Would it be great if it did 2/3" cropping

Doyle

Member
I wish the AF100 had an option to run 2/3" B4 lenses on it (without using the 2X extender). And the chip could just take the portion of the image that was 2/3" and fill a 1920x1080 frame with it. Basically, it would just throw away the unused data. It would almost be like having a 2/3" broadcast camera and a DSLR all in one, with just a click of a button or flip of a switch. If you didn't want the shallow DOF of cinema...say for news or sports or events shooting, then you could just go with a good 2/3" B4 lens and run and gun. But if you did want the shallow DOF and the ability to use primes and SLR lenses, then you could switch over to the 4/3 setting. Wouldn't that be great...
 
You could do that right now, just crop in post. It wouldn't be any different than what you propose, other than, of course, having to crop it in post... but, yeah, I had asked them for something similar, I'd asked for an in-camera 720p crop for Super16 lenses, because the image circle on the sensor of a 720p extraction would be perfectly covered by a Super16 lens.

Of course, if you did what you're proposing, you do realize you'd be getting standard-def images, right? The AF100's sensor is more than 4x as large as 2/3", so if you put a 2/3" lens on there, it'll only cover less than 1/4 of the frame. And that means (at best) you'd be getting about a 800 x 500 image from the center of the sensor, and that's pretty much standard-def.
 
Good point. For some reason I was assuming that the image sensor was capable of higher resolutions and they were just using the optical low pass filter to allow it to properly scan at 1080 lines. If that was the case, then taking a 2/3" crop would not necessarily mean lower resolution than 1080 lines (if the low pass filter was bypassed in that mode)?
 
I would like to know if the optical low pass filter is an actual "filter" in front of the sensor, or if it is a term for a digital process that happens to the signal to soften it.
 
I would like to know if the optical low pass filter is an actual "filter" in front of the sensor, or if it is a term for a digital process that happens to the signal to soften it.

Would like to know too. As far I understand it is a real filter.
 
I would like to know if the optical low pass filter is an actual "filter" in front of the sensor, or if it is a term for a digital process that happens to the signal to soften it.

It is an actual filter in front of the sensor. It would be impossible to achieve the same results with a digital process since it deals with issues that arise primarily from the size and spacing of photosites which are further exacerbated by the bayer color filter/system.

I apologize if this is answered elsewhere, but as the AF100 sensor is listed as 12MP effective, is there any documentation as to whether it oversamples or line-skips for its 1080/720p? Seems like a waste to have a 12MP sensor just to line skip and then fix the aliasing with a OLPF designed for 2MP. The GH2 achieves so much sharpness without moire in normal mode and ETC (1:1) mode that it must avoid aliasing in normal mode primarily through oversampling, so I would certainly hope that the AF100 does the same. If that is the case, I dont see why the AF100 wouldnt be able to have the same ETC mode that the GH2 has, which is essentially what you are asking for, a crop that is effectively between 1/2" and 2/3".
 
For some reason I was assuming that the image sensor was capable of higher resolutions and they were just using the optical low pass filter to allow it to properly scan at 1080 lines. If that was the case, then taking a 2/3" crop would not necessarily mean lower resolution than 1080 lines (if the low pass filter was bypassed in that mode)?
What you want is easily possible on the GH2, but not on the AF100. On the GH2 there is no video-appropriate OLPF. You can window the sensor down to about 1/2" (actually it's a little bigger than 1/2" in 1080p mode) and get a full-resolution image. The tradeoff is that you can encounter severe aliasing and moire patterns when shooting with the full sensor.

On the AF100, it's not a stills camera. Everything about the imaging chain is optimized to provide 1080p video. There is no higher resolution possible. The OLPF is tuned to deliver 1080p video with drastically reduced, if not eliminated, aliasing issues. Therefore, windowing won't help any.

On the GH2, windowing is great. On the AF100, it's not possible.
 
I apologize if this is answered elsewhere, but as the AF100 sensor is listed as 12MP effective, is there any documentation as to whether it oversamples or line-skips for its 1080/720p? Seems like a waste to have a 12MP sensor just to line skip and then fix the aliasing with a OLPF designed for 2MP.
There is no documentation. We don't know what they're doing. I would doubt it is as simple as just taking a 12mp sensor and sticking a blur filter in front of it. For all we know they might have changed the whole color filter so that instead of a Bayer, it might be using an RGB stripe, or it might be using a grouped/binned pattern, we don't know.

The GH2 achieves so much sharpness without moire in normal mode
No it doesn't, the GH2 has just as much aliasing and moire in normal mode as the GH1 had.

and ETC (1:1) mode
Ah, ETC mode is very different. ETC mode is apparently a pure 1:1 scan of the sensor, with no compromises. It's a 1/2" chip at that point (or, when in 720p mode, it's a 1/3" chip). So you sacrifice large-sensor DOF, in exchange for an utterly alias-free and moire-free image.

that it must avoid aliasing in normal mode primarily through oversampling
Nope. The GH2 has extreme aliasing in normal mode. It just isn't nearly so objectionable as it is on the Sony NEX and Canon DSLR because it doesn't exhibit the color contamination and purple/orange rainbow patterns that the NEX/Canon cameras do. But the GH2 has just as much luma aliasing as they do (in 1080p mode; obviously in 720p mode the Canons are far worse).

If that is the case, I dont see why the AF100 wouldnt be able to have the same ETC mode that the GH2 has, which is essentially what you are asking for, a crop that is effectively between 1/2" and 2/3".
Doesn't work that way. The AF100 is not a stills camera. The GH2 is a stills camera designed to deliver 16MP stills, and its video unquestionably does exhibit massive aliasing when that full-sized sensor is scanned for 1080p.

