1.2 UHF DAMA Satellite Communications Network
Transcription
1.2 UHF DAMA Satellite Communications Network
UHF DAMA SATELLITE COMMUNICATIONS CAPABILITIES, ARCHITECTURE, NETWORK CONTROL STATION: AND PERFORMANCE Lee E, Taylor Frank Ganaden ViaSat, Inc. Carlsbad, California ABSTRACT To establish context, this paper briefly describes the histoiy of DAMA on UHF military satellites, beginning with experimental systems built during the 1970’s. The paper goes on to describe the technical features common to present-day 5-kHz and 25-kHz DAMA systems, and proceeds to take an in-depth look at capabilities unique to each of the DAMA waveforms, which were designed independently and took rather different philosophical design approaches. The variety of services which are offered (e. g., global and local voice, data, and packetized messages) are described, as are the techniques used to allow automatic adaptation to variations in load, individual terminal link qualities, and trafic priorities. In addition the various types of network configurations supported are described. The hardware and so~are architectures for the currently-fielded Network Control Station sites are then presented. Recently an automated software tool set that can characterize DAMA network pe~ormance in terms of call setup times and channel usage eficiency has been developed. Pe~ormance data are presented, demonstrating the general quality of service that can be delivered. Finally the paper describes evolutionary enhancements planned for the DAMA waveforms and overall system, as additional use warrants and finding becomes available. I. INTRODUCTION The combined techniques of Time Division and Demand Assigned Multiple Access (TDMA/DAMA) offer the promise of significantly more efficient usage of narrowband UHF satellite channels for all of the military services than the original, dedicated network per channel mode of operation. This paper on the subject of UHF DAMA controllers is particularly timely, since a 5-kHz network control capability was fielded and went operational in late 1997, and the controller upgrade to add 25-kHz DAMA capability will be reaching the same operational state early in 1999. Performance data for the operational system is hence just now becoming available and is the primary focus of this paper although a significant amount of system description information is included for the purpose of establishing context. Initial proposals to apply TDMA/DAMA concepts to narrowband military UHF satellite channels originated in the mid- 1970s. MIT Lincoln Laboratory built i~nd demonstrated a prototype DAMA fuily-automated controller and network terminal for use on 25-kHz channels in 1978 (Reference 1). Shortly thereafter, the U. S. Navy contracted with Motorola to develop a fieldable 25-kHz TDMA controller, which resulted in the TD-1271 multiplexer. In the ensuing years, many terminals consisting of TD- 1271s and E-Systems’ WSC-3 transceivers have been deployed, and TDMA has become the regular day-to-day operating mode for Navy 25-kHz UHF satellite communications. Recently the Navy installed the DAMA Semi-Automated Controller, a prototype channel controller developed at NRaD which replaces the manual marker-boards used until now to manage access to TDMA slots with a real-time perscmal computer application. The concepts for DAMA on 5-kHz UHF transponder channels evolved about a decade behind 25-kHz. Among the early 5-kHz DAMA development programs were the UHF Satellite Terminal System (USTS) and Mini-DAMA projects in the 1980s. These were followed by a competitive development of prototype, automated 5-kHz network controllers by Titan Systems and ViaSat for the Air Force in 1993 – 1995. The second-generation ViaSat Network Control Station (NCS) is the system presently fielded world-wide and being upgraded to add 25-kHz control capability. For purposes of ensuring interoperability among different manufacturers’ terminals, standard waveforms have been specified in MIL-STD-188-182 for 5-kHz bandwidth channels and in MIL-STD-1 88-183 for 25-kHz channels. Though the two standards were developed at different times, and with different military services being most involved, they share many common concepts and approaches; these are the focus of Section II which describes at a top level how DAMA “works”. Section III addresses the unique features of the individual waveforms, taking on a strong 5-kHz DAMA flavor since that standard specifies protocols in much more detail and was develclped more recently. Section IV describes the DAMA controller arGhitGGture,starting with the world-wide placement of the primary and alternate channel controllers, and a discussion of the concepts for transfer of control. The architecture 0-7803-4902-4/98/$10.00 (c) 1998 IEEE section delves into the next lower level of detail, describing how the functions at each site are distributed across hardware elements plus indicating the size, language, and operating system for each processing element. Section V, Performance, discusses quantitatively how well the DAMA channels function under traffic load. In the present era, before sizeable numbers of DAMA terminals are deployed, these data can only be derived using the Traffic Generator, an emulation tool that resides in the channel controller computer, along with the NCS application. The performance data are of course required in order for ViaSat to demonstrate compliance with specifications. But when more operationally realistic scenarios are used, the performance data can also help direct the DAMA development community towards the high-leverage improvements to the system in the future. In addition the DAMA operations planning community can use the same tools to predict performance under various network configurations they might consider for live missions. This paper closes with a short section describing presently planned improvements to the military standards, and the Navy’s near-term plans for adding capability to the DAMA Network Control Stations, master time base to which all users on the channel synchronize, and time slots that contain burst transmissions originating from those individual channel controller and network terminals using the channel. The slots must include “guard time” before and after each terminal’s bursts, to accommodate timing drift between instances when the terminal re-synchronizes. The use of periodic slots for communicating inherently continuous data streams requires that the data be buffered at the originating terminal, resulting in end-to-end transit delays that do not occur in continuous transmission systems. Each controlled DAMA channel has four fundamental types of slots. The first type is “forward” orderwires originated by the NCS and containing control information such as slot assignments for the population of network terminals. Slot type two is ranging slots that are transmitted and received by individual terminals and are used to measure signal propagation time from the terminal to the satellite. The third type is “reverse” orderwires originated by the network terminals and containing control information such as service requests for the NCS. Finally, the fourth category of slots contains communications bursts among pairs of users or network members. Both waveforms offer users a basic type of service; a synchronous, half-duplex circuit with destination(s) Simple restricted to the “local” satellite footprint. depictions of the frame formats for 5- and 25-kHz DAMA are included in Figure 1. II. COMMON 5- AND 25-KHZ CAPABILITIES UHF narrowband channels are themselves derived through frequency division multiplexing across the allocated band and time division is the multiple access technique chosen for handling multiple users per channel. TDMA requires a o 5-kHz Frame ● “ 8.96 sec FOW ● ~ Contention . RNG 4 j Contention ; ROW & ~ Assigned : RNG + ROW ~ ~ COM 4 0 Generic Frame . . . Forward Orderwire Frame ~ : Range Burst ! Reverse ~ Orderwire ~ ! User Communications 1.386 sec o v ● ● ” CCOW [ C:M : RNG ; ;;: [RCCOW; C:M Figure 1: UHF DAMA Frame Formats 0-7803-4902-4/98/$10.00 (c) 1998 IEEE duration ● A 25-kHz Frame ● :C:M ● .* ** *O With multiple network terminals using the. channel, and some reverse orderwire slots allocated to “random access” (use by more than one terminal) there is a possibility that two terminals will transmit in the same slot, The result is contention (mutual interference) such that neither terminal’s burst can be successfully received on the satellite downlink. Both military standards include provisions for handling this situation, consisting of retransmission protocols that will ensure that the contention is ultimately resolved. Data conveyed in the various types of slots can be protected against the effects of transmission errors by the application of forward error correction codes. In addition, cyclic redundancy check coding on selected slots provides detection of the rare errors that cannot be corrected. Various symbol burst rates can be selected; higher rates being most efficient and lowest rates providing additional robustness for terminals with small antennas. In general, the orderwires are operated at the lowest burst rates and with strongest coding, in order to guarantee the best possible reception by the most disadvantaged terminals. Finally, in the context of signal processing, the waveforms both use phase modulation, whose constant-envelope properties are suited to the non-linear satellite transponders. Orderwire data flowing in both directions are typically encrypted to prevent traffic analysis. Just as in the public telephone system, a DAMA user needs a destination number in order to place a call. The DAMA address space has 216elements, which can signify terminal port addresses, network addresses, or “guard” addresses (which can indicate an individual person, function, or mission). Calls may be placed point-to-point, multi-party conference calls may be established without preplanning, or preplanned multi-party networks may be activated, Requests for service include a numeric precedence level in the range 1 -5, so that when the overall load exceeds channel capacity, lower precedence traffic may be preempted. Each individual DAMA channel has its own forward and reverse orderwire slots, so a given terminal will typically remain on the channel it started out on – its “home” channel. In the event of high load on one channel, however, the NCS can send one or more terminals to a different DAMA channel; they can even be commanded to move from a 25-kHz to a 5-kHz channel (or vice versa). Finally, if very high priority users require a channel without TDMA buffering delays, the NCS can assign them temporarily to “dedicated” channels that operate full-duty-cycle in accordance with MIL-STD-188181. In addition to a centralized time base generator, UHF DAMA employs a centralized channel controller, which “assigns” individual users and networks to time slots, Each channel controller has one or more alternates, typically located at a different geographical site. Control may transfer from primary to alternate under a number of situations, including an unplanned equipment outage at the primary site, a primary site operator request to hand over control to the alternate, or an alternate site operator request to take control, Parameter Frame Duration Basic Time Unit I/O Rates Burst Rates Guard Band Size Modulation FEC Code Rates 5-kHz (MIL-STD-182) 8.96 sec 8.75 rnillisec “building block” 75-2400 bps 600-3000 SDS 25.208 millisec SOQPSK % 25-kHz (MIL-STD-183) 1.3867 sec 52 microsec “chip” 75-16000 bps 9.6-16 kst)s + 1.25 rnillisec BPSK ‘/2,% Figure 2: Comparative UHF DAMA Numerics 111.UNIQUE FEATURES OF EACH WAVEFORM Philosophically, the 5-kHz DAMA standard exerts much tighter control over the population of network terminals and is intended to be much more dynamic in adapting to changing conditions. Terminals are required to log in with the channel controller immediately upon acquiring the downlink and uplink timing whereas in 25-kHz DAMA a terminal may passively monitor the channel and make its presence known to the channel controller much later, using a variety of reverse orderwire types. Typically this will occur upon a status change such as putting an I/O port in service or changing the port configuration of baseband equipment. Also there is a 5-kHz DAMA protocol to -ensure that the controller’s version of each- terminal’s guard list (all of the addresses associated with each I/O port) remains consistent with the terminal’s version. When discrepancies are detected the protocol automatically brings the lists back into a consistent state, accepting either the channel controller or terminal version, under NCS operator control for each terminal. The 5-kHz channel controller reconstructs the frame format and rebroadcasts slot assignments every frame ! This allows more or less of the frame, as demand requires, to be allocated to forward and reverse orderwire traffic. It also allows immediate preemption of lower precedence user traffic to accommodate more important calls and for immediate restoration of the preempted service once the higher precedence call terminates. Basically, if a 5-kHz terminal does not receive its slot assignment in a frame, it ceases transmitting; when the slot assignments start appearing again, the terminal resumes usage of the slots. Finally, the