Distributed Power System For Large Orbital Infrastructures: Architecture, Requirements Analyses and Lesson Learned

Antonio Ciccolella

ESA-ESTEC, D-TEC/EEE, Keplerlaan 1 - 2201 AZ Noordwijk, The Netherlands

Abstract learned during the development and operational phase of Modules and payloads. Distributed power systems offer many benefits to system designers over central power systems such as I. INTRODUCTION: ISS POWER ARCHITECTURE reduced weight and size. Distributed systems also allow the designers to control the quality of power at The ISS is a joint project of six space agencies: the different loads and subsystems, since DC-DC U.S. National Aeronautics and Space Administration converters allow close regulation of output voltage (NASA), Russian Federal Space Agency, Japan under wide variations of input voltages and loads. Aerospace Exploration Agency (JAXA), Canadian Distributed power systems also provide a high degree Space Agency (CSA/ASC), Brazilian Space Agency of reliability because of the isolation provided by (AEB) and the European Space Agency (ESA). DC/DC converters; it is very easy to isolate system The space station is located in orbit around the Earth failures and provide redundancy. These systems are at a nominal altitude of approximately 360 km (220 also very flexible and easily expanded. miles), a type of orbit usually termed low Earth orbit. The actual height varies over time by several This paper addresses the DC distributed power kilometers due to atmospheric drag and re-boosts. It system of the International Space Station, which is a orbits Earth at a period of about 92 minutes. specific case of this kind of distributed system. In many ways the ISS represents a merger of previously planned independent space stations: It is a channelized, load following, DC network of Russia's Mir 2, United States' Space Station Freedom solar arrays, batteries, power converters, switches and and the planned European Columbus Attached cables which route current to all user loads on the Pressurized Module (APM). Today it represents a station. The completed architecture consists of both permanent human presence in space, as it has been the 120-V American and 28-V Russian electrical manned with a crew of at least two since November networks, which are capable of exchanging power 2, 2000. through dedicated isolating converters. It is serviced primarily by the Space Shuttle, Soyuz and Progress spacecraft units. In 2006 a new The presence of DC/DC converters required special European service vehicle, the Automated Transfer attention on the electrical stability of the system and Vehicle (ATV), will be operational. ISS is still being in particular, the individual loads in the system. This built, but it is home to some experimentation already. was complicated by complex sources and undefined At present, the station has a capacity for a crew of loads with interfaces to both sources and loads being three. designed in different countries (US, Russia, Japan, At its final integration, the ISS characteristics will be: Canada, Europe, etc.). These issues, coupled with the program goal of limiting costs, have proven to be a • Dimensions: 87 m (L) x 106 m (W) significant challenge to the program. • Weight: 425 000 Kg • Pressurised Volume: 1160 m3 As a result, the program used an impedance • Power: 110 kW (total), 45 kW for Users. specification approach for system stability. This approach is based on the significant relationship NASA is responsible for the overall system, between source and load impedances and the effect including Space Station integration and verification, of this relationship on system stability. It is limited in supported by proactive and intense interactions with its applicability by the theoretical and practical limits all the participating parties. on component designs as presented by each system The ultimate power source for the station is the segment. Consequently, the overall approach to sunlight, which is incident on solar arrays deployed system stability implemented by the ISS program as paired sets. consists of specific hardware requirements coupled Each array wing consists of two thin blankets held with extensive system analysis and hardware testing. under tension on each side of a central collapsible Highlights of both experimental and analytical mast. The entire assembly turns on a “beta gimbal”, activities will be shown, as well as some lesson which provides one axis of rotation for solar pointing.

A second orthogonal axis of rotation is provided at power directly to core system loads and to all the “alpha gimbal”, where the entire solar power payloads. module connects to the rest of the truss structure of the station (see Fig. 1). These arrays each provide a The DDCUs provide excellent isolation between total of 25 kW of power during the sunlit portion of input and output ports, ensuring that perturbations on the 90-minute, low-earth orbit of the station. Due to one branch of the primary system will have minimal solar cell degradation over time, and different impact on other branches. Downstream from each pointing angles, the optimum output voltage will vary DDCU are Remote Power Controller Modules between 160 and 140 Volts over the life of the (RPCMs), which are computer-controlled, solid-state station. circuit breakers that branch out the power to the loads.

