Home · Register · Search · View Winners · Software · Hosting · Software · Join Upload & Sell

Moderated by: Fred Miranda
Username   Password

  New fredmiranda.com Mobile Site
  New Feature: SMS Notification alert
  New Feature: Buy & Sell Watchlist
  

FM Forums | Alternative Gear & Lenses | Join Upload & Sell

  

Archive 2013 · Are NEX really 12-bits?
  
 
Mescalamba
Offline
• • • •
Upload & Sell: Off
p.1 #1 · Are NEX really 12-bits?


Well, according to DxOmark they are 12-bits cameras. According to RAW Digger program they are 14-bits. Kinda interesting..

Another interesting is that A900 apparently produces nice 14-bit files too. Wonder how thats done with 12-bit camera? Unless.. its not 12-bit. So either there is some trick, or program is wrong. Or Sony lied a bit.

Obviously those 14-bit outputs (if they are not 14-bit RAWs but just some sorcery from 12-bits) explain quite a bit about image quality of those mentioned.

Yea and why I tried that program was mere curiosity if Fuji S5 really is 14-bit or not. It is by the way. Plus it records images in pretty wierd way.


I dont know where is truth, or if I understood something wrong, so if someone can explain?



Apr 04, 2013 at 04:15 AM
Jeff Kott
Offline
• • •
Upload & Sell: On
p.1 #2 · Are NEX really 12-bits?


Interesting that you noticed this since Jim Kasson posted the same conclusion this week.

http://blog.kasson.com/



Apr 04, 2013 at 04:27 AM
itai195
Offline
• • •
Upload & Sell: Off
p.1 #3 · Are NEX really 12-bits?


I could've sworn I read somewhere that the NEX-6 is 14-bit.


Apr 04, 2013 at 04:34 AM
LightShow
Online
• • • •
Upload & Sell: Off
p.1 #4 · Are NEX really 12-bits?


IIRC, Sony has just recently started using 14bit raw, I'm not a Sony guru so I'm not sure about when exactly it happened, I'm just glad they joined the party.


Apr 04, 2013 at 05:29 AM
theSuede
Offline
• • • •
Upload & Sell: Off
p.1 #5 · Are NEX really 12-bits?


They're still 12-bit inputs. Looking further up the range, you can see that at it begins at four periodicity, at ADU 3200 it becomes eight steps, and at ADU 5708 its 16...

What's important is the number of unique values. And that's about 1400 for both the NEX6 and 7.

2log (1400) >> tone resolution of about 10.5 bit, spread with a slight gamma-like top end spread over a 14-bit value ladder.

Sony has used this for quite some time now. The RX1 starts out at a periodicity of two, and it has an additional step at 8116. Total number of unique values: ~1760, or 10.8 bits of resolution.

So, once again - 14 bit doesn't matter that much, visually the difference "down" to 10.8 bits of tone resolution is very, VERY close to zero. The difference between 16 and 14 bits is four times smaller than this, i.e four times smaller than "very close to zero".



Apr 04, 2013 at 08:15 AM
Mescalamba
Offline
• • • •
Upload & Sell: Off
p.1 #6 · Are NEX really 12-bits?


theSuede wrote:
They're still 12-bit inputs. Looking further up the range, you can see that at it begins at four periodicity, at ADU 3200 it becomes eight steps, and at ADU 5708 its 16...

What's important is the number of unique values. And that's about 1400 for both the NEX6 and 7.

2log (1400) >> tone resolution of about 10.5 bit, spread with a slight gamma-like top end spread over a 14-bit value ladder.

Sony has used this for quite some time now. The RX1 starts out at a periodicity of two, and it has an additional step at 8116. Total number of unique values:
...Show more

Yup, thats what I wanted to know! Thank you.



Apr 04, 2013 at 02:52 PM
AhamB
Offline
• • • •
Upload & Sell: On
p.1 #7 · Are NEX really 12-bits?


Mescalamba wrote:
Yup, thats what I wanted to know! Thank you.


So in the future you won't make statements like this: "Obviously those 14-bit outputs (if they are not 14-bit RAWs but just some sorcery from 12-bits) explain quite a bit about image quality of those mentioned."




Apr 04, 2013 at 05:16 PM
Mescalamba
Offline
• • • •
Upload & Sell: Off
p.1 #8 · Are NEX really 12-bits?


Theres that "if" part for good reason.


Apr 04, 2013 at 07:36 PM
 

Search in Used Dept. 



sorney1
Offline
• •
Upload & Sell: Off
p.1 #9 · Are NEX really 12-bits?


So, once again - 14 bit doesn't matter that much, visually the difference "down" to 10.8 bits of tone resolution is very, VERY close to zero. The difference between 16 and 14 bits is four times smaller than this, i.e four times smaller than "very close to zero".
thanks, this kept me from asking in my ignorance, what does all of this mean?
sorney1



Apr 04, 2013 at 07:41 PM
AhamB
Offline
• • • •
Upload & Sell: On
p.1 #10 · Are NEX really 12-bits?


