Sharpness and bokeh behavior based on different OLPFs

[Christoffer Glans]

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.

I rather have the lenses giving the image character than something fixed in the OLPF. While I find character to be charming in some of my own projects, sometimes I need to do super crisp work and if the OLPF is changing what makes a Leica 65mm look great, then that is indeed an issue. If you want character, get lenses that gives you that, the camera however should be made for perfection, not character.

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

Yes, especially how the Standard OLPF behaves compared to the LLO and STH in this matter.

 

[Phil Holland]

As far as I can tell, based on the pixel pitch of the Dragon sensor (and perhaps the reason to not go much smaller in pixel pitch), this is an effect of birefringence and variation between the OLPFs. The STH "spreads" wavelengths out more on the x and y axis, preventing some unwanted optical artifacts while inducing an almost anamorphic type of double refraction on a microscopic scale. Hard to tell where that happens. Likely on the birefringent layers or the wavelength pass through/blocking crystal. Don't think it's any of the IR or AR layers/coatings that would do this.

The LLO and Standard OLPFs have a more direct transfer of frequencies it appears. If you look at the small aperture flashlight examples hunting down the Red Orb artifact, this is likely one way it accounts for that variance.

Just a theory though.

The almond OOF image example posted earlier is a spherical lens aberration. That isn't a theory. It's a real thing and shows up differently depending on the optical design of the lens. Usually you will experience it in the field or edges/corners getting closer to where the projected image circle ends as the light is closer to the edge of the spherical or aspherical elements. For instance, Master Primes don't experience this "that much" on the Super 35mm format as they were handsomely over designed for most focal lengths. This also is why they don't exhibit as much vignetting wide open as some other lenses.

 

[Brian Boyer]

I rather have the lenses giving the image character than something fixed in the OLPF. While I find character to be charming in some of my own projects, sometimes I need to do super crisp work and if the OLPF is changing what makes a Leica 65mm look great, then that is indeed an issue. If you want character, get lenses that gives you that, the camera however should be made for perfection, not character.

I get that philosophy and it makes sense in many respects. However, coming from an audio background (which often foreshadows what happens in the video world), I don't know if it's as sound a rationale as people make it out to be.

I think about recording studios that spend hundreds of thousands of dollars on an SSL or Neve console for their A control room. They buy them for the sound they impart, not because they're the most pristine and transparent consoles money can buy.

That doesn't stop them from recording all manner of music and other material on them, each with its own final character and sonic identity. It's the same with microphones, preamps, amplifiers, etc.

I believe the camera wars are going to come down to who can produce the most pleasing looking images, not who can produce the most perfect ones. I don't think they're the same thing.

Maybe it would be better to petition RED to make a super pristine OLPF that produces razor sharp images. Of course, in doing so, you'll probably have to compromise on other aspects of the image.

 

[Christoffer Glans]

I get that philosophy and it makes sense in many respects. However, coming from an audio background (which often foreshadows what happens in the video world), I don't know if it's as sound a rationale as people make it out to be.

I think about recording studios that spend hundreds of thousands of dollars on an SSL or Neve console for their A studio. They buy them for the sound they impart, not because they're the most pristine and transparent consoles money can buy.

That doesn't stop them from recording all manner of music and other material on them, each with its own final character and sonic identity. It's the same with microphones, preamps, amplifiers, etc.

I believe the camera wars are going to come down to who can produce the most pleasing looking images, not who can produce the most perfect ones. I don't think they're the same thing.

Maybe it would be better to petition RED to make a super pristine OLPF that produces razor sharp images. Of course, in doing so, you'll probably have to compromise on other aspects of the image.

I'm not sure you really applied the right comparison. If the studio system is the camera, then the microphones, cables, amplifiers etc. would be the lenses. I'm not an analog lover so there might be something in the bolts and cables of the studio systems that you like that I don't care about, but in terms of recording audio, I would prefer to have perfection in how the sound is stored and recorded, while giving character through different microphones and setups. If the recording system in itself creates it's own sound, then why bother getting prestine old microphones if the system trash all that character? I still believe that the recording, the storage, the modern digital capturing tools both for video and audio should be perfect in order to store the real character in physical objects like membranes and lens optics.

