Blackmagic Forum

I just updated to 7.3.2 and immediately found that my standard setup, ProRes HQ, "film" profile, yielded extremely different, and for my purposes, unusable results. I went back to the version 6.9.5 on my Pocket 4k.

Please look at the jpeg I have attached and look at the RGB parade for each shot. I shot this with the zebras pretty close to the same in both images, note that the black levels came in the same and both have zero color correction but look at how much more compressed the new version is.

I need to shoot around ISO 1000 for my interiors, which is what I have this camera for, and expanding the new "film" version introduced so much noise to the footage that for me and my particular circumstances, is totally unusable. I thought I would post this in case someone else is wondering about upgrading and might find it useful.

Gen5 uses a different transfer function than Gen4, it's not a secret.

ISO 1000 introduced the noise not the color science.

Shoot the exact same settings then normalize the footage into 709. Comparing two log formats in their native space is pointless without the accompanying transforms.

Good Luck

I just updated to 7.3.2 and immediately found that my standard setup, ProRes HQ, "film" profile, yielded extremely different, and for my purposes, unusable results. I went back to the version 6.9.5 on my Pocket 4k.

That's not "unusable", this is simply the Gen5 curve, that we all have been waiting for, for months now! It's a highly desirable feature, particularly for those of us who want to get closer to the film look. It has a much smoother highlight rolloff, like the Alexa. It doesn't change anything in terms of color, only in terms of curve.

The changes are fine as part of the Gen 5 color science update. It is very usable and desirable, Howard and Eugenia said it correctly. The noise is again, from shooting at ISO 1000.

Missa like it!

As an alternative to using ISO 1000, try to see if you get better results with either ISO 1250 or ISO 3200.

The point for me is that using ISO 1000 is a necessary evil. I can seldom get by with a lower ISO so the previous version works far better for my needs.

Also please note that the previous version did not add a ton of noise at ISO 1000 and the new version, once color corrected, did.

And btw I did try ISO 3200 and feel that the results were not as good as at 1000. I do know the native ISO settings.

Woodward-Media wrote:And btw I did try ISO 3200 and feel that the results were not as good as at 1000. I do know the native ISO settings.

Along with the explanation in the manual, you can find threads here which explain why iso 1250 is much cleaner than 1000, at the expense of some highlight headroom (which you may not need).

Gen5 shouldn't have any effect on sensor noise at any given ISO, though noise might be initially more obvious in one color science version versus another, as a measure of how the log footage is mapped.

I feel like I have to defend myself a bit here. The point is not that the new color management of 7.3.2 introduces noise but that it greatly compresses the file. Whenever you do that you will introduce more artifacts when you expand the file back to a normal contrast range. Also looking at the parade of an ungraded clip is relevant and important because it shows you how much compression is being used.

In the older generation file note that the highlights are actually continuing beyond the 1023 mark and in fact can be pulled down. In essence that "film" profile uses around 90%or more of the available luma range when writing the file to ProRes. The new 7.3.2 uses approximately 64% of the available luma range and that means it must be expanded much more extensively to get to a normal/pleasing luma range. When you do that you will make the noise in your file more apparent and much harder to manage. You might also introduce banding in some sky areas etc...

I feel like there might be users like me who might benefit from knowing this. I am just trying to put information out that might be relevant to some users.

Woodward-Media wrote:I feel like I have to defend myself a bit here...

Apologies if I’m responsible for causing you to feel that way. We’re just trying to share our experiences as you are doing in order to be helpful. Sometimes it is helpful and sometimes, not so much. All is good.

There are only two isos on the bmpcc cameras. Pushing to 1000 is not different than shooting at 400 and digitally amplifying it (plus, you can do that in post too). Anything apart from 400 and 3200 are not native isos and you don't gain more dynamic range by doing so (the DR graph in the bmpcc manual is counter-intuitive for most users, thinking that they're gaining more DR by shooting in higher ISO).

The compression is the same as in previous versions too (apart the new Q3 addition), it's just that the noise might be a bit more visible now due to the curve being flatter. In the Sony cameras, to avoid that exact problem of noise with s-log2/3, is suggested by the manufacturer to overexpose by 1.5-1.8 stops. You can do the same with the bmpcc, the results will be as you had them before. And with the new exposure tools on 7.3.x firmware, you will be able to expose even more precisely.

