For operators of IP trunking or Backhauling networks, profitability is a permanent challenge: one has to keep up with the ever increasing demand for IP traffic, cope with a limited supply of satellite capacity in regions where satellite services are most needed, and deal with eroding prices of competitive terrestrial services in other regions.
While high satellite costs and fierce competition keep driving the margins down, it is essential to get the maximum efficiency of the available satellite capacity.
Over the last few years, several new technologies have been put on the market to help service operators to optimize the performance of their networks. Old DVB-S and Turbo code modems have been progressively replaced by more efficient equipment implementing LDPC Forward Error Correction (FEC) codes either in compliance with the DVB-S2 standard or as proprietary solutions. These products increase the bandwidth efficiency by as much as 40 percent and usually offer a return on investment in a matter of months. On top of these new modulation formats, other advanced technologies are offered to further optimize the bandwidth efficiency.
In this article, two of the most popular technologies available today will be examined: Signal cancellation technologies that allow sending the forward and the return channels in the same space segment, and DVB-S2 ACM technologies that allow the user to vary the transmission parameters dynamically in order to operate with much less link margin. These two technologies will be compared for a typical IP trunking network delivering services from one central hub to 10 remote sites.
The Reference Network Configuration
To perform this comparison, a reference network that delivers IP trunking services to 10 remote sites in Africa from a hub in Sudan is considered. A typical downlink station is in Dakar, Senegal. The satellite offers 36MHz Ku-band transponders, with an EIRP of 47.9 dBW and a G/T of -0.7 dB/K in the direction of Dakar. The hub uses a large 7.8 m antenna and each remote is equipped with a smaller 2.4 m satellite dish.
For this analysis, two typical cases will be analyzed: In the first case the network provides symmetrical services (10 Mbit/sec in both directions) to each remote site. In the second case, asymmetrical services (10 Mbit/sec forward, 2.5 Mbit/sec return) are considered instead. To enable the comparison, the assumption is that all implementation examples are using DVB-S2 modulation equipment (note that bandwidth cancellation compatible with DVB-S2 may not be available from all vendors. Proprietary LDPC solutions should, however, offer comparable performance). All carriers are modulated with a 20 percent roll-off factor. For fixed rate services, availability is expected to be in the range of 99.9 percent. For variable rate services, it will be expected that the service is available at maximum capacity for 99 percent of the time, and a service with a maximum data throughput degradation of 30 percent is available for 99.9 percent of the time.
Conventional Network Implementation
In a conventional implementation, each signal of each service uses a separate transponder segment and the transmission parameters are fixed in order to achieve the expected service availability over time and link condition variations. The link budget optimizes the forward and return links separately. A link budget optimization is performed to achieve the highest bandwidth efficiency, assuming the cost of the amplifiers in the ground station is less an issue than the cost of the satellite capacity.
The most critical factors of the link budget are the uplink fading in the forward and the return channels, and the Carrier to Inter-Modulation ratio in the transponder.
Running link budget calculations based on the satellite characteristics described in our reference case above shows it is possible to use the modulation parameters described in Table 1. For the symmetrical case, a total bandwidth of 105.5 MHz is necessary to carry the 10 services. A minimum of 3 transponders is required to implement the network (Figure 2). In the hub, one TWTA is uplinking the 10 carriers. With a minimum of 6dB back-off for the multiple carrier configuration, at least 275 watts of power are necessary. In the remote the power requirements are also large so TWTAs must be used. With 4.5 dB of back-off for the 16APSK modulation scheme, at least 350 watts amplifiers are needed.
For the asymmetrical case with 2.5 Mbit/sec return channels, a total bandwidth of 73.8 MHz is necessary, which is just too much to fit in two transponders (Figure 3).
Optimization By Signal Cancellation
Signal cancellation technologies allow the return channels to be transmitted in the same satellite segment as the forward channel. The signal received from the satellite is therefore the sum of the two uplink signals. A digital device in each station subtracts the signal sent by that station from the signal received from the satellite. This allows the station to recover and demodulate the other signal, intended to that station, with a minimum degradation (typically 0.6 dB).