As far as I can tell, based on the pixel pitch of the Dragon sensor (and perhaps the reason to not go much smaller in pixel pitch), this is an effect of birefringence and variation between the OLPFs. The STH "spreads" wavelengths out more on the x and y axis, preventing some unwanted optical artifacts while inducing an almost anamorphic type of double refraction on a microscopic scale. Hard to tell where that happens. Likely on the birefringent layers or the wavelength pass through/blocking crystal. Don't think it's any of the IR or AR layers/coatings that would do this.

The LLO and Standard OLPFs have a more direct transfer of frequencies it appears. If you look at the small aperture flashlight examples hunting down the Red Orb artifact, this is likely one way it accounts for that variance.

Just a theory though.

Sounds like a sound theory. It's a bit sad though, because I love the color representation and DR handling of the STH, but this focus distortion is messing up my videos. Latest work I did we shot wide open at T1.4 to get a very shallow feeling, but the distortion looks unpleasent in my opinion and the smooth bokeh of the lens is missing because of it. If the standard OLPF holds up ok with highlights and skin representation but reducing this issue, I will probably never use the STH again. All the other issues that people had with the STH, the noise etc. has never bothered me, but this definetely has.

It's like if I pay $1000 to rent a set of Leicas or Master Primes, they will not look as good as they do on other cameras without this issue. Even the Epic MX will give you a better looking image lens-wise. So it's a big waste to rent such expensive lenses just to have this OLPF issue smear their focus out unaturally.

 

[Phil Holland]

other cameras

This artifact does indeed exist on other cameras. It's magnified by a few things like resolution in this case.

If you like how MX looks, I would aim for the LLO or perhaps the Standard OLPF. They are closer to what you are after.

I'm not of the same opinion on the "waste of time" stuff. Each camera system has it's things. This is a slight thing on that is stronger on one OLPF.

 

[sam karr]

I saw it on my cam since day one. I don't see it as extra character, it looks like motion blur, it looks faulty, not organic or sexy.

 

[Brian Boyer]

I'm not sure you really applied the right comparison. If the studio system is the camera, then the microphones, cables, amplifiers etc. would be the lenses. I'm not an analog lover so there might be something in the bolts and cables of the studio systems that you like that I don't care about, but in terms of recording audio, I would prefer to have perfection in how the sound is stored and recorded, while giving character through different microphones and setups. If the recording system in itself creates it's own sound, then why bother getting prestine old microphones if the system trash all that character? I still believe that the recording, the storage, the modern digital capturing tools both for video and audio should be perfect in order to store the real character in physical objects like membranes and lens optics.

It's the right comparison. Sound flows through the console on its way to the recording medium (like light travels through the OLPF to the sensor). That medium could be analog tape, which has it's own sound that people still covet.

Or, it could be a digital recorder of some kind. The digital recorder will have ADCs that also influence the character of the sound.

It's called the recording chain and has a million different permutations. It's a balancing act. Tube mic, solid state preamp, analog console, digital recorder? Ribbon mic, tube preamp, digital console, digital recorder, with analog summing for final mixdown? It ultimately depends on how it sounds on the source.

But, you can't swap out the console in a studio the way you can lenses. It's a fixed entity, so it influences everything that passes through it, much like the OLPF. I'm saying it's not nearly as limiting as you might think.

There's only so much vintage glass out there and not all of it is accessible, physically or financially. Given the choice between owning an absolutely pristine camera that required older lenses to impart character and one that had a built-in character I love that could use many easy to find and less expensive lenses, I'd choose the latter.

No one I know of would plug a Les Paul into 10 different amps and wonder which one was producing the truest representation of the Les Paul sound (as if anyone outside a Gibson lab even knows what that is).

As long as it comes reasonably close to their expectation of what one sounds like, they mainly care about which one sounds the best to their ears, which may include hyping and/or cutting certain frequencies or adding certain types of harmonic distortion.

