Barry_Green
Moderator
Exactly.
what could af100 do that gh2 couldn't?
- Thread starter ghosttv
- Start date
roei z
Veteran
stoneinapond
Veteran
Barry, I agree with everything else you said. However, the ETC mode on the GH2 is definitely a useful feature. I am surprised the AF100 does not have it.
Barry_Green
Moderator
I love ETC. But it is impossible for the AF100 to have it. You can have aliasing, or ETC, you have to pick which one you get. The AF100 is designed to avoid aliasing, and that necessitated the sacrifice of ETC.
The GH2's great, I love it. I'm very glad I have it.
I just don't understand why people cling to the notion that they're getting something for nothing. Life doesn't work that way.
nyvz
Veteran
Barry, I apologize if I'm seeming overly-argumentative on the subject, but I am having trouble understanding why you are so sure of something you admit to having limited information about. If I am overly skeptical of some of your statements about the AF100 sensor, it is only because I know so many people look to you for your knowledge in these matters.
I, too, am waiting for Alan Roberts report regarding his evidence that he feels proves that the AF100 does no line skipping. I remain skeptical due to the math (seems like reading the whole sensor at 60fps and >12bits would require internal bandwidth in excess if 1GB/s), as he also may mean that there is not EVIDENCE of line skipping in the images, which could also mean that the OLPF removes any negative effects of line skipping.
I'd like to again post some evidence. All I am asking is whether you think there could be some doubt that the AF100 and GH2 have completely different sensors. It may be that you define "different sensors" differently from me, so I should define what I mean.
Same Sensor would have same technology at the photosite, same performance at the photosite, same size photosite, consist of silicon cut to the same size, maybe even the same ADC on chip. You are basing most of your information on product specs, here are the listed AF100 specs straight from the manual:
"Number of Pixels:
Effective pixels: 12.4 million approx.
Single Panel (16:9)
Image Size
Valid Image Range
Approx. 17.8 mm (H) x Approx. 10.0 mm (V)"
You'll notice that the image capture specifications do not list anything about the actual sensor itself, only about the "effective pixels" and the "valid image range". Other Panasonic cameras, like the DVX100 for example, list total pixels and effective pixels separately as every camera I am aware of has extra photosites that are not used during normal operation, in some cameras these pixels are used to set black level or for lookaround and in others they may just be unused because they are not active for whatever reason.
Also you mention the Sensors "look" similar in the GH2 and AF100 when looking into the mount. I might expect the difference between a 1.44:1 and 1.77:1 aspect sensor of (slightly) different sizes would be visible?
We have established that both sensors have the same pixel size (The first evidence I would look for if they were different sensors would be for different pixel size, especially since the two cameras are designed for different purposes). Of course pixel size does not prove anything about them being the same sensor but if it is the same pixel size, they still could be the same sensor.
I am not saying that they are the same sensor, I'm just saying there is NO evidence put forward as of yet that says one way or the other. Active area or effective pixel count could mean that some photosites are just disabled so it is no proof one way or another. Nor is field of view, which is directly related only to active imaging area. Do you agree?
I, too, am waiting for Alan Roberts report regarding his evidence that he feels proves that the AF100 does no line skipping. I remain skeptical due to the math (seems like reading the whole sensor at 60fps and >12bits would require internal bandwidth in excess if 1GB/s), as he also may mean that there is not EVIDENCE of line skipping in the images, which could also mean that the OLPF removes any negative effects of line skipping.
I'd like to again post some evidence. All I am asking is whether you think there could be some doubt that the AF100 and GH2 have completely different sensors. It may be that you define "different sensors" differently from me, so I should define what I mean.
Same Sensor would have same technology at the photosite, same performance at the photosite, same size photosite, consist of silicon cut to the same size, maybe even the same ADC on chip. You are basing most of your information on product specs, here are the listed AF100 specs straight from the manual:
"Number of Pixels:
Effective pixels: 12.4 million approx.
Single Panel (16:9)
Image Size
Valid Image Range
Approx. 17.8 mm (H) x Approx. 10.0 mm (V)"
You'll notice that the image capture specifications do not list anything about the actual sensor itself, only about the "effective pixels" and the "valid image range". Other Panasonic cameras, like the DVX100 for example, list total pixels and effective pixels separately as every camera I am aware of has extra photosites that are not used during normal operation, in some cameras these pixels are used to set black level or for lookaround and in others they may just be unused because they are not active for whatever reason.
