Blackmagic Forum

This makes no sense as the file sizes are larger and should be higher quality. I found this from doing my dynamic range test between my Sony A7RIV and my BMPCC4K on a sunny day with dark shadows. I recorded at 12:1 on the P4K because it's the best file size I can work with considering it was 20GB for around 10 minutes of footage at 4K24.00. Any reasons for this? Because if I have to shoot at a much lesser compression to match the lack of blocky leaves of an 8 bit Sony (near bottom of the barrel for color grading) then I might need to look for another camera with a more efficient compression process. I do like the colors out of the P4K but I really don't like what it's doing with these bushes. Take a look.

In the comparison shot, Sony leaves on left, Blackmagic on right.

Nobody can draw any conclusions at all from those screen shots! smh

Braw needs NR Spacial or Temporal or both, depends your taste and distribution objetives. Even, could need some sharpness or contrast. If you have skills with color corrections there is multiple ways to get the results.

Easy way, not accurate, Luts.

If you want some image with pre processing you can try Pro Res flavors directly from camera.

8 Bits Sony Avchd mp4 has pre processing into the camera, NR, limiting alternatives in color post processing, but give you a final image. The advantages of Sony sensor are low light capabilities.

Yes, Sony is doing quite a bit of filtering in the camera.
Compare a RAW still out of the A7S with a video: at close examination it's worse.
And then, BRAW is an I-frame only codec, while the Sony is doing GOPs. You can't easily compare the file sizes.

javier forza wrote:Braw needs NR Spacial or Temporal or both, depends your taste and distribution objetives. Even, could need some sharpness or contrast. If you have skills with color corrections there is multiple ways to get the results.

Easy way, not accurate, Luts.

If you want some image with pre processing you can try Pro Res flavors directly from camera.

8 Bits Sony Avchd mp4 has pre processing into the camera, NR, limiting alternatives in color post processing, but give you a final image. The advantages of Sony sensor are low light capabilities.

I did sharpen the BMPCC4K and Sony footage as I have sharpening in the Sony camera set to -7. The sharpening only made the macroblocks more clear.

JonPais wrote:Nobody can draw any conclusions at all from those screen shots! smh

Alright... I will get better stills from Resolve and also show the macroblocking that is happening on my hair.

Uli Plank wrote:Yes, Sony is doing quite a bit of filtering in the camera.
Compare a RAW still out of the A7S with a video: at close examination it's worse.
And then, BRAW is an I-frame only codec, while the Sony is doing GOPs. You can't easily compare the file sizes.

IDK how filtering would make macroblocks look like detail. This is right out of DaVinci Resolve. No rendering. In the render the same phenomenon shows up. Will post more stills.

Alexrocks1253 wrote:

JonPais wrote:Nobody can draw any conclusions at all from those screen shots! smh

Alright... I will get better stills from Resolve and also show the macroblocking that is happening on my hair.

Are you sure this is 'macro blocking' and not aliasing / moire? The P4k is an 8mp sensor with no OLPF.

John Griffin wrote:

Alexrocks1253 wrote:

JonPais wrote:Nobody can draw any conclusions at all from those screen shots! smh

Alright... I will get better stills from Resolve and also show the macroblocking that is happening on my hair.

Are you sure this is 'macro blocking' and not aliasing / moire? The P4k is an 8mp sensor with no OLPF.

This is macroblocking. Moire would bring false colors in and weird patterns that aren't like this. I will post more images, though PNGs are too large for here I guess.

It's as Uli explained, BRAW compresses each frame individually and the Sony calculates the differences between frames (an oversimplification for the sake of brevity) to achieve higher compression. There's tradeoffs to both methods once you decrease the bit rate. You should always use the appropriate bit rate for what you shoot.

Here are some better frame grabs. Full PNG quality. In the comparison one, Sony left, Blackmagic right.
https://drive.google.com/drive/folders/1WWcULoyJNSK0aiyND8KXcc0BGmI95CQ0?usp=sharing

roger.magnusson wrote:It's as Uli explained, BRAW compresses each frame individually and the Sony calculates the differences between frames (an oversimplification for the sake of brevity) to achieve higher compression. There's tradeoffs to both methods once you decrease the bit rate. You should always use the appropriate bit rate for what you shoot.

