Recent Comments:
Engadget's Black Friday giveaway, part two: win a pack of PSP Go accessories! {Engadget}
Nov 27th 2009 11:07AM how bout a PSP go?
Welcome to the next Engadget {Engadget}
Nov 19th 2009 10:56AM WOW comment type is friggin TINY!
Welcome to the next Engadget {Engadget}
Nov 19th 2009 10:55AM It's blocky. Like late 80's blocky.
But I guess that's retro.
Palm Treo 650 boots into Android, lives a fulfilling life (video) {Engadget}
Nov 13th 2009 10:31AM I haven't found a "To Do" replacement that's as simple, as fast and as usable as Palm's.
Caption contest: Big Four's CEOs finally united by a very large check {Engadget}
Oct 8th 2009 10:08AM They're all standing right thee and e still can get them to standardize on a short, "my voice only and beep" greeting on our cell phones.
Panasonic Lumix DMC-GF1 hands-on, with 720p test footage! {Engadget}
Oct 3rd 2009 12:20AM Well, maybe not super obvious, but there. And once you've been "sensitized" to all the different ways CMOS sensors muck up the image (lots of examples at TechThoughts.org) you'll be able to see it (and wince) very easily. It's the main reason my main HD camcorder is still a 3 CCD model. I just can't lay money down on a CMOS-based camcorder.
For a better example of it, follow the link on Tech Thoughts to The Foundry to see their software that corrects for rolling shutter- and not just for when the camera pans.
Panasonic Lumix DMC-GF1 hands-on, with 720p test footage! {Engadget}
Oct 3rd 2009 12:12AM And, interestingly, the whole point if the 4/3 system is no mirror / prism system to make it smaller.
How would you change Panasonic's Lumix DMC-GH1? {Engadget}
Sep 20th 2009 1:56PM In response to Temple...
Temple, interesting concept, except you need to address several issues you overlook:
- The broadcast industry was building 1/2" and 2/3" sensors with oversampling (way more than 720x480 pixels) back in the SD days. So it's not hard to make a large sensor.
- You say "for a dSLR-sized sensor reading out the entire frame at once is very difficult for video" which makes no sense. It's well known that at least the canon's don't use the entire sensor. They use every third line (creating aliasing artifacts) which means, from the outset, they are throwing away 66% of the image data from the imager. Since they are recording ATSC HD, the processing chip and codec is dealing with 1920x1080. No more. If a consumer camcorder for a few hundred dollars can do this, how is this hard for a DSLR? Those engineers are smart, they're not going to choke a HD processor with more data than it needs. So it's no harder for a DSLR than any other HD video camera.
The physical size of the sensor is irrelevant if the video portion (processor and codec) only compresses and records 1920x1080 pixels. It's data to everything else but the glass. The size of the sensor is completely irrelevant. In fact, if they made a "full frame" 1920x1080 sensor, there'd be so much space between the pixels they could make one hell of a lens-on-chip array (like Sony's Hyper HAD chips) and dramatically increase light gathering capability.
- Moreover, I am well aware of CMOS rolling shutter image distortions. I've blogged about it extensively at http://TechThoughts.org. The rolling shutter has to feed a buffer that hands off full frames of video to the compression engine. Codecs need to work on whole frames because they examine the entire image looking for large blocks of image they can heavily compress, 8x8 pixels, 16x16 pixels. Wavlet compression utilizes even larger areas. They can only do that if they get the entire frame as a single, solid, image, like a CCD presents it. Codecs can't compress an image a line at a time. So CMOS cameras use a separate buffer to ensure that the rest of the hardware sees frames, not lines.
- By and large, CMOS chips are less expensive to manufacture than CCDs.
- They offer more framerate and sub sampling capability than CCDs because of the way pixels are addressed. Thusfar, that feature has only been utilized for high frame rate images for a short period of time (before the camera's internal buffer- which I just discussed- fills up) and then it records it to your flash media/hard drive.
By offering a lower price and more "features" the manufacturers entice the market to buy new camcorders.
I own a CCD HD camcorder. The Sony FX1. Looks great. Canon's CCD camcorders are able to spit Full HD our HD-SDI. JVC continues to use CCDs. No, these aren't DSLR sized. But, as I already said, manufacturers have long been making big chips. It's just not cost effective for them to make a DSLR-sized CCD with only 1920x1080 pixels on it. It has nothing to do with any of the reasons you proffered, despite the fact that your reasoning is factually flawed in several places.
Anthony Burokas
http://TechThoughts.org.
How would you change Panasonic's Lumix DMC-GH1? {Engadget}
Sep 19th 2009 1:43PM Put in a CCD.
Really, the CMOS distortions in the GH1 are far worse than the canon DSLR cameras. But every CMOS chip distorts reality, and the only way to get an accurate representation recorded is to use a CCD till they start incorporating software like The Foundry has coded to correct for CMOS distortion. You can see all kinds of examples of CMOS distortion at http://TechThoughts.org.
Unless there's a CCD in it, I don't buy it. No matter how much I really dig the small size, tilt-swivel screen, audio input, interchangeable lenses, autofocus during video, etc. I'd get the Canon SX20 over this purely because it uses a CCD.
How would you change Panasonic's Lumix DMC-GH1? {Engadget}
Sep 19th 2009 1:37PM AVCHD is a very efficient codec, but not one I like at all. My Panny shoots 720p30 and will put 2 hours on a 16GB card with AVCHD, only 8 minutes with motion JPEG. I've not been able to comapre the image quality from two identical cameras shooting side by side with different codecs, but what I do know is that the MJPEG footage can be easily viewed and edited on the computer, so that's what I shoot on.
Offering the option to shoot with a high-bitrate MJPEG codec is crucial to a faster post workflow.
IEBA.com
