F55: Looking for 2K 240fps raw MXF to look at

But red is resolving 3.5k and the oplf is too strong - chop that in four and, well you have 720, throw in a native 320 ISO chip masquarading as 800 and not usually ettr exposed and you have mud

maybe the same would be the case but I think not
 
I did a quick test - a wall of bricks at a school nearby.

I shot a few seconds in 120fps (HD XAVC), then switched to 24fps and shot a few seconds. There is a very slight aliasing in the 120fps footage (look at the roof top) but at the same time it looks slightly SHARPER than my 24fps footage. I did not change the focus between HFR and regular, but I did have to stop down from f4 to f5.6 and a half to get the same exposure.

Below are some frame grabs and original files for download from tonights tests. The night was coming fast so I ran out of light but I will do some more tests tomorrow. Seeing a slightly sharper image in the 120fps footage is somewhat backwards to what I was expecting and the only variable was stopping down for the 24fps footage but in generaly stopping down would give you a sharper image not a softer one.

120fps frame grab at f5.6 and a half.
120fps_brick_zps5baaccf1.jpg


24fps frame grab taken at f4
24fps_brick_zps744db01f.jpg


Difference matte image actually shows more of the difference in sharpness between frames, not necessarily aliasing although it looks that way.
diff_brick_zpsd350fc9b.jpg


Original Frame Grabs extracted using Davinci Resolve:
120fps brick DPX
24fps brick DPX
 
Um, did you move the camera at all?

It's def one of those things you have to see in motion. Still frames look fine.
 
Yes that was test # 2 :)

Benjamin Franklin helped me on with a 120fps pan and there was definitely a lot of jumpy little artifacts around highly detailed areas. The exact same shot in 24fps did not have the artifacts at all.
 
starcentral said:
Seeing a slightly sharper image in the 120fps footage is somewhat backwards to what I was expecting
Click to expand...

Aliasing gives the illusion of more "sharpness", but what you're seeing is very sharp edged errors. When you shoot resolution trumpets with a camera that is lacking a strong enough OLPF, it's pretty clear what goes wrong. And it *really* jumps out at you when there's small amounts of motion at the aliasing crossover. That's why the texture of your bricks goes bonkers. In my tests it was grass.

Circa 2005-2006 I was involved in a string of published and non-published camera tests, most of them including Adam Wilt, who taught me a lot about imaging science. Adam was really good about explaining what was going on during the tests to everybody present. Anyway, that's why I feel I know a little on it.

You know, if the OLPF fixes it well, I'll sigh and say "fair enough". I kind of expected Alexa level IQ at this price, but maybe it was too much to ask. Truth is, the Alexa has a lot more processing onboard and it sucks down batteries like no tomorrow. Easy to forget sometimes there's a good reason for it.
 
Nate Weaver said:
You know, if the OLPF fixes it well, I'll sigh and say "fair enough".
Click to expand...

The friggin OLPF should be FREE. This is a serious blow.



I'm uploading my Benjamin Franklin $100 US dollar bill test and will post once it's done. I will also upload the original 500MB MXF file if anyone wants to see it without any added compression.

I'm over on the Sony boards, sheer disappointment by a ton of people over there. Wait until the word gets out even more... not good.
 
This is exactly what I would have expected, given that they're trying to maintain the field of view and therefore skipping rows and columns.

Seriously the windowed mode is the only way to avoid that. Image degradation (as compared to 4k) is not actually happening at the pixel level; instead it's in that you have to technically blow up the image 4x for it to fill the same monitor space (thus exaggerating noise and softness). Thing is, you would have to do the same magnification (and experience the same softness and visibility of noise) with Sony's current implementation. The only difference should (SHOULD) be aliasing vs field of view. The windowed mode should be every bit as sharp as the current method, but with no aliasing.
 
Barry_Green said:
The windowed mode should be every bit as sharp as the current method, but with no aliasing.
Click to expand...

Maybe in theory. I've had enough Red 2K mode to not use it unless it's a last resort. But I understand what you're saying.

I think this all comes down to how well the OLPF works.

Also, I've been watching the raw 2K @240 I have from Peter Corbett, and I think I'm getting over it. I would have hoped Sony could read out the whole chip at 120 like Arri, but I guess it's just not meant to be. The Alexa has to go into an HFR mode too and to do so load a software module...maybe they just used all that onboard processing power to smooth out issues like this if they could clock their sensor to 120hz.
 
Nate Weaver said:
Maybe in theory. I've had enough Red 2K mode to not use it unless it's a last resort. But I understand what you're saying.

I think this all comes down to how well the OLPF works.

Also, I've been watching the raw 2K @240 I have from Peter Corbett, and I think I'm getting over it. I would have hoped Sony could read out the whole chip at 120 like Arri, but I guess it's just not meant to be. The Alexa has to go into an HFR mode too and to do so load a software module...maybe they just used all that onboard processing power to smooth out issues like this if they could clock their sensor to 120hz.
Click to expand...

Nate, looking at the Raw garden pan again, I mentioned the noise in the darker areas. Is it that the raw image contains no noise reduction and hence the warts-and-all image is sent to the DSP before processing? Maybe with some noise reduction in Resolve this would help the subjective picture quality? I just don't have enough experience with Raw yet to offer an objective opinion.

