Zebra/ISO independence?

Hi-

First post, thanks for having me.

I've got the BMC on rental for a project, and I'm noticing that zebras do appear to respond to ISO settings. I was under the impression from reading the forum that this would not be the case, and that zebras are simply an indication of what's going on at the sensor level, not the metadata level. However, I've got the camera sitting in front of me at the moment and ISO/zebra interaction is indeed what I'm seeing.

This is my first day with the camera so it's quite possible I'm missing something. Just to check my math here, here's the setup: I've got the camera on RAW/Film and the display set to Video. ASA is set to 1600, frame rate 24p. Framed on a low-wattage practical table lamp with a Tamron 28-300 EF mount zoom at 28mm and wide open at f/3.5.

So: Cranking up the shutter angle to see when clipping sets in, this doesn't happen until 360 degrees. (Impressive!) Jumping into the menu and dropping the ASA to 800 causes the zebras to disappear. Alternatively, maintaining the ISO at 1600 and closing down the aperture one stop also causes the zebras to disappear.

Is this the expected behavior?

Thanks!

From what I understand, when shooting in raw, which iso you set the camera too is irrelevant and doesn't actually affect the image you're recording. In which case, the zebras changing between different iso values is just for a frame of reference depending on how you're planning to grade it.

Zebras are all about exposure..

They appear according to how you set them up. For example.. if you set your zebras to appear at 100%. as you iris up the camera.. the zebras will appear when the sensor indicates you're exposure is at 100% IRE..

So what you are experiencing is that you are at 1600 ASA and your shutter is at 360 degrees.. as you open up the aperture of your lens.. the zebras will appear as you hit 100% ire of exposure.. they disappear when you close your aperture or lower the ASA and the exposure falls below 100% IRE.

Thanks for the input!

Yes, I have the zebras set at 100%. Both the manual, and John Brawley's comments here on the forum seem to suggest that this is the optimal way to determine when the senor is clipping. If you're shooting RAW, it makes perfect sense that the zebras would reflect what's going on with the senor, and not what's going on in the metadata.

However I'm now at my desk here in the office with the camera pointed at a trio of large sunny windows, with a more practical shutter angle of 180 degrees, and seeing the same results: 1600 ASA zebras, 800 ASA no zebras. This appears to be the case whether the LCD is set to display film or video, or whether you're recording in ProRes or RAW.

If this is typical behavior, how would you determine when the sensor itself is actually clipping?

The 100% zebra shows sensor clipping (which you usually want to avoid) period.

No camera is such a no brainer, when it comes to perfect exposure as the BMC.
Just make the 100% just disappear and you are golden - it's as simple as that.

For sure -- but let me phrase the question another way. Why does reducing the ASA make the zebras disappear? Is ASA a metadata setting, or is it actually controlling the sensor electrical gain?

For sure -- but let me phrase the question another way. Why does reducing the ASA make the zebras disappear? Is ASA a metadata setting, or is it actually controlling the sensor electrical gain?

ASA for RAW doesn't mean anything, it's just metadata. It'll always be at 800ASA.
ASA for ProRes gets imbedded into the file.

I hear you. But if ASA does not control sensor gain, and zebras only reflect the sensor, then changing ASA should have no effect on zebras. But it does -- at least on the camera sitting on my desk. Is it possible that this behavior was different in an earlier firmware? Can anyone confirm whether zebras appear/disappear when ASA is changed on their particular camera?

I hear you. But if ASA does not control sensor gain, and zebras only reflect the sensor, then changing ASA should have no effect on zebras. But it does -- at least on the camera sitting on my desk. Is it possible that this behavior was different in an earlier firmware? Can anyone confirm whether zebras appear/disappear when ASA is changed on their particular camera?

It's just changing to help you meter what you're shooting.
It doesn't actually DO anything to the image other than give you a reference.
Think of it this way.

When shooting in raw, think of it at as 800ASA film.
You're always shooting at 800ASA.
You can set your camera to 200ASA, and it'll meter for that, but you're still shooting on 800ASA.
If you were to bring the image into post, at 800ASA, it would look much brighter than you intended.


Set it to 800ASA and forget about it. Then you'll have a better range on either end to push or pull your image.

Edit: I still don't have my camera so I can't check for you. Sorry! Hope someone else chimes in.

So do we think the zebras are really only accurate for RAW capture when set to 800 ASA, since that's where the sensor itself is set? Further, do we think the zebras are accurate at all ASA settings when shooting ProRes since that's what's getting baked in?

