MULTI-CARRIER NETWORK-CENTRIC SATELLITE COMRlCTNZCATIONS MODEM DESIGN

MULTI-CARRIER NETWORK-CENTRIC SATELLITE COMRlCTNZCATIONS MODEM DESIGN Dr. Heywood I. Paul Femme Comp Inc. (FCI) Sterling, Virginia Im woo~i.pml!rilceim-rieconiv. om Abstract This paper presents concepts for Network-Centric communications that alae matched to the planned Wideband Gapfiller Satellite System (WGSS) and other transponder based satellite communications. Gapfiller satellites will digital& transpond uplinks and downlinks between X- and Kaband beams using multiple 125-MHz transponders that can be subchannelized into as many as forty-eight 2.6-MH: band segments. The proposed concept is a multi-carrier networkcentric modem (MC-Net) design, dubbed the McNet modem. Several key features of the McNet modem design include a simplijied method of matching the dynamic trafic variations of IP networks to the digital transponder operations (DTO) pianned as part the WGSS, but the implications of the design are also applicable for analog transponder bused satellite systems. One of the key design driversfor the development of the McNet modem is to match the needs of the more networkcentric terrestrial segments of the DoD Global Information Grid (GIG). The basic premise of the McNet modem is to use an IP LAN and select a simple modulation format such as QPSK or 8PSK with multiple carriers, each at a single fixed rate of nominal!y 3.3 Mbps. The selection of the nominal 3.3 Mbps rate is based on 155 Mps / 47 subchannels within 125 MH: band segment of the WGSS (note one subchannel is reserved for controlfiinctions). Rates zp to 155 Mbps atjixed sites are then supported in steps of 3.3 A4bps by adding or deleting zp to 47 3.3-Mbps carriers that can be set up or torn down according to the rate of IP packets presented by the IP router at the Ji-ont-end of the modem. The proposed McNet modem considers the use of up to 47 3.3-Mbps carriers because the each of carriers are matched to the minimum independently routable and gain adjustable hub-channel bandwidth planned for WGSS. In addition to providing the capability to quickly increase or decrease bandwidth between X- and Ka-band terminals, the McNet modem avoids the need for- baseband patch panels to connect data sources to modems, thus providing additional flexibility for remote control operations. Finally, the McNet modem is also consistent with other concepts for a more flexible Ka-band (or any band) terminal design as discussed in a companion paper [2], Concepts for a Flexible Ka-band Terminal Design Matched to Gapfiller Satellites. 1 Network Centric Modem Goals I. 1 Rationale: on the need for Network-Centric Efficiency The planned DoD WGSS will introduce a new capability for wideband SATCOM to tacticad and fixed users. Current plans also include the introducl ion of Bandwidth Efficient Modulation (BEM) modems that, in addition to current BPSK, QPSK, OQPSK modems with convolutional coding, will incorporate higher order modlulation formats such as SPSK, and 16-ary modulation and the use of Turbo codes or Turbolike product codes. The planned BEM modem formats will be spectrally efficient waveforms, however, the need for network-centric communications means that modem bandwidth efficiency cannot be measured solely by waveform and coding spectral efficiency {typically defined as e, = (bitdsec /Hz)). Instead, network-centric efficiency requires that link efficiencies be measured by the effective spectral efficiency, es*dnet. where dnet is the network-centric ftaction, i.e.. the useful portion of the svstem path capacity, that actually supports user communication requirements. In other words, use of highly bandwidth efficient, but static rate link carriers that may have long periods of stuff bits (or fill bits ), is inherently inefficient from a network-centric perspective since d,,, << 1 implies es << 1. It is clear that the DoD is moving towards a network-centric IP-based (or packet-based) terrestrial GIG iis depicted by the TAC-LANE systems shown in Figure 1. Hence, it is critical that wideband DoD SATCOM systems also imove in a compatible direction to achieve network-centric efficiency and compatibility. 1.2 Definition of Network-centric SATCOM The capability to connect time-varying SATCOM traffic with minimal setup time delay and a specified error quality and message latency between any user to any other user, or between any switch node (collocated with a satellite terminal) to any other switch node. More specifically, network-centric SATCOM is the capability to connect time-varying traffic from users (defined by their source IP address, or alias) or IPswitch nodes or routers that are connected to a router or switch associated with a SATCOM terminal. The switch or router associated with the SATCOM terminal must have a processor capable mapping a transmit modem IP address to the IP address of a receive modem at the destination terminal, or a 1 To include, as examples. amplitude and phase modulation formats such as 1dQAM or 2-level phase constellations such as (12, 4) with four phases at one amplitude level and 12 phases at a second amplitude level. 0-7803-7625-0/02/$17.00 02002 IEEE. 12

