III. Experimental Results and Discussions
The measured AM-VSB CNR at the cable-TV receiver was 51.8 dB with an average CSO distortion of 56.8 dBc and CTB distortion of 60 dBc in the 256-QAM band. The CSO distortion at 572.5 MHz is the dominant nonlinear distortion in the QAM channel band. The laser transmitter was operating at AM modulation index of 3.5% per channel with a clipping index γ = 4·10-4.6 In contrast, the CTB distortion is typically the dominant distortion in the QAM band in externally modulated laser-transmitter-based video lightwave systems.7
Figure 2 shows typical 100-μs time traces of the average (trace A) and the peak (trace B) CSO distortion at 572.5 MHz on a spectrum analyzer (SA) in a zero-span mode. The SA trigger was set up to start the sweep for any event of the peak CSO distortion, which is above the solid line. The line represents the threshold level for impulses with higher amplitudes that would degrade the coded BER. The observed bursty behavior of the CSO distortion, which has nonGaussian statistics, can be explained as follows: Unlike CW carriers, the peak envelope power of modulated video signals can vary by as much as 18 dB, depending on the picture content.8 The synchronization pulses of the modulated video signals may temporarily align with each other, causing the corresponding video carriers to be at their maximum power at the same time, resulting in increased CSO/CTB distortions.9 Thus, the use of CW carriers to simulate modulated video carriers does not accurately represent the time dependence of the peak CSO/CTB distortions.
Figure 2 A typical 100-μs time traces of the average (trace A) and the peak (trace B) CSO distortion at 572.5 MHz. The solid line is the single-sweep trigger for the peak CSO distortion.
The use of CW carriers from a multitone generator instead of modulated video carriers also affects the laser-clipping distortion. In particular, the Automatic Gain Control (AGC) circuitry in the directly modulated (DM) DFB laser transmitter maintains the average RF input power at a fixed level. Under these conditions, the AGC circuitry reduces the AM modulation index when CW carriers are used by a factor of √β, and the CNR is lowered by a factor of β (about 3.4 dB) compared with modulated video carriers, resulting in reduced CSO/ CTB distortions.10
In order to combat the generated burst errors, a variable interleaver in the QAM modem with depth I up to 204 symbols, and J=204/I symbols, was used with R-S T=8 (204,188) code. The maximum burst length that can be corrected by the interleaver and the deinterleaver combination is given by the following equation:11
where I and J are the interleaver parameters, N = 204 symbols is the R-S block size, Rs = 5.056 Mbaud is the transmitted symbol rate, and T = 8 symbols2. Figure 3 shows the measured 256-QAM coded BER versus the maximum interleaver burst tolerance τ(μs) for QAM channel-center frequencies of 571.25 MHz and 569.5 MHz. The 256-QAM SNR was set to 30 dB, corresponding to coded BER of 1.5x10-9 with no AM-VSB channel loading.
Figure 3 The measured 256-QAM coded BER versus the maximum interleaver burst tolerance τ(μs) with and without QAM channel-frequency offset at QAM SNR of 30 dB and CSO distortion of 56.8 dBc.
Since the interleaver was used with I·J = 204 symbols, increasing I according to the above equation results in increasing the burst tolerance as shown in Figure 3. For the I=204, J=1 interleaver, the nearly four times reduction in the 256-QAM BER is obtained by the QAM channel frequency offset alone.
For burst duration of 30-μs, the 256-QAM coded BER slowly reduces as the interleaver depth is increased, since the average burst length is larger than the maximum burst tolerance of the interleaver and deinterleaver combination. However, as the maximum interleaver burst tolerance becomes significantly larger (≈3X) than a burst duration of about 30-μs, most of the generated burst errors are being effectively corrected by the interleaver, resulting in a steeper reduction in the coded BER. In fact, the 256-QAM coded BER is reduced by almost three orders of magnitude compared with the no-interleaver case. Shifting the QAM channel-center frequency down to 569.5 MHz, Figure 4 shows that the dominant CSO distortions at 572.5 MHz and at 566.5 MHz (trace 2) are now located outside the downshifted QAM channel band. The other nonlinear distortions in the downshifted QAM band, namely the CSO distortion at 570 MHz and the CTB distortion at 571.25 MHz, have smaller magnitudes relative to the QAM carrier. If the CTB distortion becomes the dominant one in the QAM band, a 3 MHz downshift of the QAM channel frequency to 568.25 MHz is required to improve the 256-QAM coded BER. The fixed QAM channel-frequency offset method can easily be applied to multiple contiguous QAM channels, where the dominant CSO or CTB distortions will fall in the small gaps between the QAM channels.
Figure 4 The RF frequency spectrum through a 6 MHz bandpass filter with (trace 1) and without (trace 2) of the transmitted QAM channel at 569.5 MHz. Notice that the dominant CSO distortions (trace 2) at 566.5 MHz and 572.5 MHz are located outside the shifted 256-QAM channel band.
The Figure 3 result also indicates that using the I=68, J=3 interleaver with QAM channel frequency offset provides the same coded BER as a three-times-longer interleaver (I=204, J=1) with zero QAM channel-frequency offset. From an implementation-cost point of view, it can be significantly cheaper to keep the depth and the symbol delay (I, J) of the convolutional interleaver as short as possible, since it requires the addition of static random access memory (SRAM) in the QAM receiver.
Does an I=204, J=1 interleaver in the QAM receiver work as well at higher CSO distortion levels? To answer this question, the AM modulation index per channel at the laser transmitter was changed while keeping the power ratio between the QAM and the AM-VSB channels constant. Figure 5 shows the measured 256-QAM coded BER with QAM channel at 571.25 MHz versus the average CSO distortion at a frequency of 572.5 MHz for (A) no interleaver, (B) the I=68, J=3 interleaver, and (C) downshifted QAM channel to 569.5 MHz with 30 dB SNR and I=204, J=1 interleaver. Notice that an I=68, J=3 interleaver (trace B) works best at CSO distortion levels < 60 dBc. At higher CSO distortion levels, the interleaver is being overwhelmed with burst errors, resulting in a BER performance that approaches the no-interleaver case. In this case, using an I=204, J=1 interleaver with the QAM channel-frequency offset method can provide robust transmission, even in the presence of large CSO distortions (≥ 55 dBc).
Figure 5 The measured 256-QAM coded BER with QAM channel at 571.25 MHz versus the CSO distortion at 572.5 MHz for (A) no interleaver, (B) an I=68, J=3 interleaver, and (C) QAM channel at 569.5 MHz with 30 dB SNR and an I=204, J=1 interleaver.