AF101 BBC/EBU test results posted
- Thread starter nugat
- Start date
- Status
Hmm.
Not a glowing report by any means. I wonder how the BBC will react to this camera?
I especially found these passages interesting:
However, to achieve a smaller depth of field in
this camera, relative to, say, a ½” camera, then the lens must be opened by at least 1.5 stops; using an F/2.8
lens on this camera, wide open, will give the same depth of field as on a ½” camera with a lens opened to
F/1.6. This camera will not necessarily always deliver short depth of field, large aperture lenses must be
used to achieve that.
Not a glowing report by any means. I wonder how the BBC will react to this camera?
I especially found these passages interesting:
This camera does not perform particularly well at HD. Clean resolution is limited to about 1210x680 by the
presence of high-amplitude spatial aliasing.
However, to achieve a smaller depth of field in
this camera, relative to, say, a ½” camera, then the lens must be opened by at least 1.5 stops; using an F/2.8
lens on this camera, wide open, will give the same depth of field as on a ½” camera with a lens opened to
F/1.6. This camera will not necessarily always deliver short depth of field, large aperture lenses must be
used to achieve that.
Last edited:
Barry_Green
Moderator
Not a glowing report in terms of overall sharpness, no, but seeing as Alan Roberts is tasked with hardcore evaluation and pulls absolutely zero punches, it should be noted that he also said:
and as far as rolling shutter goes, when set to 1/1000:
He does say that on the zone plate, that there is strong aliasing, something that I've seen in chart testing as well, but ... it simply doesn't show up in actual shots. It seems like there's something they're doing, when combining the colors, that just makes the aliasing not visible in actual shots. This is something I've been exploring but can't quite figure out yet.
His conclusion is certainly not flattering -- he says that:
I think we all agree that it's not delivering 1000 TVL of res like the EX1/EX3 do; Alan's numbers here would seem to imply a solid 720p of res. Adam Wilt showed 700 lines, Panasonic claims 800. I have tested against my HMC40 (the only "full raster" camcorder I have), and the AF100 is not quite as sharp as the HMC40. That we knew. However, it is quite interesting to hear him say that there is no OLPF or an inadequate OLPF, as that's something Panasonic has trumpeted.
While it's not a flattering report, it's not condemning either (such as his report on the 5D, which was an open-and-shut case and he he said the 5d "effectively fails" and didn't even bother to conduct full testing). Compared to language such as that, the AF100 report is certainly much milder, but I'm sure that we were all hoping for better.
Interesting that after Alan Roberts completed his testing and submitted it to the BBC, CreativeVideo.co.uk tweeted that the BBC had approved the AF100 when used with the NanoFlash at 50mbps or more. I would like to hear that direct from the BBC, of course.
and as far as color goes:
and as far as rolling shutter goes, when set to 1/1000:
He does say that on the zone plate, that there is strong aliasing, something that I've seen in chart testing as well, but ... it simply doesn't show up in actual shots. It seems like there's something they're doing, when combining the colors, that just makes the aliasing not visible in actual shots. This is something I've been exploring but can't quite figure out yet.
His conclusion is certainly not flattering -- he says that:
I think we all agree that it's not delivering 1000 TVL of res like the EX1/EX3 do; Alan's numbers here would seem to imply a solid 720p of res. Adam Wilt showed 700 lines, Panasonic claims 800. I have tested against my HMC40 (the only "full raster" camcorder I have), and the AF100 is not quite as sharp as the HMC40. That we knew. However, it is quite interesting to hear him say that there is no OLPF or an inadequate OLPF, as that's something Panasonic has trumpeted.
While it's not a flattering report, it's not condemning either (such as his report on the 5D, which was an open-and-shut case and he he said the 5d "effectively fails" and didn't even bother to conduct full testing). Compared to language such as that, the AF100 report is certainly much milder, but I'm sure that we were all hoping for better.