And regarding noise: a super-clean digital image is not always desirable. Maybe you think that having a squeaky clean image will sell your properties, or your products, but in reality, the way the human brain works, it usually prefers some analog randomness in it. That's why people add grain back to their videos and pictures, or why people still buy vinyl. Noise is not that much of a problem. It definitely wasn't when watching the dark scenes on Indiana Jones, trapped with snakes under a pyramid, for example.

Woodward-Media wrote:I feel like there might be users like me who might benefit from knowing this. I am just trying to put information out that might be relevant to some users.

What some of us are trying to tell you is, you're mistaken in your understanding of the difference between gen4 and 5. The sensor is capturing the same data in both cases. The difference you've noted is not a difference in the underlying data but in the way to the two color sciences present the image in log form. As an aside I was just grading some Canon c70 footage -- you might be shocked at how restricted the log waveform appears. And yet there's a whole lot data and dynamic range there.

Assuming exposure for middle gray, the noise characteristics of various isos, as well as the difference in distribution of stops, is also reasonably well understood by now. But you reject this information as well.

Might shooting in BRAW be an option for you? You can switch to Gen 4 (the old color science) in post there and check if that helps you.

Woodward-Media wrote:I feel like I have to defend myself a bit here. The point is not that the new color management of 7.3.2 introduces noise but that it greatly compresses the file. Whenever you do that you will introduce more artifacts when you expand the file back to a normal contrast range. Also looking at the parade of an ungraded clip is relevant and important because it shows you how much compression is being used.

In the older generation file note that the highlights are actually continuing beyond the 1023 mark and in fact can be pulled down. In essence that "film" profile uses around 90%or more of the available luma range when writing the file to ProRes. The new 7.3.2 uses approximately 64% of the available luma range and that means it must be expanded much more extensively to get to a normal/pleasing luma range. When you do that you will make the noise in your file more apparent and much harder to manage. You might also introduce banding in some sky areas etc...

I feel like there might be users like me who might benefit from knowing this. I am just trying to put information out that might be relevant to some users.

I understand your point, but I would like to know more about your grading workflow.
Do you grade prores manually? Cts? LUT? What color space?

New gen5 is very different about contrast curve, but also be cause curve is the same for all iso, in past was different, and if you delog in a wrong way, may be that you cause more noise enhancement than you want.
I understand that prores is not 12bit raw, but is a 10bit file, to create banding on sky mean that could be something of wrong in delog workflow.
Anyway is better (if you need more cleaner picture) to go in 1250 which is second gain, more cleaner than 1000 and eventually go down in post.


Inviato dal mio iPhone utilizzando Tapatalk

Thanks for the input. I will test ISO 1250.
I grade manually. I have many years of professional still photo retouching experience, including pre press, so I am pretty good at grading etc. In fact I wish many of the tools available in video editors were available in Photoshop or Capture One!

I appreciate the input and will test again when I have more time. I am in serious promotion mode right now.

Eugenia Loli wrote:There are only two isos on the bmpcc cameras. Pushing to 1000 is not different than shooting at 400 and digitally amplifying it (plus, you can do that in post too). Anything apart from 400 and 3200 are not native isos and you don't gain more dynamic range by doing so

Thats how I see it too. And not only for the pockets. And thats why I only shoot either in 400 or 3200. Any other need for exposure I do with an ND, lens iris, etc.

I will test ISO 1250.

ISO 1250 belongs to the second set of ISOs, and its highlight performance is very bad compared to the first set of ISOs. You should be shooting in iso 400, and then overexposing a tiny little bit instead if you want to avoid noise.

I think it goes without saying, if he's got enough light to over-expose at 400, it would make no sense to set ISO 1000 and expose for it.

The presumption (mine, anyway) was that he was @1000 for lack of light. In those circumstances, 1250 will be much less noisy, at (again) the expense of a reduced number of stops above middle-gray.

Beyond what others have advised (which is also all valid), I'd encourage you to use/try Blackmagic RAW if possible since it has 12bit non-linear encoding and won't fall apart as quickly as the 10bit ProRes options when pushed around in grading.

The log curves in Gen 4 and prior were optimised around 10bit workflows (and hence changed per ISO and per camera), but since Blackmagic RAW is now seeing far more use on our cameras than ProRes we could use the extra bit depth with Gen 5 and have a single log curve that covers higher dynamic range for the various ISOs/cameras and simplify workflows/grading.