Canon 5d mark iii improved- olpf removed for higher resolution

[Cosimo Bullo]

My 5D3 is currently at MaxMax getting the upgrade, they have had it since early last week. We'll see what happens. I have not heard from them concerning when I might get the camera back, hopefully this week.
I also have the C300 so I will be able to see how these 2 might compare.
Bill

Bravo, and thank you for doing this. I can't wait to see the results. If they continue to be as expected, I wouldn't at all be surprised then to see Canon offer an OLPF free version down the line as Nikon is doing.

No, the post sharpening is bunk if avoidable as it also sharpens the out of focus stuff, a very undesirable side effect.

Thank you again and keep us posted.

 

[dustylense]

You are brave to spend all that money voiding your warranty when the evidence for the effectiveness of removing the OLPF for improving video quality is about as strong as that for improving the sound quality of CDs with a green marker pen.

I'm gonna have to agree with this. We've seen 2 videos, both without much movement with camera. Andrew at Eoshd has removed the OPLF and has shown nothing. But he did do a really nice piece in London with the OPLF in. Post sharpening is not hard to do. But once you make that conversion removing the OPLF, if your shutter blows out, circuit board fries, mirror cracks, LCD panel burns, top panel burns, a control dial becomes unresponsive, or a number of other simple things covered under warranty goes wrong, you've lost that warranty. Now if Canon offered this service, different story...

 

[nigelbb]

nigel
Why do you care?
But since you asked, I'm curious and I like what I see from Mr.Miller's efforts in this area. Contrary to what most people are saying about sharpening in post, I am not that happy with the overall look once I've sharpened footage. Just looking for a different path.
However, people's efforts to wring something out of their gear should not elicit anger from folks who don't find it to their tastes. I guess, I think it's interesting to see what might might be inside this camera. Which might be why any of us continue to use a still camera to produce video, not the best tool in the bag, but it's interesting that it can do what it does.

I'm not angry just flabbergasted that you should spend $500 or whatever on hacking your 5D3 on the flimsiest of evidence that it will actually improve anything. Even Andrew of EOSHD has gone quiet since he voided the warranty on his camera. Actually he hasn't gone totally quiet as he is still moaning about all the new Canon products but he has gone awfully quiet about the virtues of removing the OLPF on the 5D3. If there were good news I would expect him to be shouting it out to contradict all of us non-believers.

I look forward to seeing some side by side comparisons of your hacked 5D3 with an unmolested model.

 

[dustylense]

I'm not angry just flabbergasted that you should spend $500 or whatever on hacking your 5D3 on the flimsiest of evidence that it will actually improve anything. Even Andrew of EOSHD has gone quiet since he voided the warranty on his camera. Actually he hasn't gone totally quiet as he is still moaning about all the new Canon products but he has gone awfully quiet about the virtues of removing the OLPF on the 5D3. If there were good news I would expect him to be shouting it out to contradict all of us non-believers.

I look forward to seeing some side by side comparisons of your hacked 5D3 with an unmolested model.

What we have are a known 2 mk3 users that have done this. One has what, 4 minutes of static shots to show and the other has not shown anything. But it seems that both have disappeared from the scene entirely on the subject. Until we see hard evidence about the removal of the OPLF, I'll stick with my warranty and taking the 5 seconds it takes to apply sharpening in post.

 

[Kholi]

nigel
Why do you care?
But since you asked, I'm curious and I like what I see from Mr.Miller's efforts in this area. Contrary to what most people are saying about sharpening in post, I am not that happy with the overall look once I've sharpened footage. Just looking for a different path.
However, people's efforts to wring something out of their gear should not elicit anger from folks who don't find it to their tastes. I guess, I think it's interesting to see what might might be inside this camera. Which might be why any of us continue to use a still camera to produce video, not the best tool in the bag, but it's interesting that it can do what it does.

Good response, especially when you didn't have to respond at all to someone questioning or commenting on the way you spend your money. Seems "entitlement" is rampant in various ways.

I'm sure you'll post examples when you get it back, hope it works out. Still waiting on Andrew to post his three way comparison. Like Pappas, I'm surprised he hasn't capitalized on it sooner.

 

[Cosimo Bullo]

3hats, you'll have the prestige of Galileo by the time this is all done, so many naysayers attempting to block your path to clarity. Good luck, and I have a strong feeling you're making the right move. The samples from James were anything but subtle.

EDIT- But DO post samples asap!

 

[jcs]

3hats- testing against the C300 is a cool idea, however if you can get access to another 5D3 body and test side-by-side with the same lens, then show a split screen comparison in post (with & without post sharpening) that would be very helpful.

