Sharpness and bokeh behavior based on different OLPFs

[Christoffer Glans]

When I upgraded to Dragon from my Epic MX I noticed something. There seemed to be something off with the bokeh close to the sharpness plane. After a lot of testing, even sending it in to Red to have "it fixed", no one seemed to see this error or notice that something was off. I started investigating it myself in order to have sufficient proof that my camera was a bad apple. Instead I found out that the OLPF is to blame for this, not my camera, but all skin tone highlights. Worth mentioning, I've showed this to experienced cinematographers and technicians with more years in the business than me on this planet, so this is NOT something about backfocus or lens problems and whatever reason I've been given. This IS an OLPF issue, no question about it.

The issue was that my Dragon showed a slight distortion in the bokeh, most of the time close to the area of focus, the sharpness plane where everything is in focus. When focusing "through" and past the subject, the nature of the bokeh switch from vertical distortion to a horizontal distortion instead of being uniform, round and smooth. It's most notable at wide open and I tried a lot of lenses to see if it was the lens itself that was the problem. These lenses included the Sigma 35mm ART, the Canon 24-70 II zoom, the Canon 85mm f1.8, the 32mm Ultra Prime, the 50mm Ultra Prime and in this test the Leica Summilux 65mm, which based on some tests is the sharpest cine prime of them all. Non of them was to blame for this issue and because I used different mounts, the mount itself was not the problem either.

The only thing I could think of was that it's either the OLPF or in worst case scenario, there was something wrong with my sensor installation when upgrading. Sending my Dragon in to Red I recieved notice that there wasn't anything wrong with it, but no one seemed to actually understand the issue, even if I sent them my own tests. I got my camera back without any changes and no fix to this issue at all. So I had a thought that this may be the OLPF, since it's the only thing I couldn't change and the only thing that's a changable part apart from the mount and lens.

Moving on, I conducted this test, using the Leica Summilux 65mm, which is super sharp and in my experience one of the sharpest lenses wide open of all cine primes.

I used my own Dragon, another Dragon which we changed the OLPFs on and a Scarlet MX to really iron out where the problem was. At first, I was supposed to do this to get evidence that my own camera was off, but I was stunned to see that the other Dragon had the same issue. I remembered that I did not encounter this on my Epic MX so we tried the Scarlet and found non of the distortion at all. We then changed the OLPF on the borrowed Dragon and used the LLO and here's where things got interesting. We could see a slight distortion, but not at all like the skin tone highlight. We decided to remove the OLPF all together and voila the issue disappeared.

So the issue is clearly OLPF related and most noticable on the skin tone highlight. Now this isn't any of the issues people have talked about before, it's not about flares or noise, it's about the nature of the bokeh. It's about the bokeh being distorted instead of uniform and smooth. After thorough testing, I'm quite able to spot this issue even on full frame shots, not just crops. It affects the image as a whole and the character of the bokeh. Just like different lenses have different characters, this issue affects everything on the image, non-reversible.


Here's a list of videos and photos of the different setups, all cropped at 100% from the original R3D files.

Dragon STH camera 1 (notice the distortion in the bokeh when pulling focus)
Dragon STH 1

Dragon STH camera 2 (notice the distortion in the bokeh when pulling focus)
Dragon STH 2

Dragon LLO (slightly less vertical/horizontal distortion than the STH)
Dragon LLO

Dragon NO OLPF (No vertical/horizontal distortion, uniform and round bokeh)
Dragon NO OLPF

Original MX (No vertical/horizontal distortion, uniform and round bokeh)
Original MX


Ultra HD split-screen video of them all
Splitscreen

Timeline with them all
All

Images of the bokeh
Dragon STH
Dragon LLO
Dragon NO OLPF
MX


I recieved word that this issue might also have been encountered when using the motion mount. Based in this text, it sure reads like it's the same issue, but I can't figure out why this is happening. When looking at just the OLPFs, I would like to know how the new standard OLPF handles the bokeh characteristics, because frankly, I don't like the distortion at all, it makes even the best lens look cheap. If the standard OLPF works better, I will through away the STH.

If you have both an MX and a Dragon STH, please, take a look into this and see if you encounter the same.

 

[sam karr]

Great finding Christoffer. I also had the feeling that something was wrong with my dragon focus compared to my ex MX. Thanx.