II. DESIGN ISSUE: IMPEDANCE MATCHING

The Power System of the ISS is distributed rather than centralised, i.e. the power is processed by multiple cascaded stages of switching regulators. The canonical arrangement of distributed power systems consists mainly of:

• A stage of regulators, acting as line conditioners, which take the unregulated input voltage and convert it to a regulated bus voltage. This bus distributes power to

the system. • Cascaded stages of parallel regulators, Fig.1 – Power Architecture of ISS (one channel) acting as load converters, which take the power from the above bus and generate the These are remotely controlled switching boxes, used appropriate output voltage required by the to route power between redundant channels, and to loads. direct power to Dc-to-Dc Converter Units (DDCUs) and local loads. The adoption of a Distributed Power System for the ISS is dictated, inter alia, by the difficulty in The first switching converters in the power string are maintaining regulation at the load location with a the BCDUs, which regulate charging of the Nickel- centralised power distribution, due to the possible Hydrogen batteries during sun exposure, and serve long distance between load and source. By placing the dual role of discharge control and bus regulation intermediate converters closer to the loads, the during eclipse. The units are bidirectional, stepping requested voltage is regulated with higher accuracy. the voltage down to the battery level of Another advantage of the Distributed Power System approximately 72 Volts when charging, and stepping for large spacecraft is that cascading regulators the battery voltage up to the main bus voltage when inherently enhance isolation between source and discharging. loads. Furthermore, the use of parallel load regulators implies modularity, which allows the spacecraft Switching between charge and discharge is design to flexibly manage power reconfiguration and automatically triggered by the bus voltage set-point, reliability by adding redundant modules. even during sunlight, if load demand exceeds the ability of the solar arrays to provide full power, the These features of the distributed power system are batteries will switch in to pick up the slack. achieved at the expense of higher system complexity, which may render the analysis and the design of the ARCUs allow interconnection between the US-built system as impracticable if it is considered as a whole. 160/120-Volt system and the Russian-built 28-Volt However, the intrinsic modularity suggests the use of system. Several of the Russian modules have their an approach for both the design and the analysis, own independent solar arrays, providing important which foresees the complete system as partitioned in levels of redundancy for power availability. When smaller power subsystems that can be independently power is being transferred between systems these and individually considered. converters play a role in the overall bus impedance of both the systems. The system is then built through the appropriate The DDCUs provide the interface between the interconnection of the previously mentioned primary 160-Volt system and the more tightly subsystems, whose interaction needs to be carefully regulated 120-Volt secondary system, which provides

analysed eventually to generate compatible their input filter loaded by a negative resistance, specifications. whose magnitude is |R|=V2/P. This is not valid when the input filter is integral part This approach, widely used in the frame of the ISS of the power cell, which can occur for some [1], is based on the underlying assumption that the topologies. The above impedance matching concept interacting subsystems are linear, thus small signal was applied to determine realistic boundaries to levy stability is the objective to achieve. requirements on loads and subsystems in flight elements (i.e. APM, JEM). Namely, we consider (Fig. 2) a source block with It implied analysis of multiple loads with a known forward gains FS and a load block with forward gain source and the requirement were responsibility of FL. We designate ZS as the source impedance and ZL NASA, with the necessary attribute of being not as the load impedance. overly conservative or impractical to implement NASA initially identify two load’s category:

• Complex (load assemblies or equipment racks): interface B • Simple (individual loads): interface C

