GH5 Game Over, Panasonic

combatentropy said:
Yes, and 10-bit log should actually hold up better than 12-bit linear, http://www.normankoren.com/digital_tonality.html#Human_vision, so important is logarithmic over linear.
Click to expand...

There's quite a bit of wrong packed into this statement:

If we were comparing the same formats, then yes, a log encoding would pack more range into the same bandwidth, at the expense of resolution. (Which resolution is mostly sacrificed in areas the human eye is not particularly sensitive to, as you indirectly point out.)

But we are *NOT* comparing the same formats. A RAW image is *FUNDAMENTALLY* different than a 4:2:2 YUV image. Put simply, the latter is already significantly manipulated, and severely degraded. LOG encoding is a clever hack, but it's nothing like working with RAW.

This article about ProRes RAW does a pretty nice job at explaining RAW in general: https://www.newsshooter.com/2018/05/07/understanding-prores-raw-is-it-prores-or-is-it-raw/
 
joe12south said:
Shoot a still recording, under exposed 2 stops, as both a JPEG and a RAW image. Lift that image in the image editing app of your choice (hell, your NLE if you want.) Tell me which one you'd rather work with.
Click to expand...

...But if you correctly expose your shot for both formats and the scene has a normal contrast range ( i.e. no more than 12 stops of DR ), you may find that the finished result is close to identical between RAW and LOG / HLG. This is what I'm expecting to find with this new BMPCC 4K camera, where I will likely only use it when I have to shoot under bad lighting ( mixed color balances in the same shot ) or when the shot contains an extreme contrast range. Otherwise LOG or HLG will probably produce near identical results for me with a simpler workflow.
 
Hey Thomas, if I could be bothered to quote Andrew Reid at EOSHD, he claims there is no difference between 8 and 10 bit. So if you need proof that there is a diff between 10 and 12 bit, would you consider me foolish thinking there is no diff between 8 and 12 bit when grading tricky stuff ? just sayin'
 
Last edited:
TheDingo said:
...But if you correctly expose your shot for both formats and the scene has a normal contrast range ( i.e. no more than 12 stops of DR ), you may find that the finished result is close to identical between RAW and LOG / HLG. This is what I'm expecting to find with this new BMPCC 4K camera, where I will likely only use it when I have to shoot under bad lighting ( mixed color balances in the same shot ) or when the shot contains an extreme contrast range. Otherwise LOG or HLG will probably produce near identical results for me with a simpler workflow.
Click to expand...
Certainly. If you like a camera's baked-in look, aren't going to do much manipulation in post, and value workflow speed over wringing the last bit of PQ out of the file, then there is little reason to shoot RAW video.

That is *NOT* the same as arguing that the RAW data isn't empirically better (less compromises, more flexible) which some people here seem to be edging towards.
 
Ned Kelly said:
Hey Thomas, if I could be bothered to quote Andrew Reid at EOSHD, he claims there is no difference between 8 and 10 bit. So if you need proof that there is a diff between 10 and 12 bit, would you consider me foolish thinking there is no diff between 8 and 12 bit when grading tricky stuff ? just sayin'
Click to expand...

In their GH5 review they said the difference is noticeable when shooting log but not in other profiles.
 
I figured some may find it amusing that I just walked out of B&H with my second GH5. (NOT GH5S). The camera you can get paid for using is better than the camera that you can't buy yet.
 
joe12south said:
That is *NOT* the same as arguing that the RAW data isn't empirically better (less compromises, more flexible) which some people here seem to be edging towards.
Click to expand...

...But for many typical shooting environments RAW may show little improvement over LOG or HLG profiles, so it may not be worth using under these conditions.

Let's wait for the actual BMPCC 4K camera to be released before we start arguing about image quality because right now there is no image to compare.
 
TheDingo said:
...But for many typical shooting environments RAW may show little improvement over LOG or HLG profiles, so it may not be worth using under these conditions.