The link budget is optimized to balance the uplink and downlink power, and to minimize the intermodulation effects on the transponder. The most critical factors are the uplink fading of the forward link, and the uplink fading in the return channel.
For the symmetrical case, the results are summarized in Table 2. The combined Input Back-Off for the sum of the forward and return channels is 10.8 dB. The total bandwidth to carry the 10 symmetrical services is 79 MHz and, therefore, 3 transponders are still needed.
This example illustrates that because power and back-off issues, signal cancellation requires more bandwidth than just putting the original forward and return signals on top of each other.
In the hub, a single TWTA uplinks the 10 carriers. With 6dB back-off, at least 400 watts are needed. In each remote, 275 watts TWTAs can be used with 3.7 dB back-off for the QPSK signal. The link budget results for the delivery of asymmetrical services are presented in Table 3 on page 83. The combined Input Back-Off of the sum of two signals is 10.2 dB. The 10 services can fit in 70.4 MHz, which can be distributed over two transponders (Figure 6). In the hub, a TWTA of at least 450 watts is needed to transmit the 10 carriers with 6 dB back-off . The remotes can be operated with 20 watts SSPA and a minimum back off of 1.7 dB for the QPSK signal.
Optimization By Adaptive Coding + Modulation
In ACM mode, the receive site of a transmission link continuously monitors the instantaneous receive conditions and reports this information to the uplink site in real time. The uplink site is able to change the modulation parameters dynamically without loosing the synchronization with the receiver and without loosing any data in the process. Thanks to this mechanism, the transmission system can operate with a minimum link margin at all times: whenever the link degrades for one reason or another, the system automatically increases the level of error protection and/or uses a more robust modulation constellation, so the demodulator remains locked and the transmission remains error-free. When the link improves again, less error correction and higher modulation schemes are automatically reactivated. Since statistically rain fades are short and happen quite rarely over time, higher modulation parameters can be used most of the time and the average throughput of such a system is much higher than the fixed throughput of a conventional system.
In a DVB-S2 ACM system, the modulation parameters can be changed dynamically on a frame by frame basis. This gives the possibility to send several services on the same satellite carrier continuously while selecting different modulation and error coding parameters for each service, and making these parameters vary dynamically independently for each service. Using a single large carrier instead of multiple smaller carriers allows saturating the transponder and is a very efficient way of sending all the services from the hub to the remote sites. Thanks to the Multistream mode of DVB-S2, the various streams are logically separated from each other with a unique Input Stream Identification (ISI) number, so from the enduser point of view, the level of separation among the services is the same as with the multiple carrier configuration.
ACM systems are best used with the Automatic Level Control (ALC) mode activated on the transponder. This enables the use of the most efficient MODCODS in the forward link. The link budget balances the uplink and downlink power, minimizes the nonlinear distortion of the single carrier in the forward channel, and minimizes the intermodulation of the return channels. The most critical factors of the link budget are the distortion and noise levels in the downlink of the forward link, and the intermodulation in the uplink of the return channels. The parameters of the link budget are summarized in Table 4.
For the delivery of symmetrical services, a total bandwidth of 65.46 MHz is needed, which fits in less than 2 transponders (Figure 7). The uplink power needed in the hub is quite limited and with 3.1 dB back-off for the 32APSK modulation schemes, a 40 watts SSPA is sufficient. In the remotes 100 watts SSPAs are necessary, with a back-off of 3.1 dB for 32APSK.
For the asymmetrical services, only 39.58 MHz are needed to carry all the services (Figure 8). The uplink of the hub can be operated with a single 40 watt SSPA while each remote requires 25 watts SSPAs with 3.1 dB back-off for 32APSK.
Ideally, the ACM technology is used in systems where the data throughput may vary over time, according to the link condition variations. This is the case of most IP networks, as protocols such as TCP/IP can adapt automatically and dynamically to the available bandwidth. In addition, an ACM system can cope with very deep rain fades and the link availability is much higher than fixed transmission systems.
However, when the transmission format or the business model requires a fixed transmission rate or a permanent high availability, a method exists to avoid the throughput variations in the forward link while keeping most of the efficiency gain of ACM. This method relies on the fact that rain fade will not affect two stations at the same time within the statistical model for the requested availability (assuming two remote stations are not in the same city or area).