Also you mention the Sensors "look" similar in the GH2 and AF100 when looking into the mount. I might expect the difference between a 1.44:1 and 1.77:1 aspect sensor of (slightly) different sizes would be visible?
We have established that both sensors have the same pixel size (The first evidence I would look for if they were different sensors would be for different pixel size, especially since the two cameras are designed for different purposes). Of course pixel size does not prove anything about them being the same sensor but if it is the same pixel size, they still could be the same sensor.
I am not saying that they are the same sensor, I'm just saying there is NO evidence put forward as of yet that says one way or the other. Active area or effective pixel count could mean that some photosites are just disabled so it is no proof one way or another. Nor is field of view, which is directly related only to active imaging area. Do you agree?
Last edited:
nyvz
Veteran
But if the AF100 does not do line skipping, then it does not need a strong OLPF since aliasing is avoided simply by doing proper downscaling from the oversampled sensor. And if they do not need a strong OLPF designed for 2MP video, then there would be no problem implementing ETC.
Think of it like RED, it has a sensor with a similar number of pixels and does not line skip and does not alias because it has an OLPF designed for its pixel size. Yet it can also record cropped images in its 2k and 3k modes.
The only reason ETC would be impossible in an AF100 is if it has a 2MP OLPF, and the only reason it would need a 2MP OLPF is if it line skips.
Barry_Green
Moderator
Do you have examples of that? I've stated my sources. If my own observations, and the company's official statements, and Alan Roberts' reviews aren't enough -- what additional are you looking for?
Clarity, crispness, noise performance, 3200 ISO performance, aliasing performance. I keep meaning to put them through a comprehensive side-by-side, I just haven't because a) there's too much to do, and b) it's a rather pointless exercise. As I've said a million times, it's like comparing a motorcycle to a pickup truck.
Yes, both.
It reduces the false detail. One can convince themselves that the GH2 looks sharper, until they try panning. Then the sparkly shimmering of aliasing reveals that the actual detail is quite a bit lower, and the appearance of sharpness is due to the false detail of aliasing.
Not exactly sure why you've decided that we need to have a pissing-contest here -- why not just use both cameras and examine the real-world evidence, instead of trying to play some big game of "gotcha"?
I don't necessarily disagree with Emil Martinec's findings, as they do correlate with accepted wisdom and real-world observation. Statements such as "The conclusion is that noise is resolution dependent. If one compares two cameras having the same sensor size and technologies but different resolutions (pixel densities), the one with the lower resolution (lower pixel density) will have less noise at the pixel level, as the above exercise again demonstrates" are perfectly reasonable.
What I do disagree with, is taking these general conclusions and trying to apply them to a specific situation with uncontrolled variables. It is not reasonable to say "Martinec says bigger sensors are better, therefore a GH2 is less noisy than an AF100." It's simply not a rational conclusion. These cameras do not exist as sensors in isolation. You cannot compare sensor to sensor. The only thing you can compare is video to video, after whatever filtering, digital signal processing, noise reduction, sharpening, and whatever else has all been done.
If someone wanted to say "bigger sensors are better", they could argue that the 5D must be better than the GH2. Yet a simple observation of actual results would show that the GH2 is substantially sharper, and cleaner from artifacts, than the 5D could ever hope to be.
Just look at the output.
The only person to whom the actual sensor output could possibly matter, is that person who is building his own camera. For everyone else, it doesn't matter -- it only matters what the camera builders DID with that output.
I do feel that the GH2 and the AF100 create images with at least as much latitude, and less noise, than Canon's APS-C and FF35 cameras, yes. Art Adams tested a 5D and found it to deliver 10 stops on a DSC Labs 17-step calibrated dynamic range chart. I tested an AF100 on a DSC Labs 13-stop calibrated dynamic range chart, and observed 10.5. Those are the facts. Those are simply-observable, obvious, and easily-repeatable facts.
Don't just look at the studies and at the theories of what a particular pixel well can hold in relation to the dark current, etc. Just look at the footage. You cannot compare three-year-old technology against three-month-old technology and think that they're going to be the same! Canon's been using pretty much the exact same technology since the 5D was released in 2008. Panasonic's sensor technology is at least one, if not two, generations more recent.
Going back to Martinec's study, the absolute key line that somehow keeps being left out of discussions like this is:
Bold added for emphasis.
