Welcome to our community

Be a part of something great, join today!

Register Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Dragon with no OLPF

Will Keir said:
Gavin. I thought your views on 4k were enlightening and appreciate your challenge of Michael Cioni's post for the Expendables 4k workflow.

Now how do you come up with the number 1/4 of a stop of light from the New OLPF to none? A guess?

The New OLPF is less sensitive in lowlight than the old. Has anyone actually measured the difference? It's dramatic enough for RED to consider the old OLPF to be the "lowlight" option in their modular OLPF design. I'm just looking for some numbers here man, it seems obvious the image will suffer. Diffusion filters lower measured light levels correct? So do sunglasses.
Click to expand...

You should re-read Gawins post...
 
Who is Gawins? Which line specifically of Gavins post do you feel I need to reread? In your estimation, what am I not getting?

Gunleik Groven said:
You should re-read Gawins post...
Click to expand...

This one?

Gavin Greenwalt said:
I'm going to be the voice of dissent here and say that while I agree uprezzing in software is superior to uprezzing in the projector, let's not lose sight of the fact that we're still talking about up-rezzed content.

The two options you're proposing Michael are essentially:

1) Shoot 4k -> Scale 200% in Software -> Project 4k.
2) Shoot 4k -> Crop to 2k (no scale) -> Scale in Projector 200% -> Project 4k.

Yes the software option can be a slightly better choice since you can employ more sophisticated algorithms but that's not a universal true statement. There are some superb hardware solutions these days at uprezzing content to 4k. And there are some superb software solutions to uprezzing content. I wouldn't say "Software uprezzing looks better" what I would say is that "If you deliver a 2k master in 2k then you have no idea what algorithm will be employed to uprez it for 4k, if you uprez your movie yourself it will have full control over what algorithm is employed."

However, if you just use something like a vanilla bicubic then your projector and your software will probably look identical pixel for pixel since they're literally doing the exact same thing to your footage.

Ultimately I agree with the sentiment. I've done 720p YouTube spots in 5k and ran out of resolution. But since it was a 720p delivery I didn't scale the entire project up to 4k. One nearly universal truth is that if you can't deliver a 4k display it's better not to do what you just described and simply crop? Why? Number of processes.

Let's say you are delivering 1080p for TV. In this case you're going to do one of the following:

1) Capture 5k -> Crop 50% -> Delivery 1080p Master
2) Capture 5k -> Scale 266% -> Scale 37% -> Delivery 1080p Master


Option 1 is going to look WAY BETTER. One of the really important things when manipulating images in VFX is that you concatenate your operations wherever possible, so if you transform, skew and scale an image you have to be very very careful to ensure that you don't perform multiple transform operations out of sync. The problem with scaling up to 266% and then scaling back down for a 1080p delivery is that both are destructive operations. Stacking multiple destructive operations on top of each other and lose quality at each step. MOAR PIXELS isn't always true. I'm glad it worked well for a film where you were confident most of your projectors could display a 4k master and of course if you're going 4k because 50% of your shots are 4k then you'll be better off, but this advice does not carry over to TV where 0% of your displays will be in 4k and 0% of your shots will end up in 4k.



In this example I started with a 0.25k image, scaled it up to 200%, then scaled it back down by 50%. As you can see you've lost substantial sharpness by running it through 2 scaling algorithms as opposed to no scaling at all.

Similarly if you are able and have time your 2k masters should not be from a 4k timeline for this reason other wise you are doing:

1) 4k Master -> 200% Scale to 4k zoom -> 50% to 2k DCP -> 200% scale to ?4k projector?
vs
2) 4k Master -> Crop to 2k DCP -> 200% Scale to 4k projection.

You've eliminated one scale operation and your image will look way sharper than a 2k master from a 4k master.
Click to expand...

Or this one?

Gavin Greenwalt said:
No. OLPFs are like very very weak diffusion filters and an IR hot mirror plus a slight color filter. The actual loss of light is probably pretty small. Most of the light loss is in the OLPF color filters. At best you'll probably get 1/4 of a stop without an OLPF. The biggest difference is shaprness, you trade off resolution for aliasing and possible moire. You would also have substantially more IR sensitivity and would have to use a hot mirror in front of the lens (and lose at least 1/4 of a stop anyway) and the debayer color science would be off since it would be purely neutral.

The new OLPF added extra red filtration to cut out near IR flaring. Without it the red channel would be cleaner since it would be less underexposed but it will be too bright unless RED changes your color science to not apply the corrective digital gain.
Click to expand...
 
hahahaha
The post of Gavin...

He says:
IF you remove the OLPF, you'll need to IR cut and the IR cut will lose you 1/4 stop.
Has nothing to do with the rest of the chain... :)
 
My question is how is this measurement quantified? Why 1/4th, how did he come up with this number? Or is it a theory... a guess based on his experience and intelligence?

Gunleik Groven said:
hahahaha
The post of Gavin...