The AF100 is not a repackaged GH2. The sensor is not the same, it's not the same size, it's not the same shape, it's not the same resolution, and it's been optimized solely for video. The combination of the sensor design itself, and the OLPF, produces 1080p video. That's all there is to work with. If you tried to ETC that down to 1/2", you'd end up with standard-def video.
 
we dont know how af100 makes the full sensor picture into 1920x1080. Jan confirmed it doesnt pixel skip. the olpf is there, but that prevents aliasing. if the whole sensor all 12,2 mpix (panasonic figure) were read and dowsampled at say 60fps, that should fry the plastic body and power consumption would be much higher. the actual method is not revealed, but chances are we'll find out soon.
 
No it doesn't, the GH2 has just as much aliasing and moire in normal mode as the GH1 had....

...The AF100 is not a repackaged GH2. The sensor is not the same, it's not the same size, it's not the same shape, it's not the same resolution, and it's been optimized solely for video...

Gotcha, I probably overstated the lack of aliasing in the GH2 as it does appear to be less than the 7D which I am familiar with, but I'm sure there is still quite a bit as it is clearly skipping a lot of the sensor, and probably more than some people feel would be acceptable.

You mentioned there is no documentation regarding the optical and digital formula used by the AF100 sensor, but there is documentation that the sensor differs between the AF100? I dont doubt that they developed a separate sensor for the AF100 but I didnt know it was confirmed, and it wouldnt have surprised me if they used the same sensor and used a different OLPF.
 
The specifications for the GH2 call the sensor 16 megapixels, in a 17.3 x 13 size (and I think it's 18.9 wide in 16:9 aspect ratio). The AF100's sensor is 17.8 x 10, in 16:9 shape, and 12.4 megapixels. The OLPF in the GH2 is clearly tuned for stills, the one in the AF100 has been described as having been specifically optimized for 1080p video.

I have confirmed that the field of view is not the same between the two cameras, I have both a GH2 and an AF100, and the differences in field of view are minor but they're unquestionably there.

I don't know if there is any common ancestry between the GH2 chip and the AF100 chip, but I do know that by the time they get shoved in the cameras, they're not the same.
 
What you want is easily possible on the GH2, but not on the AF100. On the GH2 there is no video-appropriate OLPF. You can window the sensor down to about 1/2" (actually it's a little bigger than 1/2" in 1080p mode) and get a full-resolution image. The tradeoff is that you can encounter severe aliasing and moire patterns when shooting with the full sensor.

On the AF100, it's not a stills camera. Everything about the imaging chain is optimized to provide 1080p video. There is no higher resolution possible. The OLPF is tuned to deliver 1080p video with drastically reduced, if not eliminated, aliasing issues. Therefore, windowing won't help any.

On the GH2, windowing is great. On the AF100, it's not possible.

I'm not sure I'd call the aliasing on the GH2 severe at all. Especially compared with our 5d or with our GH13. In fact after a fair bit of shooting now, we've yet to get hit with bad moire, something that usually pops up at least a couple of times a shoot day with the 5d. And, before people start defending the 5d, don't. We still LOVE ours and use it all the time. I'm just talking about aliasing and moire here.
 
we dont know how af100 makes the full sensor picture into 1920x1080. Jan confirmed it doesnt pixel skip. the olpf is there, but that prevents aliasing. if the whole sensor all 12,2 mpix (panasonic figure) were read and dowsampled at say 60fps, that should fry the plastic body and power consumption would be much higher. the actual method is not revealed, but chances are we'll find out soon.

So you are basically saying Jan says it doesnt line skip but that is probably not true? I would expect it either line skips and has something to the effect of a 2MP OLPF or it reads all 12MP for each frame and has a 12MP OLPF (which would mean ETC would be possible). Downsampling a 12MP image with a 12MP OLPF to 2MP would give a very sharp oversampled image with no aliasing, so there would be no need for a heavy OLPF if it is not line skipping. Seems like reading a 12MP frame without line skipping from a sensor with a 2MP OLPF would be just wasting 8MP/frame of bandwidth in the camera...
 
she says it doesnt skip pixels and I believe her. how exactly the sampling works must be some clever designs from panasonic. I'd rather hear it from them, but if not we'll know soon enough anyway.
 
The specifications for the GH2 call the sensor 16 megapixels, in a 17.3 x 13 size (and I think it's 18.9 wide in 16:9 aspect ratio). The AF100's sensor is 17.8 x 10, in 16:9 shape, and 12.4 megapixels. The OLPF in the GH2 is clearly tuned for stills, the one in the AF100 has been described as having been specifically optimized for 1080p video.

I have confirmed that the field of view is not the same between the two cameras, I have both a GH2 and an AF100, and the differences in field of view are minor but they're unquestionably there.

I guess it depends how one defines "different sensors". I agree with what you said, but it still seems possible that they both use essentially the same design, with essentially the same silicon and pixel pitch just with different active areas available. Interestingly, if you do the math on the sensors, they appear to have exactly the same pixel pitch...

GH2 [16:9] 4976x2800(L) *from spec list
GH2 pixel pitch = 18.9mm / 4976 pixels = 0.003798 mm/pix
17.8mm / 0.003798 mm/pix = 4686 pixel
4686 pixels / 1.77 aspect ratio = 2648 pixels
4686*2648= 12.4MP
 
Well -- Panasonic has said it's not the same sensor, and they've said it doesn't pixel-skip. Alan Roberts has now confirmed that through his testing, he knows for a fact it doesn't pixel-skip. Is it the exact same sensor, except that it's a different size, different res, different OLPF, different field of view, different "blue band" performance, different aliasing, different color filter, different pixel-reading method, different moire, and whatever else is different about it? Maybe. But that all sure adds up to me to be something that sounds, well... different. :D
 
Back
Top