Maximizing Data Volume for Direct to Ground Satellite Systems David Carek Satellite Networks and Architectures Branch NASA Glenn Research Center David Andrew Carek, P.E.
Overview Study initiated as part of ACAD (Advanced Communications Architecture Demonstration) Direct to ground communication system for ISS Reliable transmission of latency tolerant payload data Objective –maximize data volume David Andrew Carek, P.E.
ACAD Link Initially examined ACK based protocol - TCP Requires one ACK for every other packet Large bandwidth asymmetry requires large packet size on downlink E.g. 622Mbps down; 622Kbps up => 1000:1 asymmetry Required large downlink packet size to prevent ACK feedback congestion on uplink > 20Kbyte downlink packet required for 40byte uplink ACK packet Needed to determine implication of bit errors on large packet transfers David Andrew Carek, P.E.
Factors Affecting Data Volume Contact time (satellites) Satellite altitude and orbit inclination (fixed) Ground station minimum elevation angle (design parameter tied to link budget) Ground station latitude (design parameter) Transmission Rate Function of link budget transmitter power, antenna size, etc.(design parameter) Protocol Efficiency David Andrew Carek, P.E.
Factors Affecting Data Volume (Contact Time) STK Simulation Ground Contacts with ISS ISS Orbit Inclination: 51.6 deg Average Alt: 380 km Ground Station Lattitude: 45 deg Min Elevation Angle: 10 deg. David Andrew Carek, P.E.
Maximizing Contact Time GRC White Sands David Andrew Carek, P.E.
Factors Affecting Data Volume (Transmission Rate) Function of Link Budget with many interrelated factors Ground Segment Antenna size – Goal for small transportable ground terminal (1.2 meter dish) Space Segment Transmit antenna type/beam width (gimbaled horn) Frequency/Bandwidth (~27GHz/500MHz) RF amplifier power David Andrew Carek, P.E.
Factors Affecting Data Volume (Protocol Efficiency) Protocol algorithm (e.g. TCP) Congestion control degrades efficiency when actual loss is corruption Acknowledgment feedback congestion high bandwidth asymmetry degrades efficiency Information efficiency Amount of end user data carried over link relative to total data transmitted David Andrew Carek, P.E.
Factors Affecting Data Volume (Protocol Efficiency) Information Efficiency Packet delivery efficiency (driven by bit errors) Function of error free packets received Increased packet size = decreased efficiency Header efficiency (for fixed size header) Function of data allocated to header vs. user data Increased packet size = increased efficiency David Andrew Carek, P.E.
Types of Bit Errors Gaussian Bit Errors Burst Errors Systematic Errors Random RF Noise Burst Errors Random occurrence of multiple bit errors E.g. rain, snow, particles, etc. Systematic Errors Often caused by internal electronics Can be periodic distribution of single bit error or burst error Pattern Sensitive Bit Errors Form of Systematic Errors Influenced by data pattern within stream *Reference: An Introduction to Error Location Analysis, Are all your errors truly random?, Application Note 1550-2; Agilent Technologies, 2000 David Andrew Carek, P.E.
Packet Delivery Efficiency Data Stream with BER = 1x10-1 Bit efficiency = 90% Random - Packet Efficiency = 33% - Packet Efficiency = 67% - Packet Efficiency = 0% X Packet 3 Packet 2 Packet 1 Burst X Periodic X David Andrew Carek, P.E.
Header Efficiency Quantities for Illustrative Purposes Only Burst Error Example BER = 1x10-1; Packet Size = 10 bits; Header Size = 5 bits Bit efficiency = 90% Packet delivery efficiency = 67% Information efficiency = 33% X H I Packet 1 Packet 2 Packet 3 X = Error H = Header bit I = Information bit (user data) David Andrew Carek, P.E.
Deterministic Efficiency Probabilistic Efficiency Assumes n can be determined n = average # bit errors per errored packet Probabilistic Efficiency Information Efficiency Information Efficiency David Andrew Carek, P.E.
BER 10-5 10-4 10-3 David Andrew Carek, P.E. Deterministic Efficiency Header loss dominates Packet loss dominates Deterministic Efficiency Probabilistic Efficiency David Andrew Carek, P.E.
Packet Size vs. Information Efficiency Probabilistic Equation (random single bit errors) (93, 27%) 2% Actual Difference (71, 25%) Deterministic Equation (worse case periodic error distribution; n=1) (71, 19%) BER = 10-3 Sh = 40 bytes David Andrew Carek, P.E.
Deterministic Efficiency Probabilistic Efficiency Assumes n can be determined n = average # bit errors per errored packet Probabilistic Efficiency Information Efficiency Information Efficiency Optimal Packet Size Optimal Packet Size David Andrew Carek, P.E.
Information Efficiency – Periodic Error Distribution (Sp = variable, Sh = 40 bytes; n = 1) David Andrew Carek, P.E.
ISS Subsystems and Internal Payloads ACAD Processor/Storage Segment VME Bus Mass Storage Device Main Processor Analog/Discrete I/O TBD R/W IF Dual LVDS HRDL/1553/100BT IF 100Mbps APS COR to HRFM to Ku Payload HRDL 1553 PL/MDM Ethernet RS232 Gimbal Assembly ISS Subsystems and Internal Payloads ACAD Processor/Storage Segment L-band Up Data Flow Unreliable Data Flow Reliable Data Flow Reliable NAK’s OMT Modulator B Modulator A Upconverter A Upconverter B HPA A HPA B RHCP Ka LHCP Ka Antenna Pointing Controller Ka Down LNA Uplink Antenna Downconverter Demodulator David Andrew Carek, P.E.
Conclusions BER alone is not enough to design a link Bit error distribution must be accounted for Upper layer protocol must be considered Properly sizing packet sizes can maximize information efficiency and extend link availability Auto-tuning protocols based on packet error ratios could extend link availability and efficiency David Andrew Carek, P.E.