Technology Advance in Post-Production & Broadcast

Abstract

This article will discuss the recent technology advances in the areas of compression, disk technologies and associated RAID technologies, editing systems and the impact on the post-production and broadcast markets. It will also dwell on the shifts being seen in the video industry from analog to digital, from analog tape to digital tape, from tape to disk and digital tape formats. Recent advances in the areas of compression, with special emphasis on the brand new MPEG2 professional 4:2:2 profile, and its impact to the news, broadcast, archival and the post-production markets will be discussed. As more and more computer systems are being used in the video industry, the challenge to transport this video information faster (than real time) from one system to another becomes apparent. The various network topologies and technologies such as ATM, FDDI,Fibre Channel, SDH etc., will be positioned.

Compression Update

In the recent years Motion JPEG has become a stable and proven form of compression in - digital disk recorders, non-linear editing systems, etc. Motion JPEG is a compression technique that reduces information contained within a frame of video by treating it as an individual photograph, and this process is repeated 25 times a second. MPEG takes into consideration the inter-frame changes and reduces content based on changes between frames. The first implementations of MPEG2 storage systems, primarily in the video server market for the video and demand/NVOD applications and broadcast over satellite. However, the initial implementations of MPEG2 do not address the quality, flexibility and frame access needs of the post production market. Tektronix is taking a leading role and is driving a new standard namely MPEG2 4:2:2 profile, through the standards bodies and will become a defacto standard shortly that will address various applications from ENG to editing. The data rates in the MPEG2 4:2:2 main profile range from 4 Mbits/sec to 50 Mbits/sec. The future MPEG2 based system will not be just fixed rate encoders decoders, but will have variable data rates depending on the applicat-ion, which proves a major advantages to end users. 

The other significant announcement at NAB’96 was that five companies (SCI, Panasonic, HP, Avid & Tektronix) announced that they would work on interoperability across the various compressed file formats as part of the Fibre Channel announcement mentioned later in this paper. This is a significant step that will great]}- benefit the user community and encourage multivendor interoperability over networks and with the advent of open file format formats such as OMF2, true integration is becoming a reality. The bottom line though is that the cost of MPEG2 encoders are still in the range of 10X the cost of Motion JPEG Encoders, and requires a different chip for a de-coder, whereas in the case of Motion JPEG, the encoder and decoder re-mains the same and is bi-directional. So for some time to come, users should expect the cost of MPEG2 4:2:2 profile base system will be significantly higher than that of a Motion JPEG based system. One of the other important uses of MPEG2 is in broadcast technology with the associated standards such as DVB (Digital Video Broadcasting). To take an example of cable TV networks, the bandwidth necessary to transmit one analog video channel, can now with the help of digital compression, transmit 6-10 channels, thereby increasing the number of channels substantially. 

The two charts below describe the channel bandwidth management in a hybrid Fibre co-axial cable TV network using digital technology. One can clearly see that the 750 MHz. bandwidth that used to transport approx.50 channels, can now transport 100+video channels, plus audio, voice as well as switched data services. 

The use of digital compression also greatly benefits satellite transmission where transponder bandwidth management becomes critical and architectures such as DTH (direct to home) can now offer bouquet of channels that 80+(unthinkable in the analog domain!). 

Storage Technology

With the industry slowly but surely drifting from the analog tape to the digital disk recorder era, let us see the specific area in the disk storage area. SCSI has become a defacto disk controller technology, starting at 5 Mbytes/sec. Then came the fast SCSI with a bandwidth of 10 Mbytes/sec and eventually in the last couple of years Fast Wide Differential SCSI (20 Mbytes/ sec), becoming the defacto controller technology. We are beginning to see the advent of Fast-20/ Ultra SCSI, which doubles this bandwidth and will be the standard for the next couple of years. The usual practice for most vendors/users is to connect disk drives to a SCSI controller. The problem in doing this is two fold. The first is bandwidth and the second is availability. 

