SpaceWire — high-speed serial link standard for spacecraft

5.5.27 · Coding › Embedded Systems & Real-Time Software

SpaceWire Actually Hai Kya

Core Architecture Components

Physical Layer: LVDS differential pairs (Data + Strobe signals), low power (~10 mW per link)
Character Layer: 8-bit data bytes + parity bit, Data-Strobe transitions ke roop mein encoded
Packet Layer: Variable-length packets with destination address, data payload, EOP/EEP markers
Network Layer: Wormhole-routed switching fabric (header propagate hone ke saath packets path establish karte hain)

Figure — SpaceWire — high-speed serial link standard for spacecraft

SpaceWire Kaise Kaam Karta Hai: Signal Encoding

Formula Data-Strobe (DS) Encoding — toggle rule

SpaceWire alag clock nahi bhejta. Yeh do signals, Data aur Strobe, transmit karta hai, is rule ke saath (ECSS-E-ST-50-12C ke anusaar):

Data wire par actual bit value carry karta hai.
Strobe tab toggle karta hai jab do consecutive bits same hoon, taaki exactly one of {Data, Strobe} har bit period mein change kare.

Equivalently, transmitter yeh invariant maintain karta hai: $Data (n) = b_{n} Strobe (n) = Strobe (n - 1) \oplus \overline{(b_{n} \oplus b_{n - 1})}$ Kyunki exactly ek line har bit per toggle karti hai, receiver do lines ki transition activity ko XOR karke clock recover karta hai: $RecoveredClock (n) = Data (n) \oplus Strobe (n)$ aur $b_{n}$ pane ke liye har recovered clock edge par Data line sample karta hai.

Yeh step kyun? DS ka point data+strobe ko bit value mein XOR karna nahi hai — bit value sirf Data line hai. DS guarantee karta hai ki har bit mein kahin na kahin exactly ek edge hogi, isliye do lines ka XOR ek self-generated clock hai. Alag clock wire nahi → no clock skew, aur radiation glitches ko shared clock se behtar tolerate karta hai.

First Principles se Derivation:

Problem: Ek alag clock line space mein cables ke across data ke against drift karta hai (thermal expansion, radiation damage) → skew sampling corrupt kar deta hai.
Requirement: Guarantee karo ki kam se kam ek, aur sirf ek, transition har bit period mein ho taaki timing khud wires se recoverable ho.
Insight: Agar Data sirf tab change kare jab bit value change ho, toh identical bits ki run mein Data flat hoga — timing lost ho jaati. Isliye hum Strobe add karte hain, jo exactly tab toggle karta hai jab Data nahi karta. Ab har bit period mein exactly ek transition hoti hai (ya toh Data par ya Strobe par).
Recovery: Data aur Strobe ki transition activity ko XOR karo → ek edge per bit = the clock. Us edge par Data sample karo.

Character Transmission

Har data character 9 bits ka hai: 1 parity bit + 1 data-control flag bit + 8 data bits… zyada precisely, SpaceWire characters ko ek parity bit + control flag + payload se build karta hai; ek data character wire par 10 bits ka hota hai sirf leading parity+flag count karte huye, lekin payload 8 data bits hai jo 1 parity bit se protected hai, jo useful ratio $8/9$ deta hai. Efficiency accounting ke liye hum use karte hain: 8 data bits ko 1 parity bit protect karta hai → 9-bit effective data character.

$P = \overline{D_{0} \oplus D_{1} \oplus \dots \oplus D_{7} \oplus (flag/prev bits)}$

SpaceWire mein parity odd parity hai jo previous character ke data bits aur current control/flag bit par compute ki jaati hai (interleaved parity). First-principles understanding ke liye, parity ko aise treat karo: "ek bit jo is liye choose ki gayi hai taaki protected group ki fixed parity ho."

Parity kyun? Cosmic rays se single-bit errors detect karta hai. Agar parity fail ho, receiver error state mein enter karta hai aur link re-establish hota hai.

Packet Structure aur Routing

Formula SpaceWire Packet Format

[Destination Address] [Cargo bytes...] [EOP/EEP]

Destination Address: 1+ bytes (path address ya logical addressing, 0-255 per byte)
Cargo: Arbitrary data (RMAP jaisi protocols upar ride karti hain)
EOP (End of Packet): Normal termination
EEP (Error End of Packet): Transmission error signal karta hai

Yeh design kyun? Variable-length cargo matlab koi bandwidth waste nahi (fixed-frame protocols ke unlike). Address-first ordering wormhole routing enable karta hai: switches header bytes ko immediately forward karte hain bina poora packet buffer kiye — low-latency control loops ke liye critical.

Wormhole Routing Mechanism

Wormhole routing ek circuit establish karta hai jaise packet header propagate hota hai:

Header byte switch par arrive hota hai → routing table se output port lookup karo
Switch us output port ko reserve karta hai aur header byte forward karta hai
Body bytes reserved path follow karte hain (switch unhe inspect nahi karta)
EOP reserved path release karta hai

Latency formula:

$L_{total} = L_{header} + n \times L_{hop}$

jahaan $n$ = number of hops, $L_{hop}$ = per-switch forwarding delay (~1 μs typical).

