View Full Version : A1U review
Radek Svoboda June 24th, 2005, 01:49 AM http://videosystems.com/mag/video_scaling_down_cmos/
14 bit ADC, 5% reduction in chip pixel count in Super Steady Shot mode, CMOS does not need ND filters, digital still camera mode. Sony makes 12.4-megapixel CMOS sensor for Nikon's top-of-the-line digital SLR, has been shipping CMOS sensors since 2003. Spot focus and spot metering anywhere on the screen, same viewfinder as Z1, only 123K LCD, available early fall.
There is 5% pixel rediction in pixel count in SSS. This is very interesting. IMHO is used for horizontal only. There are already plenty extra vertical pixels; the vertical pixel count is 1440 and only 1080 are needed for image. Can someone comment on this? Is 5% horizontal pixel count for anti-shake enough?
Radek
Mike Farrington June 24th, 2005, 06:18 AM Is 5% horizontal pixel count for anti-shake enough?
Since this camera is geared toward the professional and not the home user, perhaps they felt 5% was the best compromise. On one hand, you would assume that a professional would do a better job at steadying their camera. On the other hand, the reason they might get this version over the FX1/Z1 might be to have an smaller 'run-n-gun" camera.
I feel like I should add a signature line for myself that says: Warning -- These are the ramblings of a complete newbie. While I'm quite the technonerd, the HC1 will be my first video camera since using my family's betamax camcorder in the 80's.
-Mike
Radek Svoboda June 24th, 2005, 07:51 AM This occured to me. The total is 5% reduction. There could be big horizontal reduction and big vertical addition, making total 5% less. What is interesting is that camera lens is 10:1 zoom with 40-480 range, from what remember. Maybe without video stabilization it is 40-400 and with stabilization 48-480. That would mean about 17% reduction in field of view. Could someone comment with opinion?
Radek
Michal Laskowski June 24th, 2005, 01:17 PM From article:
which requires 5 percent of the available CMOS real estate around the edges for image movement, leaving fewer than 1920×1080 pixels to form the HDV image
I still have hope that it's not true. It will be for me unbelievable that Sony could do something like that (is it really so hard to add additional 5% of pixels at the edge of sensor). I hope that there is 5 percent of pixel edges that are not counted in 1920x1080.
If the statement form article is true it'll be the bigest dissapointment at camcorder's market since many years ;) and I won't buy this camcorder. I'm already collecting money for HC1 - I need more that my 3 months salaries (PAL version :( ). It would be too much for decreased quality.
Regards
Thomas Smet June 24th, 2005, 02:20 PM Would it really make a huge difference going down to 1828 pixels for steady shot when HDV only uses 1440 of those anyways?
I mean a lot of people have a hard time telling the Z1U from a cinealta camera let alone telling how 92 missing pixels will look when they are sampled down to 1440.
Besides couldn't you turn off the steady shot? I have never used a camera with electronic steady shot so I do not know if that is something that can be turned off.
Even with steady shot reduction that is still a heck of a lot more pixels than most other HD cameras.
Steven White June 24th, 2005, 02:21 PM I wouldn't worry about the stablization. The 1920x1440 chip has plenty of buffer pixels in the vertical dimension, and a 5% loss would barely be noticable. Remember, it's got to down-sample its 1920 horizontal to 1440 anyway. That's a whopping 33% compared to the 5% they're talking about. Also, 1440-5% is still greater than the 1280 DVCPRO-HD will give.
The interesting comment to me was:
"With the very high dynamic range of CMOS and EIP, the A1 does not need the ND filter"
I'm also waiting to see what it's 24P capabilities are. Everything I've read about CMOS says they don't need to cheat if they don't want to.
-Steve
Mike Farrington June 24th, 2005, 02:56 PM For professional audio, an optional dual XLR input module attaches to a shoe above the lens.I was assuming that the XLR input box was standard. So the $1500 increase in price over the HC1 is mostly for the optics and DVCAM capabilities? The XLR audio inputs cost extra?
