The New F

[Andy Shipsides]

Originally Posted by Nate Weaver

Andy, does the 350 kinda fold out and unhinge itself so you can use it without the eyepiece?

I see the hinges on the mirror box, but I can't see if the bracketry lets it flip out 90º for direct viewing?

Also, I'm starting to guess the 350 is going to be the cheaper one and the 100 OLED is going to be the step up. Yes? No? Can't say?

The LCD doesn't move, but the mirror box hinges up for viewing on the side.

The OLED evf will be pretty pricey, The LCD version will be cheaper for sure.

Andy

[Dennis Hingsberg]

Anyone want to comment on the F55 sensor and if they think it is the new Q67 filter pattern or just regular bayer? I did ask Sony directly on their Facebook FanPage and was told they would not divulge any information about its sensor. Seemed like a tardy reply for a company looking to heavily compete in the super35mm market.

Sony's press release had 3 paragraphs of reference to the F65 and in the end only said the F55 would "deliver the same color filter with ultra wide color gamut as the F65 for true color reproduction" - but that could mean anything.

The press release indicates each camera features a new type of 4K Super 35mm image sensor with a 4096 x 2160 resolution (11.6M total pixels).

[David Heath]

Originally Posted by Dennis Hingsberg

I did ask Sony directly on their Facebook FanPage and was told they would not divulge any information about its sensor. Seemed like a tardy reply for a company looking to heavily compete in the super35mm market.

Even more odd since it's already been divulged!!

Brian posted a very good link which explains it well in another thread - Sony’s PMW-F5 and F55: Defining CFA | CineTechnica

Basically, no, the F55 is not Q67 (as the F65 is), it's a normal Bayer 4k in geometry, so 3840x2160. [EDIT That should be 4096x2160, though it will also do quad-HD.] (Q67 has twice as many photosites, so 3840x2160 green, and as many again shared between red and blue. You really need to look at a diagram, but it means a green photosite for every output pixel, and easy reconstruction of red and blue from surrounding sites. Rows are at 45 degrees to the horizontal.)

So in that respect (4k Bayer) it's like the F5. How it differs is regarding the gamut of the filtration, it's theoretically capable of defining a greater range of colours, which makes it more suitable for high-end digital cinema, but may be overkill for display on such as LCD or OLED monitor screens. The other way it differs from the F5 is in that it's a global shutter. But do note that that comes at a price - reduced sensitivity compared with the F5.

[Dennis Hingsberg]

David, thanks for your post and for pointing me to that article by Mitch. I felt the way that press release was worded it was easily going to be misconstrued as meaning something else but I didn't know enough about the CFA differences in-camera to be sure.

Yes I've seen the Q67 diagram and noted that pixels (photosites) are oriented in 45 degree rows. The diagonal on the F65 was 8000 pixels I believe while the horizontal and vertical counts were somewhere around 6000 x 3000.

So with these new 4096x2160 F5 and F55 sensors do you think we can expect a "theoretical" resolution of around 3.2k lines or perhaps slightly more? I read posts by Alister Chapman and Graeme Natress indicating bayer filter pattern sensors typically yield between 70-80% in sensor resolution as a rule of thumb. I'm just curious how these F's might perform on the charts.

[David Heath]

Originally Posted by Dennis Hingsberg

Yes I've seen the Q67 diagram and noted that pixels (photosites) are oriented in 45 degree rows. The diagonal on the F65 was 8000 pixels I believe while the horizontal and vertical counts were somewhere around 6000 x 3000.

No, not true. Likewise it's misleading to call the F65 an "8k sensor" in my opinion. That infers something with dimensions of (about) 8,000x4,000 photosites, or 4x the number of a 4k sensor. In practice, it's got 2x the number and because of the Q67 nature can't really have a number put on it in the same convention as gets used for Bayer.

Think of it this way. Imagine black and white tiles on a bathroom floor, laid corner to corner across the width and length of the room. Now imagine the white tiles are green photosites, the black tiles are red and blue photosites. If you count the green tiles, you get 4096x2160 (laid tile corner to tile corner) and they are interspersed with another grid of 4096x2160 red and blue. Ask how many in total, and all you can realistically say is "2x4096x2160". A figure of 4096x4340 is just as valid as 8192x2160 - see why I don't like saying "8k"? :-)

It also follows that for a diagonal which goes the full height of the sensor the no of photosites must be 2160xsq rt 2, or about 3,050.

So with these new 4096x2160 F5 and F55 sensors do you think we can expect a "theoretical" resolution of around 3.2k lines or perhaps slightly more? I read posts by Alister Chapman and Graeme Natress indicating bayer filter pattern sensors typically yield between 70-80% in sensor resolution as a rule of thumb. I'm just curious how these F's might perform on the charts.

That's as good a figure as any. Practically, it's impossible to directly compare a Bayer with a theoretical 3 chip design, as the resolution will behave in a different manner. It will also depend on the image being looked at, and especially things such as whether it's monochrome or saturated colours. So if you look at a scene of saturated red, the definition is only going to be about 2k, if black and white, may be more like the 3.2k you suggest.

And yes, the F65 will be better, but we must be getting into a law of diminishing returns. I can certainly tell the difference between such as 720 and 1080 resolutions, and I'm willing to agree there is a place for 4k for digital cinema. (Though only the further forward rows will notice any difference compared to 1080.) But you may have to sit in the front one or two rows to really see the difference between 4k Bayer and 4k Q67. :-)

[Dennis Hingsberg]

Yes I agree calling it 8k was misleading, and it lead some people to think that the F65 measured 8000 pixels diagonally including some published articles about it! But to measure 8000 pixels diagonally would have meant it was approximately a 7000 x 3700 sensor by traditional means which would have meant it was a 26MP sensor - which the F65 is definitely not.



