Annex L to the general DVB-RCS guidelines [i.3] presents a feasibility analysis on the use of EN 301 790 [i.1] on mobile scenarios. As suggested there, few if any special considerations for mobility are necessary in the forward link.
There are however some considerations for the return link; these are addressed in general terms in [i.3] and in more detail in the following.
6.1.1 Carrier Frequency Doppler Shift
6.1.1.1 Log-On
Annex L [i.3] derives lower limits for the symbol rates that can be supported without Doppler compensation. These are determined by the carrier frequency tolerance of the hub, and are in some cases quite high. On the other hand, upper bounds the symbol rates are set due to the EIRP limitations in combination with the anticipated space segment
characteristics. In some cases, there is only a narrow range of possible rates left. This is clearly unattractive in terms of system design flexibility. It is difficult to increase the receiver's frequency tolerance substantially beyond the 3 % assumed in [i.3]; hence, we recommend the use of Doppler pre-compensation in the return link where appropriate. This can be implemented with relative ease, provided that navigation information is available. With such compensation, bound on the symbol rate are set e.g. by phase noise considerations, just as they are in fixed implementations.
The Doppler shift in the return link can be determined from the link geometry and the velocities of the terminal and satellite, using conventional solid geometry. The Doppler shift is proportional to the relative speed of the satellite and terminal. It should be noted that the total frequency error at the hub receiver is determined not only by the Doppler shift in the return link, but also has a contribution from the forward link, resulting from the Doppler shift on the NCR clock.
As shown in [i.3], the worst-case Doppler is experienced by aeronautical terminals. For these, the terminal motion dominates the Doppler shift. In the worst case, the Doppler shift is therefore given by fv/c, where f is the frequency, v is the terminal speed and c is the speed of light. With an initial frequency tolerance of 3 % as suggested in [i.3], a
100 kBaud carrier can tolerate an offset of 3 kHz. In turn, this corresponds to an error in the terminal's speed estimate of 30 m/s or 108 km/hr. This is well in excess of the error that can be expected, for example, from satellite navigation equipment. Estimating the initial frequency offset due to Doppler with sufficient accuracy should therefore not be a problem in practice.
6.1.1.2 Carrier Frequency Synchronisation Maintenance
Following logon, the terminal's transmit frequency can be controlled by closed-loop mechanisms already present in DVB-RCS. The worst-case Doppler rate identified in [i.3] is 1 700 Hz/s. A closed-loop frequency control can therefore maintain the carrier frequency within 3 kHz, provided the latency (update interval plus two-way propagation delay) of the loop is less than approximately 1,75 s.
For typical implementations, these tolerances are relaxed in proportion to the return link symbol rate.
6.1.2 Symbol Rate Doppler Shift
According to [i.3], the worst-case relative Doppler shift, which is experienced by the aeronautical terminal, is 1 ppm.
This is considerably smaller than the 20 ppm symbol rate tolerance allowed by EN 301 790 [i.1] (clause 6.1.4), and is therefore not an issue for system synchronisation.
6.1.3 Burst Timing
6.1.3.1 Log-On
The ordinary log-on process for DVB-RCS RCSTs relies on accurate knowledge of the position of both the RCST and the satellite, in order to allow transmission of the CSC burst within a window that is at most one or two milliseconds long. In certain mobile applications, it is highly desirable to allow a much wider acquisition window. Reasons for this include a desire to be independent of for example accurate satellite ephemeris data and/or positioning information. It can be noted that the timing uncertainty applies to all types of mobile terminals, including for example maritime terminals that may not need the navigation information for correction of carrier frequency Doppler shift.
Large initial timing uncertainty may also arise where accurate ephemeris data is readily available. This includes, but is not always limited to inclined-orbit satellites. It can be noted that use of inclined-orbit satellites is particularly attractive for mobile applications, where the RCST most often needs a tracking antenna anyway.
The present document has some limitations that impede a straightforward extension of the initial timing tolerance. The leading-edge guard interval (burst_start_offset) of the CSC slot is limited to ~1,42 ms, because the corresponding field in the Timeslot Composition Table is only 16 bits wide. While the time slot can be up to 364 ms long, the leading-edge guard interval limitation effectively limits the tolerance. In addition, the maximum time correction that is possible using the Correction Message Descriptor is 127 × 27 PCR ticks, or ~439 μs.
Assuming that any NCC/satellite path delay variation is handled either by making use of the optional NCR packet payload or by appropriate adjustments of the NCC timing, the uncertainty in the absence of positioning or ephemeris information is determined by the path delay variation between possible terminal locations and that caused by satellite motion. The one-way delay variation between a satellite directly overhead and one on the horizon is ~20 ms. The daily delay variation of a satellite in a 10° inclined orbit is around 7,5 ms (for a user location that has the satellite on the horizon). The worst-case uncertainty is therefore around 28 ms - i.e. much bigger than what can be handled by the current method.
In the past, TDMA systems have employed techniques such as dedicated log-on carriers with completely open acquisition windows in order to overcome the initial uncertainty. Adding such a feature is unattractive for DVB-RCS, both in terms of modification of the standard and equipment, and because it requires dedicated bandwidth, which does not generate revenue. Instead, the method described in the following has been adopted. Details are defined in
clause 7.3a of EN 301 790 [i.1].
The method is illustrated in Error! Unknown switch argument.. The system is configured with a number of
contiguous, normal-length CSC slots that together cover an interval at least as long as the anticipated uncertainty. When operating in this mode, the RCST always aims for a CSC slot as close to the middle as possible. Due to the uncertainty, the burst may however be received in another slot. To resolve this ambiguity, the Correction Message Extension descriptor (clause 8.5.5.10.20 of EN 301 790 [i.1]) is sent in the logon TIM. This descriptor identifies the slot in which the CSC burst was received. It is sent to the RCST in addition to the usual Correction Message Descriptor. This information enables the RCST to compute the overall timing correction required before proceeding to coarse or fine acquisition.
In order to prevent RCSTs that require this large initial uncertainty from attempting to log on to systems that do not support it, NCC support for this feature is signalled in the Superframe Composition Table. The feature can thus be made available on a per-superframe basis.
Figure 14: Log-on method for terminals with large initial uncertainty
The method is completely backwards compatible. RCSTs that do not need the extra tolerance can log on in exactly the same way as they currently do, and will simply ignore the Correction Message Extension descriptor that tells them which slot they hit.
It may happen that the CSC burst is transmitted across a slot boundary. Depending on the NCC receiver
implementation, this may result in the loss of the burst. This situation is handled by the regular CSC retransmission scheme; however, the RCST should ideally use a back-off that is not an integer multiple of the slot duration, in order to avoid continually selecting locations that cross slot boundaries.
6.1.3.2 Timing Synchronisation Maintenance
Once initial acquisition has been performed, the closed-loop timing of DVB-RCS can maintain synchronisation. The worst-case timing drift identified in [i.3] is 1 100 ns/s. For a 4 Msps carrier and a loop latency (update interval + two-way propagation delay) of 1 second, the maximum drift is approximately 4 symbol periods. Uncertainties of this magnitude can be accommodated by minor adjustments of the guard intervals, without dramatic effect on the spectral efficiency. Expressed in symbol periods, the uncertainty is proportional to the data rate, so it is correspondingly smaller for carriers with symbol rates lower than 4 Msps.