If so, these are important details worth mentioning in the documentation.

Just double-checked it. When I set my zebras to 100% they never change, no matter what ASA, no mater if raw or compressed, they are always the same.

Which makes perfect sense, since they show sensor clipping at hardware level, that has nothing to do with the recording format or metadata (ASA) settings. When the sensor clips it clips.

Super, thanks for checking that -- very helpful. Something must be going on with the camera I've got. Either that, or I've failed to operate the simplest camera out there. I never underestimate my capacity for user error!

If you have a Zebra set to 100% on the BMCC then it will always show sensor clipping.

Some people have seen slight differences between 1600 and 800 (but not 400 and 200) in terms of the way the DR works.

When I've pressed BMD about it, the engineers have told me there is a TINY difference in the way they map the 16 bit linear of the camera into the 12 BIT LOG of DNG @ 1600 ISO. So while there is the same DR, they allocate more bits to the bottom end (shadows). The idea being that any noise reduction later might be cleaner.

What this seems to sometimes mean is that there can appear to be a difference in where something ZEBRAS @100% in highlights, but I think all your seeing is a slight difference in the bit allocation.

If you take a bare lamp to JUST on the point of clipping, you shouldn't see any difference on the zebra between 800 and 1600 though. If there is, I'm going to go out on a limb and say it's wayyyyy smaller than a *stop* of difference.

What is amazing with this camera, and I've true;y yet to see this on other cameras, is that something that is near clipping still can look fine and be recovered. A lot of other cameras will have something up near 90% and they might as well be at 100%/

jb

Thanks for the details John. I spoke to BMD tech support about this yesterday, and you are correct. They confirmed that there is indeed something going on with recorded exposure when ISO is adjusted. Following the "Zebra And Back Off" method, bumping ISO from 800 to 1600 can drive hot parts of the frame back into zebra if they're close enough to clipping. For me it was one or two aperture clicks, a third or a half stop probably. I've done the bare lamp test and the open window test, and it's easily repeatable.

So... if the camera is ZABO'd (ha) at ASA 800, and then bumped up to 1600, it's a good idea to check the zebras again to make sure nothing's over the line. As a new user of the camera, it was confusing to watch an entire window jump from no zebra to completely blown out (overcast day -- the window was basically a giant grey card), and if for no other reason I think this is worth including in the documentation.

Thanks!

For the record, I understand what you're saying.

The zebras, at least before, were not supposed to change in RAW mode if at 100 percent, if you change ISO but I've observed that they do. I can understand that for ProRes, curious if something changed in the firmware as well.

You can easily check this by clipping something at 1600 then switching to 200.

However, I believe that if you simply rate the camera at 800 when shooting RAW, zebras at 100 percent will always tell you what's clipping. I've personally begun to rate the camera at 400 and 800, and meter at 1000 or 1250 for anything extremely lowlit.

EDIT: looks like John responded while I got on about typin'.

With that in mind, it's interesting that there is a change in the RAW when rating at 1600 and that it may be worth simply using 1600 for lowlight. Hmmm~

With that in mind, it's interesting that there is a change in the RAW when rating at 1600 and that it may be worth simply using 1600 for lowlight.

This is good news then for trying to get the best out of a low light situation or at least get more detail or less noise in the shadows when that is important while shooting raw. I had thought it was best to leave the ISO at 800 all the time for raw, but now it's best to judge if you want to protect more highlights (ISO 800) or reveal more in shadows (1600). It may not be a full stop difference as if it was a real film camera, but even a half stop is significant. Still expose to the right to avoid clipping highlights is the general rule, but shift to a higher ISO when you want as much in the shadows as you can get without clipping highlights.

Rick Lang
Sent from my iPad using Tapatalk HD

Thanks for the in-the-trenches corroboration Kholi. I think they should update the docs with some more details on ISO vs sensor, if only to eliminate confusion for new users.

I just returned the rental, but after a couple days with the cam I generally get where it's coming from exposure-wise. Great image when well lit, but gets grungy fast in the shadows. Given the sensor size this is understandable. We were doing nighttime interiors, and even when the subject was exposed, backgrounds descended into noise and colors faded pretty quickly. We ended up pumping the whole room a couple stops with some tungsten flood bounce, raising key/fill/hair lights accordingly, and bringing everything down in post to get the darker vibe back.

Also, everything looked worse in ProRes, which was notably softer than RAW. Whenever possible I would shoot RAW with this camera, and give it as much light as possible.