50p on the F35
- Thread starter macgregor
- Start date
Jusgorilla
Well-known member
Im hopefully getting a 7Q finally this week and I'm keen to see any difference with the 10bit 50P compared to the Cinedeck 8bit 50P.
I'll def post my finding, although I will be in Amsterdam on a shoot for a week, so maybe the shoot will be a good test.
I'll def post my finding, although I will be in Amsterdam on a shoot for a week, so maybe the shoot will be a good test.
Jusgorilla
Well-known member
Im testing the 50P on the 7Q as we speak.
F35 reports 600% DR on 0db, and
800% on 3db,
450% with range normal at 0db (strangely said 340 ISO?)
650% with range norma at 3db (strangely said 480 ISO?)
But my scene was very soft lit, limited dynamic range, so I used normal range (don't use light meters, so can't say how many stops DR needed to be captured, but eye balling the scene, no where near 12 stops needed, so range normal is better to fit less bits in the bit bucket)
I can say that the footage in 10 bit 422 on 7Q looks ALOT better than with my previous cinedeck 8bit 422 cineform. It is less contrasty, even less noise it appears.
F35 reports 600% DR on 0db, and
800% on 3db,
450% with range normal at 0db (strangely said 340 ISO?)
650% with range norma at 3db (strangely said 480 ISO?)
But my scene was very soft lit, limited dynamic range, so I used normal range (don't use light meters, so can't say how many stops DR needed to be captured, but eye balling the scene, no where near 12 stops needed, so range normal is better to fit less bits in the bit bucket)
I can say that the footage in 10 bit 422 on 7Q looks ALOT better than with my previous cinedeck 8bit 422 cineform. It is less contrasty, even less noise it appears.
macgregor
Veteran
Alright, here we go again.
I've spent hours trying to figure this out, and now I'm even more confused than ever. Lets see if you guys can help. Latest tests:
Sony F35 on CineEi mode. Camera is set to the default Slog and wide Drange.
A given scene. Constant lighting.
- 25fps 1/50 of a sec. f4. All good. This will be our target shot for future refs.
- switch to 50fps, 1/100, f2.8. The footage comes out darker. Noise level is increased. DR on display shows 600% instead of the usual 800%. I need to open an extra half stop more to make mid gray as bright as the same shot at 25fps and still not quite the same. Wait, what? Definitely the quality of the image is worse than 24/25/30p.
- Switch to 50i, 1/100, f2.8. Overexposed shot by almost one stop.
- Switch to 50i 1/100, f4. Pretty much the same as our target shot. Wait, what? How is this possible? If the camera is 450ISO at 24/25p/30p, does it mean it's 900ISO in the interlaced modes? The display still shows ISO450.
Nothing makes sense!!!!!!!!!!
I've spent hours trying to figure this out, and now I'm even more confused than ever. Lets see if you guys can help. Latest tests:
Sony F35 on CineEi mode. Camera is set to the default Slog and wide Drange.
A given scene. Constant lighting.
- 25fps 1/50 of a sec. f4. All good. This will be our target shot for future refs.
- switch to 50fps, 1/100, f2.8. The footage comes out darker. Noise level is increased. DR on display shows 600% instead of the usual 800%. I need to open an extra half stop more to make mid gray as bright as the same shot at 25fps and still not quite the same. Wait, what? Definitely the quality of the image is worse than 24/25/30p.
- Switch to 50i, 1/100, f2.8. Overexposed shot by almost one stop.
- Switch to 50i 1/100, f4. Pretty much the same as our target shot. Wait, what? How is this possible? If the camera is 450ISO at 24/25p/30p, does it mean it's 900ISO in the interlaced modes? The display still shows ISO450.
Nothing makes sense!!!!!!!!!!
Last edited:
Viddovation
Veteran
Here's a theory - line-skipping. I can't prove this one iota, but it fits better than any theory given to date.
Hear me out. All the documents I've read on the F35 indicate that each pixel point in a compiled video image comes from six photosites - two red, two blue, two green. The documents I've read indicate the reason for two of each is that for each color there is a high and a low sensitivity photosite, and this extends the overall dynamic range of the camera, when the two values are internally combined.
A visual representation might look like this:
[TABLE="class: grid, width: 300"]
[TR]
[TD]Red - High
[/TD]
[TD]Blue - High
[/TD]
[TD]Green - High
[/TD]
[/TR]
[TR]
[TD]Red - Low
[/TD]
[TD]Blue - Low
[/TD]
[TD]Green - Low
[/TD]
[/TR]
[/TABLE]
So with that in mind, let's say that the camera has a bottleneck somewhere that can't read the entire sensor more than 30 times per second.
