PXW- FS5 - A very "soft" 4k image?

[Cliff Totten]

I love my FS5 and don't have much to complain about. However, when I compare it's image to other Sony cameras like the X70, RX10-II, A7s-II and Z150, it's image is noticeably softer and renders significantly less detail.

I was thinking about this earlier and I suppose it's sensor numbers completely explain the situation. All these other Sony cameras start with a pixel count that is considerably higher than a 1:1 readout ratio. Many Sony cameras today take a 6k image sample and artfully scale it down to a very detailed 4k image. (It's probably so high that it approaches 4K's maximum Nyquist frequency) Unfortunately, the FS5 cant do this because of it's 1:1 Bayer pattern read out. Let's look at the numbers:

A typical 1080 3 chip ENG camera like the Sony EX1 or EX3 and a dozen other shoulder models all have 3 full rater sensors for each color channel:

Red = roughly 2 million pixels.
Green = roughly 2 million pixels.
Blue = roughly 2 million pixels.

As we all know, this produces a very nice HD (1080) FULL RASTER image. but what about the FS5 or any other single Bayer pattern 4k sensor with a 1:1 RGGB pixel read out? The numbers now don't look so good. A 4k (UHD) 8 million pixel RGGB sensor delivers:

Red = roughly 2 million pixels (1080 resolution or equal to one quarter 4k raster)
Blue = roughly 2 million pixels (1080 resolution or equal to one quarter 4k raster)
Green = roughly 4 million piles (Only HALF raster 4k resolution)

So yeah, I guess this easily explains why the FS5 just doesn't have the same resolving power that other Sony Bayer sensors do that heavily over sample their image. So the FS5 gives you basically a 1080 image with "double HD" on the green pixels.

Some might not care and some might. For me, the FS5 is a nice camera overall, I just wish it had a 10-12 megapixel (or more) sensor and over sampled like other Sony 4k cameras do.

What do you think? Anybody care? I think this is an interesting topic to toss around.

CT

[Noa Put]

I know you like to talk numbers but have you actually placed the 4 camera's you mentioned next to your FS5 and shot the exact same scene at comparable settings and then visually compared the footage?

[Andy Wilkinson]

Cliff,

All I know is that I had an EX3, and still have a PMW-300 and C100 (3 chip HD and 1 chip 4K resolution downscaled in cam to HD respectively). By comparison (and I've done it) my FS5 produces FAR more detailed images than either if I shoot in 4K with a well set up pictute profile, obviously. But it is my only 4K cam so I have not got the ones you seem to have to compare.

It may just come down to camera picture profile set-up, especially edge enhancing "detail" settings in the various models you list.

But also, unless you deliberately shoot with a huge depth of field with a large sensor cam then, by default, there will be less of the image throught the frame that is in tack sharp resolution than on smaller chip/multi-chip cams that typically have a much greater DOF for reasons you will well understand.

I recently tested this huge depth of field approach on my FS5 to see exactly how sharp it could be with the lowly, but growing on me, kit lens. So typically a high F-stop, static shots in high contrast light conditions to eliminate motion blur etc.

Looked pretty sharp to me in the highest resolution (sub 4K) I am able to watch it on, see my 'North Norfolk in Spring' 4K Test Video in this recent thread here:

http://www.dvinfo.net/forum/sony-xdc...ml#post1913994

So I challenge you in your statement that the FS5 4K image is soft - watch the film (especially the Vimeo version) and tell me it's soft!!!

[Cliff Totten]

Yes, did it this weekend. That's why I got the idea about this topic.

It's not that I like "numbers". It's that the "numbers" explain actually explain the results.

Take a look at 4k from and A7R-II or A6000. The detail is downright striking. You will never see that on an FS5. But yes, the FS5 has other fantastic qualities like great ergonomics, wonderful highlight handling, high dynamic range and 12 bit raw sensor data output. (soon)

CT

[Noa Put]

Why don't you show some framegrabs to illustrate the difference and also list the exact camerasettings of each camera, including all profile settings for each camera like for detail etc? That would say a lot more then the earlier calculations you made. Once you have done that you can talk numbers that "might" explain the differences you see.

[Cliff Totten]

Originally Posted by Andy Wilkinson

All I know is that I had an EX3, and still have a PMW-300 and C100 (3 chip HD and 1 chip 4K resolution downscaled in cam to HD respectively). By comparison (and I've done it) my FS5 produces FAR more detailed images than either if I shoot in 4K with a well set up pictute profile. But it is my only 4K cam so I have not got the ones you seem to have to compare.

It may just come down to camera set-up, especially edge enhancing "detail" settings in the various models you list.

But also, unless you deliberately shoot with a huge depth of field with a large sensor cam then, by default, there will be less of the image throught the frame that is in tack sharp resolution than on smaller chip/multi-chip cams that typically have a much greater DOF.

