Big announcement coming for PMWF3???

[Ron Wilk]

You have my vote.

[George Griswold]

Originally Posted by Doug Jensen

Oh ya, well then I'm going to wait for the F9!

Ok.. I hear there is a F11. Think Spinal Tap! Cheers Doug.

[David C. Williams]

Originally Posted by David Heath

I think those figures are very pessimistic.

As evidence, look at the F3. It's known that the effective sensor count is somewhere close to 2456x1372 (the actual count including the masked pixels is 2468x1398), yet it delivers performance very good in terms of 1080 recording. Extrapolating what you say about a 4k sensor only giving "more likely 2.5k in general use", that would imply F3 performance equivalent to about 1500x875. In practice it's much better.

I'd heard somewhere around 80% as a ballpark figure, and the F3 results back that up. So horizontally the C300 sensor would be predicted to have around "3.2k" when fully deBayered. Maybe not "true 4K", but surely wrong to describe it as "no gain"?

Likewise remember that in the C300 the sensor does not use a considerable number of photosites round the edges - it windows 3840x2160 of a somewhat bigger sensor. Those could be used to give a somewhat bigger inprovement than the above suggests.

I used your figure of 80%. That's 3072. 3K. Best case. Look, all you've written will not make that 4096. The C300 chip will not shoot 4K, it is not suitable for 4K I can't even see how you can prevaricate on that?.

[David Heath]

Originally Posted by David C. Williams

I used your figure of 80%. That's 3072. 3K. Best case.

"4k" is normally taken to refer to 4096, and by my maths 80% of that is 3276.8 - not 3072. Compared to 1920 that's an increase of just over 70% on each axis.

Originally Posted by David C. Williams

Look, all you've written will not make that 4096. The C300 chip will not shoot 4K, it is not suitable for 4K I can't even see how you can prevaricate on that?.

And if you go back to my earliest posts you'll see no claim it will make "true 4K". Look right back at post 34 as example:

Originally Posted by David Heath

So it's true enough that the "4k" that the C300 sensor should give won't be as good as what would be expected from a 3 4k chip design, or a 5k deBayered and downscaled design.

But what will be the case is that the difference between it's current use for 1080p and using it deBayered with 4k recording is likely to be significant and worthwhile - of the order of 60% on each axis.

Is that clear enough?

(The 60% is by comparison with Quad-HD - 3840 - 70% is by comparison to 4096, which the C300 chip is capable of.)

Many cameras over the last few years have claimed "1080 recording" but have not been able to manage 1920x1080 resolution (eg the HVX200 with 960x540 chips, and res of about 1150x650). Should it have been the case that if "true 1080" couldn't have been achieved everybody should have just stayed with SD in the meantime? I don't think so.

In this case, a 60-70% res increase per axis over 1080 seems well worth the effort - even if it's not the full 100-110% that "true 4k" would achieve. And the reason for ending up with a 4k raster recording (as opposed to a 3.2k recording) is that the chip deBayers easily to that resolution, no downscaling is needed, and it ends up a common standard. Hence the reason for "plonking the 3.2k into the 4k bucket" as you put it. Very similar reasoning to the design of the HVX200 and the choice of 960x540 chips. Easy processing to a "1080 bucket", even with the pixel shift.

[David C. Williams]

The C300 only has 3840 pixel horizontally. Hence 80% is 3072. Good lord. Take your shifting goal posts elsewhere please.

If you think something that could resolve 3K with a very strong tail wind is suitable for 4K, good luck to you. I'm going shooting.

[David Heath]

Originally Posted by David C. Williams

The C300 only has 3840 pixel horizontally. Hence 80% is 3072. Good lord. Take your shifting goal posts elsewhere please.

If you're going to say I'm shifting goalposts, it's worth checking facts first. The C300 itself only uses 3840 photosites horizontally - but there is a big degree of windowing happening, the chip has far more, 4206 in total. (See http://cinemaeos.usa.canon.com/asset...-EOS_Specs.pdf )

Imaging sensor
Effective pixels: 3840x2160 ; Approx 8.29 megapixels
Total pixels:4206x2340 ; Approx 9.84 megapixels

It's usual to have a border of blanked photosites around the edge - but 366 horizontally is a far higher proportion than normal. The belief is that a second camera will follow soon, and to expect 4k. (Oh, OK, expect about 3.2k actual resolution in a 4k wrapper if we have to be precise.) In that case 4206 total makes sense - it's what would be expected if 4096 were actually used.

Similar story vertically. The total number is 2340, and the expectation is that 2304 would be active in 4k mode, as opposed to 2160 as used in the C300.