Interesting that after Alan Roberts completed his testing and submitted it to the BBC, CreativeVideo.co.uk tweeted that the BBC had approved the AF100 when used with the NanoFlash at 50mbps or more. I would like to hear that direct from the BBC, of course.
Last edited:
Barry_Green
Moderator
That part doesn't make any sense, either, because earlier in the report he said "the camera noise level is quite low."
I wonder which of those contradictory statements sums up his real feelings?
Plus, if you set the camera for low noise, it is basically *extremely* low noise. Put the detail level on -4 or lower (and v-detail-level on -4 or lower), crank up the coring to +2, turn the color saturation down a bit (chroma level -3 or so) and use the cine-V gamma and the norm2 matrix, and it's nigh unto noiseless. Mabye not 100% noiseless, but extremely low noise.
I wonder which of those contradictory statements sums up his real feelings?
Plus, if you set the camera for low noise, it is basically *extremely* low noise. Put the detail level on -4 or lower (and v-detail-level on -4 or lower), crank up the coring to +2, turn the color saturation down a bit (chroma level -3 or so) and use the cine-V gamma and the norm2 matrix, and it's nigh unto noiseless. Mabye not 100% noiseless, but extremely low noise.
Barry_Green
Moderator
Also, another odd thing - he says that to match the DOF of a 1/2" camera, you have to open up by 1.5 stops. That's just totally wrong -- to match the DOF of a 1/2" camera to a 4/3" camera, you'd have to open up by three full stops. Seems like the report could use another pass to clarify some of these things (what he means by noise, whether it's low or high, and that there's a three-stop difference between 1/2 and 4/3).
Alpha Beta Omega
Veteran
The AF101 is not a perfect camera but I haven't noticed any aliasing in my footage.
Remember, just because BBC or some other high tech cameraguy thinks one thing about it, dosen't mean that everyone should feel the same. AF100 is a really good camera when you use it the correct way.
By the way, if you manage to film something really big and important for the news, BBC won't ask you what you've shot it on, they would be satisfied with iphone footage if so needed.
By the way, if you manage to film something really big and important for the news, BBC won't ask you what you've shot it on, they would be satisfied with iphone footage if so needed.
Last edited:
I certainly appreciate the report, or any effort to conduct decent tests. But really? In the end, use your eyes, people. I don't much care what claims are made for which camera, if it works for my purposes, it works, if not, then not. There are many factors that determine the final image, and shooting charts does not always capture it in a lab/test environment. The only issue in this particular case, is what the BBC decides to do, as that may impact what cameras some people can use. Are we still waiting for official word from the BBC re: AF101?
Barry_Green
Moderator
Well, as far as I know, yes... but even then, it's odd, because after Alan submitted his report, CreativeVideo.co.uk tweeted that the AF101 had been approved by the BBC (as long as you used the Nano at 50mbps+, because the BBC HD requirements demand 4:2:2 color sampling). Since then, I haven't seen anything official, so ... I don't know where creative got their information from. I haven't seen any official announcement from an official BBC source.
olof
DVXuser Sponsor
This kind of confirms my feelings about the AF100.
I still think it is a great camera and I am now using it in production. But I found that my EX3 and EX1R are much sharper, I spent a lot of time trying to make the AF100 look as sharp, and I never really could (lots of different lenses and lots of trying settings). The less than 800 (680) vertical resolution seems consistent with my findings.
There are also some really strange things that happen with banding, but I have not had a problem with aliasing at all.
I find that I can create beautiful images with the AF, just always a little soft. This is not a bad thing, you just have to be ready for it. The DSLRs are way to (artificially) sharp for my taste, this can be dialed down and fixed in post.
At this time I think in this price class the AF100 is the best you can do for image quality and SDOF, with reasonable working ergonomics and features. It is a big step above the DSLRs. And I bet that Pany will improve some of the codec problems with future FW updates.