-Mike
Michal Laskowski June 24th, 2005, 03:15 PM Besides couldn't you turn off the steady shot? I have never used a camera with electronic steady shot so I do not know if that is something that can be turned off.
I've got Optura 40 camcorder with electronic stabilization. The algorithm is almost perfect, diffrence is huge with stabilization on and off. For me it's not a go without stabilization, it must be.
I wouldn't worry about the stablization. The 1920x1440 chip has plenty of buffer pixels in the vertical dimension, and a 5% loss would barely be noticable. Remember, it's got to down-sample its 1920 horizontal to 1440 anyway. That's a whopping 33% compared to the 5% they're talking about. Also, 1440-5% is still greater than the 1280
I think we should a bit concered. I don't think it's only 5%, it's as much as 5%. 5% from left, 5% from right gives 10% = 192 pixels. To keep proper aspect ratio we get vertical resolution (1920 - 192) * 9 / 16 = 972. While horizontal resolution is not a big deal, because it's squeezed to 1440, the vertical resizing from 972 to 1080 is completely unaceptable - the content is interlaced, so the decrease in quality is enormous. It would be OK if only it would cost 600$, but for 2000$ (or 3000$ for PAL version) I think it would be a joke ("would" because I still hope it's not true ;) ).
Regards
Thomas Smet June 24th, 2005, 06:32 PM The vertical chip has 1440 pixels and not 1080 so there is plenty of buffer room for even 10%. Come on people even 1745 (10%) x 1080 (no loss since only the middle 1080 of 1440 pixels is used) is better than a 960 x 1080 chip in the FX1/Z1 using pixel shift to jump up to 1440x1080. No recording format except for HDCAM SR can actually record that many pixels. This is the highest pixel count we will find in a camera in this price range. The only two camera that will have a higher pixel count will be the other JVC HD camera that will have 3 1920x1080 2/3" CMOS chips and the SONY 950 Cinealta camera. I know this camera is only a single CMOS chip but this is a lot more resolution than we have ever hoped for in a HDV camera. Heck I would have been just happy with a natural 1440x1080 camera since the recording format can only use 1440x1080 pixels. Anything above 1440 is kind of lost anyways and only helps for a very very very very slight amount of sharpness. Good luck being able to tell the difference between 1745 and 1920 on a HD display. Especially considering that there are only a handfull of displays that can actually show 1920 pixels.
Thomas Smet June 25th, 2005, 01:36 AM Try this little test for those who cannot understand how much this slight reduction in resolution doesn't really matter.
Scan an image at 1920x1080 and scan again at 1750x1080. The first image scale down to 1440x1080. The 2nd image scale down to 1440x1080. Now try to see the difference between the two. As much as I would also love to have a true 1920x1080 image the fact is that at that high of resolution a small loss doesn't matter as much as it did with SD.
Radek Svoboda June 25th, 2005, 01:42 AM Relize that camera employs RGGBGG pattern Bayer filter, because it is single chip. If was 3-chip camera with alligned pixels, you get 1920 horizontally. Bayer filter image has to be processed, you lose resolution by so. FX/Z1 employ horizontal pixel shift in green channel, pixel shift gain resolution.
Radek
Michal Laskowski June 25th, 2005, 03:44 AM Try this little test for those who cannot understand how much this slight reduction in resolution doesn't really matter.
But why you not include aspect ratio? There is plenty room in vertical direction (1440, we need only 1080), but if we decrease horizontal resolution from 1920 to 1728 then we have to change also vertical resolution (not because of stabilization, but because of aspect ratio - 16:9)! So we can't squueze 1728x1080 to 1440x1080, but we have to squeeze 1728x972 to 1440x1080. And here is the problem - read my previous post about interlacing.
