Kevin Randall
1
COW Coin
Forum Replies Created
-
Pixel shift is not the same as uprezzing; they are two entirely different things. The HPX170 takes the image off the CCD block as analog data, and it is quantized at 1080; all processing is done at 1080, and recording is downsampled from that. “Uprezzing” normally refers to taking a digital image and interpolating new data to reproduce that image at a higher pixel rate.
It has been reported, repeatedly, by people such as Barry Green (who literally “wrote the book” on the HVX) that recording at 1080 gives sharper images, with less compression, than recording at 720. This would probably be why some people have said that they get better 720 results by recording at 1080 and converting to 720 themselves, vs. recording at 720.
I’m not doubting that you’re a fan of the cameras, Michael. But it is not helpful to give out erroneous information.
-
Michael: The “chip” is not a 720p chip. The CHIPS each have about 520,000 photosites (960×540), with the green chip slightly offset from the red and blue. The image is taken from the CCD block as a 1920×1080 image. It is recorded at either 1280×1080, 960×720, or 720×480, depending on the recording format chosen. There is no “uprezzing” done to record a 1080p image. ALL images start out as 1080p, then are processed according to the recording format chosen (1080i, 1080p, 720p, 480i or 480p).
-
Be aware that deciding whether to record to 1080 or 720 on the HVX200 or HPX170 has absolutely nothing to do with the pixel size of the CCDs. ALL image capture on these cameras is at 1080p, no matter what format/resolution you are recording to. Repeat–ALL image capture on these cameras is at 1080p. If you are recording at 720, the original image is actually being DOWNSAMPLED to 720. I have heard on some discussion boards that some people have even had better results recording at 1080 in camera, then downrezzing to 720 in their NLE, vs. recording at 720 in camera.
-
Are you sure that progressive is what you’re really looking for? What shows on HGTV are you thinking of? Although I have very rarely watched anything on that network, I have a feeling that most of the programs are shot as interlaced, not progressive.
-
1080 and 720 will both give the same “look”. The only difference in the resulting image is resolution. You will get higher resolution in the 1080 image. If the higher resolution is not necessary, you can get more record time on your recording media by going with 720 native. On the other hand, I have heard some people report that they get slightly better results at 720 if they record at 1080 and then render at 720 in in postproduction. I am supposing that is because whatever they are using for rendering their final product does a better downconvert than the camera’s internal downconversion. (Bear in mind that the camera is working with a native 1080 image coming off the chips; when recording at a lower resolution, the camera is downsampling the original image to give you the desired recording format.)
-
Just to clarify: The frame rates that end in “p” (e.g., 1080i/25p) do record true progressive images. They are only encoded in DVCPro HD as interlaced fields. Don’t worry thinking that the camera is giving you interlaced footage; it isn’t. It’s true progressive footage, just in an interlaced package. Any NLE capable of handling DVCPro HD footage should be able to treat it as progressive.
-
It’s called 1080i/30p because it is 30fps Progressive recorded in an interlaced stream. DVCPRO HD is an interlaced only format in 1080, recording interlaced streams of 60 fields per second; there is no such thing as 1080p in DVCPRO HD. However, the images being recorded are truly progressive; they are merely being laid down to tape or card or disc in an interlaced stream. Each frame is recorded as two or more fields, depending on the frame rate. For 30p, every frame comprises two fields; for 24p, some frames are three fields.
(This explains why you can’t get 60p in 1080 DVCPRO HD. Each progressive frame requires two interlaced fields, so you would need 120 fields per second to record the 60 progressive frames; there are only 60 fields available in the format.)
-
Misinformation about the so-called “native format” of the HVX200 is constantly being given. Dan Brockett had it right: 1080 will give you the highest resolution possible from the camera. That is the resolution that ALL images–repeat, ALL images–start out in. When you record in 720, you are recording a down-rezzed image from the original 1080 image the camera captured.
-
Please resist repeating the misinformation about the resolution of the images created/recorded by the HVX200.
*Everything* output by the camera begins as a 1920 x 1080 progressive image. Doesn’t matter if you are shooting progressive or interlaced, or at standard def, 720, or 1080. It all starts out at 1920 x 1080 progressive. The image is derived from a block of three CCD chips, each 960 x 540, with the green chip slightly offset in relation to the red and blue chips.
The image is derived from the chips as a 1920 x 1080 progressive image, and then is processed for recording according to the format chosen by the user. If the user is recording in 720, the image will be downrezzed to 720; if the user is recording standard def, the image will be downrezzed to the appropriate vertical resolution in NTSC or PAL. Some users have done tests comparing the 1080 and 720 recordings from the camera and found they were able to get slightly better 720 footage by recording 1080 and downconverting in other software, versus recording 720 in-camera.
(All of the above is totally separate from the question of how well the images from the HVX compare to images generated from other cameras, or whether the images from the HVX would be better if the chips had 1920 x 1080 photosites versus 960 x 540 photosites. The reality is, however one judges the quality of the images, for better or worse, when they begin their “lives” as digital images assembled from data coming off the chips inside the camera, they are 1920 x 1080 images. Any other resolution chosen by the user is a downrez.)
If the user has chosen to shoot interlaced, the images will be recorded as interlaced. If the user has chosen progressive, the images will be recorded as progressive, but in most formats laid out in an interlaced stream with pulldown. (They are true progressive images, “re-assembled” from the interlaced recording format, in the NLE.)
A detailed discussion of this matter can be found on the DVXUSER site: https://www.dvxuser.com/V6/showthread.php?t=96274
-
Vince Becquiot said:
“Some have argued about the benefits of 1080 over 720. I tried both and I can’t see much of a difference. In fact the pixel shifting may hurt you on the edges in the end (that one’s not coming from me).”
While I have not done chromakeying myself, it has been reported on DVXUSER site that 1080 works best. And this only makes sense, since when recording in 720, the original 1920 x 1080 image is downsampled to 960 x 720. When recording in 1080, the image is downsampled only in the horizontal (from the original 1920 x 1080 to 1280 x 1080).