Let us address bandwidth first. There are three bandwidths that become important.

a) The total sustained data rate of the incoming + outgoing video streams (often compressed) 

b) The controller band-width and

c) The disk media transfer rate. This is best explained with an example. Let us take two incoming and two outgoing motion JPEG compressed video streams that are of digital Betacam quality. 

Total sustained data rate = 2 x 48 Mbits/s (Incoming) + 2 x 48 Mbits/s (outgoing) =24 Mbytes/sec. 

In order to handle these four video streams, let us look at how many disk controllers would we need. A FWD SCSI controller has a peak data transfer rate of 20 Mbytes/sec, so quite clearly one controller would not suffice, so we would need a minimum of two controllers. 

The third area is the disk transfer rate. This is an area quite often misunderstood. Disks attach to SCSI controllers, and in this case to a FWD SCSI controller, and is assumed that the disk transfer rate too is 20 Mbytes/s peak. This is far from true. Each disk drive has many spindles and has a finite data rate at which it can write onto the magnetic media and read from it. This is often in the range of 2-10 Mbytes/s. As we can clearly see, we therefore cannot use a single drive (with a bandwidth of 2-10 Mbytes/s) to handle a data rate of 24 Mbytes/s. To solve this problem we resort to a technology called RAID - Redundant Array of Inexpensive Disks. 

RAID Technology

RAID technology has been under development for many years. It is generally discussed in the context of RAID types, the most common types of RAID are called RAID 0 - RAID 5 (there are other variations but they are usually proprietary to a vendor). There are three common, mainstream RAID types: RAID-0, RAID-3, and RAID-5. Each has specific design goals and is most applicable to the application for which it was intended. 

PAID C

Description: This refers to “stripping” data across a series of disks to achieve higher disk bandwidth. For example if a drive has a data transfer rate-of x Mb/s then stripping data across two drives would give you a data transfer rate of 2xMb/s. 

Benefit: Provides highest performance at a minimal cost. 

PAID 3

cription:D 3 stripes data across a series of drives and in addition calculates a parity check and stores it on a parity drive. For example, data is stripped across 4 drives and parity is written to a fifth. A RAID 3 array is a parallel access array with all disks operating in unison. It allows a balance between the speed of both read and write operations. 

Benefit: Provides excellent performance and protection for data transfer-intensive application (i.e. large files like video clips). 

RAID

Description: Is similar to RAID 3 except that there is no dedicated parity drive. The parity data is spread across all drives. A RAID 5 array is an array in which disks can operate ndependently of each other. RAID-5 allows data to be read from the array faster than it can be written to the array. 

Is most applicable to applications that require high read I/O rates (i.e. lots of transactions of smaller data files such as databases and airline reservation system). 

As you can see in this chart, RAID-0 and RAID-3 are the obvious choices for a video disk recorder or video server. The read /write balance and parallel access architecture are turned to handling large files. 

In the diagram shown on the next page, to the left is conventional disk drive where information is written onto disk in a sequential manner. What RAID 0 accomplishes is to split the information (say a video stream), into many parts and writes each part onto a separate disk. What in effect we have achieved is for a given time, we have occupied a significantly lower bandwidth per drive (with RAID 0) as compared to writing onto a single drive. This therefore enables us to write more information onto this virtual disk (disks A through E in the diagram) - either in the form of higher quality video or additional number of channels (either read or write). 

Typical disk drives used in the video industry have a sustained data transfer rate of -3-5 MB/s. By using 5 such drives in a RAID 0 stripe set, the user effectively achieves a bandwidth or -15-25 MB/s. Now that we have seen how RAID 0 has helped us turbo charge the disk band- width, lets look at another important factor-namely redundancy. 

Redundancy

In RAID-3, data is “striped” across say, 4 disks parity value is written to a 5th. Using the data on 3 of the 4 disks, and the parity data from the parity disk, the data that should exist on a missing disk can be computed. So, if a drive fails, the system can continue uninterrupted play by continually computing the missing data values. If the drive is removed from the system, and replaced with a new drive, a comparable process occurs. The new drive is installed as a blank drive. Subsequently, the entire contents of the tailed drive are reconstructed on the new, empty drive, using the same method for summing data values from 

These data and parity drives are installed in the RAID array in-group of five (to nine depending on the vendor). Every five drive set consists of four data and one parity drive and constitutes one RAID array. An important factor to bear in mind is that the larger the group, more the probability of two failures occurring which would result in all data being lost. Quite a few vendors increase the group size to reduce cost, which has its repercussions. It is also important to note that “useable” capacity of a RAID system does not include the capacity of the parity drives. 