We've had over 100 years to figure out what we like in an image. A lot of that, celluloid gave us. Digital is making its contributions where film is lacking. Some combination of the two is going to win in the end, methinks. The sterility of highly pristine, almost clinical looking images will have their place, but I believe they'll be a specialty look.

 

[Christoffer Glans]

This artifact does indeed exist on other cameras. It's magnified by a few things like resolution in this case.

If you like how MX looks, I would aim for the LLO or perhaps the Standard OLPF. They are closer to what you are after.

I'm not of the same opinion on the "waste of time" stuff. Each camera system has it's things. This is a slight thing on that is stronger on one OLPF.

It might be a slight thing for some, but for me I "feel" the bokeh when looking at these pictures, it feels wrong, distorted, it doesn't look like any camera I've EVER worked with. I'm looking at 1080p offline renders of the video I'm working on and I see this distortion everywhere, it's not something only visable when cropping 100% looking for tiny details. When I shoot wide open and focus on infinity it looks even worse, like I held up a plastic wrap in front of the lens or something. I don't think this looks like a "character" of the camera, it looks like a sensor that's not alligned correctly, a faulty manufacture etc.

I really want to see what other cameras has this focusing distortion, because I have never seen it on any other camera. The resolution is not the problem since I can see this on a 1920x1080 scale down. It's affecting the entire image. If I use the LLO I will loose the light and skin properties of the STH. I rather see that Red takes a serious look at this and maybe fix it if possible. If I buy a $5000 camera I might live with the fact that things aren't going to be perfect, but for a $50000 camera I don't think I'm totally out there when demanding a certain level of quality. Especially on an upgrade that cost me $12500.
And I don't seem to be alone in thinking that this actually looks bad. As Sam said, there's nothing sexy about this, it looks like a bad sensor issue.

 

[Sergio Perez]

Have to say this is not exclusive to the sth olpf. It also happens on the llo.

 

[Christoffer Glans]

Have to say this is not exclusive to the sth olpf. It also happens on the llo.

Yes, that is true, as mentioned, it is less apparent on the LLO but it's still there. To get rid of it entirely you need to remove the OLPF all together. But the STH shows the most of it and for those, like me, who love the STHs color representation and DR handling, it's sad that it behaves like this.

 

[bevan goldswain]

I agree with you Chris, this does not look pretty.
As I think most would agree CA is not pretty.

This OLPF distortion definitely looks like a flaw in optics.

But, as others have said, I'd love to see some real world side by side test to see how often this is really a problem, at slightly higher stops an not just at 1.4

As much as I do think the distortion is an ugly one, I feel the trade off is worth it for the better colours, better tonal response and absence of Red orbs.

 

[Naim Sutherland]

Thanks for posting and pointing this out. Hopefully the standard OLPF tackles this issue. Even better would be a trade in program since the Highlight OLPF will definitely be retired if Standard can get away from these issues.

 

[Christoffer Glans]

This OLPF distortion definitely looks like a flaw in optics.

If you mean a flaw in that the OLPF is affecting the light transmission then yes. Optics however is not the problem since I've checked this out with Ultra primes, the Leica 65mm in this test, the CN-E series and numerous Canon still lenses. It occurs on my camera, a randomly picked other Dragon from the rental house where we tested it and not on an MX while also switching to LLO OLPFs on one of the Dragons which reduced the amount a little.

To see it in real world examples, it can be seen close to the plane of sharpness in your shots. Most notable, or when I have the most problem with it is when I have a shot of a persons face and the focus is shallow. If any of the skin is outside of the sharpness, that bokeh smears the details in the skin to the point that it almost looks like some anamorphic aberrations, but not nice looking ones. Problem is that it affects the "roll off" into bokeh, giving, based on if it's closer to us on the Z-axis or past the focus on the Z-axis, an image that has sharp details on one axis while being blurry on another axis. It then doesn't look like a nice optical roll off but an optical problem that just looks weird.

 

[bevan goldswain]

If you mean a flaw in that the OLPF is affecting the light transmission then yes. Optics however is not the problem since I've checked this out with Ultra primes, the Leica 65mm in this test, the CN-E series and numerous Canon still lenses. It occurs on my camera, a randomly picked other Dragon from the rental house where we tested it and not on an MX while also switching to LLO OLPFs on one of the Dragons which reduced the amount a little.

