Uwe Lansing
Veteran
I`d say the sensor of the 5D is four times larger than that of the GH1:
http://www.wrotniak.net/photo/43/sensor-size.html
Panasonic unveiled G1-HD...
- Thread starter Uwe Lansing
- Start date
- Status
I`d say the sensor of the 5D is four times larger than that of the GH1:
http://www.wrotniak.net/photo/43/sensor-size.html
Uwe Lansing
Veteran
Huh, do you have any other problems or is this the only one? ;-)
Uwe Lansing
Veteran
Ah, I see. Then I´ll continue to be precise 
)
I`ve heard that the 5DII encodes with a data rate of 38mbps (inclusive audio = 41mbps). Do you think, that could be right?
)I`ve heard that the 5DII encodes with a data rate of 38mbps (inclusive audio = 41mbps). Do you think, that could be right?
it has four times the area, sure, but sensors are always measured in one dimension. not only is this the norm, it's also the best way to determine field of view as well as dof and other optical properties.
/matt
Uwe Lansing
Veteran
Isn´t it true, that the full area of the sensor determine the size of each pixel. And the larger each pixel the better the lowlight abilities for example...
that's not the point. the point is that the video image isn't as high quality to begin with, for various reasons.
/matt
DavidNJ
Carbonite Member
The 5D Mark II slides upto about 45Mb/s...Canon lists it at 42.5. Quality wise it appears to be way ahead of on camera recording compared to the GH-1...but we have to wait for back to back tests.
There are lots of stories about what is processed. It is thought by some very knowledgeable people (Red forum) that it scans every third row. That it would be impossible to scan a 18M photo sites 30 times a second. That would make it 5760x1080 but with a loss of light and vertical noise. Canon hides the gain control and uses have reported noise with examples on vimeo.
We don't know what the GH-1 is scanning but it is widely believed to be full frame for lots of reasons: no easy subsample, 4 channels from the sensor, dual processors, and support for contrast auto focus. Therefore even though 1/4 the area, it is likely/possible the GH-1 will capture a better image. Full frame would be 3840x2160 probably. That would mean that each pixel in output is sampled once for red and blue and twice for green, effectively removing any advantage of a 3 sensor system.
More info in the other thread though...
There are lots of stories about what is processed. It is thought by some very knowledgeable people (Red forum) that it scans every third row. That it would be impossible to scan a 18M photo sites 30 times a second. That would make it 5760x1080 but with a loss of light and vertical noise. Canon hides the gain control and uses have reported noise with examples on vimeo.
We don't know what the GH-1 is scanning but it is widely believed to be full frame for lots of reasons: no easy subsample, 4 channels from the sensor, dual processors, and support for contrast auto focus. Therefore even though 1/4 the area, it is likely/possible the GH-1 will capture a better image. Full frame would be 3840x2160 probably. That would mean that each pixel in output is sampled once for red and blue and twice for green, effectively removing any advantage of a 3 sensor system.
More info in the other thread though...
Bitrate is not a measure of quality-- the reason you've seen great images from the 5D is that Canon hired an expert to light and shoot some great demo footage. If fed the same source data, the GH1's 17Mbps stream would result in higher quality final images than the 5Ds 40Mbps.
The 16:9 image frame for the GH1 is 4352x2448 pixels, which is downsampled to an HD frame. Think of the GH1 is a 4k red camera that downscales to HD before storing the footage, rather than giving you RAW frames.... that's the image quality advantage it has over the 5D, which only takes a 2k sample to produce its frames.
The 16:9 image frame for the GH1 is 4352x2448 pixels, which is downsampled to an HD frame. Think of the GH1 is a 4k red camera that downscales to HD before storing the footage, rather than giving you RAW frames.... that's the image quality advantage it has over the 5D, which only takes a 2k sample to produce its frames.
DavidNJ
Carbonite Member
There are wonderful images from just about everyone who shot with the Canon...they are all over the Internet. Everyone gets great images with the Canon. And it is simple to understand why: Great lens x great sensor x great processing x great recording. Note you can get very good images from any camera by adding great recording. Evidience: SDX-900 (50Mb/s SD MPEG-2 I-Frame), HR11 via uncompressed HDMI to Cineform (100-160Mb/s HD wavelet), etc.
As a rule of thumb, you can NEVER win an argument my telling someone else what they have seen or know. It just detracts from your argument.
Bit rate counts for a lot, the rest is the quality within that bit rate. You have NEVER substantiated any of your claims. And all analysis of published information (see chroma artifacts of 24p pulldown on 60i) indicate Canon was smarter than Panasonic. Differences between AVCHD and MPEG-2 are debated. The difference between two H.264 High Profile@Level 4.1 protocols are not going to exceed a 2.5x difference in bandwidth.
