Читать книгу Position, Navigation, and Timing Technologies in the 21st Century - Группа авторов - Страница 142

As specified in [37], the ATSC‐8VSB DTV signal stream repeats a frame structure as illustrated in Figure 40.2. Each frame consists of two fields, marked as “Field 1” and “Field 2,” respectively; each field has 313 segments, and each segment has 832 symbols. In all, there are 520,832 symbols per frame and 260,416 symbols per field. The baseband signal has a symbol rate of 10.76 mega symbols per second (Msps). At such a rate, a segment lasts 77.32 μs, a field 24.2 ms, and a frame 48.4 ms.

Figure 40.1 Distribution of digital broadcasting systems for terrestrial television (Wikipedia [34]).

Source: Reproduced with permission of DVB.

Figure 40.2 Frame structure of ATSC‐8VSB DTV signals.

Each field consists of 1 field sync segment and 312 data segments. As shown in the lower‐right plot of Figure 40.2, most of the symbols in the data segment (828) are amplitude‐modulated into eight levels, ±7, ±5, ±3, and ±1, except for the first four symbols, which make up a four‐symbol binary sequence of {5, −5, −5, 5} known as the segment sync. Data segments carry video and audio data, which are fed through a randomizer, a Reed–Solomon encoder, and an interleaver before synchronization data are inserted. The data symbols are embedded with a watermark (a spread‐spectrum transmitter ID signal) down by 30 dB.

In the lower‐left plot of Figure 40.2 is shown a field sync segment, which starts with the same segment sync as in data segments. It is followed by a maximal‐length pseudorandom number (PN) sequence (m‐sequence) of 511 chips, denoted by PN511, three PN63 m‐sequences, and other control data. The segment sync and PN sequences are represented in binary levels ±5. The control data include a 24‐symbol VSB mode, 92 reserved symbols, and a 12‐symbol precode. The VSB mode and reserved symbols are also presented in a binary format (±5), but the last precode symbols are presented in 8VSB format.

The longest PN sequence, PN511, is originally designed for estimation of the channel impulse response. The middle PN63 reverses its sign so as to distinguish between Field 1 and Field 2. The polynomial generators for PN511 and PN63 sequences are G₅₁₁(x) = x⁹+x⁷+x⁶+x⁴+x³+ x+1 and G₆₃(x) = x⁶+x+1 with the initial states being 010000000 and 100111, respectively.

The acquisition of one PN511 and three PN63s in the field sync segment can be used to determine the TOA of DTV signals. It is interesting to compare the DTV PN code with the GPS P(Y)‐code. The symbol rate of 10.76 Msps is slightly higher than the GPS P(Y)‐code chipping rate of 10.23 Mcps. But the DTV signal bandwidth of 5.38 MHz is narrower than that of the P(Y)‐code of 10.23 MHz. However, the DTV signal is much stronger in power. Assuming a timing accuracy of 10% of a symbol duration, the expected ranging accuracy with ATSC‐8VSB signals is about 4 m.

There are three distinct features to consider when designing a software receiver for ATSC‐8VSB signals for the PNT purpose. First, the signal has a vestigial single sideband (VSB) spectrum. Second, it comes with a strong pilot signal. Third, the binary pseudorandom code is not continuous but appears in 1 out of 313 segments (a duty cycle of 0.32%). Since we are not interested in data segments (audio/video data), there is no need to implement a full‐blown DTV receiver, a reference design of which is given in [38–40]. A software receiver with a rather simple architecture for ATSC‐8VSB signals is presented below.

Consider a tunable RF front‐end that performs a single stage I/Q down‐conversion of DTV signals from RF to the baseband. The software receiver starts with setting the LO frequency to the center frequency of a chosen DTV station’s frequency band as shown in the left plot of Figure 40.3. The RF spectrum (a real signal) is translated into the IF spectrum (a complex signal). Pilot detection is done on the IF signal.

As shown in the middle plot of Figure 40.3, the software receiver applies a frequency shift, which is the sum of a fixed pilot offset (which may be different from station to station) and a small amount of frequency error, thus converting the signal to the baseband. The baseband spectrum is shown in the right plot of Figure 40.3. Code acquisition and tracking is done on the baseband signal.

Figure 40.3 Recovery of single sideband signals.

Figure 40.4 Architecture of an ATSC‐8VSB baseband signal processor with TOA tracking.

Figure 40.4 is the block diagram of a DTV software receiver. It has two major systems. The top portion shows the acquisition search system, which includes two major steps, namely, pilot detection and field sync segment detection. The pilot detection step provides an estimate of the frequency error off the nominal pilot offset. The field sync segment detection determines an estimate of the code phase error and an estimate of the symbol rate error. These estimates are then used to initialize the code tracking delay lock loop (DLL) and pilot tracking phase lock loop (PLL).

The bottom portion of Figure 40.4 shows the pilot and code tracking loops. After removing the nominal pilot frequency offset, the signal is low‐pass‐filtered to select the pilot signal while filtering out the wideband video signals. A phase error discriminator is applied to obtain an estimate of the phase error, which is processed by the loop filter. The estimated frequency error is used to adjust the carrier NCO, which in turn drives the carrier replica generator.

