Does the F3 Have 5X the Light Gathering Ability of the RED EPIC?

You can't compare differing technologies directly. You could compare a Mysterium-X on the Epic against a Mysterium-X on the Scarlet, and say that the Epic will have better performance than the Scarlet due to larger pixel size, but Sony vs. Red and "Mysterium-X" vs. Q67-whatever, well, that's something that you're not going to be able to answer by looking at spec sheets.

When it comes to F3 vs. Epic, that's really a question that can only be answered in side-by-side tests. Both sensors are rated at 800 ISO. The Epic should be substantially sharper. The Epic is reported to be quite clean, and the F3 is reported to be extremely clean too.
 
There comes a point when I don't think that increased resolution matters. With clean, beautiful images where your audience does not lose its ability to suspend their disbelief you've hit gold. Both cameras will be able to do that. There is no need to compare them.
 
Dan Browning mathematically demonstrated that it is the total area of the sensor that determines light gathering ability or sensitivity, not the size of individual pixels. That assumes of course that everything else is equal, pixel design, A/D quality, circuit noise, etc.
 
I haven't seen Dan Browning's information, but just off the top of my head I can think of a few things that make that obviously wrong.

If you take a fixed area and fill it with 4 times as many photosites as another you loose far more area to photosite edges making your total capture area less. Photosites don't link edge to edge, they have large gaps. Microlenses help, but don't eliminate this, they also reduce light transmission.

Smaller photosites produce lower SNR due to lower photon capture ability. This is directly related to size. Fewer photon numbers to sample means lower accuracy. Smaller photosites means less total area, so less light gathering. More noise and darker. You also start to get funky effects like quantum tunneling causing all sorts of strange halation.

These issues will be solved some day, but at the moment large photosites means more sensitivity and accuracy.
 
If you reduce the Epic image to 1080p, and compare then maybe you can have a sense of a comparison....

REDMX footage is noisy, but nothing outrageous, however, when you down res to 1080 it cleans up a lot. RED claims its ISO800 Native, but many DP's shoot the MX at around a ISO400 for a clean signal, highlight clipping be damned. ISO800 is the point where it retains the most dynamic range, not necessarily its "cleanest point".

The Same is true with the F3, @ ISO800 its at its widest dynamic range. On the F3, it really hits a sweet spot where you get pretty spotless footage @0dB, and the best Dynamic Range. You lose highlight data when you drop to -3dB, but when in a controlled studio environment thats no issue.
 
David Williams said:
I haven't seen Dan Browning's information, but just off the top of my head I can think of a few things that make that obviously wrong.

If you take a fixed area and fill it with 4 times as many photosites as another you loose far more area to photosite edges making your total capture area less. Photosites don't link edge to edge, they have large gaps. Microlenses help, but don't eliminate this, they also reduce light transmission.

Smaller photosites produce lower SNR due to lower photon capture ability. This is directly related to size. Fewer photon numbers to sample means lower accuracy. Smaller photosites means less total area, so less light gathering. More noise and darker. You also start to get funky effects like quantum tunneling causing all sorts of strange halation.

These issues will be solved some day, but at the moment large photosites means more sensitivity and accuracy.
Click to expand...

That may be true for extreme differences in pixel size, especially at the small end of the range, but Dan was talking about pixel sizes over the 3-7 micron range. Within that range the sensors he was working with pretty much exhibited equal s/n ratios at the margin due to lower read noise offsetting the difference in pixel noise. I.E. 4 x 3 micron pixels in the same area as one 6 micron pixel could show equal s/n due to greater sample accuracy at levels approaching the noise floor. But a sensor of twice the dimensions and four times the total area of the smaller one would have a 6dB advantage at the same pixel size due to greater total light gathering ability at the same relative aperture.

Bottom line is you really can't predict performance based on something as simplistic as just pixel size.
 
TimurCivan said:
RED claims its ISO800 Native, but many DP's shoot the MX at around a ISO400 for a clean signal, highlight clipping be damned. ISO800 is the point where it retains the most dynamic range, not necessarily its "cleanest point".
Click to expand...

Couple notes, Red claims MX has a native ISO of 320 just like the M sensor. Lower noise allows it to be used at ISO800 which is acceptable to a lot of people it seems...and the reason people like to do this is the extra highlight protection. Because remember Red's ISO is metadata, the dynamic range is the same regardless of what ISO you use, all that changes is where that dynamic range is...at low ISOs that dynamic range is more in the shadows, and at high ISOs that dynamic range is more in the highlights. (This is assuming you're exposing properly for each ISO setting)
 
Gabrobot said:
Couple notes, Red claims MX has a native ISO of 320 just like the M sensor. Lower noise allows it to be used at ISO800 which is acceptable to a lot of people it seems...and the reason people like to do this is the extra highlight protection. Because remember Red's ISO is metadata, the dynamic range is the same regardless of what ISO you use, all that changes is where that dynamic range is...at low ISOs that dynamic range is more in the shadows, and at high ISOs that dynamic range is more in the highlights. (This is assuming you're exposing properly for each ISO setting)
Click to expand...

I was under the impression they say shoot it at 800...

Either way good info. Thank you Gabrobot.
 
Gabrobot said:
Couple notes, Red claims MX has a native ISO of 320 just like the M sensor. Lower noise allows it to be used at ISO800 which is acceptable to a lot of people it seems...and the reason people like to do this is the extra highlight protection. Because remember Red's ISO is metadata, the dynamic range is the same regardless of what ISO you use, all that changes is where that dynamic range is...at low ISOs that dynamic range is more in the shadows, and at high ISOs that dynamic range is more in the highlights. (This is assuming you're exposing properly for each ISO setting)
Click to expand...


Indeed, on all accounts. M sensor also claimed 320 but it wasn't that, IMO. But, I'd definitely say 320 on MX is about right. And, to add to that, 800 ISO is ridiculously clean, even with baked in curves, when lit appropriately.

800's a good barometer for a final grade, showing what you might get when you've adjusted curves of raw footage or R3D metadata.

IMO, 1600 ISO is still clean enough for any sort of exhibition, and looks to me to have the same amount of noise @ 4.5K as the F3 @ 3200ISO.

When it's crushed, the noise is stifled.

Will probably get to shoot some F3 footage this week with the Cooke 18-100 and some Panchros. Perhaps anamorphic as well if I can grab the lenses in time.

My first personal test is going to be 3200 ISO @ 2000watts of light.
 
TimurCivan said:
I was under the impression they say shoot it at 800...
Click to expand...

They do...but technically it's still 320. Small difference, but I remember them making a point about it. Also just a side thought, I remember John Shwartzman post that he personally rated the MX sensor at 800 for daylight and 640 for tungsten fwiw.
 
Gabrobot said:
They do...but technically it's still 320. Small difference, but I remember them making a point about it. Also just a side thought, I remember John Shwartzman post that he personally rated the MX sensor at 800 for daylight and 640 for tungsten fwiw.
Click to expand...

For noise floor in raw footage, yes? I think 640 @ Tungsten might actually be pushing it a little. Haha. But it's all great looking @ 1080/2K anyway.

Even 3200+ ISO at night doesn't look too terrible when treated to a Downrez.
 
Kholi said:
For noise floor in raw footage, yes? I think 640 @ Tungsten might actually be pushing it a little. Haha. But it's all great looking @ 1080/2K anyway.

Even 3200+ ISO at night doesn't look too terrible when treated to a Downrez.
Click to expand...

I hope they do a 4k release for Spiderman so we can see how it turns out.
 
You must log in or register to reply here.
Back
Top