The formula
The whole system is a promise: pick an origin, pick axes at right angles, and every location becomes a number tuple.
2D point: P = (x, y)
3D point: P = (x, y, z)
Reading: x is how far you walk along the first axis, y along the second, z along the third — each measured from the origin, with a sign telling you which side.
Distance between two points falls straight out of Pythagoras:
2D: d = √( (x₂ − x₁)² + (y₂ − y₁)² )
3D: d = √( (x₂ − x₁)² + (y₂ − y₁)² + (z₂ − z₁)² )
Reading: the straight-line gap is the square root of the summed squared gaps along each axis. Because the axes are perpendicular, the cross terms vanish and Pythagoras just works.
Rotating a 2D point by angle θ about the origin:
x' = x·cos θ − y·sin θ
y' = x·sin θ + y·cos θ
Reading: this rotates the point itself counterclockwise by θ in a fixed frame (positive θ = CCW with the standard axis layout). To express the same fixed point in a frame that has itself been rotated by θ, use the inverse — swap the sign of θ: x' = x·cos θ + y·sin θ, y' = −x·sin θ + y·cos θ. Confusing the two — active vs. passive rotation — is a classic sign-error source in CAD and robot-controller work.
Converting to and from polar, when a problem is really about radius and angle:
x = r·cos θ r = √(x² + y²)
y = r·sin θ θ = atan2(y, x)
Reading: Cartesian and polar describe the same point; you pick whichever makes the equation shorter.
Where you meet it
- The CAD seat and the drawing. Every solid model starts with a part origin and three datum planes. When a machinist reads GD&T off a print, the basic dimensions are Cartesian offsets from a datum reference frame. Get the origin wrong and the whole feature tree floats.
- The machine axis definition. A 3-axis mill, a gantry, a coordinate measuring machine — X, Y, Z are literally the motion axes, and the controller's home position is the origin. A CMM inspection report is nothing but a list of measured (x, y, z) points compared against nominal.
- The FEA pre-processor. Nodes are (x, y, z) coordinates; elements connect them. Every mesh, every applied load vector, every reaction lives in a Cartesian frame — usually a global one plus local element frames the solver rotates between.
- The test stand and the review board. When someone asks "where's the strain gauge," the answer is a coordinate off the part datum, not "kind of near the fillet." A design review that can't state feature locations in a shared frame is a review that's about to argue in circles.
How it works
The system does one thing well: it makes geometry into algebra. A circle stops being a shape you sketch and becomes x² + y² = r². A line becomes y = mx + b. Once shapes are equations, you can intersect them, differentiate them, and hand them to a computer. That trade — pictures for equations — is the entire reason CAD, FEA, and CNC exist.
The axes must be perpendicular and the units must be linear and equal on each axis, or the clean Pythagorean distance formula quietly stops being true. Stretch one axis, skew the frame, and you no longer have a Cartesian system — you have an affine or oblique one, and distances need a fuller quadratic form. Most engineering tools assume orthonormal axes; the assumption is usually silent, which is exactly why it bites.
The mistake people make is forgetting which frame a number is in. A coordinate is meaningless without its origin and orientation. The same hole is (0, 0, 0) in the part frame, (152.4, −38.1, 12.7) in the machine frame, and something else again in the assembly frame. Datum errors, sign-flip errors, and swapped-axis errors are the bread and butter of scrapped parts and blown fixtures. Nail down the frame before you trust the number.
The other trap is orientation handedness in 3D. Standard engineering practice is a right-handed frame: point the right thumb along +X, index finger along +Y, and the middle finger points +Z. Cross products, torque directions, and surface normals all follow from that convention. Mix a right-handed CAD model with a left-handed import and mirrored parts, flipped normals, and inside-out solids follow. Many CAD and graphics packages differ on whether Y or Z points "up," and reconciling that is a standard integration headache.
Cartesian is the default, not the only choice. Anything with rotational symmetry — a shaft, a nozzle, an antenna pattern, a pressure field around a bolt hole — is often cleaner in polar, cylindrical, or spherical coordinates, and forcing it into x-y-z buries the physics under algebra. The skill is knowing when to switch frames, and every conversion is exact and reversible, so switching costs nothing but bookkeeping.
The 2D plane splits into four quadrants (I through IV, counterclockwise from the upper right, where both x and y are positive). Three-dimensional space splits into eight octants by the sign pattern of (x, y, z). The classical names for the axes survive in older texts: the x-value is the abscissa, the y-value the ordinate, and the z-value the applicate — worth recognizing even if nobody says them out loud at a stand-up anymore.
History
The idea that a curve could be a number pair is older than Descartes. In the 14th century, the French scholar Nicole Oresme drew figures using perpendicular reference lines — he called the horizontal extent longitude and the vertical intensity latitude — to graph how a quantity varied, and in doing so anticipated the core move of plotting one variable against another [1][2]. Ancient work by Apollonius of Perga on conic sections had already leaned on reference lines, but none of it turned into a general method.
The general method arrived twice, independently, in the same decade. Pierre de Fermat worked out a coordinate treatment of loci in a short manuscript, Ad Locos Planos et Solidos Isagoge — built on results he had by 1629 and circulating among mathematicians by 1636, though it wasn't printed until 1679 [3][4]. Fermat's version was, in some ways, closer to how we use axes today.
But the name stuck to the other man. René Descartes published La Géométrie in 1637 as an appendix to his Discours de la méthode, written while he was living in the Netherlands [1][5]. In it he showed systematically how to attack geometry problems with algebra, tying points to lengths measured along a reference line. The Latinized form of his name, Cartesius, is why we say Cartesian rather than Fermatian [5]. Descartes was not first, and arguably not clearest, but he published in a work that got read — and that is how mathematical credit usually gets assigned. The two men, working without knowledge of each other, are jointly the reason a modern engineer can describe a bolt hole as three numbers and have everyone in the room know exactly what is meant.
Related tools
- /tools/convert-angle
- /tools/convert-length
- /tools/projectile-range
- /tools/moment-of-inertia-shapes
- /tools/section-modulus-rect