RIP Sony a7s III

The author of the article makes all kinds of assumptions.

“In fact, a true 16-Bit sensor readout only belongs to high-end cinema cameras. Yes, that’s one of the things you’re paying for!”

The RED Komodo shoots 16-bit REDCODE RAW for only $6,000. Isn’t it an entry-level cinema camera? And isn’t the dynamic range similar to the FX3 and a7s III?

So if the argument of the author is that only high-end cinema cameras shoot 16-bit RAW, then I call FOUL.

Furthermore, the author states: “In fact, the bit-depth of the RAW file is chosen based on the actual, measured dynamic range of the sensor. If a camera sensor can capture 13.2 stops of dynamic range, for example, it doesn’t make sense to use 16 bits: in this particular situation, 14 bits are more than enough!”

Again, this could equally apply to the RED KOMODO, which measures less than 13.2 stops DR according to CineD (12-1/2 stops, to be exact). According to the author’s logic, since the KOMODO is neither expensive nor does it have magical dynamic range, therefore, it does not record 16-bit RAW.

Finally, the author confuses feelings and assumptions with facts. “Furthermore, a true 16-Bit sensor readout, paired with backside illumination technology, also meant that this new generation of Sony cameras would provide a huge dynamic range and image quality improvement over the old FS7 and FS5 models. Or at least, that’s what we were all expecting.”

Really? How can the author presume to know what everyone everywhere was expecting the dynamic range of these cameras to be? I never read anywhere what anyone expected the DR to be. As for myself, I was expecting at least 12 stops, and anything more would be gravy.

Did the author truly expect cameras costing just a few thousand dollars to surpass ARRI in dynamic range? Nobody I know thought like that. So the author’s argument here is, the dynamic range isn’t HUGE, therefore Sony is misleading the public about the number of bits. That is not even logical.

The author seems to think that “I find it hard to believe” is enough reason to call Sony a liar. Not cool.
 
You will never hear me disparage ProRes RAW, even though I've never used it. (I don't want that workflow or bitrates and I don't want to use an external recorder. That's why I just say I like the concept.)

But I like to be accurate and, frankly, there's no way that the FX3 is recording 16-bit 60 fps from a full-frame sensor. I've looked at a lot of Sony sensor data sheets, and nothing comes close to that. Sorry.

12-bit? Possibly. I think the bigger excitement about that feature is the PRR codec. The Netflix certification in no way confirms that the FX3 can do 16-bit readout. It doesn't particularly matter since you can evaluate the footage for yourself to see what quality and flexibility you're getting. But the fact is that it's not a 16-bit readout.

Where do you read that its all-pixel scan rate is only 90 fps? I'm pretty sure it can record 100 fps from the full sensor (or, at least, with a 16:9 crop). I can record 100 fps on it in PAL mode without a crop whereas 120 fps in NTSC crops to 1.1x.

You seem to take any skepticism of this camera personally. I'm not arguing about the utility of the camera. I own one and I like it. I'm just questioning some of your hyperbolic claims.
 
ahalpert said:
You will never hear me disparage ProRes RAW, even though I've never used it. (I don't want that workflow or bitrates and I don't want to use an external recorder. That's why I just say I like the concept.)

But I like to be accurate and, frankly, there's no way that the FX3 is recording 16-bit 60 fps from a full-frame sensor. I've looked at a lot of Sony sensor data sheets, and nothing comes close to that. Sorry.

12-bit? Possibly. I think the bigger excitement about that feature is the PRR codec. The Netflix certification in no way confirms that the FX3 can do 16-bit readout. It doesn't particularly matter since you can evaluate the footage for yourself to see what quality and flexibility you're getting. But the fact is that it's not a 16-bit readout.

Where do you read that its all-pixel scan rate is only 90 fps? I'm pretty sure it can record 100 fps from the full sensor (or, at least, with a 16:9 crop). I can record 100 fps on it in PAL mode without a crop whereas 120 fps in NTSC crops to 1.1x.

You seem to take any skepticism of this camera personally. I'm not arguing about the utility of the camera. I own one and I like it. I'm just questioning some of your hyperbolic claims.
Click to expand...