continuous broadcast of service assignments allows terminals that log in late to immediately join established networks. In contrast, the 25-kHz waveform sends a slot 0-7803-4902-4/98/$10.00 (c) 1998 IEEE connect in a single frame and then does not repeat it. Thus the channel controller must originate a new copy of the slot assignment in order to bring a late entering terminal into a network; and the controller must send a slot disconnect orderwire message in order to force a call off the air. As a result of its more fixed frame format, the 25kHz waveform offers precisely one forward orderwire message and one reverse orderwire message slot per frame. Similarly segments of the 25-kHz frame devoted to user communications take on a relatively fixed format; they can change but generally not from one frame to the next. So, if the current 25-kHz DAMA frame format is structured for, say, nine 1200 bps slots and one 75 bps slot (the format designated “9, B, A“) and some of these slots are assigned to active calls, the channel controller may require several frames to change the format in order to accommodate a 2400 bps call. The 5-kHz waveform is sufficiently dynamic that the controller can immediately compress the existing services into the end of the frame and create a 2400 bps slot. The 25-kHz DAMA waveform provides an end-end link quality measurement capability, At the request of either the channel controller operator or a terminal operator, the terminal can be assigned the single link test slot (see Figure 1) for enough frames to accumulate statistically significant bit error rate data. The terminal itself originates the uplink and receives the downlink, so this constitutes a true end-to-end link test. The burst rate is variable over the options available for communications services, so the data may be related directly to performance in the communications slots. Once the link test is completed, bit error rate results are displayed at the terminal and reported to the channel controller using a reverse orderwire message. At present the channel controller does not automatically use the link test data in selecting a communications slot with appropriate burst and code rates, since bit error rates rather than carrier to noise density ratios are reported. The capability to report C/No is a feature of the next-generation 25-kHz standard (“183A”) and will ultimately be used by the 25-kHz controller. In contrast, 5-kHz DAMA network terminals derive link quality information by processing forward orderwire bursts over many frames. This technique has the limitations of providing information only on the terminal’s downlink and only at the very robust 800 sps rate 1/2 coded mode. Nevertheless, results are periodically sent to the channel controller in reverse orderwire status reports, and the controller uses the results to automatically select burst and code rates which will ensure adequate performance for the weakest terminal in the service. The 5-kHz controller strives to achieve 10-3 bit error rate (BER) for voice services and 10-5 BER for data services. In addition to the generic half-duplex, synchronous, localfootprint circuit service type offered by both waveforms, 5-kHz DAMA distinguishes between voice circuits and data circuits for BER purposes. Also, MIL-STD-182 includes the structure for handling start, stop, and fill bits necessary for operation of asynchronous data I/O. Beyond circuit services, the 5-kHz waveform includes packetized, simplex “message” services. When the length of a data communication is precisely known a priori, such as in the case of a computer-to-computer file transfer, the message can be divided into small packets and sent at a variable I/O rate per frame, in the background relative to the circuit services. Exactly enough packets are allocated to just fill up each frame, after all active circuit services have been accommodated. This feature allows highly efficient usage of the available channel capacity. At present message length is limited to 114688 bits because of possible cumulative timing drift between source and destination. Point-to-point message services include acknowledgements of each burst and of the entire message, providing an extremely reliable form of communication. Another service-oriented capability unique to 5-kHz DAMA is global connectivity, which applies to all point-to-point (vs. network) calls. The worldwide complement of channel controllers intercommunicate and have the intelligence to find a user anywhere globally and to automatically relay traffic through multiple satellite hops until it reaches the proper destination. 5-kHz DAMA terminrds are required to specify in each service request whether the traffic is voice or data and whether the I/O is synchronous or asynchronous. A quite different approach is used with 25-kHz DAMA, where the terminal originating the service specifies one of 99 “configuration codes”, the specific values of which are left open to be operationally assigned. Each code will correspond to a complete string of I/O equipment, so in addition to voice/data and sync/asynchronous it conveys whether the port is attached to an encryption unit, if so what model, and what variety of I/O device is currently attached. A unique feature of the 25-kHz DAMA waveform is the ability to assign slots in a second channel. This is termed a “frequency-switched slot connect”, and requires the participating terminals to tune their radios twice per frame, such that they access orderwires on the home channel and handle communications bursts on the other channel. A second unique 25-kHz DAMA capability is a call waiting alert similar to that offered in commercial telephony. If a user is denied connection to a second party because the second party is busy with higher precedence traffic, the originating user may manually send a “paging” reverse orderwire and the channel controller will originate a “call waiting” forward orderwire which may be displayed at the busy party’s terminal. One noteworthy unique feature of 5-kHz DAMA is the ability for users to notify the channel controller that their 0-7803-4902-4/98/$10.00 (c) 1998 IEEE use of a slot is completed, using an in-band signal rather than taking the time and bandwidth to send a reverse orderwire. The terminal simply inserts a special 12-bit “tear-down” code in the circuit burst kind (CBK) field of the last transmitted burst. Since the channel controller monitors usage of every slot, upon receipt of the teardown CBK, the controller broadcasts a teardown forward orderwire and immediately puts the slot back into the available pool. Another unique 5-kHz concept is the “future forward orderwire”. The first few fields