Thus, it would help us understand, how close can we get to no-OLPF with post-sharpening, and are any post-sharpening artifacts hurting ultimate image quality. While it's true that post sharpening affects blurred regions (subtle areas through bokeh), when adding noise grain I don't see a huge problem so far.

As long as aliasing and moire aren't significantly increased (e.g. similar behavior to the C300), the maxmax mod might make sense for some people (removal of both OLPFs + functional dust system). It's clear that for shooting people/faces/clothing up close, the 5D3 with OLPF works great. When shooting outdoors, with objects in the distance (including people at medium range), extra sharpness+detail would be helpful.

 

[Lammy]

No, I have an iPhone 4S and the video quality is nice for a phone however the quality is not even as good as the Mark II, especially when there is significant motion. A line chart test for the iPhone 4S might be entertaining.

I'm not sure if your IPhone 4S is faulty or if my eyes are not as trained but from my experience, not so. Here are some examples:

https://vimeo.com/37871187
https://vimeo.com/32418623
https://vimeo.com/37537384


The key here is proper 1080 dumping and using a decent manual control app like Filmic Pro to also gain 48Mbps. Of course, functionality, low light, and highlight roll off much better on the 5D. And that we'd look like a tit using an iPhone in a professional environment, but still, very funny when put in perspective about how resolution lol.

I look forward to seeing non OLPF footage with some post sharpening added in to make the 5D3 as best looking as it can. The deal breaker for me on the 5D3 was I couldn't make it look adequately sharp enough without breaking apart the codec. Even with adding noise and grain to disguise that.

 

[jcs]

I'm not sure if your IPhone 4S is faulty or if my eyes are not as trained but from my experience, not so. Here are some examples:

https://vimeo.com/37871187
https://vimeo.com/32418623
https://vimeo.com/37537384


The key here is proper 1080 dumping and using a decent manual control app like Filmic Pro to also gain 48Mbps. Of course, functionality, low light, and highlight roll off much better on the 5D. And that we'd look like a tit using an iPhone in a professional environment, but still, very funny when put in perspective about how resolution lol.

Thanks for the tip on Filmic Pro, however 48Mbps won't fix excessive rolling shutter. Might be useful for low-visibility shooting / C-camera work (static/low-camera-motion shots), but even if it has slightly more resolution, has limited use compared to a DSLR (even the 5D2).

Regarding 'breaking apart the codec': I found that to be true with ALL-I, however IPB can be sharpened enough to look decent. Now that I am working on my film and not doing tests, I have backed off on sharpness for many shots as it affects the feeling (some shots can look OK with sharpness turned up, however it does bring out noise grain that at times I prefer not to have).

 

[Rick Burnett]

Thanks for the tip on Filmic Pro, however 48Mbps won't fix excessive rolling shutter. Might be useful for low-visibility shooting / C-camera work (static/low-camera-motion shots), but even if it has slightly more resolution, has limited use compared to a DSLR (even the 5D2).

Regarding 'breaking apart the codec': I found that to be true with ALL-I, however IPB can be sharpened enough to look decent. Now that I am working on my film and not doing tests, I have backed off on sharpness for many shots as it affects the feeling (some shots can look OK with sharpness turned up, however it does bring out noise grain that at times I prefer not to have).

Have you done any testing of sharpening and the appearance of juddering? This is one of the areas that I think using too much sharpening looses some of the motion characteristics at 24p. I observed this with people shooting the AF100 with default sharpening and seeing moderately high contrast edges looking steppy on a pan shot (like blown out curtains on a doorframe).

I never saw this on my AF100, and then I realized, I had my sharpness set to -4 and never post sharpened.

Others have noticed this as well on other cameras, and it always seems to be sharpening that causes the effect more times than not. Which, makes sense, as a higher contrast edge might have just a little blur on it, but then, you go sharpen that edge and then it looses the exact blur that makes it looks smoother in 24p.

This would be my concern with adding sharpening. It's one of the things that has me just reducing sharpening altogether in cameras because I think the motion renders better myself. I'd rather have a little softer and smoother myself (of course, not always, as I do shoot high shutter speed content as well for fight scenes) but I like having the option of getting better motion capture.

Something to think about.

I'll repeat, we really need to see some charts just to know what is happening to the camera with and without the OLPF.