Let's wait for the actual BMPCC 4K camera to be released before we start arguing about image quality because right now there is no image to compare.
Click to expand...

BMD 12 bit RAW is compressed log encoding, with not all frame rates available to all resolutions.
 
joe12south said:
If we were comparing the same formats, then yes, a log encoding would pack more range into the same bandwidth
Click to expand...

Yes, then we agree.



But we are *NOT* comparing the same formats.
Click to expand...

(sigh) Then you should not have said "there is a meaningful difference between 10-bit LOG and 12-bit RAW," which is what I was replying to. Maybe I should have taken how you said RAW instead of linear as a warning that you don't really know what you're talking about, but I did, and here we are, and this probably should be my last reply.

If in general 10-bit log is more gradeable than 12-bit linear, then 10-bit log RAW is more useful than 12-bit linear RAW, and 10-bit log MPEG-4 is more useful than 12-bit linear MPEG-4. Then you weren't really talking about 10 bit or 12 bit at all. What you really were saying is that RAW is more gradeable than MPEG-4, which is true. You should have just left out any mention of bits or logness.

A RAW image is *FUNDAMENTALLY* different than a 4:2:2 YUV image.
Click to expand...

Eh, "fundamentally"? I don't know, it depends on what you mean. Let's see where this goes.

Put simply, the latter is already significantly manipulated, and severely degraded.
Click to expand...

Okay, now you're using those hyperbolic words again, "severely degraded." You do realize 10-bit 4:2:2 YUV can be completely uncompressed? In fact, in broadcast studios it used to be a thing.

Let's back up. What is RAW? It used to mean two things: (1) not yet debayered, and (2) uncompressed. In the past few years, starting with Blackmagic I think, it can be compressed. So now all it means is "not yet debayered." That in and of itself is no benefit. Well, it is a benefit in saving bandwidth. But that's not what you're saying. You're saying it's actually higher quality (I think, the best I can tell from your wording).

Let's continue.

LOG encoding is a clever hack, but it's nothing like working with RAW.
Click to expand...

Again, this really is a red flag, because LOG and RAW are independent. You can have LOG RAW. And log is not a hack. Linear is a naive, simplistic conversion of analog to digital, while logarithmic is a wise one, because f-stops are logarithmic, human visual perception is logarithmic.

By the way, YUV isn't destructive. It can be reversed. If you don't subsample it (that is, if you keep it at 4:4:4) and don't compress it, then you can mathematically, losslessly convert it right back to RGB.

Now if you subsample it to 4:2:2, then mathematically of course you've lost something. But the human eye is so insensitive to the difference that I don't care. I've seen an example where the color information was not just halved but made 1/100th the resolution as the luminance, and it was not immediately noticeable (not that I'm suggesting 1/100 chroma subsampling, of course).

Theoretically there should be no difference between the gradability of RAW and uncompressed 4:2:2, if they are both 10-bit log (or if they're both 12-bit linear, or whatever, as long as they're the same quantization). You should be able to push or pull them in post to the same degree.

That's theoretical, of course. In practice, I am not ready to dispute whether there is some better luck you're having with RAW. But I don't think it's because it's RAW versus not-RAW. It could be because Panasonic slathered on the noise reduction in not-RAW, which will smear detail in the shadows. It could even be a difference in how your grading software handles RAW vs. MPEG-4. In fact, the article you linked to mentions that very thing, only in RAW's disfavor in that example. Rightly they reinforce that that doesn't mean RAW itself is worse than ProRes, even though in their test ProRes looked better.

This is grade-school science. In like third grade our teacher taught us the scientific method and had us do science experiments, and mine was pitiful. But I did learn the scientific method, and that is that when comparing two things you narrow it down to one variable. The rest are constants. So I was measuring what the effect of coffee was on plants (I told you it was pitiful, hey I was in 3rd grade!). So I had to have two plants: one was the control, and one was the variable. One plant I gave water. And one plant I gave coffee. The caffeinated plant died. But to make an honest experiment I had to keep everything else the same: the plants had to be the same species, in two identical pots, in the same kind of soil, in front of the same window, etc.