My theoretical resolution calculation for the F5/F55 is based on non-3CCD cameras and what has been determined by other "experts" as the 70-80% rule. For example the RED sensor in the Scarlet and EPIC is 5120x2700 and the EPIC in 5k mode resolves around 4000 lines of resolution while the Scarlet (still with the same sensor) can only shoot 24fps in 4k so it therefore only resolves around 3200 lines.

These values seem to fall in line with what one can expect resolution-wise from a bayer patter sensor so I am speculating the F5 & F55 will be around 3200 lines.

[David Heath]

Originally Posted by Dennis Hingsberg

My theoretical resolution calculation for the F5/F55 is based on non-3CCD cameras and what has been determined by other "experts" as the 70-80% rule.

As said before, it's as good a figure as any, but I would qualify it at very least by saying it's the LUMINANCE resolution. With a 3-chip sensor, the actual sensor results will be totally independent of colour, saturation etc of the input. That's not true with such as a Bayer sensor, (same was true with pixel-shifting techniques).

Practically, the resolution will drop off gradually - you won't get a situation where it will resolve 3000lpph perfectly, but 3050 will just look grey. Hence at what point do you say resolution goes up to? Which brings in the whole subject of modulation transfer function or mtf.

Sorry - I know it's great to have simple cut and dried numbers for comparisons, but real life is not that simple! Best to think of the 70-80% figure (which I basically would agree with) as a rule of thumb rather than anything too exact.

[Alister Chapman]

And in a real world image the resolution will vary across the image depending on what the scene looks like. You might have one part of the image exhibiting a higher resolution than another because of the colour. A scene of a woodland in the summer with green leaves will have a different resolution to the same scene in Autumn with red leaves. So an average range is a reasonable compromise.

As David says, we are reaching an area of diminishing returns. Most cinemas in the UK are only 2K. The new standard for UHDTV is quad HD (3840x2160 or 2160p) and the new and very expensive 4K TV's like the new Sony use 8 mega pixel panels (3840x2160 pixels), not sure whether they class a pixel as a cluster of RGB emitters or a single R, G or B emitter. So while it is nice to have some oversampling to give some wriggle room in post, at the moment there are very few real world applications where you need more than a 4K horizontal pixel bayer sensor. Originally 1920x1080 was in part chosen for HD as this was felt to be as much as you needed for normal viewing conditions in both cinemas and at home, so 4K will only really be significantly better for those with excellent eyesight sitting closer than average to the screen.
I saw the Sony 80" 4K TV at IBC, and I I've seen plenty of demo's of 4K projection. While these have often impressed, it is often because of the quality of the cinematography. When I've seen less impressively shot 4K, to me it looks no different than HD so I have to wonder how much of the wow factor is down to putting good kit in good hands rather than just the resolution increase.

[Dennis Hingsberg]

I think I'm going to be adding a 1/4 diffusion filter to my arsenal of filters for all this rez!

:p

[Alister Chapman]

Make sure the filter is suitable for 4K. Many conventional HD diffusion filters will excessively soften at 4K.

[Mark Kenfield]

Personally, I think the Quad-HD standard makes a bit more sense than the somewhat larger 4K digital cinema standard - the resolution difference between the two is minimal, but downscaling and upscaling between them should be achievable with considerably less fuss (due to the simple 2x conversion) which makes sense to me.

[Alister Chapman]

That's a big part of the reason why UHDTV has been standardised at Quad HD. It will be so much easier to convert to HD than other odd sizes.

[David Heath]

Originally Posted by Mark Kenfield

Personally, I think the Quad-HD standard makes a bit more sense than the somewhat larger 4K digital cinema standard - the resolution difference between the two is minimal, but .........

Just bear in mind that quad-HD is 16:9 aspect ratio, same as current HDTV standards (or 1.78:1). "4k" defines a horizontal resolution, generally considered to be 4096 (cf 3840 for quad-HD), and if you assume the same vertical figure (2180) that will mean a 4096x2160 chip with square photosites will have a wider aspect ratio: about 1.9:1. (Slightly wider than the common cinema standard of 1.85:1)

That may be considered more cinematic for projection cinema, and with appropriate shoot and protect guidelines whilst filming may be very easily cropped horizontally to give the 3840x2160 (16:9) format for optimum display on plasmas, LCDs etc.

[Brian Drysdale]

FD Times has a 96 page pdf feature on the F5 and F55.

http://www.fdtimes.com/pdfs/articles...4.5-150dpi.pdf

[Mark Kenfield]

Originally Posted by David Heath

Just bear in mind that quad-HD is 16:9 aspect ratio, same as current HDTV standards (or 1.78:1). "4k" defines a horizontal resolution, generally considered to be 4096 (cf 3840 for quad-HD), and if you assume the same vertical figure (2180) that will mean a 4096x2160 chip with square photosites will have a wider aspect ratio: about 1.9:1. (Slightly wider than the common cinema standard of 1.85:1)

That may be considered more cinematic for projection cinema, and with appropriate shoot and protect guidelines whilst filming may be very easily cropped horizontally to give the 3840x2160 (16:9) format for optimum display on plasmas, LCDs etc.

Absolutely, I'm thinking more in terms of TV-bound content.