Distributed Translator Systems for ATSC

Distributed transmission is single frequency network technology applied to the ATSC system for digital television. Rather than using a single transmitter to service a coverage area, multiple transmitters are used. The transmitters are synchronized in frequency and symbol emission. The ATSC Distributed Transmission was originally developed with the intention of supplying a studio to transmitter (STL) link to every distributed transmitter. Translator systems, however, receive their signals off the air and retransmit on another channel. During the analog to digital transition, there may not be enough channels available to build full translator networks for both signal formats using conventional translators. But distributed transmission allows all of the translators to use as little as just one additional channel. One essential function of the distributed transmission system is to communicate trellis code synchronization information to distributed transmitters via a dedicated packet. The packet, part of the ATSC payload, is normally sent over STL systems. However, because of causality constraints, the trellis code data must be removed from the packet before the packet goes on the air. This means that the synchronization data necessary to operate a distributed transmission network is normally lost in the off-air signal. This paper describes methods of operating single-hop and multiple-hop distributed translator networks with only minor changes to the CS/110A candidate standard for distributed transmission. With this technology, distributed translators become possible and there is no need for an STL to every transmitter site. Introduction In some areas, particularly in the west, there are many translators on the air. However, even in relatively unpopulated areas, there simply are not enough channels available to provide DTV translator service at the same level where NTSC translator service already exists. In a traditional translator system, many channels may be used in addition to the originating channel. A more efficient use of spectrum would be to use just one additional channel for the translators, by applying distributed transmission technology. The distributed transmission system is based on the use of STLs to transmit the modified SMPTE 310 data to all of the transmitters in a network. This is acceptable and even desirable for many applications. But translator systems are intended to receive signals off the air, and to retransmit them on another channel. Requiring translators to have STL systems would seem redundant – why require the same signal to be delivered to a translator site twice? Operators of translator systems would like to be able to take a signal off the air, and use that signal to drive a network of synchronous distributed translators. As originally conceived, the distributed transmission system did not support off-air operation. However, some new ideas have been developed which can make distributed translators possible. These new concepts are discussed in this paper. These methods will require some slight modifications to the ATSC candidate standard. Brief Review of Distributed Transmission In a Distributed Transmission (DTx) system, all of the transmitters in a network are synchronized in frequency, data, and timing. Figure 1 – Distributed Transmission (DTx) System Figure 1 shows a DTx system. A Distributed Transmission Adaptor (DTxA) inserts network control data into the SMPTE 310 stream. Frequency synchronization is achieved by locking the transmitter’s pilot to a 10 MHz GPS reference. Timing synchronization is also referred to GPS. Data synchronization is necessary because of the nondeterministic nature of ATSC modulators. The trellis coders and precoders present in an ATSC modulator start up with arbitrary initial conditions. There are 36 bits in the trellis coders and precoders. In addition, an ATSC modulator makes an arbitrary choice of where to inject frame sync (once every 624 MPEG packets). This results in 624*2 = 42,880,953,483,264 different ways to transmit the same signal. So without a synchronization method, two ATSC modulators fed with the same signal would have about one chance in 43 trillion of producing the same symbol sequence. (From this point on in this paper, the nomenclature convention established in ATSC documents and other engineering papers will be adopted; the trellis coder and precoder states will be referred to as simply “trellis coder states,” with the precoder states being implicitly included.) So, a distributed transmission system must have a way to synchronize the arbitrary initial conditions so that all transmitters are generating the same symbol sequence. This is accomplished by modeling the channel coding process in the DTxA, and by sending the formerly arbitrary state information to all of the slave transmitters in the network in a Distributed Transmission Packet or DTxP. The DTxP also carries data that tells each transmitter where to set its timing. The timing reference is a one pulse per second (1 PPS) clock signal derived from a Global Positioning System (GPS) receiver. Timing of each transmitter may be slightly advanced or delayed to optimize reception in populated overlap areas. In a conventional distributed transmission system, the DTxA modifies the SMPTE 310 input signal by inserting the distributed transmission packet, cadence sync, and the field rate side channel. The resulting SMPTE 310 signal is then sent to all transmitter sites in a network via Studio to Transmitter Link (STL) systems – which may include microwave, fiber, satellite, or other means. Review of Conventional Translators A conventional translator receives an analog or digital television signal on one channel and retransmits on a second channel. Where multiple translators are used, they often use different channels from one another to avoid mutual interference. This is inefficient use of spectrum. Figure 2 – Conventional Translator Figure 2 shows a conventional translator, usable for either analog or digital signals. An off air signal is shifted by a local oscillator (LO) to an intermediate frequency (IF) where it is filtered and amplified. The signal is up converted to a different channel, amplified, and retransmitted. Review of On-Channel Repeaters (OCRs) Figure 3 – On-Channel Repeater (OCR) On-Channel Repeaters (OCRs) receive an off-air signal and retransmit it within about a microsecond. Most of the delay in an on-channel booster is from the IF bandpass filter, which is usually a surface acoustic wave (SAW) filter. Figure 3 shows an OCR. The block diagram is similar to a translator, but the same LO is used for both the down conversion and up conversion, resulting in a coherent output signal on the same channel as the input signal. A high degree of isolation is required between the receive and transmit antennas. Because the delay of an on-channel repeater is intended to be short, it is usually not possible to regenerate the data and remove errors. Received signal distortions (including the short echo produced by the booster itself) are cumulative and are simply retransmitted. Because of