set of IP addresses associated with receive modems at multiple terminals in the case of a multicast, or, the equivalent as a report-back (reverse multicast). Network-centric connectivity requires establishment of routes between modems at the source and destination terminal be accomplished with a minimal setup delay to establish, expand, or reduce capacity to achieve QoS objective levels of capability. Levels of capability would include circuit-based connectivity (i.e.. similar to ATM virtual circuits) or could be on a packet switched basis, but the QoS would be based on delay in service setup, packet latency (delivery time delay between transmit terminal switchlrouter to receive terminal switchlrouter), and also based on various measures of error rate (BER, time between errors or bursts of errors, or time between packet errors, etc.). 1.3 Other Objectives of the McNet Modern (1) Consistency with WGS filter and switch routing capability. (2) Backward compatibility with DSCS and Commercial SATCOM. (3) Modulation formats compatible with quasilinear transponder operations. (4) Provision of timeresponsive signal-to-noise ratio measurements and related features for link level power control to support fast response to rain fades events 2 Digital Transponder Operation Background Wideband Gapfiller Satellites (WGS) are currently being 'developed by Boeing Space Systems under DoD contract using digital transponder operation (DTO) technologies that break the upper and lower 500-MHz portions of the 1-GHz US DoD Ka-band (30/20 GHz) band (30-31 GHz up / 20.2-21.2 GHz down) and the 500-MHz X-band (7.9-8.4 GHz up / 7.25-7.75 GHZ down) into four 125-MHz segments per antenna coverage area. Figure 2 shows the basic channelization at X- and Ka-bands and the bandwidth assignments by antenna coverage. At Ka-band, eight diplexed (receive and transmit capable) gimbaled dish antennas (GDAs) provide narrow coverage areas of 1.5-degrees each. In addition there are two Ka-band diplexed GDA 4.5" area coverages, ACAl and ACAZ. Figure 2 also indicates that each of the ten diplexed Ka-band GDAs support up to 500-MHz of bandwidth consisting of four 125-MHz bands in either the upper?4 of the 1-GHz Ka-band or the lower?4 of the band. Frequency re-use is allowed over the full 1-GHz band by selecting nonoverlapping coverage areas for each of the GDAs in the same half of the 1-GHz band. In addition, selected Ka coverages (NCA7-8 and ACA2) can be frequency re-used via reverse polarization. The X-band coverage includes an earth agile, shapeable beam, receive phased array with eight beam forming networks (SBR1 to SBRS), an earth agile shapeable beam transmit array with eight, beamforming networks (SBX 1 to SBXS), and 50-h4Hz earth coverage receive (ECR) and transmit (ECX) horn antennas. The idea of splitting the 500- MHz upper or lower portion of the DoD's planned Ka-band is also applied to the 500-MHz allocation at X-band. Essentially the 500-MHz X-band can also be split into four 125-MHz bands that can be frequency re-used by ensuring that each of the eight X-band phased array patterns are non-overlapping spatially. The highest portion of the uplink 125-MHz band is further split into a 50-MHz earth coverage band (that is not spatially re-used as it is provided by the earth coverage horns), a guard band of 28-MHz, and a 47-MHz band that can be fiequency re-used in a manner similar to that utilized for the X-band phased array 125-MHz band segments. Examination of Figure 1 shows that within the 19 antenna coverages (10 Ka-band, 8 X-band phased array coverages, and the X-band earth coverage), a total of 39-uplink band segments (each being 125, 47, or 50-MHz) and 39-downlink band segments can be selected. The frequency bands with alphanumeric designators with primes indicate that the system will accommodate either the frequency band with a "prime" letter designation or its