Interface B category (required to have their own power distribution and protection) was subdivided in function of their feeder rating (10-12 A, 25-30 A, 50- 60 A). Fig.2 - Source Block cascaded with a Load Block Interface C category was subdivided in function of their line branch rating (1.5-3.5 A, 10-12 A, 25-30 A) Working with the hybrid g-parameters one can verify Four ranges of harness length were also considered. that the input-to-output transfer function FSL of the Monte-Carlo simulation (NASA) with ad-hoc two cascaded blocks results: representative loads gave rise to impedance requirements for each category considered. An FF LS ZS example is given in figures 3 & 4. FSL = (1), with Hm = (2) + H1 m ZL Load Impedance Magnitude Limits Hence, the impedance ratio H can be considered as m 63 the open loop gain of the integrated system. 100. 56 21 Minimum Magnitude (See Note 1) 56 63 7 10. Minimum Magnitude (See Note 2) 21 When |ZS| << |ZL| for all the frequencies of interest 3.5 (e.g. f < 100 kHz), the stability of both the individual 1.75 7 1.022 1.162 source and load blocks guarantees stability for the 0.77 0.784 0.896 3.5 1. 1.75 0.966 1.036 integrated system. In general, this condition is not 0.77 0.784 0.812

realistic. In some frequency intervals we can expect Magnitude (Ohms) 0.1 |ZS| > |ZL|. Although this case does not necessarily imply a stability problem, additional analysis is required for the open loop gain H (e.g. by applying 0.01 m 10 100 1,000 10,000 100,000 the Nyquist criterion) to determine whether system Frequency (Hz) stability is achieved. Fig. 3 - Load Impedance Magnitude Limits Hence, the following ingredients:

Load Impedance Phase Limits • Impedance characterisation of the known power sources (output Z) and loads (input 90 90 90 90 90 90 90 90 90 90 90 90 Z), in all their operational mode 60 • Layout of the power sub-network under 30 Upper Phase Limit Lower Phase Limit (See Note 1) 0 consideration Lower Phase Limit (See Note 2) -30 -64 -60 -60 -60 • Harness length and inherent AWG (mainly -73 -60 -90 -100 -60 -60 -60 R, L parameters) Phase (Degrees) -90 -108 -110 -64 -118 -73 -125 -90 -120 -110 -125 allow the system designer to obtain a preliminary -150 -132 -127 -127 assessment of the impedances’ congruence within the -180 considered sub-network by circuital simulation 10 100 1,000 10,000 100,000 Frequency (Hz) For the frequencies of interest, DC/DC converter’s input impedances can be accurately modelled with Fig. 4 – Load Impedance Phase Limits

If the load impedance is below the curve of fig. 3, The method used to achieve unified power stage then the stability requirement is met. If not, we have model of the Weinberg converter was derived. to make sure that the phase is between the limits of The discontinuous mode, the boundary between fig. 4 at the frequencies where intersections occur continuous and discontinuous mode and the between requirement and effective load magnitude. continuous mode were separately and sequentially considered with the state average approach. The load requirements have been obtained Consequently, a single model wae derived that can considering worst case 3 dB gain margin and 30º- automatically detect the mode of operation as a phase margin for the open loop gain. function of the load power consumption. The model was then completed with the Current Programming Control feature and the relevant III. ESA ACTIVITIES FOR COLUMBUS STABILITY sampling effects. Also in this case, a unified model of the Current Feedback Loop, which adapts itself to the The distributed primary power system of the DCM and CCM for constant frequency switching International Space Station (ISS) provides power to was derived and implemented. The models generated some attached modules via DC-DC Converter Units have been implemented using PSPICE, making (DDCU), either in standalone or in parallel extensive use of its Analogue Behavioural Modelling configuration. Each can deliver up to 6.25 kW with capabilities. 150 % peak power capability at 124.5 V nominal. Simulation results and comparison with test data The DDCU are manufactured by Boeing and (output impedance magnitude and phase and step represent the power interface between the ISS and the load response) are given. attached modules using the 120 V bus. The electrical power downstream of the DDCU is defined as the secondary power system. It involves a wide variety of loads and operational modes that, at the time of the design, were not known in detail or can be anyway replaced in orbit. The Columbus Attached Pressurised Module (APM), which is the European contribution to the ISS, is a user of the secondary bus regulated by the DDCU. In order to define the EMC and power stability interface documents with NASA on a sound basis, ESA developed an anchored model of the DDCU for requirement analysis and verification purposes. Stability requirements were introduced at a late stage in the ISS programme: Industrial Contract and Fig. 5a – Output impedance single DDCU, 1 A, Specification were already frozen for Columbus measured between ESA and Industry 100d Changing the established baseline might have implied a severe cost impact: many units were being manufactured. To allow an industrial assessment of the problem, ESA developed a model of the power source, 0d anchored by test data. On the other hand, Industry initiated a parallel activity on the loads. This built confidence to accept the requirements in ICD. Specifying a converter to work efficiently over a -100d VP(Vu1)-IP(Vobs_1) wide range of loads can result in a design requiring -10 both continuous and discontinuous conduction mode of operations. This is the case of the DDCU, which has Weinberg topology with Current Control. -20 The comprehensive analysis of such converters calls for a single model, which can automatically detect whether the operational mode is CCM or DCM and -30 perform consequently. Convergence to the right mode in a DC analysis, SEL>> -40 correct small signal response in an AC analysis and 100Hz 300Hz 1.0KHz 3.0KHz 10KHz 30KHz 100KHz VDB(Vu1)-IDB(Vobs_1) capability of changing operation modes during a Frequency transient analysis are the key attributes of the model. Fig. 5b – Output impedance single DDCU, 1 A, simulated