The code tracking loop acts as an inner loop as shown in Figure 40.4. The early, prompt, and late correlators provide an estimate of the timing error, which is processed by the loop filter to provide an estimate of the code delay and symbol rate. The latter is used to drive the code NCO, which in turn controls the code generator.

It is important to note that the pilot tracking loop is closed at the segment rate, whereas the code tracking loop is closed at the field rate. There are 313 segments per field. The duty cycle is 0.3%. In other words, the code loop has a rather low updating rate or a long updating period over which the code may shift several symbols. It is thus critical to obtain a good symbol error rate in order to maintain code lock. The carrier to code aiding is therefore applied.

The results of acquisition and tracking of real ATSC‐8VSB signals are now presented to illustrate the functionality and performance of the software receiver shown in Figure 40.4 [11, 23, 29]. Figure 40.5(a) shows the signal spectrum (via the fast Fourier transform, or FFT), where the strong pilot signal is visible together with the 6 MHz signal sideband. Figure 40.5(b) shows the search results in terms of correlation versus search steps (per segment). When the code replica is the code for Field 1, the larger peak corresponds to Field 1 while the smaller peak corresponds to Field 2. Note that the separation between the two peaks is exactly 313 segments apart per signal spec.

Figure 40.6 shows the code and pilot tracking results. Figure 40.6(a) shows the correlation function over 7 lags, each spaced by ½ symbols. Due to the low update rate (only once every 313 segments) and jitter in symbol rate, the conventional three‐correlator structure (i.e. the early, prompt, and late correlators) may no longer be capable of maintaining lock, particularly during large motion‐induced transients.

Figure 40.6(b) shows the correlation versus frequency offset from the nominal value. Since the receiver is stationary, this large offset is largely due to transmitter and/or receiver clock drift errors. Figure 40.6(c) shows the correlation peaks over a field of 313 segments. It is this operation that detects the field sync segment and identifies the peak location.

Figure 40.5 Acquisition of DTV signals.

Figure 40.6(d) shows the in‐phase (real) and quadrature (imaginary) components of the prompt channel (complex). The quadrature component goes to near zero, while the in‐phase component holds most of the signal power. Due to an initial sign, the in‐phase is negative in this plot. The quadrature is not zero but is biased, likely due to the fact that the code used as the replica for correlation is not balanced. Figures 40.6(e) and (f) show the code error (in symbols) and symbol rate error (in symbols per seconds) from the DLL loop.

Figure 40.6(g) is the scatter plot of the quadrature versus in‐phase components of the incoming signal samples after low‐pass filtering around the nominal pilot frequency offset. After convergence of PLL, the in‐phase and quadrature components are shown in Figure 40.6(h), where the quadrature component is rendered close to zero while the in‐phase component maintains most of the signal power. Large variations of the in‐phase component are caused by the information content (data symbols of eight levels) carried by the DTV signal.

Figures 40.6(i) and (j) show the PLL frequency error and phase error, respectively, when the loop bandwidth is 5 Hz. Figures 40.6(k) and (l) show the PLL frequency error and phase error, respectively, when the loop bandwidth is 30 Hz. It is clear that with a wider bandwidth, the convergence is faster but the estimates are noisier. In contrast, a narrower bandwidth is less noisy but the transient is longer.

In recognition of the significant effects of mobile fading, ATSC has introduced the ATSC Mobile DTV Standard (A/153) for mobile and handheld users (ATSC‐M/H) [41]. It builds on the fixed reception ATSC‐8VSB (A/53) physical layer [37] to mitigate mobile fading so as to enable mobile DTV reception [42]. In addition to strong coding schemes, ATSC‐M/H incorporates longer and more frequent training sequences for effective channel equalization against severe multipath. Since the training sequences are transmitted in place of data segments, it sacrifices data throughput for mobile reception. Indeed, ATSC‐8VSB has only 0.3% of symbols for training whereas ATSC‐M/H now has 6%, a 20‐fold increase. Examples of experimental ATSC‐M/H signals can be found in [11].

The ATSC Standard A/53 [37] contains a provision for identification of DTV transmitters through the use of “RF watermarking.” The RF watermark signal is a spread‐spectrum signal, whose insertion level can be set, at any time for operation, from well below the normal noise floor of the host 8‐VSB transmitter (e.g. 30 dB below) up to higher levels only used in out‐of‐service testing. As a Kasami code sequence, the RF watermark signal is clocked at the symbol rate of the host 8‐VSB signal (10.76 MHz) and truncated to 65,104 symbols per cycle, which therefore repeats four times per data field. Serial data at a low rate (four symbols per host 8‐VSB data field) are modulated (phase inversion) on the RF watermark signal to permit separate data transmission for remote control and other purposes. The use of RF watermark signals (Kasami sequences) for timing and positioning is analyzed in [43].

Figure 40.6 Tracking of DTV‐8VSB field sync codes (a)–(f) and pilot signals (f)–(l).