Here are illustrated instructions for the IR cut filter. I think the AA filter is similar. I wouldn't attempt this myself though.
http://www.lifepixel.com/ir-tutorials/canon-5d-digital-infrared-conversion-instructions.htm
The GetDPI Photography Forum
Great to see you here. Join our insightful photographic forum today and start tapping into a huge wealth of photographic knowledge. Completing our simple registration process will allow you to gain access to exclusive content, add your own topics and posts, share your work and connect with other members through your own private inbox! And don’t forget to say hi!
anyone with a leica and eos mountable leica lens for a shoot tomoorow or Monday a.m.
- Thread starter docmaas
- Start date
http://www.lifepixel.com/ir-tutorials/canon-5d-digital-infrared-conversion-instructions.htm
Jack and I met and spent about an hour on some different targets. What I have seen so far is that the impact of the aa filter is obvious both in the original size pics where texture is more obvious with the modified and in huge pixellated blowups of irregularly shaped objects where the aa filter spreads the colors of the object over a larger amount of space effectively blurring it. The number of colors is also impacted by the aa filter as it appears to cause a secondary blendng of colors through the blurring process. A color count using irfanview between two matching shots shows 862981 unique colors in the non AA version vs 801291 in the AA filtered image. There isn't much difference though in very straight edges like letters nor as far as I could tell in repeating patterns like chicken wire. I used Adobe Raw Converter on a Mac other raw converters may reveal more and different differences.
I haven't seen any aliasing at all.
I'll post some examples of my findings later today.
Mike
My comments after just a cursory view echo Mike's as respect color. His camera seemed to produce more saturation and slightly better inter-pixel contrast --- I was not surprise at the inter-pixel contrast, but was surprised by the more intense color (positively so).
We used the 200L at f5.6 on both cameras, then the 28/1.8 prime on my camera at f8 and the 24-105 zoom at 28mm f8 on his camera. You can see more fine detail with the hot rod, but frankly I only noticed this with regularity with the 200L in fine structure --- as Mike said, no difference rendering chicken-wire at 50 meters distance with either camera, but there was on the stucco of the building at that distance. IOW, more detail is there, but it has to be pretty fine and you have to be at 100% to see it. To quantify further, I would say the 5D hot rod resolves about halfway between the 5D and 1Ds2.
I noted some aliasing in the stucco on the building we shot with the 200 on his camera, which was also where I saw the most prominent detail gain. IMO the aliasing was so slight as to be a non-issue, while the detail gain was more notable. (Whether it would show significantly in a print is debatable, and I will print the crops up tomorrow and comment afterward.) I did notice some gains in detail on higher-contrast features with the lesser lenses, such as corrugated metal siding on a building, but this was slight and would probably never render in a print.
IMO, the big gain is in tonality and color fidelity --- the hot rod looks better out of the box --- and a little bit of "crispness" that Canon files generally lack. I was hoping for enough more crispness to make the 5D file look more like a DMR, but sadly, that wasn't the case --- the hot rod gets maybe 25% of the way there from the regular 5D.
I'll try and work up some crops to post later tomorrow. I again want to thank Mike for allowing me the opportunity to meet with him and compare our cameras!
We used the 200L at f5.6 on both cameras, then the 28/1.8 prime on my camera at f8 and the 24-105 zoom at 28mm f8 on his camera. You can see more fine detail with the hot rod, but frankly I only noticed this with regularity with the 200L in fine structure --- as Mike said, no difference rendering chicken-wire at 50 meters distance with either camera, but there was on the stucco of the building at that distance. IOW, more detail is there, but it has to be pretty fine and you have to be at 100% to see it. To quantify further, I would say the 5D hot rod resolves about halfway between the 5D and 1Ds2.
I noted some aliasing in the stucco on the building we shot with the 200 on his camera, which was also where I saw the most prominent detail gain. IMO the aliasing was so slight as to be a non-issue, while the detail gain was more notable. (Whether it would show significantly in a print is debatable, and I will print the crops up tomorrow and comment afterward.) I did notice some gains in detail on higher-contrast features with the lesser lenses, such as corrugated metal siding on a building, but this was slight and would probably never render in a print.
IMO, the big gain is in tonality and color fidelity --- the hot rod looks better out of the box --- and a little bit of "crispness" that Canon files generally lack. I was hoping for enough more crispness to make the 5D file look more like a DMR, but sadly, that wasn't the case --- the hot rod gets maybe 25% of the way there from the regular 5D.
I'll try and work up some crops to post later tomorrow. I again want to thank Mike for allowing me the opportunity to meet with him and compare our cameras!
W
workingcamera
Guest
Very interesting reading… look forward to seeing some samples
I posted full size pics from both cameras along with a couple of crops illustrating some of what I found.
http://www.pbase.com/masimo/5d_hotrod
I replaced the rock images with correctly sized images.
Mike
http://www.pbase.com/masimo/5d_hotrod
I replaced the rock images with correctly sized images.
Mike
Last edited:
Forgot to mention that the pics were converted with ACR (latest version) and that all sharpening and noise reduction are set to 0.
Mike
Mike
Part 2 --- the images.
Here are my images with little commentary. First, all were processed in ACR at the same settings except where indicated, the lens used was the 200 L.