100d

0d

-100d VP(Vu1)-IP(Vobs_1) -10

-20

Fig. 6a – Output impedance single DDCU, 26 A, -30 measured 100d SEL>> -40 100Hz 300Hz 1.0KHz 3.0KHz 10KHz 30KHz 100KHz 50d VDB(Vu1)-IDB(Vobs_1) Frequency Fig. 7a – Output impedance parallel DDCU, 1 A, simulated 0d

-50d VP(Vu1)-IP(Vobs_1) -10

-20

-30

SEL>> -40 100Hz 300Hz 1.0KHz 3.0KHz 10KHz 30KHz 100KHz VDB(Vu1)-IDB(Vobs_1) Frequency Fig. 6b – Output impedance single DDCU, 26 A, Fig. 8a – Transient due to a 80 A-12 A step load, simulated measured

Fig. 7a – Output impedance parallel DDCU, 1 A, Fig. 8b – Transient due to a 80 A-12 A step load, measured simulated

The simulations match very well the test data and the Also, this activity gave ESA confidence to procure model has been a useful tool for ESA-NASA to the Electrical Ground Support Equipment (EGSE), define the power interface specifications between with impedance characteristics reproducing the flight COLUMBUS APM and ISS. configuration. Two test campaigns were carried out with this The integration of the COLUMBUS APM flight objective. The first test campaign addressed the model with the DDCU in Bremen, required by the verification of the dynamic characteristics of the bilateral verification and integration test between power bus declared by the Russian Industry in the ESA and NASA showed stability was achieved with preliminary Interface Control Document (ICD), in very good margin. Figures 8a has been derived in that both Differential Mode (DM) and Common Mode frame. (CM). ESA reconstructed with high accuracy the impedance measured at the interface, synthesising the equivalent network of lumped elements giving IV. ESA ACTIVITIES FOR ATV STABILITY identical results. Severe discrepancy in CM was found between the ICD and measurements. ESA The Automated Transfer Vehicle (ATV) is an considered the CM impedance was seriously affected automatic, unmanned space transport vehicle being by capacitors present in the Service Module EGSE. developed by the European Space Agency, carrying The second test campaign confirmed the above cargo and supplies from Earth to the International hypothesis and, moreover, it defined the boundary of Space Station (ISS). It has also the potential to the impedance expected in flight configuration and perform the re-boost manoeuvre, to lift the whole showed a strong correlation between CM impedance Station into a higher orbit. The ATV will be docked and CM voltage transients. Also, this activity gave a to the Russian Segment of the ISS, from which it will deep insight in the EMC of the Service Module, draw part of its power required. providing useful information for the EMC and Power The Russian Service Module provides a 28V Quality specification. regulated power bus that is floating from the structure In addition to providing impedance characterisation potential. Hence, particular attention is required to results of the Russian Service Module's 28V power design an interface for the ATV, which has four bus, this paper focuses on the experimental approach separate internal power busses referred to structure. and on the modelling methods that have been The design and verification of the ATV power extensively used to elaborate the measurements, system interfacing with the Russian Service Module, leading us to finalise the activity. needed crucial information on the source impedance to: Particular emphasis was given to the method of synthesis of the lumped element network that • Ensure by design that small signal stability reproduces simultaneously both the DM and the CM between the Russian Power source and the impedance. ATV load is achieved. • Define the conducted EMC requirements We applied the same concepts outlined in paragraph giving the levels expected at ATV actual 3, where the load block is a stage of four independent power interface connecting node and their parallel regulators, each designed as a Power Control inherent test set up. and Distribution Unit of the ATV. • Represent the source impedance seen by ATV when it is docked to the Russian An accurate ground emulator of the Service Module, Service Module through an equivalent comprehensive of all the relevant loads, is available network, for verification purposes at system at Energia's premises, in Russia. It was declared to be level. fully representative of the flight hardware.

