Conclusions • Goal is to determine: • if Bessel or RRC filters better for DCS? • can -25 dBC sidelobe requirement be met? • Still a work in progress, found it difficult to compare filters, many variables • For more linear amplifiers, RRC filters marginally better, for less linear amplifiers, Bessel filters marginally better. • 25 dB sidelobe requirement is driving factor in HPA back-off, not filter BW • RRC/Bessel filter differences in filter BW at -20 dBC (276 vs. 306 Hz) small compared to 100 Hz allowed frequency drift(+- 200 Hz spacing). • Either filter can easily support half current spacing for 300 bps links. Filter sharpness not an issue!
Tentative Recommendation • Best filter depends upon SSPA non-linearity. No single best filter (depends on SSPA linearity), but differences very small. Bessel filter harder to optimize. • Either filter set can easily support 2x current spacing for 300 bps links. • Neither filter set can support 2x 1200 bps spacing. • Either filter could might barely support 3x current spacing (300 Hz filter @ 20 dBC BW + 200 Hz for frequency drift + Doppler).
What we did • Using Simulink (from MathWorks), modeled Tx/Rx link using: • Uncoded 450 bps, 8 PSK links. This gives correct bandwidth coded 300 bps but Eb/No high by 1.8 dB. • Used Simulink RRC Tx and Rx filters which optimize filter BW. • Set RRC order to be twice Bessel order (gives same number of computations) • Modeled RRC link with no amplifier. Bandwidth set by Simulink. Bandwidth narrower than other simulations. • Using link above, looked at scatter plots of Tx and Rx while adjusting filter order, over sample rate and Bessel BW to find good combinations (27 combinations examined). Bessel 3 dB BW approximately adjusted to be same as RRC. • Selected Bessel filter order 8, RRC filters, order 16, alpha = 1, BW about= 130 Hz • We have not modeled the satellite HPA
Methodology To determine performance of each combination: • Used filter parameters given in previous slide. • Added HPA, adjusted back-off so that sidelobes were about -26 to -28 dBC • Obtained BER curves for variety of OPBOs. • Repeated above for a three amplifiers of increasing non-linearity cubic polynomial (P3, Saleth1, and Saleth2)
BER Plot RRC and Bessel Filters for P3 Linear AM/PM degrees/dB
Difficulties • Setting order of filters • Bessel function can be calculated in different ways, most common for digital involves twice as many computations as RRC. Therefore we set the order of the RRCD filter to be twice that of Bessel filter. But performance varied very non-linearly on over sampling, filter order and filter bandwidth. We chose best combination. RRC filter well behaved. • Setting BW of filters • Bandwidth of Bessel filter set by approximately matching ~ 3 dB bandwidth of both filters. Not accurate • DCS SSPA used model of SSPA • AM/PM not realistic. AM/AM linear, should penalize RRC filter. • Suitable Bessel function filter not available in Simulink, MathWorks developed one for us.
Reduced spacing • 400 Hz from band center, out-of-band signal is > 55 dBC for 450 bps, 8 PSK. • 300 Hz from band center, out-of-band signal is > 25 dBC. • This would allow 5x current spacing for 300 bps link and by scaling, 2X for 1200 bps link. • However 200 Hz bandwidth allocated to accommodate frequency drift reduces capacity increases to 2x for 300 bps and less than 2x for 1200 bps links
Remaining Work • Model SSPA better if provided with data • Determine impact of satellite amplifier on performance