To see it in real world examples, it can be seen close to the plane of sharpness in your shots. Most notable, or when I have the most problem with it is when I have a shot of a persons face and the focus is shallow. If any of the skin is outside of the sharpness, that bokeh smears the details in the skin to the point that it almost looks like some anamorphic aberrations, but not nice looking ones. Problem is that it affects the "roll off" into bokeh, giving, based on if it's closer to us on the Z-axis or past the focus on the Z-axis, an image that has sharp details on one axis while being blurry on another axis. It then doesn't look like a nice optical roll off but an optical problem that just looks weird.

Christoffer - You misread me. When I said it looks like a flaw in optics I did not mean to say that I think it is. I agree with you the problem is the OLPF. It 'looks' like an optical flaw.

But some test on real images are always the way to show if this is really showing up outside of test charts.

 

[Christoffer Glans]

But some test on real images are always the way to show if this is really showing up outside of test charts.

It all started with me thinking that it was my Sigma 35mm that was faulty, so here's clips from that thread.

Sigma download

Here's a homemade chart that is shot from the side in order to se the focus plane.
Chart 2

Here's a screen from the CN-E test I did:
Screen
You can check it out in 4K on youtube through my links. The wide open shots in the end.

In the end of this clip you can see a shot outside showing the bokeh distortion.
Focus test 2
Notable primarily in the Red highlight reflection of the window. It should be round, not like it behaves.

If you want to see if for yourself, shoot wide open with a lens under T2.0 and look at how the bokeh behaves when pulling focus.

 

[Brian Boyer]

In the end of this clip you can see a shot outside showing the bokeh distortion.
Focus test 2
Notable primarily in the Red highlight reflection of the window. It should be round, not like it behaves.

This one is password protected.

 

[bevan goldswain]

Yes, I see it. It's not nice, makes the bokeh ugly for sure. But I sill think the trade off is worth it for not having red orbs + getting better colour.

If you shoot wide open most of the time then sure this is an issue for you and you may want to go for the LLO because of this. . I don't think there are a huge number of people shooting wide open for most of their work though.

 

[Phil Holland]

Yes, I see it. It's not nice, makes the bokeh ugly for sure. But I sill think the trade off is worth it for not having red orbs + getting better colour.

If you shoot wide open most of the time then sure this is an issue for you and you may want to go for the LLO because of this. . I don't think there are a huge number of people shooting wide open for most of their work though.

Remember the primary reason the OLPF is in there outside of blocking evil things is to reduce/eliminate moire, false color, and aliasing.

 

[Christoffer Glans]

This one is password protected.

Sorry pass is "sharp"

Yes, I see it. It's not nice, makes the bokeh ugly for sure. But I sill think the trade off is worth it for not having red orbs + getting better colour.

If you shoot wide open most of the time then sure this is an issue for you and you may want to go for the LLO because of this. . I don't think there are a huge number of people shooting wide open for most of their work though.

Nope, I don't shoot wide open, mostly on a T1.4 lens I shoot at T2.0, but I can see the distortion even then and all the way up to T4.0 even if it's not as present then.

Remember the primary reason the OLPF is in there outside of blocking evil things is to reduce/eliminate moire, false color, and aliasing.

Yes, but if there's a distortion that puts off all high speed lenses it means that it doesn't matter if I use sharp Leicas I still get distortion that looks like cheaper lenses. It becomes a strange situation where the OLPF puts character into the optical quality, not the lens. And what if you use a lens with already special characteristics, then these OLPF distortions will affect these. Then there's VFX and so on not getting the exact bokeh. I don't mind light loss or other aberrations of flares and alike for the OLPF, but when it's tampering with the image so that it never gets a a really nice bokeh, I think it's kinda off.

Point of me posting is primarily for Red to look into why this is happening on the STH and not so much on the LLO and nothing on the MX. If there's something in the process of making the OLPFs I think it would be good to look into it. Another theory is that it's due to the fact that the OLPF is detachable. Is it the same on Dragons with fixed OLPFs?