of a Future FOW are well defined so that the message can be recognized by old terminals and ignored, whereas terminals fielded since the remainder of the Future FOW fields were defined can go on to process the rest of the message and hence offer advanced capabilities. This feature greatly facilitates interoperability among old and new terminals; the channel controller must, of course, keep records of which terminals support which versions of the military standard. IV. DAMA ARCHITECTURE The DAMA “system” is comprised of the entire suite of NCS equipment at all four sites worldwide: Norfolk, Naples, Guam, and Hawaii. Each site is termed an NCS “segment” and has both an east side and a west side for controlling channels on satellites in the two directions. For 5-kHz global traffic relaying purposes, the east and west sides are connected by a local area network. The systemlevel architecture is depicted in Fig. 3, which indicates CONUS, Atlantic, Pacific, and Indian Ocean satellites plus two Network Terminals. CONUS PAC The combination of NCS segments, links, and satellites presently form an eight-channel ring backbone network. Individual network terminals connect into the ring using a single channel on their local satellite. Once terminals have logged-in with the channel controller, the system automatically provides them service accesses on demand. Typically, each satellite channel will be configured with a “primary” channel controller (PCC) and an “alternate” (ACC) located at the other NCS site within view of the same satellite. Geographically dispersed alternates provide considerable robustness in the face of natural or military disasters. An ACC can take over channel control in the event of planned or unplanned PCC outages. In the planned case, the PCC operator can manually initiate handover to the alternate; similarly, the ACC operator has the ability to request channel takeover, but this is allowed only with approval by the current PCC operator. Once control has been handed or taken over, the “old” PCC equipment can be placed off line for maintenance or other activities. In the event of unplanned PCC failure, and if an ACC is actively logged onto the channel, that ACC will automatically take over control of the channel and begin transmitting forward orderwires after missing the forward orderwire bursts from the PCC for four frames. The nominal configuration of PCC and ACC assignments at the four NCS sites is shown in Fig. 3, although the situation will be dynamic. Each NCS segment has been installed with eight fullduplex processing “threads” (i. e., a ViaSat MD-1324 modem and an E-Systems RT- 1771 radio) per side; a straightforward growth path to 16 channels per side has been demonstrated, but not yet fielded. Each side of an NCS segment consists of three basic elements as shown in Fig. 4: an antenna array, a set of unattended racks containing radio-frequency and modem equipment, and an operator workstation with personal computers (PC’s), monitors, and I/O equipment. )- 4 Ak AL Ab Antenna Array Ah } RF , , 1 1 1 1 I 1 1 R/T’s 1 , I 1 ,1 Modems 1 1553 BUS 1/0 ACC 10 Figure 3: UHF DAMA World-wide Architecture Modems t Pcc LAN LANT RF 1 I , 1 I 1 I I I R/T’s 1 I 1 t # 1/0 Racks } } Work station Figure 4: NCS Segment Architecture The antenna array for the current eight-channel configuration consists of two transmit and one receive 0-7803-4902-4/98/$10.00 (c) 1998 IEEE antennas per side. A photograph of the Hawaii antenna array is included as Figure 5. The photograph shows all three west side antennas, the east side receive antenna, and one of the east side transmit antennas which is shown in the foreground. In situations with multiple closely spaced satellites, the NCS antennas are typically pointed to access all spacecraft, rather than being optimized for a particular satellite. Figure 6: NCS Equipment Racks Figure 5: NCS Antenna Array Figure 6 is a photograph of the four NCS racks installed at the Hawaii site. Doors on the two east side racks are open, showing the modems (gray), transceivers (black), and other RF and AC power equipment front panels. A Rubidium frequency standard provides 10-11accuracy to facilitate synchronization of network terminals to the overall DAMA system. As an option, each side of arI NCS may be configured with a 5-kHz hot standby modem-radio “thread”. Then, in the event of the failure of a single PCC modem or radio, that side of the NCS will switch in the standby thread if the channel PCC fails to receive three of its own forward orderwires. This avoids takeover by the ACC that would occur after the fourth missed FOW, effectively keeping channel control at the local site. The “cost” of activating the standby thread is that one less channel can be actively controlled in this configuration. NCS operators are provided a DAMA control personal computer (PC) and workstation for each side of the segment, through which they can configure the system, make manual service assignments, monitor system usage, and be notified of any failures detected in the rackmounted equipment. The 5-kHz and 25-kHz applications execute simultaneously on the same PC. Additionally, the NCS operators have three I/O devices at their workstations that may be used to communicate with the population of network terminal operators: a voice terminal, a “red” (with encryption) data terminal, and a “black” (without encryption) data terminal. Each is attached to a specific modem, but “in-site routing” over a local 1553 bus automatically allows any I/O device to communicate using any one of the channels being controlled. The voice terminal is a KY-99 MinTerm, which implements standard 2.4 kbps Advanced Narrowband Digital Voice Terminal (ANDVT) speech coding and encryption. The data terminals consist of a software application on a desk-top personal computer. Black and red data terminals use separate PCs, and the red PC is interfaced to two KIV-7 data encryption units, providing full-duplex capability using KG-84 encryption techniques. The data terminal application is called the Integrated Data Terminal (IDT) and operates similarly for both red and black data, though the screen background colors are different in order to remind the operators of the sensitivity of the red data. Operators may either compose and send “notes” from the keyboard/monitor, or they can load files from diskettes into the PC and then send entire files. This application includes an end-to-end error detection code, so transmission errors are flagged to the receiving operators. The Hawaii East workstation is shown in Figure 7. The larger monitor supports the channel controller computer, while the smaller monitor is shared between the two data