Welcome to our community

Be a part of something great, join today!

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 :)

Red and the infraRED

Mike Prevette

Mike Prevette

Well-known member
Joined
Apr 19, 2007
Messages
1,692
Reaction score
0
Points
36
Location
New York, NY
Take a quick look at this frame.

2292888274_a45a538290.jpg

larger: http://farm3.static.flickr.com/2268/2292888274_ccb682857d_o.jpg


First off it's about as ugly of a subject as you could possibly want (me) I'm sorry this test didn't involve pretty girls, next time. After you read the whole test you will see why it really will be in the best interest of the group to shoot a test with a girl ;)

It's a frame "RedAlert default" (ie no color correction) grab from some tests DeadMike and I did earlier this week. Shot in the mid afternoon sun, 1/48th sec, a ND 1.2 and ND .9 stacked on top of a Schneider-Kreuznach-Xenon 75mm t2.2 lens.

At first when it popped up on my screen, my initial reaction was that the purple/magenta skew was due to the ND stacking and older lens. No ND filter has perfect colorimetry, and it's common for stacked NDs to appear purplish. Older lens coatings usually shift color temperature over time.

Then I took a closer look.

And there it was. In reality my hoodie is black to the eye, just as black as the drawstrings (see examples at the end of the post). It happens to be a 100% cotton hoodie, and the drawstrings are all nylon. So it was more than a simple color balance problem.

What we have here is a classic case of IR (infrared) pollution. Natural fibers (in this case cotton) don't reflect a lot of IR. But the dies used to make them black do! The dyes of the drawstrings in this case do not reflect much IR.

Silicone based sensors are natively very sensitive it IR, much in the same way that they aren't very sensitive to the Blue spectrum of light. This is usually compensated for by adding a "hot mirror" IR blocking coating to the OLPF. Physics limits how well these work, and there is always some IR that slips through. The RED has this filter and it works in normal situations.

Due to the significant amount of ND, the ratio between IR and the visual spectrum has been reduced. Therefore allowing the IR to make up more of the "image" that the Red sees. ND's only reduce the amount of light in the visual spectrum, they don't effect the IR range (very much at least).

SIDE-EFFECTS: Not only does this become a problem for color correction, but infrared light focuses at a slightly different point than visual light. So in a IR rich environment without much ND there will be (very faint) but subtle out of focus IR image contaminating the visual light image. This obviously becomes more noticeable the more you reduce the ratio between visual light and IR. You will also see it in some IR rich highlights, this is very often misdiagnosed as Lens chromatic aberration, or sensor based electron bloom. In a pure IR environment of course your lens markings will be incorrect.


Some more tests today:
All of these were shot with a 18mm Super speed, wide open. The only thing that changed was the shutter speed and ND. It was lit with a 1k Mickey a few feet camera right. There were some overhead floro's that caused the weird banding in the background of the high shutter speeed shots. NO REDALERT settings were changed between these images.

1/2000th sec exposure
No ND

1/250th sec exposure
ND: N9 (-3 stops)

1/15th sec exposure
ND: N9 + N1.2 (-7 stops)

EDIT: the txt is wrong on frame 2, it was shot at 1/250th of a sec.
2292888110_9004fccef9_o.jpg


Here are a few quicktimes with me passing a 486 IR blocking "hotmirror" filter over the scene.

No filtration t4 at 1/24th of a sec:
http://www.freedomfromgravity.com/work/red/ir/A001_C011_080125_nofilter_24.mov

T1.3 and ND9 at 1/250th of a sec:
http://www.freedomfromgravity.com/work/red/ir/A001_C006_080125_nd9_250.mov

T1.3 and ND9+ND1.2 at 1/24th of a sec:
http://www.freedomfromgravity.com/work/red/ir/A001_C008_080125_n9_n12_24.mov

And One quicktime using a 87 IR PASS filter over the scene.
Still with full ND (n9+n1.2) at 1/24th of a sec.
http://www.freedomfromgravity.com/work/red/ir/A001_C009_080125_87irpass.mov


Again I don't think this is a major issue for most situations, but think twice about lot's of ND, and if you see anything weird try pulling some ND, and steepening down the shutter angle a bit if you can.