RAID 5 is a variant of the RAID 3 system, where the parity information is striped across the user disks and not on a dedicated parity drive as in the previous case. RAID 5 systems are ideal for transaction processing applications where the request rate is very high and the size of data requested is very small, whereas in the video world, the size of the data is a very large requiring large bandwidth and the number of requests are not large. Therefore majority of vendors implement RAID 3 storage systems for video systems and not RAID 5. 

The Advent of Networking

With the invasion of the digital era into studio, the digital disk recorder is the most cost effective method available for the storage, editing, and caching of video data. The advantage of using a single, highly reliable, digital disk recorder to replace multiple, maintenance intensive, VTRs are generally understood. We are beginning to see a profusion of non-linear editing system, graphics workstations and high-end special effects workstations. Eventually, users want to transport video information across these platforms in a seamless fashion. The traditional method has been to use the VTR as a go between, which is time consuming as well as introduces quality issues. Since most of these devices are essentially computer systems, various attempts have been made to use traditional network that used in the data communication environment. This is where most users bit a roadblock. 

The Bandwidth

Traditional data communications networks as the name suggests, are designed to carry data (which is not delay sensitive but error sensitive) and not isochronous data types such as video and audio(that are delay sensitive). The bandwidths are dramatically different, for example, a CCIR 601 digital uncompressed video stream is 270 Mbits/sec. The most popular data communication network today, Ethernet, boasts of a theoretical peak rate of 10 Mbits/sec. Such is the gap. Most studios use compression techniques to reduce the data rate, and going by motion JPEG standards a Digital Betacam quality signal can be in the region of 48 Mbits/ sec, still clearly a high figure. Elternate technologies such as Fast Ethernet (100 Mbits/sec),FDDI (100 Mbits/ sec), ATM (25 Mbits/sec to 155 Mbits/sec) are alternatives that have been suggested. In NAB ’96, five companies realized this problem and announced that they would work on a technology that would be used as the defacto intra facility network and this is Fibre Channel with speeds upto I Gigabit/sec! 

Fibre Channel represents an efficient means for connecting two or more systems. By using this type of data transfer technique it is possible to rapidly move clips and clip lists from one system to another; in effect creating a video superhighway. For example, in a Post Production house, it is common to edit video in one suit and perform audio sweetening in another. By Fibre Channel connecting a pair of these systems, it would be trivial to move video from one suite to the other with almost no time lag. It would also be possible to rapidly move video data from, for example, a video archive system to a play-to air cache machine without occupying channels of a routing switcher. 

Channels and Networks

Before proceeding to a discussion of Fibre Channel per se, we must first understand two important background concepts; channels and networks. Fibre channel is an “interconnection mechanism” not just a” network”. Fibre Channel draws from both the world of networks and the world of channels. 

Let’s begin with channels. The term “channel” refers to a method of transporting data from one device to another by means of a direct connection. That direct connection is called a “point-point-link” because it directly joins “point A” to “point B” and carries only the data, which is intended to pass between those 2 points. For example, a serial RS. 232 connection from a minicomputer to a graphics terminal is a channel based connection. One end of the RS.232 cable connect directly to a connector on the minicomputer while the other end attaches to a connector on the terminal. The only data present on the wire is that which is required for the communication between those 2 devices. A parallel printer cable, from a PC to a color printer is a multiple wire, multiple bit, version of exactly the same thing. 

A “network”, on the other hand, is an aggregate of many devices, interconnected by a common transmission medium. It is a means of transporting data from device to device over a shared connection. One wire (or optical fiber) will carry data intended for any one or more of the attached devices. Because of this, each device must contend for the wire, recognize and correct errors, and determine whether or not action should be taken on any particular piece of data. Ethernet is the most common example of a network and token ring is likely the best second illustration. 