PS: you linked the same STH image to both dragon STH and LLO.

 

[Anton Shavlik]

Looks like astigmatism.

 

[Joshua Cadmium]

It's spherical aberration due to the extra thickness of the OLPF.

 

[Anton Shavlik]

edit: nvm

 

[Christoffer Glans]

It's spherical aberration due to the extra thickness of the OLPF.

Obviously, but my point being that the aberration is so visible that it hurts the image. I don't mind the image being more blurry than without an OLPF, but a uniform bokeh creates a smoother look overall. This "astigmatistic look" looks awful when shooting wide open, when it doesn't have to be if shooting with another OLPF.

My hopes are that the standard OLPF won't affect the bokeh like this. If it does, I might even get the LLO just for this sake. Reason I think this hasn't been that visible in productions and film releases is that most projects focus on a T4.0 "standard", so it's noticeable but only if you look closely. I tend to shoot as wide open as I can and this creates a very noticeable aberration that looks like crap.

 

[Brian Boyer]

Since there was so much difficulty in spotting what you're talking about when you sent the camera to RED, even with your findings, it would be way more helpful if you illustrated your stills in Photoshop and highlighted (drew a circle around) specific areas with annotations describing what you deem to be the problem.

It would also help to show how it negatively affects a non-test chart image, for practical reference.

Since you see it so clearly, you're assuming we'll all see it just as clearly and, to be honest, I see differences but I'm not sure I know what you're talking about.

 

[Ulf Krentz]

Brian, if you look at the circles in the corners, you see the horizontal lines less diffused (nearly sharp) and the vertical lines more diffused. Not sure how that translates to a normal shot, but I have to agree it is there.

Cheers, Ulf

 

[Christoffer Glans]

Since there was so much difficulty in spotting what you're talking about when you sent the camera to RED, even with your findings, it would be way more helpful if you illustrated your stills in Photoshop and highlighted (drew a circle around) specific areas with annotations describing what you deem to be the problem.

It would also help to show how it negatively affects a non-test chart image, for practical reference.

Since you see it so clearly, you're assuming we'll all see it just as clearly and, to be honest, I see differences but I'm not sure I know what you're talking about.

Some people accept some things more than others, in my view, I think the bokeh looks like it's distorted almost as it looked like a motion blur and it also increases the chromatic aberration, so even if you use a lens that's almost completely free of CA as the Leicas are, you get it and it affects the colors in details and high contrasts. If you look at the images and compare the Dragon with STH and the one without an OLPF, it's quite clear that the STH increased the CA while the bokeh looks like the chart was moved sideways giving it motion blur. It's not uniform and pleasent looking as on the LLO or without an OLPF. Look at the nature of the bokeh between them, the details blends together unaturally on the STH while on the other it blooms out evenly.

This is further increased at infinity sharpness and makes even the sharpest lens incapable of getting into focus, even if it does so on other cameras. Other aspects is that if you look at a short depth of focus shot of someones face, the skin in the bokeh close to the sharpness plane looks smeared, which is quite distracting compared to a smooth out of focus on the LLO OLPFs or other cameras. Mainly it blends details together making the bokeh blend unnaturally and not smoothly from unfocus to focus. It creates sharp lines on details moving along the distortion and blends everything else into a mess.

If a shot is slightly out of focus on any other camera it's "ok looking", but with this distortion it looks weird as soon as the focus is just slightly off.

Horizontal skew

Vertical skew

These show the nature of the distortion and the increased CA. This CA is not really that heavy on any other camera, without an OLPF and is less apparent on the LLO. If this isn't an issue for you, great, I however think this is a significant flaw that is affecting to the point of making the most expensive lenses like the Master Primes perform well under their price range. That's why I hope the standard OLPF doesn't have this much distortion, but it could be that it has even more, hard to know before testing, especially since not even the Red technicians even cared to acknowledge that I found this issue.

 

[Patrick Tresch]

I must say THANK YOU for (finding and) tackling this issue!

Most of DP's (me included ;-) ) have past by this issue. There was a conssensus that this was the len's fault.
Your attitude help this community (and RED's camera) to continually be improoved and pushed beyond mainstream thinking.

