Intuition The ONE core idea
To say which way a vehicle is pointing , we compare two sets of arrows: a fixed set stuck to the ground and a moving set stuck to the vehicle — and we describe the twist that lines one up with the other. Every symbol on the parent page is just a name for one of those arrows, the twist between them, or the bookkeeping that makes the twist reversible.
This page assumes nothing . If the parent note (the topic note ) used a symbol, we build it here from a picture first.
Definition Vector — an arrow with length and direction
A vector is an arrow. It has a length (how long) and a direction (which way it points). It does not care where you drew it — slide it around, it's the same vector. We write it with a little list of numbers, like [ 1 , 0 , 0 ] , once we pick axes to measure against.
Intuition Why we need numbers for an arrow
A picture of an arrow is great for humans, but a computer on an aircraft needs numbers . To turn an arrow into numbers we need rulers to measure it against — three perpendicular rulers. That set of three rulers is a frame .
Definition Frame — three perpendicular measuring axes
A frame is a set of three arrows, all at right angles (9 0 ∘ ) to each other, meeting at a corner (the origin ). We call them the x , y , z axes. Any other arrow can be described by "how far along x , how far along y , how far along z " — those three distances are its coordinates .
The parent uses two frames:
Definition Navigation frame
{ n } and Body frame { b }
{ n } (the navigation frame , e.g. North-East-Down ) is glued to the ground . x n points North, y n East, z n Down. It never moves.
{ b } (the body frame ) is glued to the vehicle . x b out the nose, y b out the right wing, z b down through the belly. It rolls, pitches and yaws with the aircraft.
Why two frames? The whole subject is the answer to "how is { b } twisted relative to { n } ?" You cannot ask that with only one frame — you need something fixed to compare the moving one against.
The curly braces { n } and { b } are just labels — read "{ n } " as "the nav frame" and "{ b } " as "the body frame."
Definition Subscript = "which frame are these coordinates measured in"
The same physical arrow v has different number-lists depending on which frame you measure it against. We record that with a subscript:
==v n == means "the coordinates of v read off against the nav frame's axes."
==v b == means "the coordinates of the same v read off against the body frame's axes."
The arrow never changed — only the rulers did. Keep this straight: the subscript labels the frame of expression , never a different arrow.
Definition Angle — the amount of turn
An angle measures how much you rotated , opening between a start arrow and an end arrow. We measure it in degrees (a full circle is 36 0 ∘ ) or radians (a full circle is 2 π ). A quarter turn is 9 0 ∘ .
Definition The right-hand rule — which way is a POSITIVE turn
A turn needs a sign : is + 3 0 ∘ clockwise or counter-clockwise? We fix this once with the right-hand rule . Point your right thumb along the axis arrow (e.g. along + z ). Your curling fingers show the direction of a positive rotation. So a positive turn is counter-clockwise when you look back down the axis toward the origin (i.e. with the axis pointing at your eye).
Applied to the three vehicle turns (each about its own axis):
Positive ==yaw ψ == (about + z , which points Down ) → nose swings toward the right/East .
Positive ==pitch θ == (about + y , the right wing ) → nose goes up .
Positive ==roll ϕ == (about + x , the nose ) → right wing dips down .
Why state this at all? Without a fixed positive direction, "ψ = 3 0 ∘ " is ambiguous and the sin /cos signs in the matrix could go either way. Every formula below assumes this right-hand-rule convention.
The parent's three named angles are all just "signed amount of turn about one axis":
Symbol
Name
Picture (what the nose does, positive sense)
Axis it turns about
ψ (psi)
yaw
swings toward the right (East)
z (down)
θ (theta)
pitch
nose lifts up
y (right wing)
ϕ (phi)
roll
right wing dips
x (nose)
ψ , θ , ϕ are Greek letters — treat them purely as names for these three signed turn-amounts. Nothing more mysterious than calling them "the yaw number, the pitch number, the roll number."
Intuition WHY sine and cosine appear at all
When you turn an arrow, its shadow onto the x -axis and its shadow onto the y -axis both change. Cosine and sine are exactly the names of those two shadows for a unit-length arrow. Because rotation is made of those two shadows, every rotation formula is built from sin and cos — that is why no other function shows up.
Definition Cosine and sine on the unit circle
Take an arrow of length 1 , starting along the x -axis, and turn it counter-clockwise (positive sense) by angle α .
==cos α == ("cosine") is its shadow on the x -axis — how far along x the tip lands.
==sin α == ("sine") is its shadow on the y -axis — how far up y the tip lands.
So the tip sits at the point ( cos α , sin α ) . At α = 0 : tip at ( 1 , 0 ) , so cos 0 = 1 , sin 0 = 0 . At α = 9 0 ∘ : tip at ( 0 , 1 ) , so cos 9 0 ∘ = 0 , sin 9 0 ∘ = 1 .