unprimed letter designation. In other words, the frequency re-use capability is limited by the number of inputs and outputs of the channelizer where each of the 39 inputs or 39 outputs may be used in part at X-band to support thirteen 125-MHz inputs, three 47-MHz inputs, and one 50- MHZ earth coverage input with twenty-two 125-MHz band inputs at Ka-band. Figure 2 shows the basic fi-equency channelization and antenna connectivity to coverage area at Ka-band and X-band. In the WGS design, each of 125-MHz band segments is A/D converted and digitally filtered into 48 2.6-h4Hz sub-segments consisting of inphase and quadrature time-samples at the A/D rate and quantized to n-bits per sample. (Note that the 50-MHZ earth coverage is divided into nineteen 2.6-MHz subchannels and the 47-MHz bands are divided into eighteen 2.6-MHz subchannels. The 48 subchannels per each of the thirty-nine 125-h4Hz segments (a total of 1872 2.6-h4Hz subchannels) considering multiple coverages and some overhead sub-channels, can be individually routed by a digital switch known as the 13

(KwA]. channelizer (or routed together in a larger sub-segment groups up to 48 sub-segments), between uplink and downlink bands at Ka-band or X-band depending upon the configuration of the channelizer. longer, typically). In essence the WGS acts a circuit switch; i.e., a connection-oriented DTO satellite. In summary, At Ka-band, in any uplink beam, any (of up to 4) 125-MHz segments can be connected to Spectrum is divided - Any of 4 125-MHz into 125,50, & 47 MHz downlink Ka-band segments channels, each of which in any of 10 Ka-band are divided into 48 sub- downlink beams, or to channels - Any of the 3 125-MHz 22 * 125 MHz @ Ka 13 *I25 MHz @ X-band segments or 47- MHz X- 3 47 MHz @ X-band band segments in any of the 1 * 50 MHz @ X-band eight-downlink phased array beams, or to the 50-MHz Channel bandwidth 7.25GHz X-band downlink 7.7 5.4GHz can be flexibly EC beam. applied to both uplink NC Similarly at X-band, and downlink coverage NC any of UP to 3 Uplink 125- areas as needed MHz or up to one 47-MHz band in each receive phased Sub-channels can be NCA array beam, and the routed, combined, or 50- NCA broadcast in MHz receive EC band, can combination from any be connected to coverage area to any - Any of the 4 125-MHz other coverage area Ka-band segments in any Ka-band uplink HZ Ka-band downlink Letter designations indicate DA, or using switch matnces - At X-band, to any of the 3 (e, A can be move to A,, of i 730 ipif& (X&kj & /ti ~~~~~~~~ 22 125-MHz segments or the 47- Figure 2. Gapfiller Satellite Channelization The routed communications signal consists of the complex (I and Q) samples from the A/D converter output (which are effectively down-rate sampled and filtered to achieve the equivalent 2.6-MHz subchannel bandwidth). These samples are routed by the channelizer and time-multiplexed combined to form up to thirty-nine equivalent digital outputs (35 at 125- MHz, 3 at 47-MHz, and one at 50-MHZ). Next, each of the thirty-nine outputs is D/A converted, and then up converted to the desired portion of the Ka or X-band downlink segments for RF transmission. One can consider this A/D, filtering, routing, and D/A process as a form of onboard processing, but because the satellite channelizer filter and switch router never demodulates the signal to recover the original digital data, the process is still considered to be a transponding technique, MHz segment in each downlink phased &ray beam, or to the 50-MHz earth coverage segment But, due to limitations on the channelizer and specific traffic connectivity requirements, not every subchannel in every 125- MHZ (or 50, or 47-MHz) segments can be connected to any other 125-MHz (or 50, or 47-hIHz) on the downlink In other words, while a Mc:Net Modem with a single 125- MHz output could, theoretically connect a user transmit modem to a receive modem in any downlink beam; it may not be possible to connect it to any specific 125-MHz band within any beam depending on the traffic scenario. Hence, the solution, for complete flexibility, is to allow any of 48 McNet Modem outputs to be pipelined to any one of four 125-MHz L-band outputs., based on the IP address of the intended receiver. The following section depicts the McNet Modem configuration showing the L-band interfaces. albeit a digitally processed transponder operational (DTO) 3 McNet Modem DAMA Access Concepts version of an analogue transponding device. Because the 3.1 Basic McNet Modern Concept Figure 3 shows a LAN-based hlcnet Modem digital switch (or channelizer) is ground controlled in the current design, the state of the switch will typically remain static for extended periods of time, (minutes to hours, or * As on the uplink at X-band. one of the 125-MHz bands is further separated into a 50-MHz earth coverage band, a band with earth coverage beacons, a 47-MHZ band provided by an earth agile phased array. and a guard band between the earth coverage and agile phased array band. 14