There are IR blocking filters out there. It might be a good option for the 72mm thread on the Red 18-50, but they don't come in large enough sizes (4.5" or 138mm) to be useful on many other lenses. They are also very expensive.

In my experience it can be VERY HARD to correct for IR pollution, as it happens to affect skin tones, as well as acting wildly different on differing materials.


ERRATA: I was one of the very early owners of the Leica M8. It was decided by Leica that in order to preserve quality and sharpness they would leave off the OLPF, the side effect of that is that the IR filtration couldn't be as strong in camera. After this was discovered people had a shit fit, and Leica was forced to provide two free IR filters to every M8 owner. At first everyone tried to make fixes in photoshop, or other software packages. Even creating custom profiles that desaturate the magenta hues, none of this ever worked well. In the end it was proven that filtration was the only solution.



The control shots, done with the exact same hoodie, and a Leica M8 both with and without a 486 filter. Notice the increase in "shadow" detail and color shifts.

Here is and example of the difference the 486 filters make:
(I shot these on the Leica M8, ZERO enhancement)

2292101135_c61dfb57f6.jpg




I hope this info is useful to someone, and that it might solve some on set head scratching about why the lead actors tux looks purple, or why you can see through the actresses skirt. I also It can save someone some money in the color suite trying to fix an issue caused by IR. Who knows it could actually be useful if you were shooting a BW film. It certainly smoothes out the skin tones in BW.


These threads also picked up on this today:

http://www.reduser.net/forum/showthread.php?t=9383

http://www.reduser.net/forum/showthread.php?t=9388

"This next point may be the lens we used, but on sky footage or a solid colour we did notice a magenta hue in the centre of the image and a green cast to towards the outer edge. I'll report again on this when I shoot with another lens."

This is typical of IR filters. They only work when light is passing through them at oblique angles, as the light starts to enter from more extreme angles the filter starts to effect the red in the visual spectrum, leaving you with a more cyan look. This is very common among the M8 crowd with wider lenses. I'm guessing he was shooting with a lens that had it's rear element very close to the internal IR filter.
 
As you've mentioned, using alot of NDs basically turns the stack into an IR pass filter so you end up with way more IR than normal. I'd highly recommend using an IRcut/hot mirrors with stronger NDs. Insist on them if you can.

This is not a RED specific thing but it does affect some cameras more than others and we do already have a conservative/strong IR cut filter.
 
Deanan,

Thanks for the quick response. I tried to point out in the story that this is indeed just a trait of silicone based sensors, and not just the RED.

That being said have you guys spot checked your Hot mirror supplier to make sure it meets your specs? They are hard and expensive to make so I fear a supplier might skimp to save a penny.

Unfortunately hotmirror filters in 138mm or 4x5 or even 6x6 are damn near impossible to find, and will undoubtedly be a custom job. and VERY expensive.

Also take a look at my quicktime shot with NO ND, and I can still see the effect of my 486 filter getting passed over the scene. That tells me that the internal IR filtration isn't quite strong enough.

http://www.freedomfromgravity.com/red/ir/A001_C011_080125_nofilter_24.mov


Deanan said:
As you've mentioned, using alot of NDs basically turns the stack into an IR pass filter so you end up with way more IR than normal. I'd highly recommend using an IRcut/hot mirrors with stronger NDs. Insist on them if you can.

This is not a RED specific thing but it does affect some cameras more than others and we do already have a conservative/strong IR cut filter.
Click to expand...
 
Thanks guys.

I should give some thanks to DeadMike for his RED and to Koerner Camera (in seattle and portland) for letting me do the test today.


I am just afraid that since the IR was instantly viewable even with no ND on the camera, that this is going to mess with the color on so many shots.

Not to mention the focus!
 
Mike Prevette said:
Also take a look at my quicktime shot with NO ND, and I can still see the effect of my 486 filter getting passed over the scene. That tells me that the internal IR filtration isn't quite strong enough.
Click to expand...

By adding an external IR filter in combination with internal you always get more IR cut than you can practically get for a given internal only design. This because there are limits you run up against in the formulation of the IR layers.

We're constantly looking for ways to improve the IR/LP filters but at the moment we have not found a better IR/LP that meet all the design requirements (optical performance, cutoff, manufacturing tolerances, etc).
 