 

[vision_filmz]

I'm not sure if your IPhone 4S is faulty or if my eyes are not as trained but from my experience, not so. Here are some examples:

https://vimeo.com/37871187
https://vimeo.com/32418623
https://vimeo.com/37537384


The key here is proper 1080 dumping and using a decent manual control app like Filmic Pro to also gain 48Mbps. Of course, functionality, low light, and highlight roll off much better on the 5D. And that we'd look like a tit using an iPhone in a professional environment, but still, very funny when put in perspective about how resolution lol.

I look forward to seeing non OLPF footage with some post sharpening added in to make the 5D3 as best looking as it can. The deal breaker for me on the 5D3 was I couldn't make it look adequately sharp enough without breaking apart the codec. Even with adding noise and grain to disguise that.

Footage looks great, don't know how well movement will hold up but I'm impressed. Does anyone know if this "Filmic Pro" can be used with Canon dslr footage?

 

[jcs]

Have you done any testing of sharpening and the appearance of juddering? This is one of the areas that I think using too much sharpening looses some of the motion characteristics at 24p. I observed this with people shooting the AF100 with default sharpening and seeing moderately high contrast edges looking steppy on a pan shot (like blown out curtains on a doorframe).

I would need to see a before/after/split-screen example depicting a suspected issue with judder. The only effect I am certain can occur temporally from over-sharpening are aliasing artifacts at a relatively small scale. Judder is typically an entire frame out of time, or with a static camera shot, a complete moving object. For objects which are moving, the blur is relative to motion, and sharpening will be the same on every frame. Blurred regions are effectively very low frequency information, and sharpening will only bring up noise in those regions, perhaps more visible on the boundary regions, but won't remove the blurred region. Even if the blurred region were to become "solid" it would cause the object motion to overlap, preventing strobing. Typically when I see judder on a PC, it's due to an issue with the playback software. To see max quality, using something like VLC or playing directly to my Samsung HDTV (which does not drop frames) using Win7's "Play To" feature is very helpful. Windows Media Player and QuickTime tend to drop frames with certain H.264 encodings, and in general don't perform as well as VLC or an HDTV.

As noted, I have been backing off sharpening for an actual production as it look more filmic & organic. It's nice having control in post to sharpen on a per-scene basis. For outdoor, high detail shots, more information would be helpful (full 1000+ lines, etc.). A natively sharp camera would be welcome as Gaussian blur could always be added in post to soften the image for people/mood.

 

[Kholi]

As noted, I have been backing off sharpening for an actual production as it look more filmic & organic. It's nice having control in post to sharpen on a per-scene basis. For outdoor, high detail shots, more information would be helpful (full 1000+ lines, etc.). A natively sharp camera would be welcome as Gaussian blur could always be added in post to soften the image for people/mood.

This is pretty much wht the other side of the fence was trying to get at. You can always soften an image and still have it look pleasant, but the opposite isn't really true (at least to me.). The OLPF removal could really prove to be beneficial, I'm just SUPER curious as to why neither James nor Andrew have posted ANYMORE footage. Andrew hasn't even posted a frame.

 

[Rick Burnett]

I would need to see a before/after/split-screen example depicting a suspected issue with judder. The only effect I am certain can occur temporally from over-sharpening are aliasing artifacts at a relatively small scale. Judder is typically an entire frame out of time, or with a static camera shot, a complete moving object. For objects which are moving, the blur is relative to motion, and sharpening will be the same on every frame. Blurred regions are effectively very low frequency information, and sharpening will only bring up noise in those regions, perhaps more visible on the boundary regions, but won't remove the blurred region. Even if the blurred region were to become "solid" it would cause the object motion to overlap, preventing strobing. Typically when I see judder on a PC, it's due to an issue with the playback software. To see max quality, using something like VLC or playing directly to my Samsung HDTV (which does not drop frames) using Win7's "Play To" feature is very helpful. Windows Media Player and QuickTime tend to drop frames with certain H.264 encodings, and in general don't perform as well as VLC or an HDTV.

As noted, I have been backing off sharpening for an actual production as it look more filmic & organic. It's nice having control in post to sharpen on a per-scene basis. For outdoor, high detail shots, more information would be helpful (full 1000+ lines, etc.). A natively sharp camera would be welcome as Gaussian blur could always be added in post to soften the image for people/mood.

I tried to get the people having the problems to post more examples at the time with different sharpening, but they didn't. I also didn't have the time to try and get my camera to do it (since I sold my AF100) so I just let it go.