causality, on-channel repeaters can only add delay. Therefore, there is little flexibility in adjustment of timing. The possibility of regeneration puts a limit on power, further limiting flexibility of OCR systems. Distributed Translator Systems Figure 4 – Comparison of Distributed Transmission with Distributed Translators Figure 4 compares a channel 8 distributed transmission system (on the left) with a distributed translator system (at right). On the left, all of the transmitters are on the same channel, synchronized in frequency, data, and timing. The system on the left cannot use the off-air signal of a “main” transmitter to feed the slaves, because of a minimum time delay of approximately 8 milliseconds associated with receiving, correcting, and re-encoding. This delay is far beyond what any receiver equalizer can correct. The emission from the outer “slave” transmitters would be delayed at least 8 milliseconds with respect to the “main” transmitter. The input to the slave transmitter needs to arrive at the slaves at least 4 milliseconds before the main transmitter emits. This system can therefore use STLs and OCRs, but not off-air signals. On the right is a distributed translator system. In this system, the slave transmitters are all on a different channel than the main transmitter, so it does not matter that they are all delayed by 8 milliseconds or more. All that matters is that they synchronize with each other – not the originating transmitter. This system may use STLs or off-air receivers to supply data to the slave transmitters. Since the translators are an off-channel network, an on-channel repeater is not applicable. Objectives for Distributed Translators Figure 5 – Multihop Distributed Translator System A basic distributed translator system should be capable of receiving an off-air ATSC signal at multiple sites, and retransmitting the signal on a different channel, with all translators synchronized to one another in frequency, data, and timing. A further requirement in some applications will be for multihop transmission. Such a system would be able to receive a distributed translator’s signal, and retransmit it on yet another channel using multiple synchronized translators. Figure 5 shows an example of a spectrally efficient two-tier distributed translator system. A main transmitter operates on channel 8. A first tier of distributed translators is on channel 41. The channel 41 translators are all synchronous with one another, and timing is adjusted to minimize skew in overlap areas. Beyond the channel 41 translator zone, the original transmitter’s channel 8 spectrum is reused with another tier of translators. Distributed transmission techniques might seem simple to apply to translation applications, but several problems must be addressed. This paper looks at some details of the distributed transmission system in order to describe the problems, and to propose solutions. Modifications to the SMPTE 310 Signal for Distributed Transmission The distributed transmission adaptor, or DTxA mentioned above, makes three kinds of changes to a SMPTE 310 signal. These are: 1. Insertion of data into a distributed transmission packet (DTxP). 2. Periodic inversion of the MPEG sync byte (cadence sync). 3. Insertion of data into the MPEG error bit (field-rate side channel). Figure 6 illustrates the modification to the SMPTE 310 data and its restoration at the slave transmitter. The first modification to the signal inserts data into the dummy placeholder packet generated by the station’s service multiplexer. This packet carries information to the slave transmitters, telling them where to set their timing, what identification sequence to use, the injection level for the identification sequence, the trellis coder states, etc. This packet ends up on the air, but the trellis coder state data must be removed. The second modification of the signal inverts the MPEG sync byte every 624 packets. This periodic inversion is used to identify the insertion point for frame sync. The third modification appropriates the MPEG packet header error bits (normally all zero) to communicate EVSB data to the slave transmitters. This data is called the “field rate side channel.” The slave transmitters, having received this data, restore the bit to its zero state before airing the packets. Figure 6 – Insertion of Data into SMPTE 310 and Its Removal Problems with Broadcast of DTx Modified SMPTE 310 One of the problems associated with applying this technology to translator systems is that some of these modifications of the SMPTE 310 data are lost when the signal is transmitted over the air. First, the trellis coder states must be removed from the DTxP before that packet goes on the air in a slave transmitter. This is to satisfy causality constraints, which is explained below. Second, the MPEG sync bytes are not transmitted in the ATSC system, so it is not possible to invert them where they do not exist. Third, the MPEG error bit in each packet must be restored to zero before transmission; otherwise receivers will reject the packets where the error bit is set, causing severe picture and sound breakup. Fortunately, there are ways around each of these problems. Removal of Trellis Coder State Data Figure 7 shows the changes made to a Distributed Transmission Packet (DTxP) as it is processed at different points in a system. First, the DTxP is formed with dummy data in the station’s service multiplexer. When the DTxP enters the DTxA, all data except for the trellis coder states are inserted into the DTxP. The trellis coder states cannot be inserted at this point because they do not yet exist. The trellis coder states are created by a coding model in the DTxA. The coding model contains the same scrambling, interleaving, forward error correction coding, and trellis coding that exists in each transmitter. The purpose of the coding model is to accurately generate the same symbol sequence that will be produced by all of the slave transmitters. Only after the DTxP emerges from the coding model can the trellis codes be inserted. At this point, the DTxP is fully formed. When the DTxP reaches a slave transmitter, the trellis codes must be removed before the packet proceeds into the coding process. This is because the coding process in each slave transmitter must exactly match the coding model in the DTxA. Not one bit of the input sequence can be different. Since the trellis codes did not yet exist when the DTxP entered the coding model in the DTxA, the trellis codes must be replaced by dummy data before the DTxP enters the slave transmitter’s coder. Only in this way will the slave coders produce the same symbol sequence generated by the DTxA. Identical symbol sequences are essential for system synchronization. This creates a problem. If the DTxP from an off-air signal transmitted from a slave transmitter is to be used for translators, the trellis codes are already gone. Trellis codes are essential for synchronization; without them, two transmitters will generate completely different symbol sequences, turning them into mutual jammers. Figure 7 – Changes to the Distributed Transmission Packet as It Passes through a System