For reference, here are the full shots. I used "as shot" WB on both. Interestingly the 5D Reg was 5500 +7 while the 5D Hot Rod was 5450 -5 --- This is very close on base color temp at only 50 degrees apart, but significantly different on tint at 12 points different. (Note: Don't pay too much attention to the sky as there were fewer background clouds when we shot the Hot Rod.) The exposure was 1/2000 @ f5.6, ISO 200 for both cameras:
5D Reg:
5D HR:
Here are the first two crops. The first is the 5D reg and second is the 5D HR which received identical post processing of just light sharpening to extract detail. Note the difference in interpixel contrast and color saturation. Then I added some post to the 5D Reg crop to add some interpixel contrast which also added a bit of saturation to make it look more like the Hot Rod. The two files now look pretty close, except you can still see more detail in the Hot Rod crop. I have circled the light artifacting I mentioned earlier in the Hot Rod crop (which I feel is of zero consequence).
Reg:
Hot Rod:
Reg, additional post:
Lastly, here are a pair of crops from the top of the image, again these were identically processed with WB "as shot". I include this to show the difference in red and blue hues. IMO removing the AA filter seems to generate better color, at least to my eye. The added clarity of the Hot Rod can be seen in the stucco, in the roof tiles and along the high-contrast edges of the peeling paint:
5D Reg:
5D Hot Rod:
~~~
Now for the true test, a print --- but keep in mind this is purely subjective and I have no way to "show" these subtleties online...
I printed the crop ares out on glossy paper for direct side-by-side comparisons at both 360 PPI and 240 PPI. My assumption was the Hot Rod might show better at 240 than the Regular due to no AA filter and corresponding cleaner pixel edges. The bottom line is that even when viewing the prints under a loupe, the detail differences seen at 100% above are essentially undetectable in a print at EITHER resolution. I say essentially because in the high contrast paint chips, I see what may be the impression of crisper edges, but it is not definitive. IOW they all look pretty identical from a resolution standpoint. Also, as expected, the slight artifacting in the Hot Rod file is undetectable.
However, color is a different story --- and a minor surprise as well. In the prints, the most accurate looking file as processed (and to my eye) was the regular 5D with my added post processing, followed by the regular 5D's. For whatever reason, there was a slight --- and I mean slight --- muddiness in the colors of the Hot Rod file, even though they look better in the online sRGB crops (again, at least to my eye). Please note the difference I'm describing here are *very* slight and I only mention them because I noted them when viewing the print. Moreover, I am confident a more selective raw white balance accompanied by a simple print output curve tailored to the hot rod would alleviate this minor difference.
~~~
Conclusion for me: If the 5D were my only camera AND if I was going to be posting/presenting jpegs, I would do the conversion. However, for my uses and the way I process files for print output, I don't see enough added benefit to warrant the conversion.
As always, this is my opinion only and YMMV
Here are my images with little commentary. First, all were processed in ACR at the same settings except where indicated, the lens used was the 200 L.
For reference, here are the full shots. I used "as shot" WB on both. Interestingly the 5D Reg was 5500 +7 while the 5D Hot Rod was 5450 -5 --- This is very close on base color temp at only 50 degrees apart, but significantly different on tint at 12 points different. (Note: Don't pay too much attention to the sky as there were fewer background clouds when we shot the Hot Rod.) The exposure was 1/2000 @ f5.6, ISO 200 for both cameras:
5D Reg:
5D HR:
Here are the first two crops. The first is the 5D reg and second is the 5D HR which received identical post processing of just light sharpening to extract detail. Note the difference in interpixel contrast and color saturation. Then I added some post to the 5D Reg crop to add some interpixel contrast which also added a bit of saturation to make it look more like the Hot Rod. The two files now look pretty close, except you can still see more detail in the Hot Rod crop. I have circled the light artifacting I mentioned earlier in the Hot Rod crop (which I feel is of zero consequence).
Reg:
Hot Rod:
Reg, additional post:
Lastly, here are a pair of crops from the top of the image, again these were identically processed with WB "as shot". I include this to show the difference in red and blue hues. IMO removing the AA filter seems to generate better color, at least to my eye. The added clarity of the Hot Rod can be seen in the stucco, in the roof tiles and along the high-contrast edges of the peeling paint:
5D Reg:
5D Hot Rod:
~~~
Now for the true test, a print --- but keep in mind this is purely subjective and I have no way to "show" these subtleties online...
I printed the crop ares out on glossy paper for direct side-by-side comparisons at both 360 PPI and 240 PPI. My assumption was the Hot Rod might show better at 240 than the Regular due to no AA filter and corresponding cleaner pixel edges. The bottom line is that even when viewing the prints under a loupe, the detail differences seen at 100% above are essentially undetectable in a print at EITHER resolution. I say essentially because in the high contrast paint chips, I see what may be the impression of crisper edges, but it is not definitive. IOW they all look pretty identical from a resolution standpoint. Also, as expected, the slight artifacting in the Hot Rod file is undetectable.
However, color is a different story --- and a minor surprise as well. In the prints, the most accurate looking file as processed (and to my eye) was the regular 5D with my added post processing, followed by the regular 5D's. For whatever reason, there was a slight --- and I mean slight --- muddiness in the colors of the Hot Rod file, even though they look better in the online sRGB crops (again, at least to my eye). Please note the difference I'm describing here are *very* slight and I only mention them because I noted them when viewing the print. Moreover, I am confident a more selective raw white balance accompanied by a simple print output curve tailored to the hot rod would alleviate this minor difference.
~~~
Conclusion for me: If the 5D were my only camera AND if I was going to be posting/presenting jpegs, I would do the conversion. However, for my uses and the way I process files for print output, I don't see enough added benefit to warrant the conversion.
As always, this is my opinion only and YMMV
robsteve
Subscriber
It may be that the DMR's qualities are related more to the CCD sensor and the camera's processing than the lack of an AA filter.
Robert
Guy Mancuso
Administrator, Instructor
It seems the AA filter also degrades color and reason many of us think Canon files are flat looking.