IMPORTANT UPDATED INFORMATION AT END OF THIS POST!
I’ve spent some time shooting test subjects with my C300 trying to replicate the green/red pixel issues getting reported elsewhere. So far the only real problem I have found is some pretty extreme LCA (longitudinal Chromatic Aberration) which is significantly worse when I use the ND filters. I have seen CA on my F3, it’s down primarily to lens performance and quality. But I have never seen anything as extreme as this. Below are a couple of frame grabs for you to look at, one full frame and the other a crop. What you can see is that anything closer to the camera than the chosen focal point has a distinct red/purple fringe and anything behind the focal point (further away) has a distinct green fringe. The lens used was a Canon 24-70mm f2.8 L series. One point to consider is that DSLR lenses were never designed to have extra glass between the back of the lens and the sensor other than perhaps the OLPF (Optical Low Pass Filter) and combined IR filter, which normal sit right on top of the sensor. The C300 has a clear glass filter and 3 ND filters. Each of these filters is at a different distance some way from the sensor. I suspect that these filters are at least in part causing the CA to be more severe than is typically seen in a DSLR as the extra glass will alter the point at which all the colours are brought into focus. With the filters being quite a distance from the sensor any dispersion of the colours through the filter will be quite well spread out by the time it gets back to the sensor, exaggerating the effect. On the Sony F3 the ND filters are extremely thin and very close to the sensor, but still I should try the same lens on the F3 to see what happens. It’s also been noted that the problem is worse when you shoot interlace. Now the C300 sensor is quite different from anything we have seen in a camcorder like this before. Some of the image processing needed to create the interlace fields from this progressive scan sensor is done on the sensor chip itself and I suspect this may be why we are seeing more artefacts in interlace than progressive as the processing is possible quite rudimentary compared to what can be done in a dedicated DSP (Digital signal processor). That is speculation on my part. Every now and again I do see some aliasing artefacts on horizontal and vertical high contrast edges that manifest themselves as small (1 or 2 pixels) red and blue blocks. Again this is probably a result of the sensor design. The aliasing artefact is very minor and not something I am concerned about, the LCA is more of a concern.
UPDATE: I tried this same test today with my 24-70mm L series lens on my F3. The results are very close, lots of strong LCA. Now, I’ve never seen this extreme LCA on my Canon DSLR. A bit of digging reveals that when you put this lens on a Canon DSLR, the camera activates a tailored compensation circuit to minimise the effects of the CA, this circuity is not present in the c300. I also tried my trusty Tokina 28-70mm ATX Pro lens on the F3 and the C300 and in both cases I get less LCA with the Tokina (which cost £200) than the brand new 24-70mm Canon lens. The Tokina, being a Nikon mount lens has a real iris ring, so no nasty stepping iris. You know I’m really disappointed by this. One of the big selling points of the C300 has been that you can use Canons “L” series lenses, yet this is a real let down as the lenses won’t appear to perform as well on the C300 as they do on a DSLR and I don’t think many people are aware of this, I wasn’t. This is not a fault or issue with the C300 itself but in my opinion it does further lessen the supposed advantage of having a camera that takes Canon lenses without an adapter.
When I owned a Canon 24-70mm f2.8L and a 5D MII, I found the 24-70 produced a lot of CA on objects like tree branches, in stills . I think the lens is the problem, enhanced by the optical path on the C300. Has anyone got this problem on a prime lens?
I’ll try some different lenses tomorrow. Certainly CA is not unusual, but this is a pretty extreme example and the same lens on my 550D does not look anywhere near as bad as this. I’m not the only one seeing this abnormally high amount of CA. There’s a rippling water clip doing the rounds that looks terrible as every specular glint off the otherwise flat water looks like a bright green flash of light.
I have seen this chromatic aberration in just about every camcorder I have ever owned to a greater or lesser extent, the Sony NX70 is quite bad for this at certain focal lengths.
I look forward to your test with the F3 and the same glass. I am using 720 50p with my C300 which as you point out is why I do not see the same problem…very interesting.