the 47-MHz band segment or one of the 19 2.6-MHz subchannels of the 50-MHz EC band. Figure 3. McNet LAN-based Modem Figure 4 shows a VME-based version of the McNet modem; either version LAN-based or VME-based is acceptable. Figure 4. McNet VME-based Modem As indicated previously, the McNet modem assumes that each of the 47 communications modems (the 48 is for control) has a bandwidth requirement that is matched to the 2.6-MHZ subchannel bandwidth3 of the WGSS payload and that each modem has an L-band output that is tunable over a 1-GHz range, 1.5 f 0.5 GHz. At Ka-band, the operation of the modem assumes each modem IF output is switched to be centered within any of the 48 subchannels of any of four 125- MHz band segments in any Ka-band beam (either the lower or upper four 125-MHz segments). Similarly, at X-band it is assumed that the controller sets the modem IF output to be centered within any of the 48 subchannels of each of the three 125-MHz bands or the one of the 18 2.6-MHz subchannels of As shown later. each modem can easily support 3.3-Mbps using rate 2/3 coded I-PSK for a total transmit rate of 155-Mbps using 47 modems. The basic McNet modem concept requires that an IP LAN switchhouter (or VME controller) in or collocated with the terminal, routes packets of data to a modem operating within a subchannel connection that provides the required uplink beamsubchannel to downlink beam subchannel connectivity. That is, it is assumed that the WGS payload is commanded by the network control terminal4 to operate with a nominal set of subchannel connections that essentially pre-configures an amount of bandwidth connectivity between beams. It is possible to think of the configured channelizer connectivity as a wiring switchboard with 1872 (48*39) inputs and 1872 outputs, each of which will have a gain-state setting (i.e., a transponder gain) and a fixed bandwidth 2.6-MHZ. The network control operator would, based on previous scenario planning, develop an initial state of connectivity; i.e., the expected number of 2.6-MHz subchannels between Ka-band coverages, between X-band coverages, and between X- and Ka-band coverages. As discussed later, the traffic flow measured by the IP LAN switch at each terminal and control monitoring of spectrum and other control information available to the WSOC, can be used to update the subchannel connectivity placing more or less subchannels between beam areas. This would basically be a traffic forecasting system that would alleviate some of the expected planning burden at the system operator level. The planning function would also preselect a limited number of gain states, (nominally three or four), that would be assigned to sets of subchannels preconnected between beams. The subchannel gains would be set to standard levels based on the G/T of the receive terminal and the coverage area of the downlink beam. Figure 5 shows the basic subchannel and beam-to-beam connectivity for the McNet modem over WGS. At the Wideband System Operation Center (WSOC). Due to space limitations. this paper cannot develop the details of the plan. but one can envision a limited number of gain-states based on the G/T of the receive terminal, the satellite transmit beam coverage size, and the fixed data rate of every subchannel (3.3- Mbps). The use of fvted data rate connections is expected to allow a simple table look-up function that would simplify the required planning processes. 15

