Visual walkthrough — Reaction control system — thruster selection, plume impingement limits

This page is the picture-companion to Reaction control system — thruster selection, plume impingement limits. Prerequisite ideas we lean on live in Cross Product & Rigid-Body Torque, Attitude Dynamics — Euler's Equations, and Control Allocation & Pseudo-inverse.

---

Step 1 — A thruster is an arrow that lives at a place

WHAT. Picture the spacecraft as a solid brick. Somewhere near its middle is one special point: the center of mass (CoM) — the balance point, the spot the whole body rotates around when you spin it in free space. Bolted to the outside is a tiny nozzle. When it fires, hot gas shoots out one way and the ship is pushed the opposite way.

WHY two facts, not one. A push is not fully described by "how hard." It also matters where it is applied and which way it points. So each thruster $i$ carries two arrows:

• ${\mathbf{r}}_i$ — the position arrow: starts at the CoM, ends at the nozzle. It is the lever arm.
• ${\hat{\mathbf{u}}}_i$ — the direction arrow: a length-1 arrow showing the way the force acts (the little hat $\hat{}$ means "length exactly 1", a pure direction with no size).

PICTURE. The brick, the CoM dot, and one nozzle with its two arrows.

---

Step 2 — Force is just "hardness times direction"

WHAT. Multiply the push-hardness $F_i$ by the direction arrow ${\hat{\mathbf{u}}}_i$:

$\underset{\text{full force arrow}}{\underbrace{{\mathbf{F}}_i}}=\underset{\begin{array}{c}\text{how hard}\\ \ge 0\end{array}}{\underbrace{F_i}}\underset{\begin{array}{c}\text{which way}\\ \text{length }1\end{array}}{\underbrace{{\hat{\mathbf{u}}}_i}}.$

WHY this shape. A unit arrow has length 1; scaling it by $F_i$ stretches it to length $F_i$ while keeping its direction. So ${\mathbf{F}}_i$ is an arrow that points the way the thruster pushes and is as long as the push is strong. Simple, but it is the atom everything else is built from.

PICTURE. The unit arrow ${\hat{\mathbf{u}}}_i$ next to the stretched arrow ${\mathbf{F}}_i$; same direction, longer.

---

Step 3 — Why a cross product turns a push into a twist

WHAT. The turning effect (torque) of that force is

${\mathbf{\tau }}_i=\underset{\text{lever arm}}{\underbrace{{\mathbf{r}}_i}}\times \underset{\text{force}}{\underbrace{{\mathbf{F}}_i}}=F_i({\mathbf{r}}_i\times {\hat{\mathbf{u}}}_i).$

WHY the $\times$ (cross product) and not ordinary multiplication. Ask the real question: how much does this push try to spin the body? Two things decide it. (a) A push far from the CoM twists more than a push right at it — so the length of ${\mathbf{r}}_i$ matters. (b) A push aimed straight at the CoM cannot spin anything; a push aimed sideways to the lever arm spins the most — so the angle between ${\mathbf{r}}_i$ and ${\mathbf{F}}_i$ matters. The cross product is precisely the tool built to answer "lever length times force times how perpendicular they are," and it hands back an arrow along the spin axis. No other single operation packages all three facts. (Full derivation of the cross product lives in Cross Product & Rigid-Body Torque.)

Reading the result arrow. ${\mathbf{\tau }}_i$ points along the axis the body spins about; its length is the twisting strength; point your right hand's fingers from ${\mathbf{r}}_i$ toward ${\mathbf{F}}_i$ and your thumb gives the arrow's direction.

PICTURE. Lever arm, force, and the resulting torque arrow popping out of the plane, with the swept angle shaded.

---

Step 4 — The two degenerate cases you must never forget

WHAT. Look at the extreme geometries so no case surprises you later.