terminal computers. The classified encryption units are behind the paper covers in the photograph. Figure 7: NCS Workstation 0-7803-4902-4/98/$10.00 (c) 1998 IEEE DAMA Network Control Station application software consists of five computer software configuration items, resident in the Channel Controller, Relay Bus Controller, and Integrated Data Terminal personal computers and the modem. The modem firmware CSCI consists of a “Common” component and a separate component for each UHF MIL-STD waveform (181, 182 and 183) only one of which is resident in memory and operating at any given time. The orderwire encryption CSCI resides on a separate processor on a circuit card assembly within the modem. With a suitable antenna, radio and modem controller, the modem can function as a stand-alone network terminal. Figure 8 provides the characteristics of the NCS software elements, including size in terms of source lines of code ;LOC). CSCI Operating Language Size System (SLOC) Channel Windows NT Ada 106000 Controller Relay Bus DOS c 1800 Controller Integrated c Windows 95 4800 Data Term’1 Modem Custom c 78000 Real-time Exec Orderwire Custom Ada 5700 Encryption Real-time Exec Figure 8: NCS Software Characteristics V. PERFORMANCE The NCS Traffic Generator is one of two fundamental tools used to determine 5- and 25-kHz DAMA network performance and quantify the NCS ability to control multiple channels under heavy traffic conditions. The Traffic Generator is file-driven and runs as a software application on the same channel control computer as the NCS software. Written in Ada, the Traffic Generator reads an input data file to determine the traffic parameters to be used for generation of service request and service termination messages. The input data allows specification of uniform random or exponential distributed message or circuit point-to-point service requests and specification of uniform random or exponential distributed service durations. The input data provides the Traffic Generator valid terminal addresses to be used in generating service request messages and provides all terminal configuration data, including I/O data rate and baseband equipment type, for the required orderwire messages, Furthermore, the Traffic Generator inputs allow for a percentage of the total requests originated by a source address to be directed to a specific destination address. Only one active service request from an address is allowed at any time. However, for 25-kHz DAMA the Traffic Generator does allow requests to be generated while the address is active in a service. This accommodates terminals requesting retransmission of the slot assignment or requesting another user to be added to an existing service. The Traffic Generator implements the service request re-transrnission protocols described by the 25-kHz DAMA standard. Running concurrent with the NCS application, the Traffic Generator accesses the NCS service request queue to determine whether a generated service request was received by the NCS. If the request was not received, the Traffic Generator calculates the backoff interval for the request, re-transmitting the request at the designated time. A future upgrade of the Traffic Generator will implement the re-transrnission protocols described by the 5-kHz DAMA standard based on the number of ROW random access slots provided in a frame. The channel controller application maintains an “Event Log”, which records all forward and reverse orderwire traffic, with time-of-day tags, in a text file, During extended running periods, the Event Log can fill the allocated PC memory, at which point it is automatically archived onto the NCS hard disk. The Event Log is naturally useful to NCS developers and operators since it details “what actually happened”. It can also be postprocessed by our Performance Analysis Tool (PAT), an application written in the C language which tracks each service request from “life to death”, noting: origination time, number of request retransmissions, time placed in the NCS service queue, time removed from the service queue (i. e., assigned a slot on the channel), and time completed. As it processes an archived Event Log file, the PAT outputs a cumulative networking performance summary report record at the conclusion of each minute of NCS run time. This record includes true call arrival rate, reverse orderwire contention delay, service queue delay, true service duration, and blockage rate. The mean value, variance, and maximum value for each parameter is derived. In addition the 95%-tile service queue delay is reported. This performance investigation starts by defining, in as comparable a way as possible given waveform and Traffic Generator limitations, a priori “operating points” for individual 5-kHz and 25-kHz channels. Ideally these two operating points should relate to a worldwide population of network terminals and satellites and should constitute a relatively heavy, but theoretically supportable, traffic load. Step two is to drive the actual Network Control Station software application with the Traffic Generator configured to generate the desired load, and to determine network performance under the operating point load, using the following parameters: queueing delay and average channel use efficiency, (Other standard networking performance measures including reverse orderwire contention delay and call blockage rate were considered, but are presently not 0-7803-4902-4/98/$10.00 (c) 1998 IEEE modeled precisely in our Traffic Generator.) After interpreting the performance data for each operating point, we will adjust call rates, in an attempt to determine the effect on the pair of performance parameters. Constituents of the load are: the number of active terminals, the average call rate per terminal (Poisson distributed), and the average call duration (Exponentially distributed) per terminal, for various types of services (i. e., voice, data, messages). With their global, reliable message service capability, 5kHz DAMA channels might be considered most valuable for users that originate primarily data traffic with occasional voice communications requirements. Similarly 25-kHz DAMA channels, with their relatively quick voice turn-around performance (due to the short frames), are in some sense best suited for users who contribute a sizeable volume of voice traffic. So for this investigation we will split terminals into “data” and “voice” categories, assigning the former to 5-kHz home channels and the latter to 25-kHz home channels. Consider a time in the intermediate future, when 16000 two-port DAMA terminals (5090 of the total available address space) are deployed world wide, comprised of 6500 Army, 6500 