~ ~ 3 h c q l s d up&daunm evely beam Figure 5. McNet Subchannel Operation Via WGS The figure shows that connections between terminals consist of one or more subchannel modems, each operating at 3.3- Mbps. The McNet modem can be used with, or without its own unique DAMA control. One potentially simple DAMA control technique is to assume that for each WGS beam, at least one of the 2.6-MHz subchannels is connected such that the uplink coverage and downlink coverage are directly connected. In other words, at least one subchannel is in each beam is reserved to permit round-trip reception of each terminal s transmission. Figure 5 indicates that up to 47 16- kbps BPSK control carriers can be transmitted in the upldown control subchannel. If a terminal were accessing any of the remaining 47 communications subchannels, that terminal would also transmit a 16-kbps BPSK carrier (or even simpler, a tone) in one of the 47 tone spacings within the control subchannel. Because all terminals with a beam can see the 47 possible tones (or even 16-kbps BPSK) carriers, every terminal would be able to determine (independent of the WSOC) whether or not a specific subchannel was available for transmission; i.e., DAMA reservation scheme6. Autonomous knowledge, of which subchannels are available, would be possible using this technique even if the desired subchannel had its downlink in another coverage area. ln other words, in this arrangement. the user terminals would not request a subchannel fkom the WSOC every time a subchannel was required. Instead the WSOC would provide a pool of subchannels with a particular beam-to-beam connectivity and would then monitor usage of the subchannels in each beam. (This could be done using the planned Integrated Monitoring and Power Control System (IMPCS) or via the planned spectral monitoring capabilities available to the WSOC.) The DAMA control concept could be integrated into lmpcs or, alternatively, could be eliminated if the response time of the IMPCS system can be shown to be adequate, or if the DAMA aspects of the McNet modem are not desired. A TDMA version of the DAMA control could also be considered, but this has not been explored in depth. 3.2 Modem Design One of the key ideas that spurred the idea of the McNet modem is that it can be a very simple low cost design. Rather than taking the current BEM path of many data rates, many modulation formats. the concept is to develop a very simple modem that can be replicated forty-eight times and achieve any data rate in the range of 3.3-155-Mbps in steps of 3.3- Mbps. The concept is that rather than establishing a high rate connection between two terminals that may, or may not, be filly utilized initially (depending on the initially required traffic load), the local switch (at or collocated with the terminal) pushes packets to one of the 3.3-Mbps modems (or, more generally, ntl modems at time t,,), but when the packet rate results in an average data rate of (say, nominally) 70% of the ntl*3.3-mbps rate of the initially established modems, then the local switch addresses some of the packets to additional modems. This presumes that the additional modem (or modems) can access the required additional 2.6-MHz subchannels with the same required beam-to-beam connectivity and subchannel gain. Hence the initial vision is for a moderately responsive capability to add modem links in data rate steps of 3.3-Mbps. For WGS and 2.4-meter and large terminals, the proposed quantization (essentially 2 Tl) appears reasonable. Further details on the demand assigned (DA) capability of the McNet modem and how it could interoperate with currently planned controlled systems is discussed briefly in the following section. but first additional details of the modem design are discussed. The modem should have moderate spectral bandwidth efficiency and be capable of operating in a multi-carrier HPA mode with moderate to lovv cost terminals, and hence terminals with moderate AMlFJvl and AMPM performance7. Based on government funded simulation results, 8PSK modulation formats appear to provide improved bandwidth efficiency over current QPSK and related modulation formats, but are less sensitive to terminal and satellite non-linearities and thus 8PSK with rate 2/3 FEC (and with Turbo-like coding) is presumed to result in reasonable performance. Hence for the purposes of this, paper, a rate 213 FEC 8PSK modulation with Nyquist filters of shaping factor a = 0.2, would result in minimal losses in the 2.6-MHz 6-dl3 bandwidth of WGSS subchannels. Simulation results in [l] show that 4.67 Mbps 8PSK with rate 213 FEC, and with 0.25 db filtering loss was achieved at the expected operating point of the WGS non-linearity specification (and Noise Power Ratio of 16.5 db; nominally,an HPA backoff of 3.5 db). More conservatively, in this paper a 3.3 Mbps rate 213 8PSK data rate was assumed to have low losses, fit within the 2.6- MHz subchannel bandwidth of WGS, and yet achieve the One benchmark on moderate A MAM and AM/PM non-linearity characteristics is the nonlinear performance of the WGS payload which is expected to operate at an NPR = 16.5 db. 16