The best way to understand what you are talking about, are the still frames "images of the bokeh" and the fast focus pull at the end of the STO video.

Hats off!

Pat


PS: I love my LLO ;-)

 

[Brian Boyer]

If you look at the images and compare the Dragon with STH and the one without an OLPF, it's quite clear that the STH increased the CA while the bokeh looks like the chart was moved sideways giving it motion blur. It's not uniform and pleasent looking as on the LLO or without an OLPF. Look at the nature of the bokeh between them, the details blends together unaturally on the STH while on the other it blooms out evenly.

Are you sure you don't have these images mislabeled, because I'm seeing plenty of CA in the NOOLPF image? Open images in a new window for larger versions.

STH OLPF:

No OLPF:

This is further increased at infinity sharpness and makes even the sharpest lens incapable of getting into focus, even if it does so on other cameras. Other aspects is that if you look at a short depth of focus shot of someones face, the skin in the bokeh close to the sharpness plane looks smeared, which is quite distracting compared to a smooth out of focus on the LLO OLPFs or other cameras. Mainly it blends details together making the bokeh blend unnaturally and not smoothly from unfocus to focus. It creates sharp lines on details moving along the distortion and blends everything else into a mess.

Do you have an example of an image of a face you can share that shows this?

These show the nature of the distortion and the increased CA. This CA is not really that heavy on any other camera, without an OLPF and is less apparent on the LLO. If this isn't an issue for you, great, I however think this is a significant flaw that is affecting to the point of making the most expensive lenses like the Master Primes perform well under their price range. That's why I hope the standard OLPF doesn't have this much distortion, but it could be that it has even more, hard to know before testing, especially since not even the Red technicians even cared to acknowledge that I found this issue.

I'm not trying to dismiss your concerns but, if it makes you feel any better, Alexa does weird things to bokeh as well. Here's an excerpt from the description below the video:

"The quality of the bokeh (out of focus highlights) in this piece is a combination of lens & the Alexa itself. Notice as the circles travel from the edge to the center as they go from an almond shape to a circular shape. After shooting on these lenses with many cameras, this feature only showed itself on the Alexa."

https://vimeo.com/22899223

 

[Andy White]

I'm not trying to dismiss your concerns but, if it makes you feel any better, Alexa does weird things to bokeh as well. Here's an excerpt from the description below the video:

"The quality of the bokeh (out of focus highlights) in this piece is a combination of lens & the Alexa itself. Notice as the circles travel from the edge to the center as they go from an almond shape to a circular shape. After shooting on these lenses with many cameras, this feature only showed itself on the Alexa."

Mmm - this is interesting, in terms of the STH being RED's 'sort of' closer solution to the Alexa look that people seem(ed) to want... looks like the STH is quite close in this regard. Personally, I have the LLO and that's fine for projects I'm involved in. Just find it very interesting you've highlighted an effect is on the Alexa - just goes to show how every element is like a piece in a really large jigsaw

 

[Patrick Tresch]

Why do you speak about the Alexa flaw?
The benchmark could be film as there is no need of OLPF. Film thikness layers could also play a role in the light transmission and bokeh?

 

[Brian Boyer]

Why do you speak about the Alexa flaw?
The benchmark could be film as there is no need of OLPF. Film thikness layers could also play a role in the light transmission and bokeh?

Because it's one I'm aware of that directly concerns bokeh. My point is they all have their flaws, if you look closely enough. As with filmmaking in general, there are always going to be trade-offs and compromises. It comes down to which ones you can live with and which ones you can't.

I assume RED, like every other manufacturer on the planet, will weigh addressing a particular issue against how likely it is to be encountered, subsequently noticed and ultimately disapproved of.

 

[Patrick Tresch]

Notice as the circles travel from the edge to the center as they go from an almond shape to a circular shape. After shooting on these lenses with many cameras, this feature only showed itself on the Alexa."

But the alomond bokeh is due to the off center light and the (small) optical path. For example, you can have some straight lines due to the mattebox shape. Not a feature (lol) only showed on the Alexa...

 

[Christoffer Glans]

"The quality of the bokeh (out of focus highlights) in this piece is a combination of lens & the Alexa itself. Notice as the circles travel from the edge to the center as they go from an almond shape to a circular shape. After shooting on these lenses with many cameras, this feature only showed itself on the Alexa."

https://vimeo.com/22899223

It's not about highlights in the bokeh, but the character of details being skewed and distorted rather than uniform round out of focus.