Where did you get your 4352x2448 number from? That is the STILL picture recording resolution. However if you took 4352 by the max vertical recorded resolution (3000) you'd have over 13M photo sites, where the camera only claims 12.1 Mp effective. However, still possible (14M total photo sites).
Otherwise either is possible. I presume the smaller since a) it is within specification error on sensor size for bokeh expectations, and b) it makes infinitely easier down converting if you don't have to interpolate between photo sites. In fact, the camera would have at least one photo site for each sensor making down converting nearly trivial. Four photo sites per pixel in 1080p mode, 9 in 720p mode.
Now, without name calling, what is your EVIDENCE that 4352x2448 is being processed for the 1920p image?
As a rule of thumb, you can NEVER win an argument my telling someone else what they have seen or know. It just detracts from your argument.
Bit rate counts for a lot, the rest is the quality within that bit rate. You have NEVER substantiated any of your claims. And all analysis of published information (see chroma artifacts of 24p pulldown on 60i) indicate Canon was smarter than Panasonic. Differences between AVCHD and MPEG-2 are debated. The difference between two H.264 High Profile@Level 4.1 protocols are not going to exceed a 2.5x difference in bandwidth.
Where did you get your 4352x2448 number from? That is the STILL picture recording resolution. However if you took 4352 by the max vertical recorded resolution (3000) you'd have over 13M photo sites, where the camera only claims 12.1 Mp effective. However, still possible (14M total photo sites).
Otherwise either is possible. I presume the smaller since a) it is within specification error on sensor size for bokeh expectations, and b) it makes infinitely easier down converting if you don't have to interpolate between photo sites. In fact, the camera would have at least one photo site for each sensor making down converting nearly trivial. Four photo sites per pixel in 1080p mode, 9 in 720p mode.
Now, without name calling, what is your EVIDENCE that 4352x2448 is being processed for the 1920p image?
That's a pretty big assumption, seeing as there's no way to substantiate that until the GH1 is released. I'll believe it when I see it, and hopefully Panasonic ups that 17Mbps to 24Mbps before the camera is released. 17Mbps just does not make sense when cheaper consumer cams are utilizing 24.
ESTEBEVERDE
Veteran
It does if they are protecting their video camera segment...
PappasArts
Veteran
It shouldn't look like their video cameras. The 5D, which I have atleast 30gigs of footage is way and beyond looking like HDV videocams. The bad 5d material is from errors of the user.
In reality GH1 video should look like it photo's down rezzed to 1920x1080 with the approiate compression settings. Did this with the 5D footage. Compared stills that get downsized to 1080P and recompressed to have the same bit rate. When compared to the video still from timeline taken at the same moment, pretty much indentical.
I suggest you go over to DPreview.com, download some G1 images, crop to 1920X1080 and save out what the bit rate per frame should be in AVCHD. You should have what the video should look like from the G1.
G1 gallery: http://www.dpreview.com/reviews/panasonicdmcg1/page34.asp
Considering the video is taken from the 14.0 million cmos ( 12 million effective ) large sensor ( unlike 1/3 video camera sensors ) and not cropped. It should for example look like Red 4K down-rezed to 1920x1080 and saved out to AVCHD. No reason for it not to. It's should, however who knows if they decide to cripple it....
Michael Pappas
http://www.pbase.com/Arrfilms
http://www.Myspace.com/PappasArts
http://www.facebook.com/people/Michael-Pappas/573417404
Arrfilms@hotmail.com
http://www.PappasArts.com
CONTACT VIA AOL INSTANT MESSENGER
AT { PAPPASARTS2 }
In reality GH1 video should look like it photo's down rezzed to 1920x1080 with the approiate compression settings. Did this with the 5D footage. Compared stills that get downsized to 1080P and recompressed to have the same bit rate. When compared to the video still from timeline taken at the same moment, pretty much indentical.
I suggest you go over to DPreview.com, download some G1 images, crop to 1920X1080 and save out what the bit rate per frame should be in AVCHD. You should have what the video should look like from the G1.
G1 gallery: http://www.dpreview.com/reviews/panasonicdmcg1/page34.asp
Considering the video is taken from the 14.0 million cmos ( 12 million effective ) large sensor ( unlike 1/3 video camera sensors ) and not cropped. It should for example look like Red 4K down-rezed to 1920x1080 and saved out to AVCHD. No reason for it not to. It's should, however who knows if they decide to cripple it....