The author of the article does not provide an iota of proof, only assumptions. Sound familiar?
 
You don't actually provide any proof either. You just parrot PR as opposed to engaging in rational deduction like CineD
 
According to your logic, Alister Chapman, Sony, Atomos and Netflix are all engaged in a big cover-up.

For those times where you want to squeeze every last drop of image quality out of the camera, you can record using ProRes RAW. The camera outputs the RAW as 16 bit linear data. The Ninja V then takes this 16 bit linear data and using a sophisticated, visually lossless process converts this to 12 bit log RAW. A 12 bit recording can contain up to 4 times more tonal values than a 10 bit recording. These finer levels help ensure that you are getting the very best from the camera. ProRes RAW is a great choice whenever you want to do any extensive grading or image manipulation.
- Shooting RAW with the FX3 (and Alpha 7S III)


BY: ALISTER CHAPMAN
Click to expand...
 
Last edited:
Outputs 16-bit: yes. Reads out 16-bit from the sensor: no. Tom Roper literally just quoted Alistair Chapman agreeing with me.
 
No, I don't know it for a fact. I just believe it to be true and would be shocked otherwise. You don't know your opinion for a fact, either. Yet you equate mine with flat eartherism lol
 
ahalpert said:
Outputs 16-bit: yes. Reads out 16-bit from the sensor: no. Tom Roper literally just quoted Alistair Chapman agreeing with me.
Click to expand...

Sony is all too willing to boast about 16 bit output, but when asked directly to confirm 16 bit A/D, they went dark. Doubts had already surfaced about the A/D because of limiting readout speeds and DR. Sony had an opportunity to clear the air but chose to let their trolls prolong what everyone rational including Alister believe to be a myth.
 
I'm all for numbers and letters - and they can matter - but at some point it's like...just use your eyes.

Look at how the Classic looks with 2K 10-bit ProRes.

Yeah, a little softer as we approach 2023, but still very nice.
 
NorBro said:
I'm all for numbers and letters - and they can matter - but at some point it's like...just use your eyes.
.
Click to expand...

This is indisputably true. And I think the 16-bit output to PRR is a great option. But I won't go around claiming that it can reverse the user's hair loss because that's just not true. (Even though I can't prove it false.)
 
ahalpert said:
No, I don't know it for a fact. I just believe it to be true and would be shocked otherwise. You don't know your opinion for a fact, either. Yet you equate mine with flat eartherism lol
Click to expand...

There is no reason in the world not to believe that the FX3 outputs 16-bit RAW. Sony is most certainly not lying or misleading anyone. Neither do they owe anyone any explanations, least of all CineD, who a famous cinematographer said was engaged in something other than honest journalism. No one but the engineers themselves know what is going on under the hood. Sony shoots 16-bit, RED shoots 16-bit, Arri Alexas prior to the Alexa 35 shoot 16-bit (stored as 12-bit log), Blackmagic shoots 16-bit (stored as 12-bit log).

”Typically log encoding means you can take (for example) a 16 bit (or more) linear data from the sensor, apply a log curve and store it as a 12 bit log file. Then you can “unpack” it back to 16 bit lin in a lossless way and it’s a good way to store high dynamic range with much smaller files, no matter if it’s raw or encoded video.” - famous cinematographer

Sony might very well be doing just as this knowledgeable DP conjectures RED is doing; in no world does this make them untruthful and there may be benefits you simply aren’t aware of. BTW, CineD tested the Sony Venice and its 16-bit X-OCN dynamic range measured in the neighborhood of 11.7 stops, or .2 stops greater than Sony a7s III ProRes RAW. Canon R5 C RAW measured around 10 stops, nearly as awful as the eight year-old Samsung NX1, the worst result ever measured by CineD at the time - and it did not even shoot log! Log encoded 12-bit RAW should yield a huge increase in dynamic range. Are we then to assume the Canon is not even outputting 12-bit? The CineD author you cite as an authority is a know-nothing - dynamic range (as calculated by Imatest) alone does not give us insight into bit-depth.