I assume RED, like every other manufacturer on the planet, will weigh addressing a particular issue against how likely it is to be encountered, subsequently noticed and ultimately disapproved of.

Perhaps, but there are a lot of people who just go out and shoot, who don't care about stuff like this. I'm not one of them, I noticed as soon as my camera got back from the upgrade that there was something strange. It just took me time to iron out the problems, since when I sent my camera to Red I got a video back of them changing the aperture on out of focus highlights... which wasn't problem at all. I explained and showed videos but they didn't get what the problem was since they just compared to another Dragon which, from what I know now, has the same thing. So their quality control camera has the same distortion and therefor "I don't have a problem". Maybe no one cares, or maybe I'm the only one actually looking, I dunno, but I clearly see this distortion, even on full size pictures.

But the alomond bokeh is due to the off center light and the (small) optical path. For example, you can have some straight lines due to the mattebox shape. Not a feature (lol) only showed on the Alexa...

These almond shapes is something I've seen on a few setups, mostly at wide open on really high speed lenses. But I don't think it has anything to do with the vertical/horizontal distortion on details. The distortion is gone once the image goes into heavy unfocus. The distortion is seen close to the focus plane. As soon as something is in focus you don't see it, but shooting high speed at say T1.4, the sharpness plane is so thin that the distortion is very noticeable, which means that any focus change or subject change will distort a lot and it doesn't look good.

 

[Nick Fury]

I have to decide on the included OLPF, for Weapon upgrade in the next few days. I was also a proponent of the LLO for Dragon. Has anyone seen a NON-Lutted comparison of the three OLPF's for Weapon? If so, please direct me to any information, in that regard.

Sorry to interject another camera, but assuming that the OLPF's are behaving the same in each camera is likely , but not sure until tested...

Thank you very much

Fury

 

[Phil Holland]

I have to decide on the included OLPF, for Weapon upgrade in the next few days. I was also a proponent of the LLO for Dragon. Has anyone seen a NON-Lutted comparison of the three OLPF's for Weapon? If so, please direct me to any information, in that regard.

Sorry to interject another camera, but assuming that the OLPF's are behaving the same in each camera is likely , but not sure until tested...

Thank you very much

Fury

That can be seen in the test footage I put up if you can live with DRAGONcolor2/REDgamma4.

 

[Jake Wilganowski]

Well I think it would be great to get an official statement from red in this thread. At least an acknowledgement.

 

[Brian Boyer]

But the alomond bokeh is due to the off center light and the (small) optical path. For example, you can have some straight lines due to the mattebox shape. Not a feature (lol) only showed on the Alexa...

Perhaps. Or, perhaps the Alexa is more susceptible than other cams. You'll have to take that point up with the guy who originally wrote it. He's speaking from his experience with those lenses on "many cameras."

Btw, I think his use of the word "feature" was simply a poor choice of words and not something he thought was intended as a selling point. I understood him to mean "anomaly."


For the purposes of this discussion, what matters is if someone thinks an anomaly looks bad and seeks to have it "fixed." I contend it depends on the anomaly.

I prefer digital images that have a very organic look to them. Granted, that's going to mean something different to me than it does to others. To some, it's going to sound like a load of crap. At any rate, I know what it means to me and, more importantly, I know it when I see it.

Now, the more organic looking the image is overall, the more it's going to have perceived flaws, technically speaking. The "fuzziness" I may think adds character and looks awesome may be the very thing that someone else thinks makes the image look horrible and wants fixed.

We're going to be at odds because I'm all for the less-than-pristine look. Christopher raised an issue that deserves to be looked at but the charts alone aren't convincing enough to me. We have a psychological expectation that black and white lines and text look a certain way.

When color is introduced (through CA) it looks worse because we know it's not supposed to be there. The STH OLPF may not be blurring the way you'd expect but that in itself isn't necessarily a bad thing, which is why I'd like to see an image of a face, with STH and without, that shows what he thinks looks bad.

I might actually agree with him. But, let's not pretend that all technical flaws are (universally) bad looking. People go out of their way to put stank on their images by tracking down old, imperfect lenses, for example.