Yeh step kyun? Traditional store-and-forward poore packets buffer karta (har hop par $L_{packet}$ delay add karta). Wormhole long packets ke liye latency ~10× cut karta hai.

Flow Control: Credit-Based BackPressure

SpaceWire buffer overflow rokne ke liye character-level flow control use karta hai:

Har receiver ke paas N-character buffer hai, aur har FCT jo woh bhejta hai, woh 8 characters ki space ke liye credit grant karta hai.
Sender sirf tab transmit karta hai jab uske paas credit ho, aur bhejte waqt decrement karta jaata hai.
Receiver space free karne ke baad FCT (Flow Control Token) characters bhejta hai → sender ka credit replenish karta hai.

Radiation Tolerance aur Fault Handling

Link Reset aur Hot-Swap

SpaceWire links in states se progress karte hain: ErrorReset → ErrorWait → Ready → Started → Connecting → Run.

Timing ka key point: standard reset/wait intervals ko character (bit) times ki count mein specify karta hai, fixed microseconds mein nahi. Unki absolute duration isliye link bit rate ke saath scale karti hai — wahi character-count interval 10 Mbps se 200 Mbps par shorter hota hai.

ErrorReset (power-on): ek fixed character-count interval ke liye dono lines ko reset par drive karo → purana state clear karo
ErrorWait: ek fixed character-count interval wait karo (worst-case propagation se lamba choose kiya gaya) → ensure karo ki remote node bhi reset ho
Ready: FCTs bhejo → buffer credits advertise karo
Started: FCTs receive karo → confirm karo ki remote node alive hai
Connecting: NULL/FCT pattern exchange karo → synchronize karo
Run: normal operation

Yeh step kyun? Kyunki wait character times mein measure hota hai, ise isliye dimension karna hota hai ki uski worst-case absolute duration maximum cable propagation delay se zyada rahe — guarantee karta hai ki dono ends ek doosre ka reset observe karein aage badhne se pehle. Yeh poora bus power-cycle kiye bina failed units ki hot-swapping enable karta hai.

Common mistake SpaceWire ko Ethernet jaisa assume karna

Galat idea: "Main SpaceWire ko standard Ethernet switches mein plug kar sakta hoon aur kaam karega." Kyun sahi lagta hai: Dono packet-switched networks hain similar speeds ke saath.

Fix: Ethernet CSMA/CD use karta hai (probabilistic collisions), clock sync ke liye preambles bhejta hai, aur koi guaranteed latency nahi hai. SpaceWire deterministic hai: wormhole routing bounded latency deta hai, DS encoding matlab koi preamble nahi chahiye, aur credits overrun rokti hain. Tumhe SpaceWire-specific routers chahiye (jaise GR-718) jo FCTs aur EOP markers samjhein.

Steel-man: Agar tumhe SpaceWire ko Ethernet se bridge karna ho (e.g., ground testing), ek protocol gateway use karo jo packets buffer kare aur SpaceWire ke flow control ko TCP ke window mechanism mein translate kare — lekin SpaceWire ke microseconds ki jagah milliseconds latency expect karo.

RMAP: Remote Memory Access Protocol

RMAP (Remote Memory Access Protocol) SpaceWire ke liye killer app hai — processors ko custom firmware ke bina remote nodes par registers aur memory read/write karne deta hai.

Command types:

Read (0x08): Address A se N bytes request karo
Write (0x09): Address A par N bytes write karo
Read-Modify-Write (0x0A): Semaphores ke liye Atomic RMW

Packet structure:

[Destination] [Protocol ID=0x01] [Command] [Address] [Data] [CRC] [EOP]

Worked example Instrument se RMAP read

Ground station spectrometer ke 0x10000000 register se 256 bytes read karna chahta hai.

Command packet:

[SpectrometerAddr=0x05] [RMAP=0x01] [Read=0x08] [Addr=0x10000000] [Len=256] [CRC] [EOP]

Spectrometer receive karta hai, CRC validate karta hai, 0x10000000 se read karta hai

Reply packet:

[GroundStationAddr=0x00] [RMAP=0x01] [ReadReply=0x18] [Data: 256 bytes] [CRC] [EOP]

Ground station receive karta hai, CRC check karta hai, apna memory map update karta hai Yeh step kyun? CRC (polynomial 0x07) multi-bit errors pakadta hai jo parity miss kar gayi. Command mein transaction ID request/reply ko link karta hai — agar replies out of order aayein toh confusion prevent karta hai.

Performance Calculations

Cable length vs. speed — asli reason:

Standard mein koi simple RC formula nahi hai. Practical cable length signal attenuation, jitter, aur inter-pair skew se limit hoti hai jaise data rate badhti hai. Faster rates → tighter timing budget → receiver ki edge sampling utna skew/jitter tolerate nahi kar sakti → shorter cables. Roughly: high rates (400 Mbps) → kuch metres; low rates (2–10 Mbps) → tens of metres. Curiosity ke cameras ~10 m par around 10 Mbps pe run karte hain; satellite backplanes 200 Mbps ~1–2 m par run karte hain.