I don't think we have to be evangelical about this camera just because we bought it, but we do have to work around the short comings and appreciate the innovative design features. I think it is the first in a whole new class of camera and a very good one at that.
Remember the BBC compares this with cameras costing many times more. This is really a very affordable camera. Personally I will always rather use the AF100 than an EXcam and a SDOF adapter, and in most cases a DSLR, I have both a 5DmII and a 7D. Most of all I love being able to use my old, and new Canon lenses with the coming Birger adapter (really great for remote focus pulling on jibs and Stedicam). And the Voightlander 25mm f.95 is almost worth getting the AF100 by itself, a wonderful lens (all I want is to now de-click its iris).
The bottom line is I like this camera a lot for many reasons, it has some problems but so does every camera. I am glad I have it and I will use it until something better comes along that makes sense for me, but for now this is my SDOF camera.
I still think it is a great camera and I am now using it in production. But I found that my EX3 and EX1R are much sharper, I spent a lot of time trying to make the AF100 look as sharp, and I never really could (lots of different lenses and lots of trying settings). The less than 800 (680) vertical resolution seems consistent with my findings.
There are also some really strange things that happen with banding, but I have not had a problem with aliasing at all.
I find that I can create beautiful images with the AF, just always a little soft. This is not a bad thing, you just have to be ready for it. The DSLRs are way to (artificially) sharp for my taste, this can be dialed down and fixed in post.
At this time I think in this price class the AF100 is the best you can do for image quality and SDOF, with reasonable working ergonomics and features. It is a big step above the DSLRs. And I bet that Pany will improve some of the codec problems with future FW updates.
I don't think we have to be evangelical about this camera just because we bought it, but we do have to work around the short comings and appreciate the innovative design features. I think it is the first in a whole new class of camera and a very good one at that.
Remember the BBC compares this with cameras costing many times more. This is really a very affordable camera. Personally I will always rather use the AF100 than an EXcam and a SDOF adapter, and in most cases a DSLR, I have both a 5DmII and a 7D. Most of all I love being able to use my old, and new Canon lenses with the coming Birger adapter (really great for remote focus pulling on jibs and Stedicam). And the Voightlander 25mm f.95 is almost worth getting the AF100 by itself, a wonderful lens (all I want is to now de-click its iris).
The bottom line is I like this camera a lot for many reasons, it has some problems but so does every camera. I am glad I have it and I will use it until something better comes along that makes sense for me, but for now this is my SDOF camera.
This is not the BBC's decision, but Alan Roberts' opinion, if I may quote him directly from DVDoctor.com:
I suspect the official position will be that it's an HD camera, but not top grade. So, it'll get approval for use in film-style shoots but not as the main camera. I guess it's footage will be classed in the 'non-HD' quota, including archive and SD footage. But, I don't make the rules or the decisions, I only measure and report
The only reason I mention the BBC is because I know Roberts does his reports for them. I've never worked for the BBC, but I do respect their guidelines.
Matt Gottshalk
Resident Preditor
I can tell you for a FACT that the camera doesn't alias on fine herringbone patterns as I just shot an interview with it where the subject wore such a pattern.
I cringed when he sat down but then looked at my production monitor and sighed in relief as the af-100 handled it in spades.
I would have been absolutely BONED if I had to shoot it with a 5d/7d.
astigmatic
Well-known member
So why is the diagonal resolution so much better. Diagonal lines are what usually causes aliasing.. Does it make a difference in real world sharpness if the 800 lines are diag rather than vert?
Barry_Green
Moderator
It is a frustrating thing to diagnose. I can't tell you the reasons why, and that irritates me to no end -- I love to figure stuff out and know the answers, I dig and dig and research for answers, but this one is escaping me.
If you judges strictly by looking at a black and white res chart, you would think there would be terrible aliasing -- but when you shoot footage, it just. isn't. there. Yes you can see a tiny little bit under the most tiny-detail scenarios, but typical shots that would send a DSLR weeping for its mommy, the AF100 just eats up for breakfast (witness McGee's post above yours).
