Sony zebras are soooo conservative (A7 II)

[Guillermo Luijk]

Today I realized my A7 II's zebras (highlight clipping warning in pre-visualization) are independent from white balance. This is weird since different WB settings mean different over exposure values applied to each cannel, so could zebras be related to RAW data? the dream of RAW shooters... I wanted to find out the relationship between the zebras, the JPEG built by the camera and the RAW data so I shot this scene:



Dissappointingly those zebras don't correlate with the RAW data nor even with the camera JPEG:




They are even more conservative than a genuine JPEG clipping warning, displaying as blown areas that are fine both in the RAW and JPEG files (look at the area on the left of the lamp for instance).

Regards

[ErikKaffehr]

Hi Guillermo,

Thanks for sharing. I see the same on the A7rII.

The Sony Alpha 99 I had was exposing nearly ETTR, and the histogram was pretty accurate, not so with the A7rII it tends to underexpose, or rather not utilise the full dynamic range.

Best regards
Erik

Today I realized my A7 II's zebras (highlight clipping warning in pre-visualization) are independent from white balance. This is weird since different WB settings mean different over exposure values applied to each cannel, so could zebras be related to RAW data? the dream of RAW shooters... I wanted to find out the relationship between the zebras, the JPEG built by the camera and the RAW data so I shot this scene:



Dissappointingly those zebras don't correlate with the RAW data nor even with the camera JPEG:




They are even more conservative than a genuine JPEG clipping warning, displaying as blown areas that are fine both in the RAW and JPEG files (look at the area on the left of the lamp for instance).

Regards

[kencameron]

The Zebra can be set to values between "75" and "100+". It looks as if those are just about how visible the zebra is rather than allowing any adjustment of the exposure values which it shows. I am wondering if there is another way of adjusting the latter, but can't think of one.

[E.J. Peiker]

When set to 100+ and shooting RAW you can go 2/3 stops brighter from the last exposure that does not show zebras without risk of overexposing the RAW file.  What I do is I reduce exposure until all hints of zebras disappear and then add 2/3 stop.  This gives a proper ETTR file.

[AlterEgo]

A7 II's zebras

to get a proper clipping in raw indication with zebras/blinkies in Sony A7* the minimal set of parameters is :

WB = UniWB
Creative Style = Neutral (contrast -3, saturation -3, ...)
Zebra = 100+

[Guillermo Luijk]

But the zebras didn't change at all when changing WB. Will that be different with UniWB?

Regards

Enviado desde mi GT-I9195 mediante Tapatalk

[AlterEgo]

But the zebras didn't change at all when changing WB.

they do

for example I set my A7R2 to M mode, I set zebras to 100+ and when I change WB I can see how the areas covered by zebras are changing when I change WB... so do a proper test yourself, it is very easy - just go to menu and start selecting various WB settings, you will see how zebra coverage varies in real real time

here is a clip = https://youtu.be/WMMfsd78UUI


.

[Guillermo Luijk]

Indeed I did a test but changes were so slight between different WB settings that I thought they were static!. I'll learn how to live with these conservative zebras

Regards

Enviado desde mi GT-I9195 mediante Tapatalk

[Jose Viegas]

One test shot viewed with RawDigger:



Only very close to clipping, so at the limit of ETTR.

Result observed with the Zebras 100+:

- UniWB: Just some tiny zebras barely visible on the lightest areas.
- Neutral Profile -3 contrast: Zebras visible on the lightest areas. Gone if exposition lowered about 1/3 stops. So, in this case, no zebras, 0% clipping.
- Standard Profile: Zebras very visible. You get a great JPG image if using this profile, exposing until just the zebras start to show. If shooting RAW, it is possible to increase exposure 1 to 1 1/3 stops after zebras start to show with this profile.

I personally use almost always the Neutral Profile with -3 contrast and I have the UniWB saved on a memory bank to use when critical exposure is required.

[Guillermo Luijk]

Log histograms are not a good idea to check RAW clipping with accuracy. Those highlight peaks are probably saturated or nonlinear in the sensor, but for some reason (RAW compreesion?) didn't reach the maximum saturation clipping value.

Anyway, I just wanted to find out if the zebras could be RAW related. If they are not I will live with it. I don't want anymore to adapt my JPEG files to have extra RAW exposure accuracy. Too bad no camera maker implements genuine RAW histograms. The first in doing so would be very welcome in the RAW shooting community.

Regards

[AlterEgo]

I don't want anymore to adapt my JPEG files to have extra RAW exposure accuracy.

why ? if you try to get better S/N in raw you screw OOC JPG anyways  ... if you are averse to a green tint in the viewfinder you can replace it by a greyscale desaturated one by addition a certain picture style on top of UniWB based settings...

alternatively for people you can use the fact that Sony A7* cameras with face detection and matrix metering expose a face (white, caucasian) with green channel in a certain area below clipping in raw, for example A7R2 = ~2.7 EV below clipping, so you can dial exposure correction based on that.

alternatively as you know yourself you can do spot metering, for example A7RII spot meters ~3.7 EV below clipping in raw (green channel, daylight) - you can run few tests for your model...

alternatively you can switch settings effects off - you will not have green (or greyscale) tint in EVF/LCD, but you still keep blinkies doing zebra work in shot review

[Hywel]

Very interesting posts, thanks everyone.

I've been using the A7RII with 100+ zebras and standard picture style, looks like I can be a little less conservative - at least I should be safe with +1/3 of a stop when the Zebras appear, and maybe even more.

As always in fluid lighting situations I'd rather play safe than sorry even at the risk of losing shadow detail- highlight detail is generally much more important to me.

ETTR is definitely the correct principle, but I'd rather be two thirds of a stop under and definitely not have any clipping than a third of a stop over and have one channel start to fry. Even at the expense of wasting a considerable amount of dynamic range and increased noise in the shadows.

One of the great things about the A7RII is that it is a low noise sensor with wide dynamic range. It's a lot more forgiving of underexposure to protect the highlight headroom as a result.

The only camera I've ever used which does this right is a RED. They have clipping "traffic lights" for R, G, B channels based on the raw data off the sensor. I really can't understand why camera makers don't implement this. Compared with processing the raw data to make a JPEG on the fly, it's EASY: just check every sensel as you read it out and increment a count if it is more than (say) 99.99% of maximum. And that's it; that's all we need. Map out any permanently hot sensels, and set a warning in the viewfinder if any channel has more than (say) 0.01% of total sensels clipping.

Or just the build the histogram channel by channel as you do the processing, rather than a posteriori based on the JPEG once you've built it.
The histogram is surely much less data than the JPEG you're building to display the picture, so what's so hard about it? Adobe, maybe add a standard format for RGB histogram data to DNG or something to encourage people to do it!

 
Cheers, Hywel.