Both of these inter-connection mechanisms rely on something we call a “protocol”. This is simply a collection of rules or conversions describing such requirements as timing, communication control, the representation and format of the data, some method for describing which devices are communicating, and the rules to follow for correcting data transmission errors. 

What is Fibre Channel?

In 1988, the Fibre Channel working group (one of the many standards working groups sponsored by the American National Standards Institute, or ANSI) began developing a new method of high speed data transfer for computer workstations, display devices, and peripherals. The objectives of that effort included support for, and interconnection of, many types of physical interfaces; high speed transfer of large volumes of data; and a separation of physical interfaces and logical protocols such that many protocols could be run on a single physical interface. In the traditional sense, Fibre Channel is neither a network nor an I/O channel; it is both. ‘Because of its hybrid nature, Fibre Channel can be confusing. First of all we cannot really talk about ATM versus Fibre Channel. Nor can we speak of IP versus Fibre Channel, HIPPI versus Fibre Channel, r SCSI versus Fibre Channel. Fibre Channel attempts to allow several protocols (originating in both channel and network environments) to’ function in a common topology. This makes it both confusing and extraordinarily versatile. 

The principal concept to understand when looking at figure above is that it contains protocols, which come from what we usually refer to as “networking” as well as those you see on a channel type communication topology. With Fibre Channel, the same physical medium can support both network and channel protocols. You could actually have two computers talking ATM protocol and a SCSI disk array on the same piece of Fibre; and address them at the same speed. When you think about it that is extraordinarily powerful. 

Fibre Channel can be deployed over various topological such as

a) Point-to-Point

b) Arbitrated Loop &

c) Switched Fabric. 

Using this combination of topologies, Fibre Channel can be implemented over a number of transmission media and at a number of speeds. At the present time, there are Fibre Channel installations that operate at speeds of 133 Mbps, 266 Mbps, 530Mbps, and IGbps. So, needless to say: it’s fast! It is possible to run a Fibre Channel connection on micrometer fiber, video coax, mini coax, and Shielded Twisted Pair. 

Archival

One of the other main areas of interest developing in the industry is the area of archival. Traditional methods have been to use standard video tapes and devices such as cart machines or flexi-carts. These devices are expensive and therefore not too widely used in the industry. Like the earlier movement from tape to disk, now we are witnessing another change i.e. the use of digital tape (that have been used as archival devices in the IT industry). Tape mechanisms such as 8mm Exabyte, DLT, etc., are making their way into the broadcast industry as they have some distinct advantages compared to the traditional tape environment. When the video streams are compressed and stored in magnetic disk drives, the digital tapes are connected to the same storage bus and therefore transfer video files in the compressed format directly from magnetic disks to tape. This has two advantages

1) Avoids multi-generation loss as the transfer back and forth is in compressed format and does not lose any quality and

2) No dropout loss with wear and tear of the tape/ head, as the storage is in digital format with error correction mechanisms. 

Typically a configuration for archival or library management, would consist of two to four drives, and 80-100 tapes (with approx. 20 Gigabytes capacity/tape), and this configuration would be capable of transferring information in real time and store approx, 120 hours of Betacam SP quality video and audio. Therefore with the advent of these devices, one would gradually see these as a replacement to cart machines or flexicarts as they give similar functionality and a substantially lower-cost. 

Summary

We have seen the various advances in the areas of compression, disk storage technology, archival and networking and all these have contributed significantly to the post - production and broadcast markets. These technologies have now thrown open new avenues of freedom to video and film professionals, to achieve the previously thought impossible. Significant reductions in costs and time have also driven users to push vendors to bring more and more digital technology into the traditional analog domain. So the future definitely is digital!

Source: http://cablequest.org/articles/broadcast-technology/item/1303-technology-advance-in-post-production-broadcast.htmlSource: http://cablequest.org/articles/broadcast-technology/item/1303-technology-advance-in-post-production-broadcast.html