Intuition Sign of the shadow tells you the quadrant
If you keep turning past 9 0 ∘ , the shadows go negative — the tip moves to the left (cos < 0 ) or below the axis (sin < 0 ). This "shadows can be negative" fact is why signs matter so much later, and why we need atan2 in Section 6.
Definition Tangent — the slope of the arrow
==tan α == ("tangent") is cos α sin α — the rise over run , i.e. how steep the arrow is. On the triangle it is "opposite over adjacent." A flat arrow has slope 0 ; a 4 5 ∘ arrow has slope 1 .
Definition arcsin and arctan — the "which angle?" questions
==arcsin ( v ) == asks: "which angle has sine equal to v ?" It undoes sine. Example: arcsin ( 0.5 ) = 3 0 ∘ because sin 3 0 ∘ = 0.5 .
==arctan ( r ) == asks: "which angle has tangent (slope) equal to r ?" It undoes tangent.
Why the parent needs these: given a rotation matrix full of sin and cos numbers, we must run backwards — from the numbers to the angles. arcsin and arctan are the "run backwards" buttons.
Definition Principal-value range — arcsin only ever answers in
[ − 9 0 ∘ , + 9 0 ∘ ]
Many angles share the same sine (e.g. sin 3 0 ∘ = sin 15 0 ∘ = 0.5 ), so arcsin must pick one . By convention it always returns the answer between ==− 9 0 ∘ and + 9 0 ∘ == (i.e. [ − 2 π , + 2 π ] ).
Why this matters for the topic: pitch is recovered as θ = − arcsin ( R 13 ) , so this convention defines pitch to live in [ − 9 0 ∘ , + 9 0 ∘ ] — exactly up to the vertical. Right at the ends, θ = ± 9 0 ∘ , is where gimbal lock strikes (Section 7): arcsin still returns an answer, but roll and yaw can no longer be separated.
arctan alone loses half the circle
Why it feels fine: slope tells you the angle. The catch: an arrow pointing up-right and one pointing down-left have the same slope — tan repeats every 18 0 ∘ , and arctan 's principal range is only ( − 9 0 ∘ , + 9 0 ∘ ) . So plain arctan can be off by 18 0 ∘ . The fix is a smarter function that also looks at the signs of the two shadows: atan2, in Section 6.
Intuition What a matrix does here
A matrix is a little grid of numbers that acts as a machine : feed it a vector (three numbers), it hands back a new vector (three numbers). The rotation matrix's job is: "take an arrow's coordinates in the nav frame and give me its coordinates in the body frame." Same physical arrow — new rulers.
Definition Reading the symbol
R n b
R n b (say "R, body-from-nav") is the rotation matrix . Read the labels bottom-then-top: from n , to b . It multiplies a nav-frame vector (Section 0's v n ) and returns the same vector's body-frame coordinates v b :
v b = R n b v n
Each of its 3 × 3 = 9 entries is built from the sin and cos of ϕ , θ , ψ — exactly the shadows from Section 2. See Rotation matrices & orthogonality and the Direction Cosine Matrix (DCM) for the same object viewed as a table of axis-alignments.
Definition How to multiply a matrix by a vector
Each output number = "walk along one row , multiply each row entry by the matching vector entry, add them up." For row [ a , b , c ] and vector [ x , y , z ] : output = a x + b y + cz . Three rows → three output numbers.
Definition Transpose, orthogonality, and
det = + 1 — the free reverse gear
The transpose R T flips the grid across its diagonal (rows become columns). For rotation matrices there is a beautiful gift: the reverse rotation is just the transpose ,
R b n = ( R n b ) T = ( R n b ) − 1 .
A matrix with R T R = I (the identity, 1 's on the diagonal, 0 's elsewhere) is called orthogonal . But orthogonal alone is not enough: it also allows reflections (mirror-flips), which are not real rotations. A proper rotation additionally satisfies
det R = + 1 ,
where the determinant det R is a single number measuring how the matrix scales volume; + 1 means "no stretching, no mirror-flip." Together R T R = I and det R = + 1 pin down exactly the rotations (this set is called S O ( 3 ) ). Why the parent cares: to send an arrow body → nav you don't compute anything new — you just transpose (Worked Example 1 uses exactly this), and the det = + 1 guarantee is what stops a "rotation" from secretly mirroring your aircraft.
Definition Multiplying two matrices = doing two rotations in a row
Writing R x R y means: first do R y , then do R x (the machine on the right acts first, because it touches the vector first). So a product reads right-to-left in time.
Common mistake "Order shouldn't matter."
Why it feels right: turning is turning. The catch: rotations do not commute — R x R y = R y R x . Yaw-then-pitch lands the nose somewhere different from pitch-then-yaw. This is the whole reason the parent insists on the fixed 3-2-1 order (z first, then y , then x ).
The figure shows the same two turns in the two orders ending at different attitudes — proof that order is not decoration.