He says:
IF you remove the OLPF, you'll need to IR cut and the IR cut will lose you 1/4 stop.
Has nothing to do with the rest of the chain... :)
Click to expand...
 
Who makes IR Cut OLPF filters?



Gunleik Groven said:
If you set in a hot mirror/IR cut, the filtermaker will usually inform you of the lightloss. 1/4 to me seems conservative... :)
Click to expand...
 
Will, remember the olpf is only part of the chain. The accompanying calibration is the second part of the chain. The old calibration was more favorable to the low end whereas the new olpf/calibration favors a shift towards the high lights. That's where we get the 1.3 less latitude in the low end in favor of swapping it up top.
 
What Matt said.

Dunno if I am particularly unclear, but I'll try again...

In an OLPF like Dragons it's baked in.

If you REMOVE the OLPF, you need to fit an IR cut in front of the lens.

THAT will according to Gavin take 1/ of a stop light.
I commented that I think that might be a conservative number, depending on which IR-cut you apply.
 
Looks like you got an edge on the custom OLPF market. Wonder if RED would let it go 3rd party which would probably produce some really cool custom results. Very similar to filters, like the popular promist 1/8th, I'd bet. But this all voids the warranty according to my Bomb Squad Rep Clark McClanathan.

"Hello Will,

At this point, removing the tamper stickers and changing out the OLPF will void your warranty on the sensor and the OLPF itself. It could possibly void warranty on the camera completely as well, if anything went wrong, which would be determined by a RED technician during an evaluation.

If you would like to have any OLPF changes made, I would highly suggest sending in your camera.

There has been no date or timeline set for the INTERCHANGEABLE OLPF. Once any information is made available I can let you know. I would keep checking REDuser.net though, as a lot of times Jarred/other RED employees will post information there before we get in for the next work day and are notified of changes or updates.

Regards,"

So take out the OLPF and put a 1/4th IR filter on your mattbox, you gain some low light sensitivity?

What is the low light sensitivity of the OLD OLPF vs the New?

Matt. 1.3 Stops of light was the measurement by Phill Holland on Epic vs Dragon.
John Marchant said:
Low volume

http://goochandhousego.com/
http://www.uqgoptics.com/
https://www.knightoptical.com/
http://www.kingsviewoptical.com/

High volume

http://www.moe.com.tw/
http://www.tradekorea.com/product/detail/P399869/Sell-OLPF-filter.html
http://www.isuzuglass.com/


And many many more with varying costs, capabilities and standards.
Click to expand...
 
Will Keir said:
My question is how is this measurement quantified? Why 1/4th, how did he come up with this number? Or is it a theory... a guess based on his experience and intelligence?
Click to expand...

Educated guess on most of the OLPFs I have seen. Apparently I was off though by a good margin for Dragon which has a super intense color filter now. Still my point is when you say "sunglasses" I think Neutral Density. OLPFs aren't just bringing down the exposure they're optically calibrating the image chain to balance out negative

As to the Hotmirror being 1/4 there are IR coatings out there which will mostly eliminate IR with only a 15% transmission loss in the visible range but they don't handle the full IR range so I figured doubling it would be safe. ;)
 
John Marchant said:
Some example OLPF / sensor filters from my stash. You can see more easily like this how the filters are more complex than just an ND. You can also see how the later Dragon OLPF cuts a lot of red light.

glass.jpg
Click to expand...


Great post! Thank you!
 
Too much talk and no r3d's. Talk is talk r3d's are r3d's! I just asked for r3d's. Let them talk. Then everybody can make his/her mind.
 
Lakis Amarantithis said:
Too much talk and no r3d's. Talk is talk r3d's are r3d's! I just asked for r3d's. Let them talk. Then everybody can make his/her mind.
Click to expand...

Agreed. This site has become a lot of talk and very little R3D's to back it up. Usually processed jpegs with NR and sharpening are the norm. Let's see the R3D's...
 
Gunleik Groven said:
As one who have shared a few...

There are issues with sharing RAW files.... :)
Unless they are specifically made for sharing.
Click to expand...

OLPF tests hardly seem like an NDA issue ;)
 
Matt Ryan said:
OLPF tests hardly seem like an NDA issue ;)
Click to expand...

That depends on the context of the test.

Tests like the one we did in Bristol are done all the time (with relevant cameras), but mostly they are not publishable as they belong to productions which pay for them.

I don't really see many popping out their OLPFs just out of curiosity. I won't... :)
 
Coming from an underwater shooting background it looks like the Dragon V2 filter blocks out a lot of the red light wavelength. This is exactly what we do NOT want for underwater shooting. I do realize that underwater shooting is a very small market.
 
Mark van Coller said:
Coming from an underwater shooting background it looks like the Dragon V2 filter blocks out a lot of the red light wavelength. This is exactly what we do NOT want for underwater shooting. I do realize that underwater shooting is a very small market.
Click to expand...

That would explain the red noise issue as well...
 
You must log in or register to reply here.
Back
Top