The solution could be not square pixels (rectangular), but because HC1 and A1 have still function we know the pixels are square :(
Radek Svoboda June 25th, 2005, 05:32 AM What would interesting to find, if indeed focal length of lens increases by 20% with stabilization. If is the case, then available pixel count in both horizontal and vertical direction be 20% higher without stabilization. If focal length does not change, resolution stays same.
Radek
Philip Williams June 25th, 2005, 09:23 AM First of all, I think its a bit weird that Sony would release a 2K and 3.5K camcorder (SD or HD) without OIS or, at the least, EIS that uses spare CCD/CMOS pixels.
Having said that, I guess we can theorize all we want to, but we just won't know for sure how much the image is hit by the EIS until the camera comes out and someone shoots some resolution charts with EIS on and off.
If worse comes to worse, then the low-budget film/doc makers seeking the best HD image from a 2K camcorder will use this with EIS off. Under controlled environments with tripods and hand held stabilizers that'll work out nicely (I'm quite certain that the average Hollywood 35mm lens doesn't have IS). With well crafted footage turning off EIS is actually often advantageous because it gets rid of that "floating" feeling sometimes associated with EIS.
Frankly, I think this might all be a mute point. I'd bet a dollar that (area 51 stuff here) Canon will release an HDV Optura Xi replacement late this year. OIS, better manual controls and probably MSRP one or two hundred less than the Sony. And I have a sneaking suspicion that Panasonic will release something in this price bracket (possibly GS400 replacement?) with 24P (not too soon though).
Hey, its all good. We've got entry level HD cams coming out at 2K. The creative types will work with the limitations and probably come up with some great footage. Exciting times.
Philip Williams
www.philipwilliams.com
Chris Hurd June 25th, 2005, 01:29 PM Bayer filter image has to be processed, you lose resolution by so.That is a misleading statement which is not at all correct. In the past, Bayer patterns *used* to cause a *slight* decrease in resolution but no longer. Advancements in DSP technology these days pretty much cancel out any resolution loss. At one time this was an issue, but not anymore.
Wayne Morellini June 25th, 2005, 10:24 PM Shoot a pure blue or red pattern. In matter of fact shoot somebody with pure blue/red pattern on their clothing. The bayer relies on things having some washout, or change from pixel to pixel, so the algorithms can correctly predict the pixel changes, cut out green and you might as well shoot with 3chip DV. But as this is 4:2:0, not 4:4:4, this doesn't really matter that much.
I think the truth is that everything is a compromise so some people will buy the next model up. But for point and shoot, great. Philip, yep for doco the loss don't matter, for film they can use handheld stabilisers and tripods.
Radek, I would like to say on the behalf of everybody, thanks for your continuing efforts to bring us news on these cameras.
Wayne Morellini July 29th, 2005, 02:08 AM I'm actually not right here, and have come up with another method I am discussing over in alternative imaging:
http://www.dvinfo.net/conf/showthread.php?p=339561#post339561
Stephen Finton July 29th, 2005, 08:25 AM The resolution of HDV is 1440 x 1080. You guys keep on writing about 1920 x 1440...
1440 is stretched to 1920. This is the HDV standard. It's how they fit it in a 25Mbps stream.
The 5% reduction in pixels would be both horizontal and vertical if we are talking about electronic image stabilization but I thought the HC1 had optical...
Steven White July 29th, 2005, 08:32 AM 1920x1440 is popping up because that's supposed to be the resolution of the CMOS sensor in the HC1/A1U.
-Steve
Stephen Finton July 29th, 2005, 12:28 PM 1920x1440 is popping up because that's supposed to be the resolution of the CMOS sensor in the HC1/A1U.
-Steve
That's weird. So are they downsampling or are they just ignoring the extra pixel count until you switch over to take a picture?
Douglas Spotted Eagle July 29th, 2005, 01:20 PM That's weird. So are they downsampling or are they just ignoring the extra pixel count until you switch over to take a picture?
The extra pixel area is ignored with video. I should have some images of this up soon.
|
|