Is the new OLPF doing something that filtration and post could do?

[Brett Harrison]

I want all the information coming through unmodified. A stop above is the same as a stop below, just move the exposure. It's all linear in the maths of it, right? That's what RED has been all about for me, so I think the old OLPF would be the go. Any gradation from signal to clip is something that occurs either before, or after the sensor, in my understanding.

I am looking to be corrected and seeking opinions, so put away the knives.

 

[Dominik Muench]

Interesting concept, Im not sure about post, but filtration possibly ? after all the OLPF is basically a filter, would it matter really if its in front of the lens or between lens and sensor like it currently is ?
Post would probably be not possible or only to a very limited extend as the look of the non existent OLPF (if thats what you mean?) would already be embedded into the footage, remember that OLPF's are also responsible for such things as moire...which can only be fixed in post with limited results. I believe there are camera systems out there that work without OLPFs but they then use a different chip technology altogether.

I just picked the new OLPF for my upgrade. I'm not sure if changeable OLPFS will really be for me...although the prospective of Infrared shooting with a swappable OLPF will be very exciting.

 

[Phil Holland]

First look at the images in this thread for a visual comparison between the two:
http://www.reduser.net/forum/showthread.php?112230-the-New-Dragon-OLPF

The Old OLPF is fine for some folks. The New OLPF is fine for other folks.

The Old OLPF is prone to more IR shadow contamination with bright sources. You can also see a slight improvement in overall color performance with the New OLPF.

The Old OLPF has "orbs" around highlights in frame, trick to fix in pox. Also the "red dotted" IR reflection at smaller apertures occurs only on the Old OLPF.

Overall you are looking at about a 1.3 stop variation between the two in terms of how far it can see into the dark and how much highlight information is retained.

 

[Brett Harrison]

Actually I'm aware of some really killer filtration profiles that work very well to replicate filtration that happened before the sensor. Think of it this way: the image hitting the post production system is the same light that hit the sensor, you then put this light through a 'post filter' and that hits the 'sensor' of the file format you're storing it in, if you will.

I'm aware that an OLPF does more than a typical filter and is made bespoke for the sensor you're using, but any of the important and detailed work that the Dragon sensor needs would be in the first OLPF I would assume.

What I am trying to get at is the difference between the two, and if the qualities of the second can't be had with modern post processes - and if so then I would opt for the faster, original OLFP.

Obviously it's all moot if you can swap them to taste, but still I would like to understand the differences from a technical point of view.

 

[Brett Harrison]

Thanks Phil!

Do you foresee a third gen OLPF that has the IR advantages of the second with the speed of the first? Or is that eating one's cake whilst having it?

First look at the images in this thread for a visual comparison between the two:
http://www.reduser.net/forum/showthread.php?112230-the-New-Dragon-OLPF

The Old OLPF is fine for some folks. The New OLPF is fine for other folks.

The Old OLPF is prone to more IR shadow contamination with bright sources. You can also see a slight improvement in overall color performance with the New OLPF.

The Old OLPF has "orbs" around highlights in frame, trick to fix in pox. Also the "red dotted" IR reflection at smaller apertures occurs only on the Old OLPF.

Overall you are looking at about a 1.3 stop variation between the two in terms of how far it can see into the dark and how much highlight information is retained.

 

[Dominik Muench]

Actually I'm aware of some really killer filtration profiles that work very well to replicate filtration that happened before the sensor. Think of it this way: the image hitting the post production system is the same light that hit the sensor, you then put this light through a 'post filter' and that hits the 'sensor' of the file format you're storing it in, if you will.
.

this would actually be pretty sweet if it could be done in post, kinda like a OLPF filter choice menu in RedCine when you do your encoding.
Im not sure how the Dragon sensor would perform completely without OLPF though. aspects like moire and IR would have to be covered.

 

[Phil Holland]

Thanks Phil!

Do you foresee a third gen OLPF that has the IR advantages of the second with the speed of the first? Or is that eating one's cake whilst having it?

I think if there is way to produce something different Red will be on top of it. OLPF design is alarmingly tricky business. The Dragon sensor is alarmingly sensitive in reality and finding a new OLPF design/tech might be possible. I know of other types of sensor tech that there are different methods that can yield more ideal results under stressful situations, but we'll be seeing those cameras in a year or two. It's pretty clear what's going to be in the next generation of cameras at this point to me.

Currently the V1 and V2 have been and are currently being used on productions word wide. So they are production ready and fine. More or less the difficulty people are having at the moment is answering the tricky question of how to get the best possible images out of Dragon. Like all cameras there are things you want to avoid in terms of exposure practices. Shot with a critical mind and care, well you guys are seeing trailers everywhere now of really nicely shot stuff really. Stressing the system at high ISO ratings in high contrast situations isn't going to work nicely for any camera out there currently.

We are certainly chasing more and more low light performance, but there's all sorts of nasty things down there that are going to cause headaches. Even lens manufacturing artifacts start revealing themselves at ISO 3200 on some glass. You'll see circular marks that are rather horrible.

What I do foresee is an interesting OLPF Module that allows for the flexibility of swapping out your own V1, V2, and whatever other thing Red can come up with. Low Con I think will happen. Don't know about a haliation version as that's down to individual's taste on a micro level. We'll see IR pass and Clear as Jarred as mentioned. Should be interesting to experiment honestly.

 

[Phil Holland]

this would actually be pretty sweet if it could be done in post, kinda like a OLPF filter choice menu in RedCine when you do your encoding.
Im not sure how the Dragon sensor would perform completely without OLPF though. aspects like moire and IR would have to be covered.

