How does white blow out at 90 IRE?

[Greg Voevodsky]

I'm looking at some ocean wave footage that I shot (I thought correctly) underexposing the whites to 90 IRE with a pretty strong role off in the settings.

Normally, this should save the detail in the whites. Instead the foam is blown out at 90IRE?

How can this happen? How come it is not recoverable in Color?

Is it the KNEE SETTINGS I HAVE AT Point 85, and Slope 85, and SAT 20 in STD 3.

Normally this setting works great - yet on this one beach (not even in full sunlight) - underexposed, it blows out my whites at 90 IRE - what the heck?

Am I misunderstanding the KNEE settings? Thanks.

[Leonard Levy]

Probably Yes - That's the danger of using excessive knee settings for general purposes. It looks good on some shot you set up to test and then ist completely screws up something else. Knees are very dangerous.
Either go with the auto knee ( which is too active for my taste as it can change your highlights in the middle of a pan) or use a conservative general knee setting. I use point 93 or so with a slope of 50. Some shots might benefit from a stronger knee, but that setting won't bite you.

[Tom Roper]

Rules:

- Leave Zebra2 on.

- If you have to go higher than +50 on the slope, point is too high.

- If you have to go lower than 70 on the point, exposure is too high.

[Gints Klimanis]

Any chance the -3dB setting was used?

[Greg Voevodsky]

Originally Posted by Gints Klimanis

Any chance the -3dB setting was used?

Yes - how does -3db factor in. I also have black at -5 too.

Thanks to all the responses. Im saving those recommendations now, and will do some more tests. Thanks. Im assuming Zebra 2, is 100 IRE correct?

Any other setting recommendations for bright sunlight and white surf and sand are highly appreciated.

So - I take it my knee compressed and blew out by whites (rather than preserving them) - evan at 90 IRE. I thought you can only blow out whites above 100+ IRE?

But I guess that is wrong?! Funny is my settings worked on 90% of the beaches I shot just fine.

[Jon Sands]

"Studying the SAW display in HDRack's waveform monitor, it is apparent to me that a gain setting of -3dB is NOT equivalent to the "real" 0. A GAIN setting of -3dB limits the superwhite IRE to a value of ~90 IRE. Only the 0dB GAIN with 3200K for a white balance yields a neutral, full scale curve"

[Leonard Levy]

Its impossible to know exactly what happened in your shot without looking at the footage so we are only guessing.

Usually we think about losing our white areas by blowing them out in overexposure by getting them over 100%.

However the other danger is too have too radical a knee setting, or perhaps too radical a cine type gamma curve which might behave a similar way. The problem here is not that you go over 100%, but that you compress your highlights so much that they just sort of lose detail and/or color quality. Excess compression is very ugly. Its a dance up the highlight area.

Lenny Levy

[Greg Boston]

Originally Posted by Leonard Levy

Usually we think about losing our white areas by blowing them out in overexposure by getting them over 100%.

However the other danger is too have too radical a knee setting, or perhaps too radical a cine type gamma curve which might behave a similar way. The problem here is not that you go over 100%, but that you compress your highlights so much that they just sort of lose detail and/or color quality. Excess compression is very ugly. Its a dance up the highlight area.

Was thinking the same thing, Leonard. The loss of detail in a highly compressed (almost horizontal) portion of the response curve in the highlights can be quite ugly. I can see how it would be mistaken for 'blown out' white.

-gb-

[Jon Sands]

Let me repeat:

"Studying the SAW display in HDRack's waveform monitor, it is apparent to me that a gain setting of -3dB is NOT equivalent to the "real" 0. A GAIN setting of -3dB limits the superwhite IRE to a value of ~90 IRE. Only the 0dB GAIN with 3200K for a white balance yields a neutral, full scale curve"

[Serena Steuart]

I think the SAW curves can be misleading, insofar that the obvious interpretation isn't always correct. My recent example of my making that error was the thread about "blowing out skin tones", where I used the SAW results to point out that the RGB paths will clip successively (except, perhaps, at 3200K). I am correct in that statement, but said that B will clip first (because the SAW test, at 5600K, shows B greater than R) -- in fact R clips first. The in-camera SAW test appears to show limits below 109 IRE for various gamma settings, but these are artificial. It is true that you can set limits below 100IRE, and cine2 limits at 100. If you add more than +25 gamma to cine4 that will limit to around 90, with the horrible highlights described by others. My general setting is -3dB cine4 (no added gamma) and that doesn't limit the response (that is, 109 clipping is available to me!).

[Jon Sands]

It's a well known fact that negative gain settings reduce dynamic range. If the sensor is receiving maximum light input and outputting its maximum voltage, but then being ran through negative gain, your 109 IRE could now very well be down to 90.

[Leonard Levy]

Jon, I don't understand the Saw curves (wish I did) but rather rely on just looking at waveform results which are probably more practical as they directly show you what your real world images are like. On that score, I have never seen that kind of reduction in output when using an EX-1 @ -3DB, nor has anyone else reported it.
In the Panasonic SDX900 that is a significant issue though and I was advised by every tech I knew to avoid -3 DB. It didn't reduce you output to 90, but it did reduce latitude in the highlights. I haven't seen that on an EX-1 but am open to the idea if you could confirm it with real world tests like waveforms of actual images.

[Serena Steuart]

Originally Posted by Jon Sands

It's a well known fact that negative gain settings reduce dynamic range. If the sensor is receiving maximum light input and outputting its maximum voltage, but then being ran through negative gain, your 109 IRE could now very well be down to 90.

Well, it isn't. I think you're assuming an analogue gain/sensor-voltage relationship with the ADC, whereas I think the gains are relative and have effect much further down the digital chain. However I have no detailed knowledge of the processing, so I guess we can each be satisfied about that -- it's the pictures that matter.