Multimedia
Blogs
Forum
Affordable HD
Whitepapers
Newsletters
Events
Advertisers
Blogcast
MillimeterReviewing Sony HVR-V1U HDV Camcorder
Mar 12, 2007 12:00 PM, By Steve Mullen
Sony shooters have long wanted an HDV camcorder that records 24p video. In December 2006, Sony began shipping the Region 60 HVR-V1U HDV camcorder. The V1 is relatively small and lightweight: 12.8"x7.2"x5.2i" and 3.6lbs. The MSRP for the HVR-V1 is $4,890.00.
Sony 3ClearVid System
The HVR-V1U employs three 16:9 3ClearVid CMOS sensors; each 1/4in. chip has 1.03 million effective pixels. The chips are always progressively scanned at either 60Hz or 48Hz (Region 60) or 50Hz (Region 50). Unlike other CMOS chips, 3ClearVid features elements that have been rotated 45 degrees.
A 3ClearVid chip has 1080 staggered rows. Each row has 480 elements, so each pair of rows has 960 elements. Likewise, the chip has 960 staggered columns. Each of the staggered columns has 540 elements, so each pair of columns has 1080 elements. The chip's diamond pattern allows the distance between elements to be smaller, thereby increasing spatial resolution. According to Sony, four elements-from four columns-are read simultaneously from each 3ClearVid chip into each of the EIP's three 2-million-cell buffers. By outputting four elements at a time, read-out speed is increased by a factor of four, thereby significantly reducing the rolling-shutter artifact.
The output from the 3ClearVid chips is an analog signal that is converted to digital by a 14-bit A-to-D converter in the Digital eXtended Processor (DXP). The DXP chip uses 14-bit words for baseband signal processing. Sony’s 3ClearVid technology is tightly integrated with Sony’s Enhanced Imaging Processor (EIP). According to Sony, the EIP works at 1920x1080 in a 4:2:2 color space. The EIP has a 2 million-pixel buffer for each of the three primary colors. Traditionally the image sensor is considered the device in which an image forms and from which it is captured. I consider the 6 million-cell EIP buffer to be the V1’s “virtual sensor.”
After each capture from the CMOS chips, each buffer has 1 million samples — leaving 1 million buffer cells empty. Interpolation is used to fill the 1 million empty cells in each EIP buffer. According to Sony, interpolation is performed by calculating the average value of the four adjacent CMOS samples. Sony’s diagram, below, shows four empty cells (red dots) filled by using the four samples (open circles) on all sides of empty cells. The result is four calculated pixel-filled circles. (A 1.33:1 pixel aspect-ratio supports the 16:9 chip aspect-ratio.)
Image Creation
After interpolation, each buffer row has 1920 data points, composed of 960 CMOS samples alternating with 960 interpolated pixels. Line 2 is shown.
Each synthetic (interpolated) pixel contains information from adjacent samples, making them spatially “wider,” thereby reducing the resolution of these 960 pixels. Nevertheless, they enhance the horizontal resolution of each video line, as they're created from a mix of synthetic pixels and CMOS samples. Before MPEG-2 encoding, each line is scaled from 1920 to 1440 data values. After encoding, 60 fields per second are recorded to MiniDV tape (or to hard drive).
Likewise, after interpolation, each column has 1080 data points, composed of 540 CMOS samples alternating with 540 synthetic pixels. Line 2 is shown.
Each synthetic pixel also contains information from adjacent upper and lower samples, making them spatially "taller," thereby reducing the resolution of these 540 pixels. Nevertheless, they enhance the vertical resolution of each video column, as it is created from a mix of synthetic pixels and CMOS samples. I'll discuss resolution more when we look at the V1’s performance.
(Note: Sixth Row has Timecode of Each Frame in 24A Mode)
Click here for a larger image
Image Manipulation
An example of EIP manipulation is 24p recording, where the camera switches to a 48Hz scanning rate. Every other capture is discarded, yielding 1920x1080p24 video. Using industry-standard 2:3 pulldown, every four progressive video frames become 10 interlaced video fields. In this way 24p is “carried within” 1080/60i video. Two “judder” frames are generated that contain one field from one frame and one field from an adjacent frame (represented by the red text in the Table below.)
When you watch 24p video, you see a mix of 24fps motion judder and 2:3 cadence judder — just as you do when watch film converted to video using 2:3 pulldown.
The V1 offers two 24p modes: one for those who plan to edit a 24p production using an NLE, and a mode for video shooters who want 60i video with 24fps motion judder. The first mode, “24A,” forces every clip to begin with an “A” frame. When playing 24A, brief pauses may occur at the end of clips. To prevent pauses, the latter mode, “24” mode, does not force clips to begin with “A” frames.
Ideally, your NLE will have the capability to remove 2:3 pulldown. Currently, only Sony Vegas 7, Grass Valley Edius 4, and Adobe Premiere Pro 2 with CineForm’s Aspect HD can do this. Vegas supports native HDV editing, while Edius and Premiere support intermediate codec editing. (I used both Vegas 7 and Edius 4 to edit 24p with the Sony V1U.) Of course, you can also edit 24p video as 60i video, but you should avoid cuts that begin with judder frames. I used this process with FCP to create a 1080i60 (with 2-3 pullup) red-laser disc that plays on HD DVD players.
To record 30p, the V1U discards every other 1920x1080p60 frame. The odd and even lines of the non-discarded frame are encoded and recorded as consecutive fields. When these fields are viewed as frames, they will have 30fps motion judder.
Continue the discussion on “Crosstalk” the Millimeter Forum.
