG4 lift dark and mid range parts of the image, improving shadow areas and dark areas and giving you more to work with in post in these darker scenes. Exposure is not just about highlights and skin tones. As an option when shooting a dark scene this would allow you to have skin tones and mid tones at a lower ire level and still get a good looking image after grading as by reducing the exposure a little you will not be compromising you shadow and dark areas as much as with CG1.
Trying to fix skin tone exposure at 70-75ire (which is too high even for standard gammas IMHO) is not the way to use Cinegammas or Hypergammas. Not only do faces vary from person to person, but the brightness of faces will change from scene to scene. On a bright sunny day a face would typically be bright and well illuminated and should be exposed as such, in a dark scene the face might well be much darker and should be exposed darker to keep the mood of the scene.


Finally you can use Cine 2 for scenes which could introduce banding in the 8bit colorspace. Since Cine gamma 2 compresses the image to 100% white clipping (which btw can be changed back to 109 if you set the gamma level to -5), You will effectively have a better "dynamic range resolution" pr stop.

That's not correct. Cine 2 is recording 12 stops into the standard 100 % range that uses bits 16 to 235. The other cinegammas record the same exposure range but now using the full 109% recording range (same stops- more range) that includes the extra bits above 235, so now your using bits 16 to 255. So MORE bits are being used by CG's 1,3 and 4 for the same range, not less, that's why they are allowed to go to 109.

Same goes for all of you who use -3 DB gain.. If you set any of the Cine Gammas to -4 (cine gamma 2 to -9) You can maintain the dynamic range at 109%

On the majority of cameras using -3db gain reduces the dynamic range of the camera by -3db (half a stop), EX1/EX3 included. It can however bring a -3db noise improvement. Using -3db gain and then adding that gain back in again buy adjusting the curve level is pretty pointless as you will also increase the noise again, you may as well just stick to 0db. Adjusting the curve level adjusts the gain level of the curve and thus noise levels as well.

Alister, for caucasian skin in a studio with black background, what ire would you use for skin highlights?

In a studio, where I have control over the lighting, for a shoot for television, I would try to use standard gamma 3 with no knee and use the lightning to keep everything within the standard Rec-709 range that standard 3 gives. This should give the end viewer the best result as the contrast range displayed by the TV should match that of what you have captured. You won't get any gamma compression on skin highlights and it should look very natural. In this case for a typical caucasian person you would be looking at 65 to 70ire. If your struggling with specular highlights and reflections blowing out then you could introduce some knee with a fixed point at 85 to 90, but as highlight go into the knee the usual inevitable knee color shift gets introduced giving that nasty yellow hue to over exposed skin tones.

If you are shooting for a higher dynamic range output, or plan to grade then I would switch to Cinegamma 1 unless it's a low key scene in which case I would most likely use CG4. I would want to keep any skin tones below the more highly compressed part of the curve, so below 65ire, typically around 60ire depending on the shoot. This will look a touch under exposed when shooting. That way you keep any highlights in the more linear part of the curve so they will look more natural once you've done your grade, overexposed cinegamma faces don't look good.

Alister, for caucasian skin in a studio with black background, what ire would you use for skin highlights?

Hey, do you mind if I take a shot at answering this question, too?

The term "caucasian" skin has no meaning. Show me two white people, and I'll show you two different skin tones. 65% might be great for one, and over-exposed on the next. By using zebras, I would set my exposure based on either a white card or gray card of known reflectance value. And in this example, the background is irrelevant. The face should be exposed the same whether the background is white, black, pink, yellow, or a green screen. The face is the subject, and that is what matters.

Thanks Doug. I mentioned the black background because I want the blacks to crush. I mentioned Caucasian because I expected the technique for judging exposure based on highlights to be different in the Caucasian range than dark skinned. And so I was wanting to draw out that technique as I was shooting that today and would adjust the technique I use.

What I've been doing is lightIng the scene using a mannequin so as to get the relative intensity between fixtures as I want them. I then adjust the iris using false colors so that the highlights on the face have no yellows indicating under 70 (I use the Marshall 7" monitor). If my iris is in the desired range, i'm done. Otherwise I adjust the lighting or nd so I can get the iris where I want it.

When the talent shows up, I adjust the iris to achieve he same false color reading on their actual skin (no yellow). So with this technique I think I'm dialing the same level on the highlight Thus compensating for skin to es (wihin a range at least) and letting the relative intensity of the fixtures handle the rest.

Unless I misunderstand how you are describing your technique, it sounds like you are saying everyone's skin tone should result in the same false color display. Is that what you are saying? Because, if so, you are wrong. If false color works on a fair-skinned white guy, it can't also be correct for a darker skinned latino, someone with a good tan, or a black person. Your technique does not take into account different shades of skin. False color and/or zebra on skin tone tells you nothing about the proper exposure because you don't know the reflectance of the subject's skin. Ultimately, you're just guessing at the exposure.

Proper exposure is more complicated than just not blowing out the highlights. If you applied that same logic to sound, then recording a whisper and someone shouting should both be just under the redline on a VU meter. Obviously that would not sound right. And exposing everyone's skin tones at the same zebra or false-color level (no matter what the level is) is wrong too. A darker person should look darker than a lighter person, and if you exposue them both to the same IRE, at least one is going to be wrong.

And what about adjusting the Gamma Level? I remember once reading that (+) darkens mid-tones and (-) lightens mid-tones, yet the opposite appears to be true when looking at my camera's LCD.

Read through all this thread which is super informative but I was really searching to an answer to what the above poster asked . The same here when I change the cinegamma to the + side it seems to compress highlights even more, I don't get the midtone modification since nothing in the midtones seems to change.
This setting is mostly visible when I have something a tiny bit overexposed and then bringing the gamma into the + side starts compressing it and at high levels it starts looking bad, somewhat to solarizing and when I bring it into the - side it gets more and more overexposed but probably the steps are more stretched, i'm guessing the gammas start clipping earlier as in 100% instead of 109% .

The gamma gain control is a gain control. Consider 3 different points in your video signal. 2%, 40% and 80%.

If you double the gain what happens to those points? They become 4%, 80% and 160%. So you will see a massive difference at the bright end of the image but very little difference in the mid and even less at the dark end. That's why the gamma gain control always appears to have the biggest effect on highlights.

The + settings increase the gamma gain, so highlights will clip earlier. The - settings decrease gain, but as the cinegammas are already tailored to retain the maximum dynamic range a negative setting does nothing more than reduce the maximum recorded signal level. This can be useful if you want to prevent your levels going above 100%, but otherwise is of little real benefit.