Air Force and 3000 Navy units. And assume that the Army and Air Force terminals split among data and voice categories at the ratio 5500/1000. Navy terminals split with datahoice ratio 500/2500. Assuming uniform geographical distribution, that 25% of the terminals will be active at any given time, and that 8 satellites are in orbit, the result would be 360 data terminals and 140 voice terminals active simultaneously on any single satellite. Each of the UHF Follow-on (UFO) satellites has 38 channels: 17 “wideband” (25-kHz) and 21 “narrowband” (5-kHz). We will assume that three of each channel type are placed into the channel controller’s (MIL-STD-188181) single-access pool. If the communications planners are successful in distributing the population of active terminals evenly across the remaining DAMA channels, the result will be approximately 20 terminals per 5-kHz channel and 10 terminals per 25-kHz channel. Finally, the complete load input scenario must specify average call rates and durations for each type of traffic. Our scenario postulates that circuit services require 30 seconds channel time on average and that (5-kHz) message services average 3.6x104 bits. Simple analysis (as summarized in Figure 9, below) shows that if each “data” terminal (those assigned to 5-kHz channels) originates 1 voice call, 2 data circuit calls and 2 message services per hour, and if each terminal has a nominal link quality of 40 dB-Hz, the combined offered communications load (i. e., excluding orderwires) should on average total 534 building blocks per frame. Since each frame consists of 1024 building blocks (bb), and if we assume the orderwire occupies a constant 223 building blocks as shown in the table, then the 534 bb comprises a 66V0load on the 801 total bb available for communications. Each circuit service requires one request and one teardown reverse orderwire transmission and each message service requires only a single request orderwire transmission; thus the average total reverse orderwire load for 20 terminals is 160 transactions per hour or 0.2 usages for each of the assumed two reverse orderwire slots per frame. The actual number of reverse orderwire slots in the 5-kHz DAMA frame varies automatically from frame to frame. Only in the rare instance when the frame is almost entirely packed with communications bursts will the number of reverse orderwire slots be as low as two. Service Type Voice Data Msg Total COM FOW RNG ROW Total Ow Call Rate I/trm/hr 2/trm/hr 2/trm/hr Call Length 30 sec 30 sec 36 Kb Burst (bb) 832 526 99 Total bblhr 66560 84160 63360 100/hr 125 32 17 160 ROW/hr Grand Total Avg bbffr 166 210 158 534 125 64 34 223 757 Figure 9: 5-kHz Channel Load Analysis Theory tells us that a slotted Aloha system remains stable at loads below 0.37 per slot occurrence. Hence reverse orderwire contention with this scenario should be minimal. Similar analysis, as summarized in Figure 10, can be conducted for 25-kHz DAMA channels assuming that Frame Format {5, 5, 3} is in force, offering four 2.4 kbps and two 1.2 kbps slots. Each “voice” terminal originates 30 voice calls and 15 data circuit calls per hour. Each individual service requires two reverse orderwire transmissions: a two-party call and a call complete. The total average communications load for 10 terminals actively using the channel is 37.5 minutes of voice traffic per slot per hour, 37.5 minutes of data traffic per slot per hour and 0.35 reverse orderwire usages per frame. The resulting 62% channel occupancy is comparable to the 66% loading on the 5-kHz channels, but the total reverse orderwire load on the single RCCOW slot per frame is very close to the theoretical maximum and would result in significant orderwire contention (not currently modeled precisely in our Traffic Generator). Since the scenario was defined on a per-channel basis the Traffic Generator was run on a single channel at a time, separately for 5-kHz and 25-kHz DAMA. The initial 5- 0-7803-4902-4/98/$10.00 (c) 1998 IEEE I Service Type Voice Data Total COM Total Call Call I Rate Length 30/term/hr 30 sec 15/terrn/hr 30 sec I Total I /slot/hr 37.5 minutes 37.5 minutes 450 calls/hr 900 RCCOW/hr Ow Figure 10: 25-kHz Channel Load Analysis kHz performance measurement was taken over a period of 13 hours and resulted in 1292 service requests. This total is slightly low compared to the expected value of 1312 requests for that period, and differs as a result of a finite population effect; the Traffic Generator creates no new service requests from a given 110 port while that port is still busy with earlier traffic. Circuit service duration averaged 38 seconds overall. This is higher than the 30 second specified value because of two phenomena. First, each service must fill an integer number of frames, so the resultant rounding up adds a half frame (4.5 seconds) to the average. Second, contention in reverse orderwire slots delays some of the teardown messages; the associated service is extended until the teardown is finally received by the channel controller. Average queueing delay for our kHz frame format have thus far precluded deriving a purely analytical result against which to compare the 37 second experimental queueing delay value. We have analytical results for the less complex 25-kHz case, however (see below). Recall that the 5-kHz operating point was designed to generate approximately 6670 average channel usage. For one frame during each minute of the scenario, the Performance Analysis Tool tallies the number of building blocks assigned to communications services and the actual number of building blocks available for assignments after accommodating the FOW and contention ranging slots. Percent channel usage was derived by averaging this ratio over a period of 100 minutes beginning three minutes into the scenario. This is sufficient time for the channel loading to reach a steady state. Data for this run converged well to a 70V0 usage value with 100 samples. The extended service durations due to frame-filling and orderwire contention mentioned above are believed to be the primary reason why measured usage exceeded the predicted value. To assess the affect of load variation on our two performance measures, we conducted additional 5-kHz runs at 25, 50, 75 and 12590 of the operating point load. Running at 125% was feasible since loading on the reverse orderwire at the 1009o operating point was well under the slotted Aloha theoretical maximum. The results for both 0,8 0.7 006 a01 a 2 z z w Cm w z 0.2 0.1 ● o0 . 