maximum large terminal rate of 155 Mbps (47* 3.3 Mbps). It is believed that a large number of manufacturers could easily build a very low cost 3.3 Mbps IP-based modem that could be replicated. Based on expected tactical user data rates, tactical terminals would operate with four such modems. (nominally 8T1) and large fixed sites would utilize the 48-modem 155- Mbps configuration. 3.3 Future Upgrade to McNet FDMMDMA The currently proposed concept is a frequency division multiple access (FDMA) technique, however, it is possible to extend to a multicarrier TDWDMA concept to enhance capability and to be more responsiveness to packet traffic burst dynamics. However, such a capability while not beyond the current state of technology will be more complex and could be a future upgrade given that standards bodies (Milstandards) would eventually work out the required waveform details. Hence. to reduce initial deployment cost and complexity, the development should be implemented in two phases - Phase I: McNet Centric FDMA Modem - Phase 11: Hybrid TDMA/FDMA 3.4 DAMA Access and No Baseband Patch Panels One interesting attribute of the proposed McNet modem is that the routing of baseband data to each modem is via a LAN or VME bus. The net result is that connectivity to modems is changed via a routing and IP addressing technique and thus baseband patch panels can be avoided and the switching can be fully automated and remotely controlled. Coupling this advantage with a terminal design that avoids the need for IF patch panels, (see the Ka-band terminal design in a companion paper [2], by the author), the possibility for a very low cost, remotely controlled terminal and modem is very feasible. 3.5 Concept for Adaptive Power Control Given the ability to round trip measure the received signal level at each terminal (via the 2.6-MHz subchannel) and via comparison s with the received satellite beacon level, it is possible to develop a very responsive power control system since the power control system is performed autonomously at each terminal. Online performance of the power control can be monitored by the WSOC, so an override control at the system level could be provided. Space does not permit further discussion of this issue here. In the companion paper [2], it is shown that the tactical terminal could operate with four, frequency multiplexed HPAs, each supporting a single carrier. Hence each of the HPAs can be operated in a saturated mode with constant envelope modulation to minimize HPA power requirements and retain low sidelobe spectral efficiency. It is also possible to utilize up to twelve modems at tactical terminals for the 98% of the time that rain is not an issue and then drop capacity to four modems for the 2% of the time when rain margins are exceeded. 3.6 Terminal IM Issues and Resolution Initial analyses and simulation data indicate the proposed multi-carrier approach of the McNet modem will indeed raise the Intermodulation noise (IM) density at the terminal HPA, however, results indicate that the IM s are manageable and do not limit the system performance capabilities proposed here. The previously referenced companion paper [2] develops a terminal design that minimizes the IM issue. 4 Other Use Concepts Straightforward extensions of the McNet modem can be shown for a low cost antijam modem and a low cost 2-way GBS mission. In its AJ mode, the modem would transmit on one of 47 subchannels for a hop period. The subchannels can be interspersed among FDMA accesses and hence would be compatible with other users. The subchannels can also be distributed in their connectivity to other beams and hence packets could be routed to multiple users in different beams on a hop by hop basis. 5 Conclusions and Recommendations The basic conclusion of this paper is that the McNet modem offers a simple solution to network-centric support over WGSS in the near-term. Further work is recommended to develop the IP mapping between packet router, the McNet modems, and the required databases required to map modems to subchannels and antenna beams. Additional efforts are recommended to explore the potential network-centric advantages of the proposed solution. References [ 11 WGS Simulation Results Briefing, Ray Cobb, 21 Sep 01, Harris Corp. [2] Concepts for a Flexible Ka-band Terminal Design Matched to Gapfiller Satellites, Dr. H. Paul (to be published). Acknowledgement This paper was sponsored by the Defense Information Systems Agency (DISA OP/4), but DISA does not take an official position on the contents. 17