Recall 12-saal ke bacche ko samjhao

Socho class mein notes pass kar rahe ho jab teacher (radiation) teri desk hilata rehta hai. Ek alag "tick-tock" clock line timing rakhne ki jagah, tum aur tumhara dost ek clever rule agree karte ho: har single beat par, exactly tumme se ek taali bajaata hai — ya toh "value hand" ya "timing hand" move karta hai, dono kabhi nahi, koi bhi kabhi nahi. Sirf dekh ke koi clap kiya, tumhara dost jaanta hai ek beat hua aur tumhari value padh sakta hai. Yahi hai Data-Strobe: do signals jahan exactly ek har bit par flip karta hai, isliye beat hamesha recoverable hai.

Ab socho das doston ko notes swap karne hain. Shor machate kamre mein cheekh-cheekh ke bolne ki jagah, har note ke aage address hota hai, aur har desk par ek "postmaster" sirf address padh ke turant next desk ki taraf point karta hai — poora note khole bina. Yahi hai wormhole routing: super fast kyunki koi poora message ka wait nahi karta.

Finally, tum ek dost ki choti si desk par 100 notes dump nahi kar sakte. Toh woh tumhe tickets deta hai — har ticket tumhe ek small batch bhejne deta hai. Jab woh space clear kare, aur tickets deta hai. Yahi hai credit-based flow control: koi kabhi bury nahi hota.

Connections

LVDS Signaling — SpaceWire ke liye Physical layer differential pairs
Wormhole Routing vs Store-and-Forward — Network switching architectures
Cosmic Ray Effects on Electronics — Kyun radiation tolerance matter karta hai
RMAP Protocol — SpaceWire memory access ke liye application layer
MIL-STD-1553 — Purana avionics bus jo SpaceWire replace karta hai
CAN Bus — Automotive equivalent (lower speed, alag fault model)
Real-Time Determinism — Spacecraft ke liye bounded latency kyun matter karti hai
Serial vs Parallel Communication — High-speed links mein fundamental tradeoff

#flashcards/coding

SpaceWire Data-Strobe encoding mein, har bit period par kya guarantee hota hai, aur clock kaise recover hoti hai?

Exactly ONE of the do lines (Data ya Strobe) har bit period mein toggle karta hai; Strobe exactly tab toggle karta hai jab Data nahi karta. Data aur Strobe ki transition activity ko XOR karne se clock regenerate hoti hai, aur Data ko har recovered edge par sample kiya jaata hai. Koi alag clock wire nahi hai.

SpaceWire Data-Strobe alag clock line ki jagah kyun use karta hai?

Ek alag clock line cables ke across data ke against drift/skew karta hai (thermal expansion, radiation). DS guarantee karta hai ki kahin na kahin har bit mein ek transition ho, isliye timing do wires se self-generated hai — no clock skew, better radiation tolerance.

Byte 0xA7 (1010 0111) ke liye, saare aath bits ka XOR kya hai aur kitne ones hain?

Bits = 1,0,1,0,0,1,1,1 → paanch ones (odd) → saare 8 bits ka XOR = 1. (Saare aath bits XOR karne chahiye, chhe nahi.)

SpaceWire data character mein 8 data bits protect karne ke liye kitne bits hain, aur payload efficiency kya hai?

1 parity bit 8 data bits ko protect karta hai → 9-bit effective data character → efficiency = 8/9 ≈ 88.9% (roughly 11.1% overhead), 80% NAHI.

200 Mbps signaling par SpaceWire effective throughput 8/9 rule se compute karo.

T = (8/9) × 200 ≈ 177.8 Mbps ≈ 22.2 MB/s.

10 m SpaceWire cable ka ~2×10^8 m/s par round-trip time kya hai, aur yeh kyun matter karta hai?

One-way ≈ 50 ns/m → 10 m ke liye ~500 ns → RTT ≈ 1 μs. Credits ko is RTT ke dauran in-flight bytes cover karne chahiye link full rakhne ke liye, isliye lambe cables ko zyada outstanding credit/buffer chahiye.

FCT kya hai aur yeh kitna credit grant karta hai?

Flow Control Token ek SINGLE control character (kuch DS bit-times) hai. Har FCT sender ko 8 data characters ka credit grant karta hai — yeh 10 bytes NAHI hai.

SpaceWire startup/reset intervals fixed microsecond values kyun nahi hain?

Yeh character (bit) times ki count mein define hote hain, isliye unki absolute duration link bit rate ke saath scale karti hai — wahi interval 200 Mbps par 10 Mbps se shorter hota hai.

Data rate badhne par SpaceWire cable length ko actually kya limit karta hai?

Signal attenuation, jitter, aur inter-pair skew receiver ka timing margin shrink karte hain — koi simple RC formula NAHI. Higher rates → tighter timing budget → shorter cables.

RMAP kya hai aur yeh kaunsa problem solve karta hai?

Remote Memory Access Protocol — processors ko SpaceWire par remote nodes par registers aur memory read/write karne deta hai bina custom per-device firmware ke (Read 0x08, Write 0x09, RMW 0x0A), network across memory-mapped I/O standardize karta hai.