• Case A — push aimed along the lever line (${\mathbf{r}}_i$ collinear with ${\hat{\mathbf{u}}}_i$ — pointing the same way or the exact opposite way): the arrows lie on one line, the "how perpendicular" factor is zero, so ${\mathbf{\tau }}_i=0$ in either case. Same-direction means the force points radially outward through the CoM line; opposite-direction means it points radially inward along the same line — both pass through the CoM, so neither can spin the body. Pure translation. The ship slides, never spins.
• Case B — push straight sideways to the lever (${\mathbf{r}}_i$ perpendicular to ${\hat{\mathbf{u}}}_i$): the "how perpendicular" factor is maximal, giving the biggest twist per newton, torque length $=F_i|{\mathbf{r}}_i|$.

WHY this matters. Real thrusters live somewhere between these two. Every column of the machine we build next is secretly a blend of "how much it shoves" and "how much it twists," and the blend is set entirely by this angle. Knowing the two ends lets you sanity-check any thruster at a glance.

PICTURE. Two mini-diagrams side by side: aimed-along-the-lever (no spin, shown both ways) vs. aimed-sideways (max spin).

---

Step 5 — Stack shove and twist into one 6-number "wrench"

WHAT. Force lives in 3 numbers (x, y, z). Torque lives in 3 numbers (spin about x, y, z). Glue them into one tall stack of 6:

$\mathbf{w}=\left[\begin{array}{c}\mathbf{F}\\ \mathbf{\tau }\end{array}\right]=\underset{\text{6 rows}}{\underbrace{\left[\begin{array}{c}{F}_x\\ F_y\\ F_z\\ {\tau }_x\\ {\tau }_y\\ {\tau }_z\end{array}\right]}}.$

WHY glue them. Every command a controller ever issues is "produce this net shove and this net twist." Carrying them as one object $\mathbf{w}$ (the wrench, from Attitude Dynamics — Euler's Equations) lets one equation handle translation and rotation together. Top 3 rows = where the ship goes; bottom 3 rows = how it turns.

PICTURE. A single tall bracket, top half tinted "force," bottom half tinted "torque."

---

Step 6 — Line the thrusters up as columns: the allocation matrix $B$

WHAT. Each thruster contributes its own little wrench. Write one thruster's unit wrench (its wrench at $F_i=1$) as a column of 6 numbers:

${\mathbf{b}}_i=\left[\begin{array}{c}{\hat{\mathbf{u}}}_i\\ {\mathbf{r}}_i\times {\hat{\mathbf{u}}}_i\end{array}\right].$

Stand all $N$ columns side by side into a grid $B$ (6 rows tall, $N$ thrusters wide). Collect the push-hardnesses into a list $\mathbf{f}=[F_1,\dots ,F_N{]}^{\mathsf{T}}$. Then the total wrench is one matrix–vector product:

$\boxed{\mathbf{w}=\underset{6\times N}{\underbrace{B}}\underset{N\times 1}{\underbrace{\mathbf{f}}},\mathbf{f}\ge 0.}$

WHY this is the whole idea. Reading $B\mathbf{f}$ left to right: each thruster's column ${\mathbf{b}}_i$ gets weighted by how hard it fires, $F_i$, and all the weighted columns are added up. That sum is exactly "fire this mix, get this net wrench." The one-sided rule $\mathbf{f}\ge 0$ is the physics reminder from Step 1: no jet can pull. This is the central result of the parent note — and it is nothing but Steps 2–5 written once per thruster and stacked.

PICTURE. Four thrusters on a body, each drawn again as a 6-tall column feeding into the grid $B$; the list $\mathbf{f}$ scales them and they sum to $\mathbf{w}$.

---

Step 7 — Worked couple: twist with zero drift

WHAT. Two thrusters on a wheel of radius $r$. Thruster 1 sits at ${\mathbf{r}}_1=(0,+r,0)$ and pushes in $+x$; thruster 2 sits at ${\mathbf{r}}_2=(0,-r,0)$ and pushes in $-x$. Both fire with the same hardness $F$.

Step-by-step.