20 40 60 80 100 Average Load (calls per hour) 120 1 u. o 20 40 Average 100 60 80 Load (calls per hour) 120 1 o Figure 11: 5-kHz NCS DAMA Queueing Performance Figure 12: 5-kHz NCS DAMA Slot Usage Efficiency initial operating point was 37 seconds as depicted in Figure 11, Since all calls were of the same precedence there could be no preemption, so some calls had to wait until preyioufi uwm finifih~d on th~ Qhannvl b~fvr~ they could be assigned channel time. Complexities with the composite message and circuit traffic plus the dynamic 5- the initial operating point and the load excursions are plotted in Figures 11 and 12. Queueing delays should theoretically tend towards zero as load diminishes. Our expaimrzmd data rdlums th~ implemvntr+tkm, howev=, tending towards a half frame (4.5 seconds) because requests are presented to the channel controller at times 0-7803-4902-4/98/$10.00 (c) 1998 IEEE throughout the frame yet assignments can only be made in the FOW slot, at the start of each frame. At the other extreme, queueing delay increases beyond proportion as load increases. This conforms to theory, as will be shown more explicitly with the 25-kHz data. Channel usage tends towards zero, in theory and practice, as load diminishes. As the load increases, channel usage efficiency generally improves although the rate of improvement decreases as more and more blockage occurs. In summary for our 5kHz operating point and excursions, queueing delay and channel usage performance were generally as expected. The initial 25-kHz performance measurement was taken over a period of 4 hours and resulted in 1386 total service requests, This total is notably low compared to the expected value of 1770 requests for that period, the difference again a result of the finite network terminal population. Recall that the total population is only ten terminals per channel so the effect should be more pronounced than in the twenty terminals per 5-kHz channel case. In any event, the effect is immaterial, if the actual load is used for data analysis and plotting. Service duration averaged 31 seconds overall. This is higher than the 30 second ~pecified value because of the same-two Using the standard “M/M/l” queueing model (Reference 2) with parameters 1. = 0.098 calls/second and p = 33.5 seconds per call gives a theoretical corresponding queueing delay of 6.7 seconds. The difference can be attributed to framing necessary for a TDMA system and to various implementation losses. Recall that the 25-kHz operating point was designed to generate approximately 62% average channel usage. For one frame during each minute of the scenario, the Performance Analysis Tool tallies the aggregate assigned baseband data rate. For example if two 2400 bps and one 1200 bps slots were assigned, the aggregate would be 6000. Percent channel usage was derived by taking the ratio of aggregate assigned rate to 12000. Data for this run converged well to a 48% value with 100 samples. The deficit in loading (1386 actual vs. 1770 specified calls) is believed to be the primary reason why measured usage differed from the predicted value. In order to assess the affect of load variation on our two performance measures, we conducted additional 25-kHz runs at 25, 50, 70 and 90% of the operating point load. Running beyond 100% is not feasible because of the reverse orderwire limitation. The results for both the initial ● 4 ● 0 0 100 Average 200 , 300 I 400 5( 100 200 Average Load (calls per hour) Figure 13: 25-kHz NCS DAMA Queueing Performance phenomena as in the 5-kHz case. Filling slots completely adds a half frame (0.7 seconds) to the average and contention among call complete messages in reverse orderwire slots extends a small number of services. The pipeline for transferring CCOWS from the channel controller to the modem should add an additional two frames. Average queueing delay for our initial 25-kHz operating point was 8.4 seconds as shown in Figure 13. Load (calls 300 400 5( ) per hour) Figure 14: 25-kHz NCS DAMA Slot Usage Efficiency operating point and the load excursions are plotted in Figures 13 and 14. Experimental data are represented with diamonds and theoretical data with the solid curve. Queueing delays should theoretically tend towards zero as load diminishes. Our experimental data reflects the implementation, tending towards 2.5 frames (3.5 seconds) because of the CCOW pipeline. This effect overshadows 0-7803-4902-4/98/$10.00 (c) 1998 IEEE the half-frame average wait for the forward orderwire slot that dominated in the 5-kHz case. Towards the other extreme, queueing delay increases with increased load, measured data differing from theoretical by roughly 1.5 seconds. Channel usage tends towards zero, in theory and practice, as load diminishes. As the load increases, channel usage efficiency generally improves although the rate of improvement diminishes as more and more blockage occurs. Once again, for our 25-kHz operating point and excursions, queueing delay and channel usage performance were generally as expected. In practice, selecting an operating point will require network planners to trade-off between channel usage efficiency and queueing delay. In all likelihood, 30+ second queueing delays as observed with the initial 5-kHz operating point would be unacceptable to the using community, so the planners would need to back off the total offered load, for example to the 50% point. This would result in approximately 10 seconds mean queueing delay and 40 percent channel usage efficiency. In determining the precise optimum operating point communications planners would also want to consider the 90* or 95* percentile queueing delay, and they would want to enable per-call precedence rankings so that preemption could be used to minimize delays for important individual users. The Traffic Generator and Performance Analysis Tools can handle these more complex scenarios, which may be the subject of additional work. VI. FUTURE DIRECTION We begin this section by mentioning upgrades to MILSTD-188-18 1, which governs single access (i. e., nonTDMA/DAMA) use of the UHF transponders, since we anticipate certain advanced technical features from MILSTD-181 will ultimately be used in the DAMA standards also. In particular, the “B” version release of MIL-STD181 aims at significantly higher channel throughput by using multi-h continuous phase modulation (Reference 3). Early experimental results show and the new standard reflects that 9.6 kbps is achievable over 5-kHz channels and that up to 12 kbps is even possible. Similarly for 25kHz channels the new standard specifies 48 kbps and includes options for up to 64 kbps. For the long term, multi-h modulation can be expected to be included in both 5-kHz and 25-kHz DAMA standards. The first major revision to MIL-STD-188-182, termed “182A”, was approved on 31 March 1997. Any entirely new network terminal