As far as judging with your own eyes goes, I’m not even sure what that means because it is extremely vague and every fibre of my being rebels against vagueness and unclear thinking. I would posit that how the footage grades in addition to visual inspection is a more reliable indicator - but as 12-bit log from 16-bit linear is a visually lossless process, I’m not sure even that can tell us anything at all about what bit-depth the camera is outputting.
 
Last edited:
In binary math, 4 bits can represent any number between 0 and 15. The same 4 bits in trinary could represent certain numbers between 0 and 40 but not all of the numbers because there would be gaps inbetween where certain numbers (grayshades) will be skipped.. For example 4 bits in trinary could represent the numbers 13, 12, 10, 9; but not 15, 14 , 11, 8, 7, 6, 5, 2. This is because each bit contains only two states, whether it's used in a binary or trinary system.

So you see, in a single data word, it's not possible to represent all the shades when there are more of them than addressable spaces. So a choice is made, that this shade is close enough to that shade such that we don't need both and can do away with one of them. That decision is made down the line, to discard certain gray shades that are similar enough to others that we don't need both. That is exactly what happens in log encoding, the available addressable spaces are used to keep the key shades and the nearby ones discarded if not needed.

If it were really possible to have this free lunch, where you could magically store 16 bits within 12 and then reconstruct them losslessly, then why not store 16 bits within 8 and save even more space? It's a hypothetical question only, used to illustrate the significance of reconstructing "losslessly" because you could do it. The matter of how successful at doing it losslessly is the process of interpolating data values that no longer exist within the addressable data spaces. There is a limit to how many and how large the gaps can exist before the onset of banding. So I do respect Mr. Brawley, and BMD and Arri have certainly made log encoding "visually lossless," I think it's still incorrect to says it's objectively lossless.

The second way to represent a number of shades larger than the addressable space is with floating point real numbers, but these numbers unlike 0 and 1 binary, are 0 thru 9 including decimal fractions. Any real number can be constructed from any number of bits, but require more than 1 data word to do so. In other words, if I wanted to store 16 bits into a 4 bit word losslessly I could do it, but it would take (4) four-bit data words.

The problem with trying to do that in a camera is that each read of a 4 bit word is a clock cycle, would need 4 clock cycles to read (4) four-bit data words. You can't speed up the filling of the bucket containing the photons of light so this slows down the frame rate. See the relationship?
 
But nobody in this entire topic has suggested the FX3 is not outputting 16 bit raw. That is a pure distortion by the thread starter trolling his own topic. What has been suggested, and supported in the remarks by Mr. Chapman, is that the sensor readout is not 16 bits. Sony was asked this directly and did not deny it. That's the end of the story.
 
Tom Roper said:
But nobody in this entire topic has suggested the FX3 is not outputting 16 bit raw. That is a pure distortion by the thread starter trolling his own topic. What has been suggested, and supported in the remarks by Mr. Chapman, is that the sensor readout is not 16 bits. Sony was asked this directly and did not deny it. That's the end of the story.
Click to expand...

Alister Chapman has no direct knowledge of the sensor's readout and nowhere does he come out and say it is not 16 bits. Whether Sony responds to someone with an agenda at CineD is proof of nothing.
 
jonpais said:
Alister Chapman has no direct knowledge of the sensor's readout and nowhere does he come out and say it is not 16 bits. Whether Sony responds to someone with an agenda at CineD is proof of nothing.
Click to expand...

Let me be sure I understand you, he has to come out say it but Sony does not have to say anything? Only a naive person would not see the double standard in that desperate argument.

​
 
Tom Roper said:
Let me be sure I understand you, he has to come out say it but Sony does not have to say anything? Only a naive person would not see the double standard in that desperate argument.

​
Click to expand...

No one in this thread said Chapman had to come out and say anything.

"Person #1 won't say whether the moon is made of green cheese.
Person #2 thinks the moon might be made of green cheese.
Therefore, the moon must be made of green cheese."
 
Last edited:
One thing to keep in mind with these Netflix Approved cameras is that their 10-bit internal codecs won’t make the cut for HDR. Which is where the Ninja V and ProRes RAW come in clutch.
 
You must log in or register to reply here.
Back
Top