Mescalamba wrote:
Theres that "if" part for good reason.


The RAWs being 14bit still doesn't make the good IQ of the files "obvious". TheSuede pointed out in the other thread that the Nikon D800 set to 12 bits still gives better looking images than the Hasselblad H4D at 16 bits.



Apr 04, 2013 at 08:16 PM
theSuede
Offline
• • • •
Upload & Sell: Off
p.1 #11 · Are NEX really 12-bits?


sorney1 wrote:
thanks, this kept me from asking in my ignorance, what does all of this mean?
sorney1


It means Sony uses a kind of file compression that is a bit like the gamma curve in a normal sRGB file. It works very well, and has very (extremely) little bearing on the available image quality, or tone resolution. It's much more important that what you feed IN to the system is accurate than just piling on more decimals.

Each Ev step is a doubling of the number in a raw file, which makes the top number needed escalate very quickly... 1Ev = 2x, 2Ev = 4x, 3Ev = 8x, and so on.
The brightness step between the raw file numbers "10" and "20" is visually the same as the difference between "100" and "200", that is visually the same as the difference between "1000" and "2000" - exactly one Ev step. But the darker zone takes up 10 data points, the brighter zone takes up 1000 data points of "storage capacity".
The reason we use higher accuracy (12-14-16 bits) files is that we quite literally waste a lot of numbers in the highlights.

Which do you think gives the better accuracy if you try to measure the length of a matchbox:
Using a calibrated vernier scale that gives 2 digits of accuracy, but you only report one of them - like "52.4mm"?
Using a crude wooden school ruler with one half digit of accuracy, but you report two digits - like "52.50mm"?

Just looking at the numbers, you might think that 52.50 is more accurate than 52.4, since it has two decimals in stead of one - but actually we know that the other number is more accurate, since it was measured more accurately.



Apr 04, 2013 at 09:12 PM
philip_pj
Offline
• • • •
Upload & Sell: Off
p.1 #12 · Are NEX really 12-bits?


Like others I have read the a99 and other recent Sonys have more robust files (for PP) due to 14 bit files, so best not to repeat that one. It would be good to know exactly why the newer cameras are much stronger in post than say, the a900. Maybe just better sensor technology.


Apr 05, 2013 at 04:03 AM
sebboh
Offline
• • • • • •
Upload & Sell: Off
p.1 #13 · Are NEX really 12-bits?


philip_pj wrote:
Like others I have read the a99 and other recent Sonys have more robust files (for PP) due to 14 bit files, so best not to repeat that one. It would be good to know exactly why the newer cameras are much stronger in post than say, the a900. Maybe just better sensor technology.


yup, sensor tech. the nex files are better in this regard than the a900 as well.



Apr 05, 2013 at 04:29 AM
RustyBug
Offline
• • • • • •
Upload & Sell: On
p.1 #14 · Are NEX really 12-bits?


But doesn't "accurate to 32 decimal places" @ 52.50000000000000000000000000000000 sound way more impressive.



Besides, when you double 52.50000000000000000000000000000001 x 2 you get an equally impressive 105.0000000000000000000002. Then when you apply that to your screen or printer ... do they really care about the .00000000000000000000000000000002?

+1 @ accuracy/precision/significant digits @ practical limits of use.
And yes, I know that 14 bit vs. 16 bit vs. 20 bit vs. 32 bit doesn't equate to 32 decimal places ... just showing that there comes a point of diminishing returns that the increased precision (absurd example) doesn't do much of anything of real value for you, just in case someone didn't think that only 1 or 2 decimal places wasn't a sufficient example.





Apr 05, 2013 at 04:36 AM
theSuede
Offline
• • • •
Upload & Sell: Off
p.1 #15 · Are NEX really 12-bits?


At the top end of the raw file at base ISO, a NEX6 has about 25ke- charge, and the A99 has about 34ke-. Error margin due to photon shot noise is 1/sqrt(charge), so:
NEX6 : 1/sqrt(25,000) = 1/158 of the raw file
A99 : 1/sqrt(34,000) = 1/184 of the raw file

14-bit file value range is 0 > 16383

The absolute minimum noise before even considering electronics and so on is then NEX6 : 16383/158 = 103 values in the raw for ONE accuracy level near white
A99 : 16383/184 = 89 values in the raw for ONE accuracy level near white

As you look further down the exposure range, this value of course decreases. At medium gray 18%, minimum noise is 44 raw value levels for the NEX6 and 38 raw values for the A99

As you approach black, the value count loss is zero. So you need most of the resolution in the shadows, at the top end you can spread the values out quite far without even taking a remote risk of losing any real accuracy. This is what Sony (and Nikon, ans normal gamma-treated files like sRGB jpg and so on) does.



Apr 05, 2013 at 04:09 PM





FM Forums | Alternative Gear & Lenses | Join Upload & Sell

    
 

You are not logged in. Login or Register

Username   Password    Retrive password