For general production use you really, really want an OLPF in there. Mostly for the anti-aliasing effect it provides. For some it will work for their style of shooting, but in reality the first wide shot of a city scape or any tricky fabric that will produce aliasing or color moire will ruin your day.

 

[Peter Majtan]

OLPF could never be to far away from the sensor - in most sensors the OLPF is almost directly on the sensor. The actual optical structure of the OLPF needs to match the pixel pitch on the sensor and be perfectly parallel with the sensor surface. This is why OLPF could never be in front of the lens.
RED has already stated that they will include a "clear" OLPF in their soon-to-come modular OLPF solution. This is the closest to what you are asking for. Just be aware that you will need to handle some of the issues addressed by the current OLPF's - such us IR pollution, aliasing, etc...

:sifone: Peter

 

[Brett Harrison]

Peter I am referring to the difference between the v1 and v2, not the difference between OLPF and none at all. It struck me that V2 is doing more - stuff that I would prefer to do in post - but Phil hasset
me straight there.

OLPF could never be to far away from the sensor - in most sensors the OLPF is almost directly on the sensor. The actual optical structure of the OLPF needs to match the pixel pitch on the sensor and be perfectly parallel with the sensor surface. This is why OLPF could never be in front of the lens.
RED has already stated that they will include a "clear" OLPF in their soon-to-come modular OLPF solution. This is the closest to what you are asking for. Just be aware that you will need to handle some of the issues addressed by the current OLPF's - such us IR pollution, aliasing, etc...

:sifone: Peter

 

[Lauri Kettunen]

Brett, your question is a good one and makes a lot of sense. The basic reasoning behind raw is the linear correspondence between light intensity and the stored digital data. Consequently, what has been explained about the new OLPF raises a (principal) question whether the linearity still prevails all the way from the bottom to top. More precisely, the practical question is, does the new OLPF compress light.

I think a difficulty for all camera manufacturers is, many people still shoot with digital cameras as if they were shooting with film. As a result, operation of modern cameras is rather much like that of cameras with film. My guess is, if film cameras never existed and the world had went strait to digital cameras, operating cameras would be different, and I believe also most people got more out of their cameras. If I may exaggerate, it's bit like if CDs had to played in the same way as old LPs. ;-)

 

[Brett Harrison]

Thanks Lauri, that's what I am getting at. If there's a 1.5 stop reach into the highlights with a loss of 1.5 stops to a lifting of the noise floor, you still have the same dynamic range and sensitivity, but the curve introduced from the OLPF forces you in a way, to preserve highlight detail by opening up the aperture to compensate for the raised noise floor, thus regaining the highlight detail. That 1.5 stops is then distributed into a roll off of the highlights. Something akin to a 'highlight protection mode'.

But that kind of highlight soft clip can be done in post. I suppose not having to do it every time is sensible. You have to have access to a really nice de-contrast process to replicate what's going on in the V2 to graduate nicely to clipping, I guess. But on a CMOS sensor, it's meant to be a linear light situation.

So does the V2 represent some kind of 'characteristic curve like with film?

I figured I'd clarify what I meant to someone who can address it. Cheers.

Brett, your question is a good one and makes a lot of sense. The basic reasoning behind raw is the linear correspondence between light intensity and the stored digital data. Consequently, what has been explained about the new OLPF raises a (principal) question whether the linearity still prevails all the way from the bottom to top. More precisely, the practical question is, does the new OLPF compress light.

I think a difficulty for all camera manufacturers is, many people still shoot with digital cameras as if they were shooting with film. As a result, operation of modern cameras is rather much like that of cameras with film. My guess is, if film cameras never existed and the world had went strait to digital cameras, operating cameras would be different, and I believe also most people got more out of their cameras. If I may exaggerate, it's bit like if CDs had to played in the same way as old LPs. ;-)

 

[Dominik Muench]

For general production use you really, really want an OLPF in there. Mostly for the anti-aliasing effect it provides. For some it will work for their style of shooting, but in reality the first wide shot of a city scape or any tricky fabric that will produce aliasing or color moire will ruin your day.

thanks Phil, thats what I figured. I decided to go with the new OLPF. we ll see what it does for me when I get my baby back in a week or so.

 

[Dr. Sassi]

One obvious thing that post filtration does, is adding time to the post.


Take a minute processing time per frame of a feature -- (~130.000 frames).
If you have only one computer it is ~90 days. Results might vary heavily, of course.

You might have a render-farm to cut the time needed.
I look for in camera filtration based on that, I'm not a render-farmer, even I wish to have the amount of computer needed for this.

 

[David Battistella]

thanks Phil, thats what I figured. I decided to go with the new OLPF. we ll see what it does for me when I get my baby back in a week or so.

Smart move.

 

[Lauri Kettunen]

One obvious thing that post filtration does, is adding time to the post.

I think Brett is simply asking, if one finds the roll-off of brights a problem, why not exposure bit less to better protect the highlights. And then one may adjust the highlights in post as one wishes. Obviously, this does not always work out as one has to often balance with the noise in the darks. But still, there are many cases in which Brett's suggestion does work, and hence, the question is a good one. Especially so, as the standard (new) Dragon OLPF left no choice for the cameraman. But this is also water under the bridge as the changeable low pass filter brings this choice back to the user.

Another more subtle issue is, whether everything is possible in post –especially so as one needs to run the debayer algorithm before any grading. For example, tackling flare in post sounds rather challenging. If remember well at some point somebody from RED said the roll-off of brights is neither that easy set in post if some part of the image is clipped. But since there is now a choice, everybody will be able to pick an approach which fits best to one's needs.