One question I would have is how would the regions overlap? If you are shooting at basically 180 degrees, then sharpening a frame would create a sharper line at a high contrast motion blur section which would give you a moving line you would not see on a more blurred edge. This is what I observed in the shots that were posted. Areas with lower spatial frequency didn't exhibit the same effect. Give 1/2 of the time the camera is not recording, the spacing relative to that edge is totally dependent on the speed of the edge moving across the frame. It could be argued that it was a combination of the sharpening AND the codec in camera that is really causing the effect, and not just the sharpening alone, however. With a 4:2:0 codec, I notice a lot of contamination on edges that have a decently high spatial frequency to each other.

I use the word judder just to illustrate what is seen, but that might not be the correct term, maybe strobing is a better word, since juddering is not a duty cycle equivalent effect and strobing is, and what I saw was consistent in terms of the pan speed. Some of these shots were done on other cameras (wish I could find the thread) and sure enough, with the same spatial frequency, these other cameras were MUCH more smooth. But, you make a good point. Was the different footage being interpreted correctly in the different playback software. This seems to be a HUGE problem over the last two years.

I would rather have a slightly softer image myself, as I do think it looks more organic. I am still trying to define to myself where and why that is. When people say things that are too sharp look like video, I understand what they are saying, and I don't understand what it is specifically that is doing that. When I watch 1080p movies, they don't look like video. But, what I do notice about 1080p movies is that the scene colors make a HUGE different to how the footage is perceived. If I run footage through Magic Bullet, people then think it's filmic looking. So it shows there is SO MUCH MORE than just resolution that matters.

Given I don't do coloring, I'm just a shooter, this stuff is more just me exploring what people perceive in footage. Also helps me to understand what I need to do with my camera settings to get better footage for the editor.

What amazes me still is I do love the footage from the 5D/7D etc right out of the camera (minus the moire/aliasing mind you) and I still struggle to define exactly why that is. I follow this thread, as I've said, in the hopes of a 7Dm2 that follows in the 5Dm3's footsteps (mostly) because it's a great fun camera.

I guess when one steps back and looks at how many variables there are to a good picture, it's quite a big list. We *say* this one thing is all that matters, but then when that alone gets fixed, we find something else to be unhappy about. And hey, I'm pointing the finger at myself as well!

In any case, thanks for all the work with the 5dm3, it's much appreciated!

 

[jcs]

One question I would have is how would the regions overlap? If you are shooting at basically 180 degrees, then sharpening a frame would create a sharper line at a high contrast motion blur section which would give you a moving line you would not see on a more blurred edge. This is what I observed in the shots that were posted. Areas with lower spatial frequency didn't exhibit the same effect. Give 1/2 of the time the camera is not recording, the spacing relative to that edge is totally dependent on the speed of the edge moving across the frame. It could be argued that it was a combination of the sharpening AND the codec in camera that is really causing the effect, and not just the sharpening alone, however. With a 4:2:0 codec, I notice a lot of contamination on edges that have a decently high spatial frequency to each other.

I'd have to see such a 'line'; not sure how that would be possible. Real motion blur acts like a Gaussian blur.

I guess when one steps back and looks at how many variables there are to a good picture, it's quite a big list. We *say* this one thing is all that matters, but then when that alone gets fixed, we find something else to be unhappy about. And hey, I'm pointing the finger at myself as well!

Coming from primarily the image-processing/math side, I see the biggest difference between film and video is complete lack of aliasing. Next is film's gamma response/curve (along with dynamic range) which keeps the image fairly continuous, especially in the shadows along with no blown-out highlights, and, amazingly at the same time, providing excellent microcontrast all over the image (can simulate somewhat in post using Unsharp Mask to perform "Local Contrast Enhancement" or with HDR simulation tools (Tone Mapping)). I've seen fairly grainless film and it looks amazing. When there is grain, it needs to be 'soft' around the edges of the grains, otherwise it looks like digital noise (video)).

Thus, if you think about what it takes to simulate video from scratch, you would:

1. Oversharpen to make sure there is aliasing, and to really drive the point home, moire.
2. Make sure highlights are peaky and blown out + clipped.
3. Crush the blacks and add crisp digital noise to dark regions along with compression artifacts.
4. Make sure dynamic range is as limited as possible.
5. Make sure the sky has lots of banding artifacts.
6. Use a contrast-enhancing S-curve instead of a log-curve (or contrast reducing S-curve) for luminance (this will help with 2 & 3, perhaps 4 & 5).
7. Don't perform any micro/local contrast enhancement.
8. Oversaturate the colors, and make sure skin tones look like shiny plastic (limited dynamic range and colorspace helps here).
9. Make sure there are no narrow depth of field shots; even better make sure everything is always in focus.
10. Add 1-pixel wide color digital noise to entire frame.

Don't do items 1-10 if the goal is a film look. Perhaps others can add to this recipe.