The GH2/AF100 are using a different generation of sensor technology, and a different generation of image processing, than Canon is. Therefore squabbling over pixel or sensor size is not a productive discussion. And even if it was -- what would it prove? What does it matter what the sensor can deliver -- you can't use that. The only thing that matters is what the CAMERA can deliver -- which combines the sensor with the image processing and the compression and the various conversions and sampling and all the other things that go into creating a final image.
I'm far more interested in what you can actually do with these things, than with whether one has an extra megahertz or a different pixel pitch.
Last edited:
stonebat
Veteran
I ordered GH2 for my hobby. If I can get paid for taking video, of course AF100. See how editman set up his AF101. He's using it as run-n-gun style for paid works. I'm not saying GH2 isn't good enough for commercial work though. When he bought 14-35 and 35-100 f/2.0 zooms, I was like.. what da... the lenses cost as much as his AF101. But it made sense later. Moving from GH2 to AF100 could mean... replacing entire gears.
Unlike GH2 rig setup, AF comes with all in one package. It will cost a lot to upgrade to next future version, but the pros will replace it cuz time is money for them.
If you wanna know more about AF100, download AF userguide and goto AF100/101 forum and learn more about it.
Last edited:
Barry_Green
Moderator
Because I own both products and owned a 7D. I've used them extensively. The results are obvious.
Maybe so. He seems very sure that it is most definitely not line skipping.
But does it really matter? And if so, how?
I've never said that they are "completely different" -- I've said that they are different. I don't know to what degree they are different. They may have some common ancestry in silicon. I don't know if they do or they don't.
It may indeed be the same basic silicon -- the pixel size argument does point to that. But if it's got a different ADC, different processing, different dimensions, different DSP, different color filter (not known, but presumed), different scanning system, and different OLPF -- doesn't that add up to "it's different"?
You can't make assumptions about how other cameras list their specs; the HVX, HPX170, HMC150, and HPX500 manuals never published specifications about their actual or used pixels. The HPX500 is quoted only as having "620,000 pixels". Eventually Panasonic published a white paper showing the HVX chipset as being three 960x540 chips with spatial offset, but no mention was made of actual, total, or effective pixels...
True. But am I truly able to eyeball the difference of .6 of a millimeter? I wouldn't claim that capability. I'm just acknowledging that when you look in the lens port, they're not obviously different.
And I'm saying that there is NO evidence of any type to say that they are. The only evidence we have is in their differences.
This is possible. It's possible that it's the same base silicon, but they use a different type of color filter array and pixel-binning of multiple pixels to get the read speed up, reducing the effective pixel count, and that 17.8 mm made for a mathematically more-appropriate size for executing that binning arrangement. But if that's the case ... doesn't that just further emphasize that they're different?
If the GH2 uses a Bayer color filter array, and reads out the pixels individually in stills, and skips pixels and rows in order to make video...
And the AF100 uses a modifed RGB Stripe or something like that, with permanently hard-wired binning which groups adjacent pixels into so-called superpixels, which allows them to read the entire sensor at a more video-appropriate resolution with no pixel skipping, and they use an OLPF to smooth out the footage to avoid aliasing complications (just like every video camera uses an appropriate OLPF) ... doesn't that make it "different"?
It does to me.
And no, I don't know that they're using an alternate stripe pattern or binning. But what I do know is that the official statements from the Panasonic people are that it isn't pixel-skipping or line-skipping, that it uses a "primary color filter", and that somewhere on Alan Roberts' site someone said it used a 1:2:1 color filter and he didn't dispute that. He says he knows exactly what it does, his testing has revealed it, and that it's not pixel skipping.
So -- doesn't that all add up to it being different? That they're not just plucking the GH2 sensor block off the production line and shoving it in an AF100 body? Sure seems like it to me.
Barry, I still don't understand why you say it can't do both? Couldn't it do the mode it does now and still have the ability to switch to ETC mode? Couldn't it have alias free video whether it was in ETC mode or not?
I don't look at the ETC mode as something for free. I completely understand that it is not the same thing as using the same lens with 2.6x the focal length. However, it does offer me a way to get a tighter frame with low ISO videos with minimal quality loss if any. I see it as the very best cropping you can get for video. My first instinct isn't to use ETC. I will use a zoom lens first. However, when my lens runs out of reach or when Auto Focus accuracy is important to me I don't hesitate to use the ETC mode in good light.
Barry_Green
Moderator
There's no need for an "if" there. Panasonic says it doesn't, and Alan Roberts says he knows for a fact it doesn't.