Although a complete hardware emulator of the This allowed us to verify by test the information Russian Service Module is available in Koriolev given by the ICD at the ATV-Service Module (Russia) at the RSC ENERGIA premises, the size of interface. Hence, ESA and Energia planned a the ATV sensibly precludes transportation of the measurement campaign for the Service Module spacecraft to Russia to test the interfaces. source impedance (i.e. the ZS), to consolidate the ICD and provide the PCDU manufacturer substance to The above constraint dictated an experimental build a stable system (i.e. to design for an appropriate characterisation of the source impedance of the ZL). Russian Service Module. This allowed ESA to derive Given the floating nature of the Russian Bus, the the necessary information to design a stable power measurements involved both the differential and the interface. common mode impedance. The test set-up for the source impedance measurement of the Russian

Service Module at the ATV interface is shown in Fig. 11 - Typical measurement for the Service

Figs. 9a andSM/ATV 9b. Interface Module common mode

SM 28V + The objective is to synthesise the network of lumped ATV elements, which reproduces the above measurements SM LOAD Resistive Load for both differential and Common Mode, which 28V RTN Electronic results to have the topology and the values of fig. 12. Load Gain - Phase Analyser HP4194A

Oscilloscope Printer

Fig. 9a Source Impedance Test Set-up Differential Mode SM/ATV Interface Oscilloscope SM 28V + ATV SM LOAD Resistive Load 28V RTN

Power Gain - Phase Analyser Amplifier HP4194A Fig. 12 - Integrated equivalent network of the Service Printer Module at the ATV interface

Fig. 9b - Source Impedance Test Set –up, The simulation of this network with PSPICE provides Common Mode the results shown in Fig. 13 and 14:

Several measurements were performed in different operational modes of the Service Module, while the EGSE configuration remained unchanged. The results exhibited almost the same behaviour for both the differential and the common mode, respectively. Figures 10 and 11 show the typical results obtained for both differential and common mode.

Fig. 13 - Results of the simulation of the equivalent network for differential measurements

Fig. 10 - Typical measurement for the Service Module differential impedance

Fig. 5 - Typical Measurement for the Service Module Common Mode Impedance

Fig. 14 - Results of the simulation of the equivalent network for common mode measurements Fig. 10 - Typical measurement for the Service Module differential impedance

The results match very well with the measurement. circuitry of the SSPCs was susceptible to a common This work was performed by ESA to understand the triggering event. exact nature of the electrical interface of the Russian Service Module. This is presently allowing Industry Subsequent measurements of the bus common mode to design both a stable power interface for ATV and a transient voltage at various SSPC inputs, while the power emulator for verification purposes, which is hot-to-chassis fault was reproduced, proved this adherent to the in-orbit condition. Realistic EMC assertion. requirement were also derived at the interface ATV/SM. We detected negative transient’s peaks as high as – 140 V between the bus return line and the chassis at V. CASE OF TROUBLESHOOTING the input of adjacent channels (Fig. 16).

