On-going MX OLPF tests

[Tyler Besh]

So I had a thought a few weeks ago. What if you took out the standard OLPF on the MX?

Also, what if you tried putting the newer OLPFs in front of the sensor? (even though they're not made for it and it would take destroying the plastic housing).

I heard of people taking out the standard OLPF - simple. I couldn't find much test videos, information, etc. so I ran a small color and range test after taking it off.

Here are the two videos:
NO OLPF - https://youtu.be/w8x7T8wAwRk
STANDARD - https://youtu.be/3QALZmfCmZE

Some interesting things I noticed:
The standard OLPF seems to cut it down a full stop
It does get rid of a bunch of that magenta color the MX is known for (out-of-box with standard OLPF even on it). This takes that purple-ish tint out of the shadows as well.
Moire seems to be worse WITH the standard OLPF on it (something the RED Tech told me would be the opposite case). Notice lines on Xrite sides.

...which leads me to the call I had with the RED tech. Since RED no longer services the Epic MX, it's worth playing with in my opinion, because you can't even send it in to them for any OLPF work, or any repair work whatsoever. The tech was very interested in my inquiry about removing it, placing newer OLPFs in front, and gave me some interesting insight if I decide to take those next steps (said he has never seen footage or heard of someone doing this).

The firmware designed for the MX does a conversion of data which directly correlates to the specific light/information coming in WITH the standard OLPF installed. Meaning, if I were to put other OLPFs in front of the MX sensor, there is no way to adjust what that data conversion is doing, because there is no software written for the MX sensor and newer OLPFs (unlike the newer cameras). So, the conversion will still be assuming the standard OLPF is in front of the sensor. He mentioned this might result in some very interesting footage.

So, my next step will be doing exactly that - running tests with the low-light filter (which I assume is the same result as just taking the standard OLPF off and might just act as a clear protective OLPF at that point).

...and the skin-tone OLPF (which I'm assuming will yield VERY interesting footage).

That will be somewhere down the line, after I destroy them :)

Will update soon!

[Jim Skinner]

What an amazing comparison. Thanks for doing this! I have a theory that you didn’t lose a stop overall from the OLPF, you lost a stop of magenta light, which unevenly affected the frame. One way to confirm this is to let us know what color the square above the yellow square actually is. There is a night and day difference between the two sample’s rendering of that square. Without OLPF it’s close to pink, and with OLPF it’s a nice saturated purple. Which is the “real” color. The version without OLPF had crazy tinting throughout, that green swath in the hallway is really a mess, as are the shadows. I can’t even imagine how I’d color time that non-OLPF footage. But did removing the OLPF actually help the “pink” square? Maybe there is a use case with a REALLY pink set to pull the OLPF.

[Phil Holland]

One big, big change is where the Anti-Aliasing filtration occurs between DSMC and DSMC2 bodies.

Previously the OLPF actually contained an Optical Low Pass Filter, but in DSMC2 bodies that has been moved to the Cover Glass in front of the sensor. Which is why you have more aliasing on MX as there is now AA filter.

Mysterium-X doesn't have a calibration profile for the Standard OLPF.

The term OLPF is basically grandfathered in on modern bodies due to previous terminology, but more accurate they are Color Filter Glass (CFG might be a new start?) with various levels of Infrared Cutting, Blocking, or Absorbing properties as well as likely some UV related filtration.

The benefit of these filters is protection from IR contamination mainly as well as helping produce better accurate color. This of course comes at a cost of light transmission, but it's a worthy loss. The AA filter now found in the Cover Glass obviously is there to help protect from Anti-Aliasing Artifacts.

Companies like KipperTie have provided "full spectrum" pass through filters which maintain the back focus as well as maintaining all light transmission possible, but you'll get some funky color. Certainly neat if you are getting experimental or even working in Black and White.

[Mike P.]

When you say "standard OLPF" do you mean the "default OLPF that was in MX cameras" or like, the swappable "Standard OLPF" that got released with Dragon/DSMC2?

Without the OLPF, the image should be quite a bit sharper. It's hard to tell with webcompressed 1080p video.

In any case, if you can balance out the magneta, it'd be a nifty little experiment for older MX cameras just sitting around (a sharper & brighter image at the cost of a few jaggies you can't see 90% of the time? Why not!) Conversely, if the magenta is too aggressive or too sporadic to be balanced out, it might make for a cool little MX Monochrome Lite. (It certainly seems in line with the benefits of the Monochrome cameras -- a stop+ brighter and jump in sharpness.)

[Tyler Besh]

Wanted to run another test with the "standard" olpf on and off (standard I meant the default yes! )

I'm here in Detroit area with pretty snow, any suggestions for an in-depth test to cover all the bases before I move on to the other OLPF experiment? (Ie color chart, things for aliasing, red clored things, etc)

[Dan Kanes]

Try an IR/UV notch filter outside the lens to back that Purple/magenta off

https://www.bhphotovideo.com/c/produ...BI%3A514&smp=Y

Or

https://www.bhphotovideo.com/c/produ...ro_filter.html

Could try a hot mirror too since you’re experimenting

https://www.bhphotovideo.com/c/produ...or_Filter.html

You could also chart with the hot mirror in and then chart with it out under the same light source and use a matching tool to try and tweak a conversion lut, however you may end up with weird branding or weird results that can’t be taken out of the image because of the sensor’s spectral response and the way the camera captures Bayer pattern data with raw compression - still worth a try. IPP2 post on MX???

[Tyler Besh]

Wasn't able to get my hands on any of those filters but I did have my ND 1.5 which gave a very interesting result when the Factory OLPF was removed. Uploading the results today!

[Tyler Besh]

UPDATE: Step 1 is done and archived. Played around outside and inside with different colors and outfits to see what the OLPF really does. Came across a very interesting result when I had the ND 1.5 on outside and the OLPF removed.

You can see all the stuff here:
trustydigitalmedia.com/olpf (best viewed on larger screen)

Step 2 will be removing the newer OLPFs from their housing!