Net force (top of the wrench): $\mathbf{F}=F(1,0,0)+F(-1,0,0)=(0,0,0).$ The two shoves point opposite ways and cancel — no drift, exactly what Step 4 Case B promised we could not get from one thruster alone.

Each torque (bottom of the wrench), using $\mathbf{r}\times \mathbf{F}$: ${\mathbf{\tau }}_1=(0,r,0)\times (F,0,0)=(0,0,-rF),{\mathbf{\tau }}_2=(0,-r,0)\times (-F,0,0)=(0,0,-rF).$

Add them: $\mathbf{\tau }=(0,0,-2rF).$

WHY they add instead of cancel. Both jets try to spin the body the same way about $z$ (fire the top jet one way, the bottom jet the other way — like your two hands turning a steering wheel). Opposite shoves, same twist. This pairing is called a couple: pure torque, zero net force, length $2rF$.

PICTURE. The wheel, the two opposing shoves cancelling, the two same-sense spins adding into one $z$-torque.

---

Step 8 — The plume: why the answer bends the geometry

WHAT. The gas that made the force does not vanish at the nozzle lip. It fans out into near-vacuum as a widening cone. The strength it delivers to a surface a distance $d$ away, at an angle $\theta$ off the cone's centerline, is

$P_{\mathrm{imp}}(d,\theta )\approx \underset{\begin{array}{c}\text{reference}\\ \text{pressure}\end{array}}{\underbrace{P_0}}\underset{\begin{array}{c}\text{spreads over}\\ \text{a sphere}\end{array}}{\underbrace{{\left(\frac{d_0}{d}\right)}^2}}\underset{\begin{array}{c}\text{tightness of}\\ \text{the cone}\end{array}}{\underbrace{{\cos }^k\theta }}.$

WHY each piece.

• $P_0$ and $d_0$ form a known calibration pair: $d_0$ is one chosen reference distance (a fixed short distance from the nozzle exit where we have measured or computed the plume, e.g. $d_0=0.1$ m), and $P_0$ is the pressure the plume delivers at that reference distance, on axis. Everything else is scaled relative to this one anchor point — you cannot say "pressure at distance $d$" without first pinning down "pressure at some known distance." So $d_0$ is not a new physical effect; it is the yardstick that turns the shape $1/d^2$ into an actual number.
• $(d_0/d{)}^2$: the same gas passes through ever-larger spherical shells as it travels out, so its intensity dilutes as $1/d^2$. At $d=d_0$ this factor equals $1$ (you recover $P_0$), and beyond it the pressure falls — halve the distance and the pressure quadruples. This is why close surfaces (deployed panels, a docking target) are the danger, exactly what the parent warns.
• ${\cos }^k\theta$: the gas is densest straight ahead ($\theta =0$, where ${\cos }^{k}0=1$) and thins toward the cone's rim. The exponent $k$ is the focusing number of the plume: $k=0$ would mean the gas spreads equally in all directions (a hemisphere, no preferred direction); small $k$ (roughly $1$–$2$) is a wide, gentle fan; large $k$ (roughly $4$–$40$ for real space thrusters firing into vacuum) is a tight, pencil-like jet where nearly all the punch is dead ahead and the rim is almost empty. Bigger $k$ = tighter, more focused plume. It is fitted to each nozzle from test data or Rarefied Gas Dynamics / Plume Modeling.

The rule. A thruster is allowed to fire only if, for every sensitive surface $s$, $P_{\mathrm{imp}}\le P_{\max ,s}$. Here the subscript $s$ labels which surface we mean (solar panel, star tracker, docking target, ship belly, …); each one is more or less fragile, so each carries its own allowable pressure $P_{\max ,s}$ — the largest impingement pressure that surface $s$ can safely tolerate before it is damaged or pushed too hard. A rugged radiator has a big $P_{\max ,s}$; a delicate optic has a tiny one. Because the limit differs per surface, we must check every surface separately. In practice this becomes a keep-out cone: no delicate surface may sit inside the nozzle's forward cone within some range.