Likewise when comparing a single ingredient in a camera, you can't just write off all the other differences. You can't draw conclusions about any single ingredient when you compare 12-bit uncompressed RAW from a camera by Blackmagic, with 10-bit, in-camera color-corrected and noise-reduced log MPEG-4 from a camera by Panasonic.

This article about ProRes RAW does a pretty nice job at explaining RAW in general: https://www.newsshooter.com/2018/05/07/understanding-prores-raw-is-it-prores-or-is-it-raw/
Click to expand...

I read the whole thing and didn't read anything I didn't already know or anything that supports what you're arguing against me about.

Shoot a still recording, under exposed 2 stops, as both a JPEG and a RAW image. [...] Tell me which one you'd rather work with.
Click to expand...

Straw man --- https://en.wikipedia.org/wiki/List_of_fallacies
 
Last edited:
Actually the x:2:2 means that the U and V (or R-y, B-y) channels have been compressed, this is essentially what Betacam was using to fit everything onto the tapes without having to go to a higher head speed and tape speed (more bandwidth). Spent a lot of time balancing this to get some recordings to look good back in the analog days.
 
combatentropy said:
(sigh) Then you should not have said "there is a meaningful difference between 10-bit LOG and 12-bit RAW," which is what I was replying to. Maybe I should have taken how you said RAW instead of linear as a warning that you don't really know what you're talking about, but I did, and here we are, and this probably should be my last reply.

If in general 10-bit log is more gradeable than 12-bit linear, then 10-bit log RAW is more useful than 12-bit linear RAW, and 10-bit log MPEG-4 is more useful than 12-bit linear MPEG-4. Then you weren't really talking about 10 bit or 12 bit at all. What you really were saying is that RAW is more gradeable than MPEG-4, which is true. You should have just left out any mention of bits or logness.
Click to expand...
You're being pedantic. For all practical purposes when refering to LOG vs RAW (and especially when comparing a camera that can record RAW vs a camera that can't) it is understood that we are talking about using LOG compression to represent a wider dynanmic range within the narrower signal of an RGB or YUV signal, not, of course, the broad mathematical concept of a logarithmic curve. That would be uselessly academic.

Okay, now you're using those hyperbolic words again, "severely degraded." You do realize 10-bit 4:2:2 YUV can be completely uncompressed? In fact, in broadcast studios it used to be a thing.
Click to expand...
Um, no. By very definition a 4:2:2 signal throws away half of the color resolution.

Let's back up. What is RAW? It used to mean two things: (1) not yet debayered, and (2) uncompressed. In the past few years, starting with Blackmagic I think, it can be compressed. So now all it means is "not yet debayered." That in and of itself is no benefit. Well, it is a benefit in saving bandwidth. But that's not what you're saying. You're saying it's actually higher quality (I think, the best I can tell from your wording).
Click to expand...
Wrong again. The RAW data retains more color information than anything but a 4:4:4 representation of the sensor data. A 4:4:4 version would be much bigger than than the RAW signal, with no quality benefit.

Again, this really is a red flag, because LOG and RAW are independent. You can have LOG RAW. And log is not a hack. Linear is a naive, simplistic conversion of analog to digital, while logarithmic is a wise one, because f-stops are logarithmic, human visual perception is logarithmic.
Click to expand...
Pointless academic distinction. Practically speaking, there isn't a need to represent RAW with log encoding as 12, 14 16 bits are enough bits to adequately represent the captured data. And yes, the actual implementation of every commercially available LOG compression scheme (V-LOG, C-LOG, S-LOG, et. all) is to hack a wider dynamic range into what was effectively meant to be an SDR signal carrier. That is a hack. it's a really clever, really effective hack, but is an after the fact hack.

By the way, YUV isn't destructive. It can be reversed. If you don't subsample it (that is, if you keep it at 4:4:4) and don't compress it, then you can mathematically, losslessly convert it right back to RGB.
Click to expand...
When Panasonic announces a GH camera that records 4:4:4 signal, this would be a meaningful distinction.