Only Panasonic's engineers know for a fact what it does. All we know for a fact is that it has a video-appropriate OLPF.
How can you say that, empirically? You don't know how the AF100 sensor works. You don't know that the 12.4mpix are individually addressible. I wouldn't find it inconceivable at all that they're arranged in a permanent binning arrangement -- and if that's the case (and that's an IF), then ETC would be impossible.
Red also reads all 12 megapixels off its sensor each and every frame. We don't know that the AF100 does that. I would find it probably that it doesn't. I think it uses some sort of hardwired binning to accomplish its results.
Red also has an OLPF tuned to its 12-megapixel sensor resolution. If they tried to scan the entire frame at 3K res, they would end up with the same aliasing issues that the DSLRs have. The AF100 doesn't have an OLPF tuned to its frame size because it doesn't have, use, or need all 12 megapixels. It makes 1080p video, so it has a 1080P OLPF.
Are you certain that is the "only" reason? Seems a rather empirical statement to make, in light of the fact that you don't know what the sensor does, you don't know what it's capable of, you don't know whether it reads its whole sensor or not, and the testimony of Panasonic themselves is that it doesn't line-skip, and Alan Roberts says it doesn't line skip.
I think the core issue in our understanding here is that you seem to be stuck on that there are only two ways it can be done:
1) it skips a bunch of pixels, or
2) it reads each and every pixel individually and pulls a 4K frame off the chip for every frame
I don't know that it has to be that cut and dried. I think there are possibly alternatives. I think it is unrealistic to think that it's reading the entire 4K sensor of every individual pixel at 60 frames per second -- not even the Red Mysterium-X can do that (can it?) And that sensor, alone, costs more than the entire AF100 package. Further evidence of it not sampling the entire 12mp frame and then downscaling, is that it delivers 800 TV Lines of resolution. I would expect more if it was downscaling a full 4K 12mpix image.
I think it's more likely that it's using some sort of binning, and probably has a dedicated color filter array that has been created specifically for this sensor to support such a binning arrangement, and that through the use of that, and a video-appropriate OLPF, it delivers a read of the full chip without skipping rows or lines. And that the net image that it generates is more like 2mpix. And that, with all that combined together, it makes it obvious why ETC can't be implemented.
Of course, I don't KNOW this for a fact, this is conjecture on my part. But it is conjecture that seems reasonable, and would seem to fit the available facts and statements without calling anyone a liar or violating the laws of physics. And I fully admit that even this explanation may not be correct, they may be doing something entirely different, that only their engineers understand.
Last edited:
Barry_Green
Moderator
No. It's not possible. The AF100 is tuned to make 1080p video, only. If you scanned a window out of it, you'd end up wtih standard-def video. It could do that, but what would be the point?
The AF100 block is designed to deliver great 1080p video. The GH2's imager is tuned to deliver 16mpix stills. They're not the same.
nyvz
Veteran
Truly sorry if you feel that way. I only meant to balance out what I felt were inconsistent statements made by you and others, for the sake of those trying to get straight answers for those. I feel it is important that if someone asks in a forum any question for instance "do the AF100 and GH2 use the same sensor" that rumors and opinions are stated as rumors an opinions and unknowns are stated as unknowns. I never intended for you to feel I was saying are wrong on any of your points, I only meant to say, please if you do not know, state that you do not know. Or state the evidence, but do not make conclusions from clearly incomplete evidence as it will greatly confuse others.
It really is ok to say, "we don't know yet the connection between the AF100's sensor and the GH2 sensor, there are various similarities, but Panasonic has said they are not the same". If that is the information available, that is the information. And that is different from saying flat out "its not the same sensor". We dont have to tell people it is definitely a different sensor just because it "seems" different or because a Rep at Panasonic was told to say so or even if Alan Roberts seems sure of it but has not released any report on why that is so.
nyvz
Veteran
You're right, I meant the only reason I was currently aware of. I had not considered or heard of the idea that the sensor might be permanently binning larger groups of pixels. Sounds like an interesting idea and certainly possible. Do you know why Panasonic might choose a method like that instead of just using larger photosites? I imagine simply using fewer larger photosites instead of binning many small photosites might be more efficient at collecting light and reducing noise since it would require fewer electronics on chip? Would 6x6 binning simply have the same effect as nearest neighbor downscaling?
