Optical-disk-based video recorders using DVD-rewritable formats have become available and video recorders using current optical-storage devices have been proposed based on formats like DVD+RW/+R, DVD-RAM and DVD-RW/-R. However, the data rate and capacity of a single-layer DVD disk are limited, and too small for storage of full-resolution high-definition TV signals, which have a data rate of 24 Mbits/second.
One of the options to increase the capacity and data rate for HDTV broadcasts is to use the advanced optical rewritable system known as DVR-blue, with its large capacity and data rate. Announced in 1999 by Philips and Sony, the system uses a blue laser with an even smaller wavelength that the red one used in DVD (405 nanometers, against 650 nm for DVD). This blue laser provides a high-density recording device with a capacity of 22.5 to 25 Gbytes and a data read/write rate of 35 to 50 Mbits/s. The high rate and capacity enable recording of HDTV and new features like simultaneous recording and playback of standard-definition TV (SDTV) material. Moreover, the system achieves seamless playback of edited material.
Simultaneous recording and playback allows the user to watch a recorded program while at the same time recording a second one. It also enables time-shifting-that is, a viewer can stop watching a program to, say, answer the telephone and pick it up later at the exact same scene. This is a function often found in personal video recorders, which use a hard-disk drive for video storage.
Using disk media instead of tape makes it possible to realize convenient features for the user. It is easy, for example, to do random access within recorded streams, enabling implementation of features such as smooth fast forward or reverse play and making it possible to quickly access any time or scene in the recorded streams. Indexing marks to the scenes help in searching the content. Optionally, thumbnail pictures can be added to the streams for easy navigation.
In addition, DVR-blue can also support improved operations for editing the recorded streams. Virtual editing can be realized where the original material is kept on disk, and the selected presentation intervals from the recorded streams are connected without copying the recorded streams.
An application format for digital video recording has been studied and is intended for use with the DVR-blue disk to take care of proper exchange of disks among recorders of various brands. The format is designed to record digital broadcast signals based on the Advanced Television Standards Committee (ATSC) standard for digital broadcast in the United States, the International Services Digital Broadcast (ISDB) standard in Japan and the Digital Video Broadcast standard in Europe. New features such as recording HDTV digital broadcast streams (24 Mbits/s for ISDB or 20 Mbits/s for ATSC) and simultaneous recording and playback of two standard-definition TV streams (10 Mbits/s) have been studied.
In particular, the system supports seamless playback between two presentation intervals edited in a playlist by storing a bridge sequence on the disk.
The video portion of a digital broadcast is transmitted in the MPEG-2 Transport Stream format, which can contain one HDTV program or many SDTV programs. The transport stream is a specific method to put the video data in small, robust packets. On disk, only the program of choice is selected from the full transport stream and is stored as a partial transport stream. Timing information is added to the packets to make it possible to restore the timing of the recorded stream upon playback.
The capacity of a single-layer disk is sufficient for recording at least two hours of HDTV broadcast (at 24 Mbits/s) and close to eight hours of SDTV broadcast material (at an average data rate of 6 Mbits/s).
The physical format of the DVR-blue disk is prepared for fragmented recording. The data of a video file is stored in sectors at an arbitrary position on the disk. This may lead to deterioration of the throughput of the streams when too much fragmentation occurs. So provisions have to be defined to guarantee sufficient throughput for uninterrupted playback (and recording). Guaranteed throughput of video data in the case of simultaneous record and playback can only be obtained if there is a balance between the time spent reading and writing data and the time spent jumping between locations on the disk. Therefore, a certain minimum amount of data has to be read or written contiguously.
Factors that determine the size of a physically contiguous block of data are data rate, access speed of the drive and the rate of the audiovisual stream. Strict requirements must be met for simultaneous recording and playback of separate programs. In that case, read and write actions are interleaved in time and proper scheduling of read and write requests is essential for the result.
For realizing the new features, a set of rules for data allocation on the disk has been defined, which takes into account the physical characteristics of both the drive and the disk. The rules should be as simple as possible for easy implementation and optimal compatibility. The application format defines the rule and the data allocation on the disk must satisfy the rule to guarantee seamless playback.
We propose defining one large, physically contiguous collection of disk sectors, called a fragment. The size of a fragment is fixed and fragments are written in an orderly structure on disk. Using the fragments for allocation of audiovisual data enables seamless, uninterrupted recording and playback. An AV file can therefore be seen as a sequence of fragments.
A typical case of a disk rate of 35 Mbits/s, a maximum AV rate of 10 Mbits/s and an access time (worst case) of 0.8 second lead to a fragment with a size of roughly 4.5 Mbytes. However, this size only takes into account straight recording and playback with fragments that are completely filled.
An additional constraint is imposed by the editing operation. For seamless playback between two presentation intervals, two conditions have to be fulfilled: First, the MPEG data between exit point and entry points must be correct; and second, the data supply should comply with the allocation rules for continuous supply. To fulfill the first condition, some material has to be re-encoded and the result has to be stored on disk in such a way that the second condition is also fulfilled. The re-encoding can be carried out in such a way that frame-accurate editing is possible.
This data is stored in a bridge that links the two presentation intervals. Such a bridge will contain some re-encoded data and some data copied from the original streams. Note that the original material remains unchanged. It is not possible, in general, to copy data to the bridge so that both the bridge fragment and the preceding and succeeding fragments in the edited playlist are completely filled. This implies that partly filled fragments have to be allowed.
Therefore the concept of a segment is introduced. A segment contains all contiguous data of the part of the fragment used for playback. As a result of editing the size of a segment can be smaller than the fragment size. Given a fragment size of 6.4 Mbytes and the allocation requirement that each segment is greater than or equal to three-quarters of the fragment size, seamless playback of an edited video stream and simultaneous recording of another video stream can be guaranteed. The allocation rule prescribes that the data is stored in fragments that contain segments at least 75 percent as large as the fragment size. Following this rule, editing can be performed repeatedly using a limited number of additional bridge fragments.
This discussion has been limited to SDTV. To be able to play back edited HDTV (24-Mbit/s) streams, the fragment size must be enlarged. Because we want one fragment size for all applications we chose to double the fragment size used for allocation to accommodate both simultaneous recording and playback for SDTV and seamless playback of edited HDTV material.
The options to define an application format for recording digital video broadcasts using an advanced optical-storage device have been studied. New features like seamless simultaneous recording and playback functionality can be guaranteed by defining allocation rules for the data on disk. The unit of allocation is a segment at least 75 percent as large as the fragment size. This even allows for the seamless playback of edited streams.
Given the reference values for the DVR-blue drive performance, the size of a fragment should be 6.4 Mbytes. The actual choice should be twice the previous value to include playback of HDTV material that has been edited.
Other contributors to this article include K. van Gelder, J.F. Blacquiere, D.P. Kelly and M. Katom, of Philips Research and Sony Corp. The complete version of this article will be presented at ICCE.