Do you think there may be a fix either by firmware or tweeking.
It’s primarily an optical problem, so it makes no difference what frame size or rate you shoot at. Perhaps you just have a better lens than the one I have.
Thanks for testing this. The images you are seeing don’t look like the problem we were describing, as you exactly point out. That looks like pretty bad lens-caused CA. The aliasing we (well I) were/was seeing (and it’s rare and virtually unnoticeable given a properly exposed frame) is a pixelated blocky green and purple digital kind of noise against mostly horizontal surfaces. Which, as you also notice, is hard to see except for extreme circumstances. Circumstances that some have been able to create when trying, as I have. I can’t say it’s anything someone would see in normal shooting, (I hadn’t seen a thing until it was brought to my attention this week). I will try some more testing this weekend, but my guess is unless I focus in on a hard contrast edge that is backlit way above 100 IRE, I won’t see anything noticeable.
I have been using Canon glass… 17-55 F2.8 IS, 100mm F2.8 L IS, Canon 50 F1.4, Canon 70-200 F2.8 IS II and haven’t seen anything nearly that bad, or as bad as what Paul Joy showed.
Interestingly, his 50i shots look far worse color aliasing wide than his 25p shots.
Hi Alister, Canon have just replied to me after looking at your sample footage and tell me… “These frame grabs are very over exposed and designed to show the limits of 10 year old optics.
The 24-70 f/2.8L was launched in 2002 and the difference in optical performance between this and the replacement announced earlier this year is huge.”
PS. My lens was the Canon EF-S17-55 f2.8 lens (2007) which might account for better CA.
Hello Alister!
I can only repeat what others before me already have said.
The 24-70 really suffers from chromatic aberration, which I have seen a lot together with my 5DmkII.
Looking forward to reading about your further investigations in this.
thank you
Well Denis, I should have listened to the others more. The 24-70 is actually quite shockingly bad when it come to LCA. The results on the F3 are similar, lots of LCA. But when I use the lens on my Canon DSLR, most of the CA is not present as the DSLR has electronic CA reduction (which I was not aware of) tailored to the lens. So, DSLR shooters beware, your Canon lenses might not look quite so good on the C300.
I’m a still photographer w/20+ yrs. exp., now doing video for the past year and a half.
I just want to backup the feedback above from other cameraman and Canon. The 24-70 v1 is not the greatest lens by any means. It has lots of CA and it has a lot of distortion especially at the wide end. The CA is visible in still images if you look for it. Video folks seem to love this lens but I always found it wanting.
Every new camera, still or video, needs to be tested with each lens. I had lenses (the 24-70 and 16-35 v1) both of which were okay on my 1DsM1 & 1DsM2 but couldn’t cut it, in my opinion, on the 1DsM3.
I am not aware of any of the pro Canon bodies correcting for CA (though the new 5DM3 and the 1DX might, and the new version of DPP that comes with them will.) Some of the pro bodies did correct for lens created vignetting (auto light optimizer) but I think that was the extent of it.
According to a message from Canon the DSLR’s that have correction for vignetting also have a degree of correction for CA. Certainly the still images from my 550D taken with the 24-70 L don’t look nearly as bad as the video clips. The CA is still there, but not as extreme.
I’m not entirely sure I buy this as just a lens issue. how come, when the image is 1920×1080 the green tinting looks that much worse than when I look at the still image at full resolution on the 5dII? Is that really just down to the 5d adjusting for the lens? surely a lens that’s designed for such a high resolution would find 1080 a breeze?
this shows up very badly on all highlights with the 50mm 1.2 prime as well. is this also a rubbish lens then?
cheers for all the info guys, this is very interesting stuff.
Ok, I’m even less convinced that it’s a lens issue after today. The guys at CVP invites me in to demonstrate the issue and get some working examples that we could send over to Canon for analysis.
We tried the 24-70 2.8 which obviously showed the problem, then tried the 50 1.2 which showed it up even more. Neither myself nor Dominic at CVP thought the green looked like Chromatic Abberation as it was too consistent across the image.