Why is this? I don't know. Why are the diagonals a solid, sharp, alias-free 800 TVL on a b&w chart, but the strictly horz and vert aren't? I don't know. It makes no sense. I have never before encountered something easily demonstratable on a chart, that I haven't been able to then demonstrate in real-world footage. It's curious as all get-out to me. Is there some magic happening when the three color channels are taken into account? I don't know. On black and white, there is evidence of significant aliasing. But in the color chart, when recorded as 4:2:2, it looks quite solid (not perfect, but really much better). So is it some sort of pixel-shift magic like on the HVX, where the combination of red, green, and blue taken in an aggregate resulted in noticeably higher resolution than any of the individual chips could deliver on its own? I don't know. And that drives me batty!
What I can tell you is that the aliasing is extremely obvious on a black and white chart, and it will rarely ever show up in an actual shot. You can read review after review of people who've used it who will tell you, there just isn't any noticeable aliasing at all in their footage. On a chart, yes, on footage, no. How is that possible? I simply don't know. But it is the results that you can easily see when you use it. Trust me, I will be having very long, pointed, and demanding discussions with the engineers next month at NAB, to see if I can get to the bottom of this. Even then I don't know for sure they'll say, because maybe it's some sort of trade secret or patented intellectual property or something, but ... I will try to find out, because it is really curious.
If you judges strictly by looking at a black and white res chart, you would think there would be terrible aliasing -- but when you shoot footage, it just. isn't. there. Yes you can see a tiny little bit under the most tiny-detail scenarios, but typical shots that would send a DSLR weeping for its mommy, the AF100 just eats up for breakfast (witness McGee's post above yours).
Why is this? I don't know. Why are the diagonals a solid, sharp, alias-free 800 TVL on a b&w chart, but the strictly horz and vert aren't? I don't know. It makes no sense. I have never before encountered something easily demonstratable on a chart, that I haven't been able to then demonstrate in real-world footage. It's curious as all get-out to me. Is there some magic happening when the three color channels are taken into account? I don't know. On black and white, there is evidence of significant aliasing. But in the color chart, when recorded as 4:2:2, it looks quite solid (not perfect, but really much better). So is it some sort of pixel-shift magic like on the HVX, where the combination of red, green, and blue taken in an aggregate resulted in noticeably higher resolution than any of the individual chips could deliver on its own? I don't know. And that drives me batty!
What I can tell you is that the aliasing is extremely obvious on a black and white chart, and it will rarely ever show up in an actual shot. You can read review after review of people who've used it who will tell you, there just isn't any noticeable aliasing at all in their footage. On a chart, yes, on footage, no. How is that possible? I simply don't know. But it is the results that you can easily see when you use it. Trust me, I will be having very long, pointed, and demanding discussions with the engineers next month at NAB, to see if I can get to the bottom of this. Even then I don't know for sure they'll say, because maybe it's some sort of trade secret or patented intellectual property or something, but ... I will try to find out, because it is really curious.
Yes, the diagonal resolution of the images on page 9 of the BBC report is excellent - virtually no aliasing at all! So perhaps...
The image sensor of the AF100 is oriented diagonally?
If so, the rectilinear gap between RGBG photosensors would be 1.414 times wider (square root of 2). Perhaps this could explain the author's puzzlement over the scaling of the sensor's 12.4 Megapixels:
Alan Roberts said:
As a rough estimate, if we multiply the FHD frame of 1920x1080 by a 1.414 diagonal factor the result is about 3 Megapixels (with a little extra for the margins). The RGBG demosaic algorithm will need a least double this resolution, both horizontally and vertically, producing a diagonal grid of about 12 Megapixels.
LOL if this is the way it actually works!
Barry_Green
Moderator
Well, the sensor isn't diagonal, you can look through the lens port and see that it's square. But maybe the pixels are arranged diagonally on the sensor?
- Status