developments commencing after that date are required to conform to the A version. The channel controller software is being updated to support both versions of the standard and will be delivered to the operational sites in early 1999, along with the 25-kHz DAMA capability. The most notable new features provided by MIL-STD-182A include subframing, global message services, and punctured codes. Subframing is a technique which allows communications bursts to begin and end anywhere within the assigned time slot. Use of this technique means that terminals no longer must accumulate a full frame of data before starting a transmission; as a result end-to-end throughput delays are significantly reduced, a gain of particular benefit to voice users. With subframing, average round-trip delays are reduced from roughly 25 seconds to 4 seconds. Global messaging is a straightforward extension of the global circuit services already offered in MIL-STD- 182. Finally, new rate Y4 and 7/s punctured codes are derived by “throwing away” (i. e., not transmitting) channel bits from the standard MIL-STD-182 rate % code. Performance is somewhat degraded but that can be tolerated by terminals with strong link quality. The result is greater channel use efficiency, since shorter time slots are required. The “A” version of MIL-STD-188-183 was approved in March 1998. In this instance, the major enhancements are 5-kHz slave channels, 8-bit channel addressing, and ephemeris orderwire. The concept of slave channels puts all controlling orderwire on a 25-kHz DAMA home channel, but permits the terminals to frequency-switch every frame to a time slot on a 5-kHz channel whose frame timing is synchronized to the home channel. This is analogous to a frequency-switched slot connect (described previously) across multiple 254cHz DAMA channels. The benefit is letting terminals on a busy 25-kHz channel use slots on 5-kHz channels without sustaining the delay of permanently moving the terminals to a 5-kHz DAMA channel, The very earliest 25-kHz DAMA design used 6bit channel addressing, limiting the channel set to 64 elements. In the interim, many additional new satellites have been put in place, and the present standard frequency plan uses 8-bit addressing. This important enhancement will allow 25-kHz DAMA to operate on any of the extant channels. Finally, MIL-STD- 183A defines the formats of several new forward orderwires, which collectively will allow the channel controller to disseminate the complete set of orbital parameters for the satellites. This feature will support “passive ranging” by terminals with orbital calculation software; these terminals will be able to precalculate range to the satellite rather than sending a probe burst and measuring range. The 5-kHz DAMA and the first-generation 25-kHz DAMA Network Control Stations were developed and installed under Air Force sponsorship, The Navy’s long term role as UHF satellite communications system developer, however, was reflected in that all NCS and Navy Computing installations are at Telecommunications Area Master Station (NCTAMS) sites. The Navy is presently directing the evolutionary Joint (UHF) Milsatcom Network Integrated (JMINI) Control System program whose charter is to upgrade both 0-7803-4902-4/98/$10.00 (c) 1998 IEEE 5-kHz and 25-kHz DAMA. Two significant early JMINI improvements will be inclusion of a Network Management System (NMS), and achievement of channel synchronization. The NMS will be a communications planning tool, instances of which will reside at Joint Chiefs of Staff, various commanders in chief (CINCS) who “own” individual satellite channels, and the NCS sites, all interconnected through a secure network. It will be through the NMS that all decisions concerning apportionment of the channels and ranking of individual terminal’s and network’s importance will be made and disseminated. The ultimate destination of the “rules” for assignment of capacity will be downloaded into the NCS computers and implemented in real-time as requests for service arrive. Presently, frame timing for DAMA channels is synchronized on a per site basis and at the time of transmission from the ground. Maximally useful system synchronization requires alignment at the satellite, and for every channel (both 5-kHz and 25-kHz) world-wide. Benefits are numerous and major, the prime ones being allowing cross-channel scheduling and precluding frame timing jumps upon transfer of control to an alternate station. This will be a big step towards “seamless” 5 and 25-kHz DAMA wherein users will not need to concern themselves with implementation details; they would simply request services and be assigned an appropriate slot. After more than 20 years of evolution and development, network controllers for both 5-kHz and 25-kHz standard UHF DAMA waveforms have been recently fielded. Large numbers of network terminals designed to the standard waveforms are just now going operational, and commanders are expected to convert more and more channels to the DAMA mode as the user population warrants. This paper addressed performance of the overall DAMA system. The approach was to use a discrete event simulator to impose traffic loads close to the theoretical maximum on the actual channel controller software, and to measure key networking performance parameters, It was demonstrated that the system provides good quality service under heavy loads. ACKNOWLEDGEMENTS The authors are most indebted to Dr. John Bacigalupi, who extended the Traffic Generator capability to include 25kHz DAMA and who conducted all of the simulation runs. Development and fielding of the DAMA Network Control Station were funded by the U. S. Air Force Electronics Systems Center, Maj. Brian Casey, Program Manager. Enhancements to channel controllers are presently being conducted under the auspices of the U. S. Navy Space and Electronic Warfare (SPAWAR) center, Mr. James Parsons, Program Manager. REFERENCES 1. L. E. Taylor and S. L. Bernstein, “TACS – A Demand Assignment System for FLEETSAT,” IEEE Transactions on Communications, Vol. COM-’27,No. 10, October 1979. 2. D. Gross and C. M. Harris, “Fundamentals of Queueing Theory,” Wiley and Sons, 1974. 3, I. Sasase and S. Mori, “Multi-h Phase-Coded Modulation,” IEEE Communications Vol. 29, No. 12, December 1991. 3. for Standard “Interoperability MIL-STD-188-181, Dedicated 5-kHz and 25 kHz UHF Satellite Communications Channels,” 18 September 1992. 4. MIL-STD-1 88-182, “Interoperability Standard for 5-kHz UHF DAMA Terminal Waveform,” 18 September 1992. 5. MIL-STD-188-183, “Interoperability Standard for 25-kHz UHF TDMA./DAMA Terminal Waveform,” 18 September 1992. 0-7803-4902-4/98/$10.00 (c) 1998 IEEE