The Columbus APM, provided by ESA, takes power from two parallel DC/DC converters (DDCU), which are provided by NASA. The PDU distributes power lines inside the Columbus Module. Solid State Power Controllers (SSPC), whose simplified block diagram is given in fig. 15, are in place for overload protection purposes. NASA and ESA agreed a sequence of bilateral tests to verify the compatibility of their respective hardware, including critical operational scenarios (such as faults). The management of power faults is related to safety and it is essential for a manned orbital infrastructure. Transients generated in fault conditions shall also be Fig. 16 – Common Mode Voltage on adjacent considered in the EMC specification. channels

The common mode voltage transient is generated by the reaction of the return power cable’s inductance to the sudden current interruption. The common mode voltage transient is seen by the auxiliary circuitry of the SSPCs, which are connected in parallel on the power bus.

This specific type of fault, associated to long cables, generates transient levels that exceed the susceptibility threshold of the auxiliary circuitry of the SSPCs. Hence, the electronics of the SSPCs switches off and all the SSPCs go in OFF status as well. This was eventually confirmed by modelling, where the waveform was accurately reproduced.

Fig. 15 – SSPC simplified block diagram The insertion of a tranzorb clamping the voltage between return line and chassis was the fixing. During faults on a power line, only the affected SSPC The tranzorb used was previously designed and should trip, so isolating the faulted channel (Pass/Fail manufactured on specification written ad hoc by criterion). All the other channels should be ON NASA. The test in matter consisted of characterising the behaviour of the EMC and power quality of the bus It was subjected to a flight qualification campaign. in presence of a representative SSPC breadboard. The tranzorb in matter is a combination of three Faults consisted of both hot-to-return and hot-to- groups of two unidirectional Transient Voltage chassis short circuits on various channels, induced Suppressor diodes mounted back to back, in parallel. through a FET box at the switch output. This ensures at least one failure tolerance. During the fault hot-to chassis on the channel A, all the SSPC hybrids tripped at once, shutting down the The tranzorb clamping voltage is between 20 and 30 power to the all the resistive loads in the test set up. V and it can withstand very high currents for short This anomaly is typical of common mode times. phenomena. We postulated that any individual

Fig. 17 – Common Mode Voltage with the transzorb

The lesson we learnt is that, in case of fault, the performances of the SSPCs depend greatly on:

• Nature of the overload. • Inductance of the operational environment (both upstream source and downstream) • Interaction of the auxiliary circuitry with the transient phenomena associated to the fault.

Hence, design requirements must account for the Fig. 18 – Space environment and effects on expected system environment rather than rely on Spacecraft general practice rules.

VI. PART RELIABILITY & SPACE ENVIRONMENT VII. CONCLUSIONS

For Space Vehicles in low inclination (< 28°) Low We have given an overview of distributed power Earth Orbit ((LEO),< 500 km or 270 nmi) in both system for space applications and some problems we northern and southern hemispheres, typical dose rates faced. due to trapped Van Allen electrons and protons are We have learned that: 100-1000 rad(Si)/year For Space vehicles in higher inclinations (28° < I < • Design requirements must account for the 85°) LEO in both northern and southern hemispheres, expected system environment rather than typical dose rates due to increased number of trapped rely on general practice rules electrons are 1000-10000 rad(Si)/year • Power instability although it is a mainly a Radiation environments for both TID (Time local issue, it can result in cost & schedule dependent) and SEE (time independent) are specified impacts and in loss of scientific or in the ISS System documentation and are applicable commercial yields for space systems to components • Components, IC, FPGA, Power Hybrids etc These two effects have to be addressed separately in shall be closely monitored and screened to the design. Available guidelines define the basic rule clearly understand what is brought into the for selection of radiation-hardened devices within system & to meet lifetime requirements specified safe limits. Cooperation and the project teams’ efforts technically concurring to achieve the objectives of a space mission is the key of success

VIII. REFERENCES

[1] Gholdston et al. - Stability of large DC Power Systems using switching converters, with application to the International Space Station - Proc. IECEC 96 - 96079