I've just another look at my new shipping port test and I think there is something like a would describe as fine "animated noise" in parts of the picture when the camera pans. I wouldn't describe it as regular aliasing, but I think it is the same effect. It's not there on static shots. I'm starting to pixel peep now so much that I don't know what I'm looking at. :huh: Have you got access to an Alexa so you can do an A/B?

Tomorrow I will go back to the port and shoot some 2K 120fps and 240fps in the same position as the XAVC HD.
 
When doing any tests I suggest shooting some regular 24p as well as the 120fps in XAVC, and RAW. This way you can see the difference in quality quite easily.

One thing I noticed is once you upload the video to YouTube or Vimeo some of the "issues" are reduced due to added compression. I uploaded my Bejamin Franklin test to YouTube and Vimeo and its not as noticable as in the original but what you see in the final video as "noise" or grain on the detailed areas WAS NOT in fact, that was the artifacting much worse on the original.

Here are the videos online. I will post links to both the MOV file I used to upload these, as well as the original MXF file straight from the camera, just waiting for the uploads to finish.

DOWNLOAD FILES HERE: www.starcentral.ca/forums/120fps_tests

https://vimeo.com/76299054 (you can download original from here too)

http://www.youtube.com/watch?v=F7BURfDukoU
 
Last edited:
powervideo said:
I've just another look at my new shipping port test and I think there is something like a would describe as fine "animated noise" in parts of the picture when the camera pans. I wouldn't describe it as regular aliasing, but I think it is the same effect.
Click to expand...

It's aliasing. Aliasing doesn't just mean "near horizontal lines that go thick and thin". Aliasing means there's detail too fine for the sensor to render, and there's errors in it's rendering because of it. Adam Wilt's writings explain this better, I don't have any specific links but I'm sure over the years there's something Google can find.



powervideo said:
It's not there on static shots. I'm starting to pixel peep now so much that I don't know what I'm looking at. :huh: Have you got access to an Alexa so you can do an A/B?
Click to expand...

I don't have immediate access to one, but I shot all week the week before last with one at 120, and I know for a fact it does not do this. I think that's one reason why I'm sensitive to it right now.

It's funny how these things go, I've got a new (amazing) client, I foresee a fair amount of high speed with them if I do more work for them (cars), so therefore I get all hyper-sensitive to how good the high speed looks. I think there's that, and my very first test shot had nightmare content for the problem, and showed it in it's worst possible light.

I've been looking at the raw you gave me again and I'm realizing that for most shots it would probably be ok. Most.
 
Nate Weaver said:
Maybe in theory. I've had enough Red 2K mode to not use it unless it's a last resort. But I understand what you're saying.

I think this all comes down to how well the OLPF works.
Click to expand...
But -- that's the point. The OLPF that's designed for 4K, is exactly appropriate for a windowed 2K. It really is just a center extraction. You should be able to test it for yourself, just record 4K raw, and crop it to the central 2048... that's about the best high-speed 2K you're going to get from this camera.

The only way to really improve on that would be if you could read the entire sensor out at high speed, and process it in a 2x2 grid the way Canon does the C300, that would get you a bit of pixel averaging on the green which should quiet down the noise a little. But obviously that would require a purpose-built high-speed chip, and -- if you could read the entire sensor out at that speed, why wouldn't you just keep it in 4K in the first place?

I don't think you can get better, higher-quality 2K out of a system like this than windowed.

I don't have a Red anymore, sold mine back in 2008, but the theory is sound, so I don't know what could have gone wrong with the Red's footage -- is the compression what's ruining it? Shouldn't be an issue with the Sony because you're getting uncompressed raw...
 
Barry_Green said:
so I don't know what could have gone wrong with the Red's footage -- is the compression what's ruining it? Shouldn't be an issue with the Sony because you're getting uncompressed raw...
Click to expand...

I dunno. Sony raw is compressed at 3.6:1 though.

Also, when the widest lens in your kit is 18, that's not wide at all in a 2K crop. It might work as a solution for some folks, but I myself was glad to leave crop thinking behind when I sold my Red. Never looked good to me.
 
Nate Weaver said:
I dunno. Sony raw is compressed at 3.6:1 though.

Also, when the widest lens in your kit is 18, that's not wide at all in a 2K crop. It might work as a solution for some folks, but I myself was glad to leave crop thinking behind when I sold my Red. Never looked good to me.
Click to expand...

The signal/noise ratio on the F55 is so much lower than it is on the MX sensor that it should hold up significantly better than crop footage on the Epic/R1 does (I imagine Dragon's crop footage will be much improved for this same reason) - because the lower noise should allow the frames to retain more detail.

Changing to a smaller image format is still a pain-in-the-A though (and entirely speculative until it actually exists as an option with these cameras).

Has the 2K OPLF been released yet? I'd be very interested to see comparisons between the two OPLFs and how the 2K footage holds up then.
 
OLPF is not out yet officially from what I know.

The online versions of the video I posted do no justice to the actual issue at hand. The compression of vimeo or YT really lower the resolution and mask the issue.

Download my original MXF and watch it if you want a good idea of what some of us are bickering about.
 
I'm quite interested in getting a S16 zoom and using the centre-cut option when it arrives. But it's something I am going to have to wait for to see how it actually looks, especially with HFR's. There are quite a few nice Canon S16 zooms available that were manufactured up until the mid-2000's, but I am more interested in the S35 servo zoom Canon that has been mentioned.
 
You must log in or register to reply here.
Back
Top