Theoretically there should be no difference between the gradability of RAW and uncompressed 4:2:2, if they are both 10-bit log (or if they're both 12-bit linear, or whatever, as long as they're the same quantization). You should be able to push or pull them in post to the same degree.
Click to expand...
Tell that to the guy trying to pull a clean 4K matte.

Likewise when comparing a single ingredient in a camera, you can't just write off all the other differences. You can't draw conclusions about any single ingredient when you compare 12-bit uncompressed RAW from a camera by Blackmagic, with 10-bit, in-camera color-corrected and noise-reduced log MPEG-4 from a camera by Panasonic.
Click to expand...
Except that's the only meaningful comparison to make. None of us really care about theoretical cameras.

I read the whole thing and didn't read anything I didn't already know or anything that supports what you're arguing against me about.
Click to expand...
Then you should read it again, more carefully this time.

FULL DISCLOSURE: This response was typed under the influence of several tasty whiskey smashes.
 
TheDingo said:
...But for many typical shooting environments RAW may show little improvement over LOG or HLG profiles, so it may not be worth using under these conditions.

Let's wait for the actual BMPCC 4K camera to be released before we start arguing about image quality because right now there is no image to compare.
Click to expand...

1. Yes, again, accepted: Some people don't care about the quality difference.
2. We don't need to wait for the camera to understand that all else being equal, a RAW recording is of higher quality than a 4:2:2 h264/H265 recording.
 
I know you're referring to the BMPCC4K (and BM has amazing CinemaDNG RAW), but some RAW isn't that good, IMO.

And some 8-bit 4:2:0 holds up really well.

---

The thing about image quality is I think we stress over it too much...like if it looks good, who cares what it is? I know post is that heavy argument, but there are so many cameras out there that look really good that have inferior specs.
 
NorBro said:
I know you're referring to the BMPCC4K (and BM has amazing CinemaDNG RAW), but some RAW isn't that good, IMO.

And some 8-bit 4:2:0 holds up really well.

---

The thing about image quality is I think we stress over it too much...like if it looks good, who cares what it is? I know post is that heavy argument, but there are so many cameras out there that look really good that have inferior specs.
Click to expand...

Sure, context is everything. We're talking about two cameras that have the same (or virtually same) sensor, but one of which records RAW and costs half what the one that doesn't costs.
 
You know what the kicker would be?...if Panasonic released a RAW firmware update for the GH5/S, lol.

No one would ever work and instead just discuss it.
 
joe12south said:
For all practical purposes when refering to LOG vs RAW (and especially when comparing a camera that can record RAW vs a camera that can't) it is understood that we are talking about using LOG compression to represent a wider dynanmic range within the narrower signal of an RGB or YUV signal, not, of course, the broad mathematical concept of a logarithmic curve.
Click to expand...

To a lesser extent it's the same problem for 12 bit linear RAW which doesn't have enough code values to represent more than 12 stops DR, very few shades for the low key, and wasteful misallocation of bits to the highlight regions. This was the foundation for widespread complaints with Sony FS5/FS7 12 bit RAW. BMD I would credit for avoiding a linear scheme into their 12 bit RAW but it also points out the folly, it's not sensor RAW either. It would be more accurate to describe as non-linear un-debayered compressed 12 bit 422.

Um, no. By very definition a 4:2:2 signal throws away half of the color resolution.
Click to expand...

With 1:1 pixel mapped frames it's thrown away before that because it was never there in a bayer CFA. Doesn't much seem to matter.

A technical rebuttal is not really merited when the premise you floated was such a distortion, that GH5 "is hobbled by a highly compressed consumer codec." No one would argue ALL-I 400 mbps was superior but it's a fact; EBU and Netflix are fine with it. Barry Green seems fine with it. Case Closed.
 
You must log in or register to reply here.
Back
Top