Barry_Green
Moderator
Yeah, sorry for my reaction -- I do think I overreacted. Sorry.
As for the evidence -- I've stated all the evidence that I have. It's enough to get me to think that they indeed are different. I don't know HOW different, but they're different.
Do you still think that there is no evidence, and that they must be the same?
That is what I've said -- but I've taken it further by pointing out the various dissimilarities that are obvious to observe. To me that adds up to "it is different." Do you think I'm premature in that conclusion?
Well, I can see what you're saying, but I ask in reverse -- do you think it's fair to say "they are definitely the same sensor" when not only are there observable differences, but the company has said they're not the same, and a rather respected individual who knows quite a bit of what he's talking about has put it to the test and confirmed that it definitely doesn't pixel-skip in the AF100 implementation?
I don't know what you're qualifying the standard of "different" at -- but to me, it all definitely adds up to "they're different." I will grant you that we don't know HOW different. The only thing we can say for certain about what the AF100 and GH2 share, is the lens mount. Any further assertions of common hardware are speculation, not facts -- and there is (at least to me) considerable evidence to point that they are not, in fact, the same.
Barry_Green
Moderator
It's just one idea I came up with. There may certainly be others.
I would have loved it if it would have been a 3-megapixel sensor with *massive* photosites on it. That would have been awesome (or, well, at least, I think it would have been awesome, but those who make the products went the path they did, for whatever reasons they chose). If I had to speculate, I'd say that the AF100's underlying silicon may indeed be shared with the GH2, and as a starting point they went with that and reworked it for video usage. But, I don't know. It just seems to be an explanation that fits the available known facts and observed evidence, but that doesn't make it definitely "the right explanation." There could be others, I wouldn't presume to guess what all techniques sensor manufacturers have at their disposal.
Of course, 3mpix would have meant we couldn't get 4:2:2 out the HD-SDI port... it probably would need more on the order of maybe 5 megapixels to really get full 4:2:2 out of a single-chip CFA at 1080p res.
As for reasons why they went the way they did -- maybe the factory that makes the silicon only had that basic wafer to go from? Maybe it was timing? Panasonic is long known for being "first" -- first with 24p, first with tapeless, first with 1080/24p, first with variable frame rates, and now first with a large-sensor videocamera. Maybe they could have developed a new 3-mpix or 5-mpix or whatever sensor, but they felt that they could get the results they were looking for from using the silicon technology that was currently available, I don't know.
Last edited:
nyvz
Veteran
Oh I never thought they were the same. I just felt it would be jumping to conclusions to say one way or the other. I just wanted to contribute to the evidence that might possibly have some bearing on whether or not they were the same sensor to balance all the evidence that was being used to convince people that they are different sensors. This of course is a bit academic since it is possible even if they are the same sensor that the AF100 could have a great benefit if it scans the whole sensor. For example a video camera with the performance of the GH2's stills mode would certainly have benefits over the GH2 in its video mode.
As someone who is myself interested in the ways a GH2, or AF100 alternatively, might be useful to me as someone who has owned a 7D and EX1 for quite a while, I'm sure I have some of my own biases that are not based on having actually used the cameras. I and a company I work for have had GH2 and AF100s on order for several months so I will have my hands on them soon. I ended up ordering a GH2 simply because of its accessibility due to price and am waiting until I can test out an AF100 because most of the samples of its gamma curves I have seen are very worrying to me due to what appear to be hard clipping and color shifting in highlights. The GH2 also worries me because of its simplistic film mode options as I am used to the extensive controls allowed by the EX1s picture style and the, in some ways less extensive but still powerful, options allowed by building custom picture style curves for the 7D.
Clearly, there is a crying need for a shootout between the GH2 and the AF100. Barry, quit stalling and do what the crowd demands 
Back in the Roman days, even the Emperors did not dare defy the Coliseum crowd, once the crowd got it in their heads that particular gladiators need to fight it out. I think the crowd has got to a fever pitch, and there is no going back. So, from the left gate, it is the GH2 with the lions against - from the right gate, the AF100 with the bears. Barry, let us have the spectacle, we will be denied no longer!!!
Back in the Roman days, even the Emperors did not dare defy the Coliseum crowd, once the crowd got it in their heads that particular gladiators need to fight it out. I think the crowd has got to a fever pitch, and there is no going back. So, from the left gate, it is the GH2 with the lions against - from the right gate, the AF100 with the bears. Barry, let us have the spectacle, we will be denied no longer!!!