Definition atan2 — arctan that knows the quadrant
==atan2 ( y , x ) == asks "which angle points toward the tip ( x , y ) ?" — and it looks at the signs of both x and y to pick the correct one of the two candidates that plain arctan confuses. It returns an angle anywhere in the full circle, in the range ==− 18 0 ∘ to + 18 0 ∘ ==.
atan2 ( 0 , 0.866 ) = 0 ∘ (pointing right).
atan2 ( 0.433 , 0.75 ) = 3 0 ∘ (up-right).
Why the parent uses it: recovering roll ϕ and yaw ψ from the matrix needs the full 36 0 ∘ range and the correct sign — plain arctan would silently drop half the answers. This is the tool built to fix the trap in Section 3.
0 ∘ –36 0 ∘ convention when you need it
atan2 hands back ( − 18 0 ∘ , + 18 0 ∘ ] , but heading (yaw) is often reported as 0 ∘ –36 0 ∘ (compass style, where West is 27 0 ∘ not − 9 0 ∘ ). To convert, just add 36 0 ∘ to any negative result :
angle [ 0 , 360 ) = atan2 ( y , x ) mod 36 0 ∘ .
Example: atan2 ( − 1 , 0 ) = − 9 0 ∘ → − 9 0 ∘ + 36 0 ∘ = 27 0 ∘ . Whether you keep [ − 18 0 ∘ , 18 0 ∘ ] or shift to [ 0 ∘ , 36 0 ∘ ) is purely a display choice; the underlying orientation is identical.
p , q , r (met later, named now)
Later the parent's neighbours talk about ==p , q , r == — the body angular rates : how fast the vehicle rolls, pitches and yaws right now (turn per second), measured in the body frame. They are the speeds of the three turns whose amounts are ϕ , θ , ψ . See Angular velocity & body rates p, q, r .
Intuition Gimbal lock, previewed with zero jargon
When the pitch reaches θ = ± 9 0 ∘ (nose straight up/down) — the very ends of arcsin's principal range from Section 3 — the roll axis and the yaw axis point the same way , so the two turns become indistinguishable — you lose a knob. That degenerate case is why Quaternions — avoiding gimbal lock exist and why Attitude estimation (Kalman / complementary filter) often avoids raw Euler angles near vertical. You now have every symbol needed to understand why the 0/0 shows up in the parent's formulas.
Vector = arrow with length and direction
Frame = three perpendicular axes
Nav frame n and Body frame b
Subscript labels frame of expression
Right-hand rule sets positive sense
Yaw psi, Pitch theta, Roll phi
Cosine and Sine = shadows of a turn
tan and arcsin, principal range
Transpose, orthogonality, det = +1
Composing rotations, order matters
atan2 = sign-aware angle finder
Hide the right side and test yourself before opening the parent note.
What is a vector, in plain words? An arrow with a length and a direction; represented by three coordinates once you pick a frame.
What is a frame? Three mutually perpendicular axes (x , y , z ) meeting at an origin, used as rulers to give a vector its coordinates.
What do { n } and { b } mean? The fixed navigation (ground) frame and the moving body (vehicle) frame.
What does the subscript in v n vs v b tell you? Which frame's axes the SAME arrow's coordinates are measured against — nav vs body.
What defines a POSITIVE rotation? The right-hand rule: thumb along the axis, curling fingers give the positive turn direction (CCW looking back down the axis).
On the unit circle, what are cos α and sin α ? The shadows of a turned unit arrow on the x -axis (cos ) and the y -axis (sin ); the tip sits at ( cos α , sin α ) .
What are sin 3 0 ∘ and cos 3 0 ∘ ? 0.5 and ≈ 0.866 .
What does tan α measure? The slope (rise over run) of the arrow, equal to sin α / cos α .
What question does arcsin ( v ) answer, and over what range? "Which angle has sine equal to v ?"; it returns a principal value in [ − 9 0 ∘ , + 9 0 ∘ ] .
What is the 3×3 form of R n b for 3-2-1? R x ( ϕ ) R y ( θ ) R z ( ψ ) , giving the matrix whose top row is [ c θ c ψ , c θ s ψ , − s θ ] .
What does R n b do to a vector? Takes its nav-frame coordinates v n and returns its body-frame coordinates v b .
Which TWO conditions make a matrix a proper rotation? R T R = I (orthogonal) AND det R = + 1 (no reflection).
Why is ( R n b ) T the reverse rotation? Rotation matrices are orthogonal, so their transpose equals their inverse.
In the product R x R y R z , which rotation happens first? The rightmost one, R z (products read right-to-left in time).
Why can't we just use plain arctan to recover ϕ and ψ ? tan repeats every 18 0 ∘ , so plain arctan loses the quadrant; atan2 uses the signs of both inputs to get the full-circle answer.
How do you turn an atan2 result into a 0 ∘ –36 0 ∘ heading? Add 36 0 ∘ to any negative value (take it mod 36 0 ∘ ).
What are p , q , r ? Body angular rates — the instantaneous roll, pitch and yaw turn-speeds measured in the body frame.