When doing 50 frames interlaced, it scans the sensor 50 times per second, but only needs to read 25p worth of data from it - so it reads 2 lines (high, low), then skips the next two lines (high, low). Therefore, no problem, full dynamic range.
But it wouldn't be able to get 50p... unless it still discards half the data to get around the bottle neck. So, rather than not give us 50p, they discard half the data. Since the image needs to be progressive, this time they line skip every other line, likely discarding the "low" sensitivity lines - read a line (high), skip a line (low), read a line (high), skip a line (low). And so forth.
Since the camera needs both lines to get the full dynamic range, this line skipping would cause a drop in dynamic range - most particularly perhaps when no gain is applied.
This would drop the rated ISO. Also, with less dynamic range and only relying on one photosite per color, noise/grain would likely be more perceptible.
Eh?
Hear me out. All the documents I've read on the F35 indicate that each pixel point in a compiled video image comes from six photosites - two red, two blue, two green. The documents I've read indicate the reason for two of each is that for each color there is a high and a low sensitivity photosite, and this extends the overall dynamic range of the camera, when the two values are internally combined.
A visual representation might look like this:
[TABLE="class: grid, width: 300"]
[TR]
[TD]Red - High
[/TD]
[TD]Blue - High
[/TD]
[TD]Green - High
[/TD]
[/TR]
[TR]
[TD]Red - Low
[/TD]
[TD]Blue - Low
[/TD]
[TD]Green - Low
[/TD]
[/TR]
[/TABLE]
So with that in mind, let's say that the camera has a bottleneck somewhere that can't read the entire sensor more than 30 times per second.
When doing 50 frames interlaced, it scans the sensor 50 times per second, but only needs to read 25p worth of data from it - so it reads 2 lines (high, low), then skips the next two lines (high, low). Therefore, no problem, full dynamic range.
But it wouldn't be able to get 50p... unless it still discards half the data to get around the bottle neck. So, rather than not give us 50p, they discard half the data. Since the image needs to be progressive, this time they line skip every other line, likely discarding the "low" sensitivity lines - read a line (high), skip a line (low), read a line (high), skip a line (low). And so forth.
Since the camera needs both lines to get the full dynamic range, this line skipping would cause a drop in dynamic range - most particularly perhaps when no gain is applied.
This would drop the rated ISO. Also, with less dynamic range and only relying on one photosite per color, noise/grain would likely be more perceptible.
Eh?
hermmermferm
Well-known member
Don't want to derail this thread, but what are the steps for getting slow motion out of this camera? When I shoot 50p dual linked to the odyssey7q/dpx and bring it into Adobe Premiere/After Effects, it shows up as 30fps. Would anyone mind sharing their process?
macgregor
Veteran
You probably have AE default image sequences as 30fps, change the settings in the program to whatever you want.
Viddovation
Veteran
By "setup", do you mean you don't think that's how the sensor is arranged? Because there are plenty of data sources that suggest the sensor is 5760x2160 photosites. 5760/3 = 1920. 2160/2 = 1080.
Further:
http://provideocoalition.com/aadams/story/non_technical_f35_guide/P2 (confirms both sensor count and the arrangement chart mentioned)
http://books.google.com/books?id=c-...AEwAQ#v=onepage&q=sony f35 photosites&f=false (confirms sensor photosite count)
http://image-sensors-world.blogspot.com/2010/12/sony-cinealta-moves-beyond-bayer.html (includes chart by *Sony* demonstrating the arrangement pattern)
Now, some of this documentation suggests equal weighting, not a high-low arrangement, but that simply having 2 separate bins reduces noise overall.
But all of this still works in favor or my bottleneck theory...
JMtheDP
Veteran
Viddo
I just meant that I believe all of the photosites on the sensor are of the same size/sensitivity/ND/whatever - I don't believe there is any special sauce applied.
Your theory could still hold some validity though as many cameras capable of genuine 25P can also do 50I, although perhaps for other reasons. One of which is data throughput - 25P and 50I are technically the same as far as most recorders are concerned, but in-camera a lot of Sony units used to double-up pixels to try and get the best res possible out of their 50I.
This is all from my hazy memory of a dodgy Sony manual from a long time ago!
I just meant that I believe all of the photosites on the sensor are of the same size/sensitivity/ND/whatever - I don't believe there is any special sauce applied.
Your theory could still hold some validity though as many cameras capable of genuine 25P can also do 50I, although perhaps for other reasons. One of which is data throughput - 25P and 50I are technically the same as far as most recorders are concerned, but in-camera a lot of Sony units used to double-up pixels to try and get the best res possible out of their 50I.
This is all from my hazy memory of a dodgy Sony manual from a long time ago!