Worked violation (from the parent, re-derived): a forward thruster during docking points dead at the target, $d=2$ m, $\theta =0$, with $P_0=500$ Pa at reference distance $d_0=0.1$ m, $k=4$, panel limit $P_{\max ,s}=1$ Pa.

$P_{\mathrm{imp}}=500{\left(\frac{0.1}{2}\right)}^2{\cos }^{4}0=500\times (0.05{)}^2\times 1=500\times 0.0025=1.25\text{Pa}.$

Since $1.25>1.0$, that thruster is inhibited — locked to $F_i=0$. The selection of Step 6 must now solve $B\mathbf{f}={\mathbf{w}}_{\mathrm{cmd}}$ using only the remaining columns, even if that burns slightly more propellant. Safety outranks fuel.

PICTURE. The plume cone, the $1/d^2$ shells, the ${\cos }^k\theta$ falloff, and a panel sitting inside the keep-out cone triggering a violation.

---

The one-picture summary

One flow: each thruster = (place ${\mathbf{r}}_i$ + direction ${\hat{\mathbf{u}}}_i$) → shove $F_i{\hat{\mathbf{u}}}_i$ and twist $F_i({\mathbf{r}}_i\times {\hat{\mathbf{u}}}_i)$ → stack into the column ${\mathbf{b}}_i$ → line up all columns into $B$ → the net command is $\mathbf{w}=B\mathbf{f}$ with $\mathbf{f}\ge 0$ → but the plume cone of any firing thruster carves the feasible set smaller (keep-out cones), so the final selection is the cheapest firing mix that both hits ${\mathbf{w}}_{\mathrm{cmd}}$ and sprays nothing it can hurt.

Recall Feynman retelling — say it back in plain words

Imagine you are floating in space on a chair with a bunch of little spray cans taped to you, each pointing a fixed way and only able to spray (never suck). Each can does two things at once: it shoves you the way it points, and if it is off to the side it also spins you. How much it spins depends on how far out it is and how sideways it sprays — that "how far and how sideways" combined is the cross product, and it gives an arrow pointing along the spin axis. I write each can's shove and spin as one column of six numbers, put all the columns in a grid $B$, and then "which cans, how hard" is just a list $\mathbf{f}$ of hardnesses (all positive). The total effect is $B\mathbf{f}$. If I want to turn without drifting, I fire two opposite cans so their shoves cancel but their spins add — a couple. Finally, the spray itself is a cone that gets weaker with distance (like a flashlight beam, $1/d^2$) and weaker off to the side (${\cos }^k\theta$); if that cone would hit my solar panel or a nearby ship harder than that surface can take, I forbid that can and re-solve with the rest. That is the whole reaction control system in one breath.

Recall Quick self-test

What operation turns force-at-a-place into torque, and why that one? ::: The cross product ${\mathbf{r}}_i\times {\mathbf{F}}_i$ — it packages lever length, force size, and how perpendicular they are, and returns the spin-axis arrow. Why must $\mathbf{f}\ge 0$? ::: A jet can only push (spray), never pull, so every thrust magnitude is zero or positive. How do you get pure torque with no drift? ::: Fire an opposing pair (a couple): opposite shoves cancel, same-sense spins add to $2rF$. Why does the plume pressure grow so fast for nearby surfaces? ::: It dilutes as $1/d^2$, so halving the distance quadruples the pressure. What is $d_0$ in the plume formula? ::: A chosen reference distance where $P_0$ is known; at $d=d_0$ the $1/d^2$ factor is 1 and $P_{\mathrm{imp}}=P_0$ on axis. What does the subscript $s$ in $P_{\max ,s}$ mean? ::: It labels the specific sensitive surface; each surface has its own allowable pressure, so every surface is checked separately. In the docking example, why is the forward thruster inhibited? ::: $P_{\mathrm{imp}}=1.25$ Pa exceeds the panel limit of $1.0$ Pa, so it is locked to $F_i=0$.