Ahh that makes sense. I know we are up against physics at some point.

What are your thoughts regarding IR filtration on wider lenses?

It's been a bitch in the M8 world because so many of the lenses are wider, and so the external IR filtration starts to effect the visual spectrum red to the point that the corners go Cyan. It was so severe Leica added internal processing to the RAW files to correct for it.
 
I've noticed these effects as well, is there any difference in stacking filters Vs. using a stronger filter? IE a 1.2+.6 Vs. a single 1.8. Seems to me it's the ratio not the stack.

The real solution is for Schneider or Formatt to come out with IR cut ND filters.
 
The M8 is a very tough design challenge because many of the wides on M mount have the rays coming in at very oblique angles and rear elements
very close to the sensor. This forced them to go with very weak IR/LP (as far I've understood their design constraints). The cyan might not necessarily be an IR problem but could be caused instead by the oblique angles of the rays and how they interact with the radially offset microlenses on the M8. Oblique rays also cause problems the outer pixels and why the offset microlens design was chosen to mitigate the artifacts.
 
Evin Grant said:
I've noticed these effects as well, is there any difference in stacking filters Vs. using a stronger filter? IE a 1.2+.6 Vs. a single 1.8. Seems to me it's the ratio not the stack.

The real solution is for Schneider or Formatt to come out with IR cut ND filters.
Click to expand...

I've noticed that the stronger filters (1.2+) seem to be worse than stacking. The pancros are better overall but have more back reflection.
 
The Cyan artifact is a result of the IR filtration. The camera does not exhibit the trait without filtration on the lens.

The common IR filters just can't handle light entering at steep angles, they begin to effect the visible spectrum if light enters at an angle steeper than 45degrees off lens axis.

The microlens/sensor vignetting is a completely unrelated issue.

Here are two examples I just shot.
Leica M8 Zeiss 21mm lens f2.8 1/30th
With Ir and without, no in camera processing.
2292489937_5bca9879fa.jpg

2292490025_e492ab8535.jpg



Deanan said:
The M8 is a very tough design challenge because many of the wides on M mount have the rays coming in at very oblique angles and rear elements
very close to the sensor. This forced them to go with very weak IR/LP (as far I've understood their design constraints). The cyan might not necessarily be an IR problem but could be caused instead by the oblique angles of the rays and how they interact with the radially offset microlenses on the M8. Oblique rays also cause problems the outer pixels and why the offset microlens design was chosen to mitigate the artifacts.
Click to expand...
 
Just theorizing here because I haven't haven't used an m8 yet (got some M2's and M3's but that doesn't count).
 
Post of the month

Post of the month

Mike,

Great post, and defenitely a concern. Now an old post from Macgregeor comes back to mind in which he wrote that all of his shots with ND's were softer than the ones without, and where nobody had a good answer . Going back to my post about magenta instead of Red (where I did not use any ND's BTW), all the shots involved had either red filtered practicals or parts of the shot were lit with red gelled lights. I had another shot where the red leds on a TC slate turned out very magenta.
 
Well I just took a look at my little Lee and Rosco swatch books, and almost all the red gels are completely transparent to IR. So they really screw with the visual light/IR ratio the same way ND on the camera does.

And most red LEDs send out a ton of IR.

This also explains why taillights seem to blow out much quicker on the RED. Fore example in the Red VS F23 tests a few months ago it was noted.
 
Mike Prevette said:
Well I just took a look at my little Lee and Rosco swatch books, and almost all the red gels are completely transparent to IR. So they really screw with the visual light/IR ratio the same way ND on the camera does.
Click to expand...


That's expected since red gels pass through red light and block green and blue light. And since red blends over into infrared..

And most red LEDs send out a ton of IR.
This also explains why taillights seem to blow out much quicker on the RED. Fore example in the Red VS F23 tests a few months ago it was noted.
Click to expand...

color LEDs (not IR LEDs) are very peaky and tend to give off alot less IR than incandescent sources which are mostly long longwave. Red Leds are usually in the 640-680nm range compared to IR leds which are from 700-1300nm.

The red vs f23 tests looked like the exposures weren't matched so it was hard to judge why the f23 did better.
 
You must log in or register to reply here.
Back
Top