I understand, but 100Mb/s hard to grade 8-bit 4:2:0 Sony IPB footage shouldn't outright beat the 272Mb/s All-I 12 bit 4:4:4 out of the P4K. I expect higher quality for higher file sizes. Plain and simple. If the higher file sizes bring worse quality, then they're not worth the higher storage cost associated with them.

I am seeing Bayer filter artifacts here too, but that is understandable given that it isn't doing any sampling and recording pixels 1:1.

Just looks like the difference between 1:1 sampling and subsampling from a much higher res to me. The obvious test of macroblocking as a compression artefact is to compare 12:1 with one of the higher bitrate options on the P4k. Ultimately the P4k is not a very suitable camera for resolving fine irregular detail like this.

Alexrocks1253 wrote:...
I did sharpen the BMPCC4K and Sony footage as I have sharpening in the Sony camera set to -7. The sharpening only made the macroblocks more clear.

Sharpening in a BMD camera is only intended to support Live Broadcast. If as ene needs enhancements, when you’re not live, best to do it with Resolve.

First, minor point but BRAW is 4:2:2 according to their own patent on it.

I expect higher quality for higher file sizes. Plain and simple.

You expect wrong ... an full Intra coder always has a disadvantage on a static (or nearly so) scene.
The advantage it has are :
- No artefact / sudden loss of quality during fast movements
- Lower latency / Computational complexity / Memory requirements (for hw implementation)
- Independent frames allowing for better editing (and no quality loss during editing due to recoding)

The disadvantages is that it achieves a lower coding efficiency ...

rick.lang wrote:

Alexrocks1253 wrote:...
I did sharpen the BMPCC4K and Sony footage as I have sharpening in the Sony camera set to -7. The sharpening only made the macroblocks more clear.

Sharpening in a BMD camera is only intended to support Live Broadcast. If as ene needs enhancements, when you’re not live, best to do it with Resolve.

I did all sharpening in post. The footage is Braw 12:1 with 0.48 sharpening and a simple grade.

smunaut wrote:First, minor point but BRAW is 4:2:2 according to their own patent on it.

I expect higher quality for higher file sizes. Plain and simple.

You expect wrong ... an full Intra coder always has a disadvantage on a static (or nearly so) scene.
The advantage it has are :
- No artefact / sudden loss of quality during fast movements
- Lower latency / Computational complexity / Memory requirements (for hw implementation)
- Independent frames allowing for better editing (and no quality loss during editing due to recoding)

The disadvantages is that it achieves a lower coding efficiency ...

My bad. I guess that all intra codecs don't seem to give too many benefits to me for what I need. My computer is beefy (I used to edit those crazy H.265 6K GH5 anamorphic files and the computer surprisingly kept up).

So to me IPB and All-I doesn't make too much of a difference other than taking up more space for a tiny quality improvement.

John Griffin wrote:Just looks like the difference between 1:1 sampling and subsampling from a much higher res to me. The obvious test of macroblocking as a compression artefact is to compare 12:1 with one of the higher bitrate options on the P4k. Ultimately the P4k is not a very suitable camera for resolving fine irregular detail like this.

I tried Q5 on the same scene and the macroblocks went away, the file size went up to the rate of a bit above 8:1, and the Bayer artifacts stayed. I wonder if the A7SIII would show the same Bayer artifacts because of the near 1:1 pixel readout?

Alexrocks1253 wrote:

John Griffin wrote:Just looks like the difference between 1:1 sampling and subsampling from a much higher res to me. The obvious test of macroblocking as a compression artefact is to compare 12:1 with one of the higher bitrate options on the P4k. Ultimately the P4k is not a very suitable camera for resolving fine irregular detail like this.

I tried Q5 on the same scene and the macroblocks went away, the file size went up to the rate of a bit above 8:1, and the Bayer artifacts stayed. I wonder if the A7SIII would show the same Bayer artifacts because of the near 1:1 pixel readout?