 

[squig]

Check out my MKIII test image http://www.dvxuser.com/V6/showthread.php?280165-5D-MKII-vs-5D-MKIII-for-video-shoots/page2 It's a jpeg compressed screenshot I took in FCP so the original prores file is higher quality. I ran the FCP7 sharpen filter over it @ 20% and ran a heavy custom magic bullet looks grade on it. IPB compression, neutral profile, sharpness 0, contrast -4, saturation -2, NR and HTP off, transcoded in compressor to prores HQ. The white balance was a little too cool in camera. Shot with a Sigma 20mm f/1.8 lens at around f/5.6 (memory is a bit foggy on that) The MKII footage @ 1250 ISO starts to fall apart in the grade, not so the MKIII footage @ 5000 ISO.

I'm totally blown away by the capabilities of the MKIII, with the results I'm getting I don't really see a need for removing the OLPF.

 

[Gene Sung]

So the guy at MaxMax who does the 5D3 mods emailed me back today. He said they did one 5D3 mod today and are doing another tomorrow, so I'm sure the users are going to post some footage soon. It's pretty expensive, about $600 if you have them do the whole procedure and then have them replace the removed OLPF with a glass filter so your auto sensor cleaning will work again. Which seems pretty important to me.

I'm still on the fence about whether to get the 5D3 or the Sony FS100 or maybe the FS700. I know the FS100 is a way better camera, but I just love the small size and quickness of the DSLR ergonomics, especially since I shoot hand held a lot.

Tough call. Going to sit back for a month or so and see all the developments.

I'm really hoping Magic Lantern will be ported.

 

[squig]

I'm still on the fence about whether to get the 5D3 or the Sony FS100 or maybe the FS700. I know the FS100 is a way better camera, but I just love the small size and quickness of the DSLR ergonomics, especially since I shoot hand held a lot.

I disagree about the FS100 being a way better camera. I don't like the way it handles highlights at all and it also has issues with sensor ghost reflections. It has advantages in some areas like 60p 1080p and HDMI out but for film image acquisition I personally chose the MKIII over the FS100 and I don't regret it at all.

 

[dustylense]

Check out my MKIII test image http://www.dvxuser.com/V6/showthread.php?280165-5D-MKII-vs-5D-MKIII-for-video-shoots/page2 It's a jpeg compressed screenshot I took in FCP so the original prores file is higher quality. I ran the FCP7 sharpen filter over it @ 20% and ran a heavy custom magic bullet looks grade on it. IPB compression, neutral profile, sharpness 0, contrast -4, saturation -2, NR and HTP off, transcoded in compressor to prores HQ. The white balance was a little too cool in camera. Shot with a Sigma 20mm f/1.8 lens at around f/5.6 (memory is a bit foggy on that) The MKII footage @ 1250 ISO starts to fall apart in the grade, not so the MKIII footage @ 5000 ISO.

I'm totally blown away by the capabilities of the MKIII, with the results I'm getting I don't really see a need for removing the OLPF.

I'm with you squig. I'm seeing great results. I've matched my mk3 and xf105 together and I'm putting some pretty hard pushes in the ALL-I codec and seeing great results. ANd the XF codec is very nice. But I swear my mk3 footage is keeping up. The removal of the OPLF interests me only if there are NO side affects in both motion and stills. My attention is on the BMD camera announced today, but I'm still very pleased with my Mk3 purchase. Great camera.

 

[Gene Sung]

Here's what the guy at Max Max wrote me about the 5D3 conversions:

He wrote:
----------------------
Thank you for your interest in our work.


We converted our first 5D III to HR today and have another one to do tomorrow.


I am not sure why video would be changed though. I would think that the blur filter shouldn't affect video because I would guess the down sampling would occur in the software, but I could be wrong.


Cost is $450 + return S&H. The camera would function normally except the automatic sensor cleaning would be disabled unless we cut and remove the piezo strip and rebond it to the new ICF ($150 extra). The article misses a few things:
- You can't just remove the AA filter because that changes the camera focal plane. You have to put something in its place to compensate. Thickness is critical. You have to be correct within +/- 0.02mm.
- The AA also has an ICF coating. You should replace with another ICF. We remove the AA and 2nd layer ICF replacing with a new ICF with correct thickness so that focus is correct.
- I wouldn't recommend that anyone try this at home. Aside from needing special tools, clean room and techniques, you need special equipment to clean the glass going in the camera. You can't just wipe microscopic dust off because of electrostatic forces. We have equipment that burns off a thin layer of glass to get to an atomically clean layer.