As all of the services are sharing the same forward carrier, it is, therefore, possible to foresee some buffer bandwidth capacity on this carrier to cope with rain fade affecting only one of the services. Through the statistical multiplexing effect of the common carrier, this buffer capacity can be used for any service and increases the availability of all services.
In the example, rain fade on one of the forward services would change the modulation parameters for this service from 32 APSK 3/4 to 16 APSK 2/3. This reduces the throughput for that service from 10 Mbps to 7.09 Mbps. So a buffer of 3Mbps excess capacity on the forward carrier would easily compensate for the rain fade on any of the services. This buffer needs to be available with the lowest modulation parameters (16APSK 2/3), so it requires 1.4 MHz of satellite capacity. Compared to the total capacity of the forward carrier, this is only 4.1 percent overhead, but it increases the service availability to 99.9 percent, the same as in the non-ACM network implementations. It is interesting to see that the larger the number of services in the forward carrier, the lower the overhead to maintain a constant throughput. Note that the 4.1 percent satellite capacity overhead has no impact on the frequency plan and link budget.
The results of the comparison among the three network implementations are shows in Tables 5 and 6.
It can be seen that signal cancellation only provides significant bandwidth efficiency gain for the symmetrical case. But in both the symmetrical and asymmetrical cases, the gains provided by ACM are much more significant and allow reducing the necessary satellite bandwidth almost by half for the distribution of asymmetrical services. In the remote stations, both the signal cancellation and ACM implementations require less uplink power.
For the symmetrical case, ACM requires much less power than signal cancellation, while for asymmetrical services signal cancellation is slightly more power-efficient.
In the hub, however, the single carrier implementation of ACM requires much less power than both the conventional and signal cancellation implementations.
Next to the bandwidth efficiency (OPEX) and uplink power (CAPEX) considerations, the systems should also be compared on basis of operational aspects: Signal cancellation systems typically have stringent constraints on phase noise, accuracy of the carrier center frequencies, and non-linear signal distortion. They require a careful power balance between the two carriers and, of course, only work on full-duplex links, where each station must be able to receive its own signal. In case of failure of one of the uplinks, it is also difficult for an operator to indentify which of the two signals is missing, as they occupy the same bandwidth.
ACM systems, on the contrary, have the ability to adapt automatically and dynamically to any link condition, without human intervention. Newtecs FlexACM implementation for IP trunking networks relies on patent-pending a feature called NoDE (Noise and Distorton Estimator) in each demodulator.
Thanks to NoDE, the receiver can automatically make the distinction between link impairments due to fading and noise on one hand, and the non-linear distortion induced by amplifiers on the other hand. FlexACM is also able to automatically optimize its performance in non-linear conditions.
In ACM systems, the traffic to report the link conditions from the receive site to the transmit site uses very little bandwidth and can be implemented on any type of medium, being satellite or terrestrial. One way system implementations using low speed terrestrial returns are also possible. Contrarily, to signal cancellation implementations, ACM implementations do not require a station to be able to receive its own signal.
As all services are uplinked on the same carrier in the forward channel towards the remote stations, frequency planning of an ACM system is much easier. The ACM system is also very flexible, as the distribution of the forward bandwidth among the stations can easily be modified without changing the modulation parameters and without interrupting the services.
The most complex aspects of an ACM system result from the possible variations in data throughput. Here a smart and dynamic traffic shaper must be used to make sure priority is given to realtime traffic and to customers paying for higher Service Level Agreements. The traffic shaper has to work dynamically with an ACM-specific data encapsulator to minimize the IP packet drops when the bandwidth is reduced and to maximize the performance in clear sky conditions.
In Newtecs FlexACM systems, a set of functions called cross-layer optimization ensures that the traffic accelerator, the traffic shaper, the encapsualtor, and the modulator, talk to each other across the layers of the OSI model to optimize efficiency and avoid congestion at all times.
This article reveals that for Point-to-Multipoint IP trunking networks, ACM can provide much more significant bandwidth savings than signal cancellation technologies, with comparable power requirements in the remote sites and large power savings in the uplink of the hub. ACM also provides a better service availability in case of deep rain fades and is easier to install and operate.