I recently tested this huge depth of field approach on my FS5 just to see how sharp it could be with the lowly, but growing on me, kit lens. So typically a high F-stop, static shots in high contrast light conditions to eliminate motion blur etc. Looked pretty sharp to me in the highest resolution (sub 4K) I am able to watch it on, see here:

http://www.dvinfo.net/forum/sony-xdc...ml#post1913994

Yes, my FS5 has more detail than any other HD camera I have ever used. I'm simply saying that Sony 4k cameras that 6k over sample do capture significantly more eye popping detail. Even Sony will tell us this and it certainly should not be a shock to anyone.

CT

[Cliff Totten]

Originally Posted by Noa Put

Why don't you show some framegrabs to illustrate the difference and also list the exact camerasettings of each camera, including all profile settings for each camera like for detail etc? That would say a lot more then the earlier calculations you made. Once you have done that you can talk numbers that "might" explain the differences you see.

I'll setup some "formal" tests to post here this weekend. You can easily see it. In the mean time, just Google "Sony 6K over sample". You can see it on Sony's official YouTube Chanel. They talk about it.

α6300 -Ultimate resolution 4K movie in Super 35mm | α | Sony - YouTube

Again, anyone that understands how RGGB Bayer pattern sensors work, this is simple pixel math.

CT

[Noa Put]

I'll setup some "formal" tests

That's all we should need to see if the fs5 produces a very soft 4k image with significantly less detail then the camera's you mention, all other references to a video from another camera with no actual comparison or talk from Sony is of no use until you lay the images side by side and tell the exact settings and picture profile settings of each camera.
It's not that I don't believe you but you know a image says more then a thousand words :)

[Cliff Totten]

Agreed. Will do a real world resolution shootout on Saturday. The results are easy to see.

It is generally understood that higher resolution sensors will produce higher resolution images. No?

A 16 million pixel Bayer image consisting of 8 million green, 4 million blue and 4 million red pixels will have more resolution than an 8 million pixel Bayer image consisting of 4 million green, 2 million red and 2 million blue pixels. (8:4:4 vs. 4:2:2 RGB million samples)

You do understand how Bayer's RGGB color filter layout works, right? It's hard physics and there is pretty much no way around it on a single sensor design. (A single 8Mp vs. 16Mp Bayer Sensor resolution shootout)

CT

[Dave Sperling]

And of course to confuse matters even more, different cameras will have different OLPF's (optical low pass/anti-aliasing filtration.) So when zooming into those pixels, take a look and see how the 'jagged edges' are on those fine detail angular lines. (Presumably a high megapixel stills camera such as the A7R would have a significantly different approach to anti-aliasing than the FS5.)

[Cliff Totten]

No doubt!

Yeah, the optical low pass filter on a photo camera is set right at it's highest megapixel rating. So for a Sony A7R-II, that's 42 megapixels. If they optimize that filter for 4K video to filter out moire, that would kill the higher resolution photographs. So they need to open up the filter for higher frequency detail to help 42mp photos but that, in turn, will allow aliasing to pass thru in the lower frequency 4k video.

Me personally, I'll take the higher detail/resolution and suffer the moire problems that brings. In post, you can always soften an image and blur out detail. However, you can never "add" detail in post. You can only "sharpen" and that's not true "detail" at all.

The FS5 does video only. So, they can set the low pass filter right at the exact resolution of the sensor in 4k video.

The sensor column read out clock speed is much faster than any Sony Alpha camera too. So, the FS5 has no real skew (rolling shutter) problem either.

CT

[Noa Put]

Originally Posted by Cliff Totten

You do understand how Bayer's RGGB color filter layout works, right?

Ofcourse I do, that was the first thing they thought us in school.

Actually, and I"m probably not the only one, I only care about what I can see with my own eyes, I don't feel the need to break it down into little digits to know why camera "a" is less sensitive or has a less sharper image then camera "b". If I like what I see I buy it and go out and shoot with it and enjoy actually creating something instead of worrying if my camera reaches maximum Nyquist frequency.

[Cliff Totten]

lol. Yes, to each his own.

I'm the EXTREMELY curious type. When I was a kid, our VHS player had problems. I, or course HAD to open it up to see how it worked. Same thing with all of my cars and trucks. I'm the guy that grabs the tools, jacks it up and takes it all apart to see how its built.

I'm a fan of the science of imaging. Guilty as charged.

You don't have to know the inner workings of a transmission to be a great race car driver. You don't have to know ANY of the science behind image sensors to make great videos. But knowing the science behind things doesn't hurt you one bit either. ;-)

CT

[Noa Put]

There is nothing wrong with digging very deep into camera characteristics but I only find that when you come out with your findings stating they are just facts because it's pixel math and that the numbers explain the results but on the other hand use vague terms like "very soft 4k" or "significantly less detail" then at least add a visual comparison so we can see how bad it really is because you make it sound that way.

[Noa Put]

Originally Posted by Andy Wilkinson

So I challenge you in your statement that the FS5 4K image is soft - watch the film (especially the Vimeo version) and tell me it's soft!!!

Can you pls explain what was in the film, It's so soft I cannot make anything out of it, am I right I see boats? ;)
Your video is a nice sample of why I am confused about Cliffs statement that the fs5 produces very soft 4K, to my eyes that doesn't look like very soft 4K, can't imagine a RX10II would be "that" better.