We then tried one of the new primes from Canon, the 50mm EF prime. Sure enough the green/purple was there. This is one of Canons new lenses so I don’t believe for minute that they would have allowed such bad aberration off the factory line!!
We’ve sent the test to Canon and I’ll let you know what they say.
Cheers,
John
Keep us posted. I have not seen anything other than LCA and minor aliasing.
It’s driving me nuts, I’ve just come back from another shoot to find that a lot of my shots are suffering terribly. The shots in question are not overexposed at all, they are of an outdoor meeting area which has chrome topped hand rails around the edges and the chrome on the hand rails awful, as the camera pans the combination of aliasing and artefacts makes the problem jump out at you and draws your attention from the subject.
I’m currently going through the support process with canon and the issue has been escalated to canon europe, no other news as yet. I really need to know if this can be fixed because I’m quickly losing faith in this camera and only have a small window for returning it now.
I just got back from a shoot with 3x C300’s in Spain shooting racing bikes. Beautiful clear skies and lovely vibrant colours on the bikes. No problems with any of the cameras of footage beyond the ergonomic frustrations, especially the location of the audio controls. I have to wonder whether your camera has a fault. Do let us know what Canon say.
I got very interested in this issue after buying a Zeiss ZF.2 35mm 1.4mm lens for my F3 and was astounded at the amount of green/magenta LCA. It seemed particularly bad around the borders of overexposed objects like lampshades. A piece of bubble wrap that came with the lens was a riot of color in front & behind of the focal point.
I had never noticed it before except on my HVX 200 where I was persuaded that it was a camera prism problem.
Subsequently I tested ever lens I could find and noticed incredible differences. The Zeiss ZF f2 had lots but not nearly as bad as the f1.4. All my still lenses had some LCA although on the Nikon 17-35 it was blue/amber which looked more natural.
Next I compared it to a group of older Zeiss f2 (or was it 2.2) standard primes – i.e. real PL mt cinema lenses . No LCA at all! – though my Zeiss ZF’s tended to be sharper. I never tested on any other camera, but my assumption was the problem was glass not the camera. Unfortunately it seems you get what you pay for.
It would be interesting to test on a DSL or another camera if I get a chance.
Lenny Levy
A friend who knows a bit about optics told me it tended to be a problem as the lenses got sharper. Part of the trade offs that lens makers have to deal with. Perhaps it also just becomes more noticeable with a camera that resolves more as well. Just a guess.
Curiously I’ve never seen it on any real world shots since then and I took that 35mm f1.4 out at night with streetlights and didn’t see it there either. I sold the Zeiss f1.4 though and stuck with f2 because it really bugged me.
Alistair – will you be demoing anything at NAB this year?
At NAB I went to the Zeiss both to ask them about LCA with their ZF lenses. On my F3 I’ve seen tremendous amounts of LCA on the 35mm ZF.2 1.4 lens and at least some on every other still lens I use. Pretty annoying once you start to see it. The Zeiss rep confirmed for me that LCA was very hard to avoid in lens design and there is a reason you don’t see it on very expensive PL glass but you will see it some on nearly every still lens made. ( He admitted it was the dirty little secret of lenses that wasn’t often talked about.) The problem gets worse at wider f stops and as the lens gets sharper. Some lenses (like the 35mm) just are particularly hard to design.
He also said that extra glass between the back of the lens and the sensor like OLPF’s (and I’m guessing IR filters and maybe even ND’s) complicate the issue because there are so many cameras around you can’t design the lens for all of them.
Lens design is a bit of a dark art and now we are complicating the issue with rear mounted ND’s etc. It was only 3 years ago that everyone was complaining about the back focus on EX1’s changing when you switched ND filters. That was with a camera that had very similar thickness ND’s in each position. The F3 and C300 have a different number of filters for different ND levels, so what hope to we have?
Seems it wasn’t just a lens issue! So glad that Canon are releasing a firmware update! Will check it out ASAP!
This after canon repeatedly insisted that it was the lens not the sensor…