Michael Pappas
http://www.pbase.com/Arrfilms
http://www.Myspace.com/PappasArts
http://www.facebook.com/people/Michael-Pappas/573417404
Arrfilms@hotmail.com
http://www.PappasArts.com
CONTACT VIA AOL INSTANT MESSENGER
AT { PAPPASARTS2 }
DavidNJ
Carbonite Member
Pappa Artts, yes and no.
The picture is a function of many things. On the Canon you are looking at three, in order of importance:
1) AVCHD long GOP at 45Mb/s is better than anything we normally see other than RED, direct Cineform capture, the XDRs 100/160Mb/s MPEG2, and AVC Intra 100, although that last one will be close, depending on the footage. Other than the HPX300 (not yet out) and the HPX 500, we don't normally use cameras with this recording capability. And most of don't use an HPX500.
2) It is an oversampled sensor. Other than the D90 and the Red One, we haven't used any other camera that is spatially over sampled. One key here is that the optical and digital low pass filters for the sensor and the sensor's Nyquist limit, at least horizontally, are way beyond our image resolution. In our 27lp/mm (540lp/ph) world, this is the first sensor capturing high contrast at the higher range of frequencies.
3) That is a big sensor, so the diffraction limit is way up there. We are used to small lens that have steep MTF diffraction limits. With the big sensor, and often good, fast glass, when we get low MTF it is called intentional bokeh.
The GH-1 seems to have thrown the first out the window, Lisa still waiting to substantiate any of her claims. We all wish she was correct, but doubt she is. Evidence Lisa, evidence!
Our sole hope is uncompressed 4:2:2 HDMI output. However, that will still cost us $2=4k extra to capture.
On the second, it probably beats the Canon. The sensor is 1/4th the size. But it is apparently scanned better than the Canon and higher resolution in the vertical axis. The frequency is high enough above the Nyquist limit that we should be in a good part of its MTF curve.
On the third, it is smaller, but not that much smaller. And the Panasonic glass is acceptably fast. Manual prime lenses with F1.2-F2 apertures are available, so we should be able to shoot with nearly as clear as the Canon.
Net, if we get the HDMI out or Lisa is right, the GH-1 will make the 5DM2 so yesterday for serious video work. The overall package prices, with HDMI capture on the GH-1 (not possible ont the 5DM2) are comparable.
P.S.
In comparing prices I didn't include transcoding 160Mb/s MPEG2 to the equivalent speed Cineform, and the Core i7 system to work with it!
The picture is a function of many things. On the Canon you are looking at three, in order of importance:
1) AVCHD long GOP at 45Mb/s is better than anything we normally see other than RED, direct Cineform capture, the XDRs 100/160Mb/s MPEG2, and AVC Intra 100, although that last one will be close, depending on the footage. Other than the HPX300 (not yet out) and the HPX 500, we don't normally use cameras with this recording capability. And most of don't use an HPX500.
2) It is an oversampled sensor. Other than the D90 and the Red One, we haven't used any other camera that is spatially over sampled. One key here is that the optical and digital low pass filters for the sensor and the sensor's Nyquist limit, at least horizontally, are way beyond our image resolution. In our 27lp/mm (540lp/ph) world, this is the first sensor capturing high contrast at the higher range of frequencies.
3) That is a big sensor, so the diffraction limit is way up there. We are used to small lens that have steep MTF diffraction limits. With the big sensor, and often good, fast glass, when we get low MTF it is called intentional bokeh.
The GH-1 seems to have thrown the first out the window, Lisa still waiting to substantiate any of her claims. We all wish she was correct, but doubt she is. Evidence Lisa, evidence!
Our sole hope is uncompressed 4:2:2 HDMI output. However, that will still cost us $2=4k extra to capture.
On the second, it probably beats the Canon. The sensor is 1/4th the size. But it is apparently scanned better than the Canon and higher resolution in the vertical axis. The frequency is high enough above the Nyquist limit that we should be in a good part of its MTF curve.
On the third, it is smaller, but not that much smaller. And the Panasonic glass is acceptably fast. Manual prime lenses with F1.2-F2 apertures are available, so we should be able to shoot with nearly as clear as the Canon.
Net, if we get the HDMI out or Lisa is right, the GH-1 will make the 5DM2 so yesterday for serious video work. The overall package prices, with HDMI capture on the GH-1 (not possible ont the 5DM2) are comparable.
P.S.
In comparing prices I didn't include transcoding 160Mb/s MPEG2 to the equivalent speed Cineform, and the Core i7 system to work with it!
PappasArts
Veteran
bronxjragon
Well-known member
ha saw that coming long time ago
- Status