File size is the price you pay for 12bit 4.4.4 ALL-I vs 8bit 4.2.0. long-gop

John Griffin wrote:

Alexrocks1253 wrote:

John Griffin wrote:Just looks like the difference between 1:1 sampling and subsampling from a much higher res to me. The obvious test of macroblocking as a compression artefact is to compare 12:1 with one of the higher bitrate options on the P4k. Ultimately the P4k is not a very suitable camera for resolving fine irregular detail like this.

I tried Q5 on the same scene and the macroblocks went away, the file size went up to the rate of a bit above 8:1, and the Bayer artifacts stayed. I wonder if the A7SIII would show the same Bayer artifacts because of the near 1:1 pixel readout?

File size is the price you pay for 12bit 4.4.4 ALL-I vs 8bit 4.2.0. long-gop

Makes sense. Would you think 10-bit 4:2:2 would be a good compromise between file size and color quality?

Shoot ProRes on the P4k if you want to keep file sizes to a minimum. ProRes vs BRAW should be indistinguishable if you nail exposure and WB in camera.

The intra vs gop is what's costing more, the coding efficiency advantage of being able to use temporaly redudant information is huge.

Try with the sony codec : put a cover blocking the lens, then very quickly remove it, then look at the footage and the quality of the _very_ first frame that's not covered. That's the frame it had to code without any temporal information.

John Griffin wrote:Shoot ProRes on the P4k if you want to keep file sizes to a minimum. ProRes vs BRAW should be indistinguishable if you nail exposure and WB in camera.

I have found ProRes to be larger than BRaw except 3:1 or Q0.

Prores HQ is about as sharp on fine detail as 3:1 and Q0, but slighter sharper (on close examination) than higher compression braw ratios. You'll see the difference between Prores and 12:1, for example, if you look for it. But as you noted, Prores HQ file sizes are generally much higher and, in theory, for less bit depth. Not a great bargain.

John Paines wrote:Prores HQ is about as sharp on fine detail as 3:1 and Q0, but slighter sharper (on close examination) than higher compression braw ratios. You'll see the difference between Prores and 12:1, for example, if you look for it. But as you noted, Prores HQ file sizes are generally much higher and, in theory, for less bit depth. Not a great bargain.

Yeah. I understand the benefits in theory of ProRes but it seems better to shoot in raw that is as large as it or just shoot something in a high bitrate H.264/5 than to worry about it especially since I can edit and playback 200Mb/s HEVC 59.94fps just fine on my machine.

Thanks for the clarification about where you applied sharpening. Good.

rick.lang wrote:Thanks for the clarification about where you applied sharpening. Good.

Yep! No matter what camera I'm using, it is the first option I turn as far down as possible, just like in camera NR (wish Sony would allow me to turn off its NR. DaVinci's is better...)

I understand, but 100Mb/s hard to grade 8-bit 4:2:0 Sony IPB footage shouldn't outright beat the 272Mb/s All-I 12 bit 4:4:4 out of the P4K. I expect higher quality for higher file sizes. Plain and simple. If the higher file sizes bring worse quality, then they're not worth the higher storage cost associated with them.

This is not an accurate statement. The Sony gives you h.265 (or h.264) which are highly optimized codecs, and require hardware acceleration to decode. Basically, you do get smaller filesizes, but to decode them, you need to sacrifice a goat. You don't see that, because every GPU today can decode these formats. So for you, you get the impression that it's a "normal", "run of the mill" codec. It's not. It's HIGHLY dependent on hardware acceleration. Without it, you will be getting 1 fps.

BRAW on the other hand, it runs with minimal hardware decoding needs (it does use the GPU, but not for everything). And it gives you 12bit and 4:4:4 (which by themselves they REQUIRE more bits in order to fit in that bitrate).

So yes, the macroblocking you're seeing, is correct. A less hardware-dependent codec at 12:1, with added bonuses of 12bit/raw (which are bit-hungry), will perform LESS well than any modern codec at a lower bitrate, should that codec is highly hardware-dependent.

Basically, you trade size for hardware acceleration. Unfortunately, you can't create a new codec (like BRAW) and still get FULL hardware acceleration on every level like h.264 does, because that requires the GPU chip to be specifically designed for your codec. And nvidia/AMD/Intel don't care about BRAW (they barely care about Prores either). So, what you're seeing, is correct, and to be fully expected.