What isometric actually means

Isometric means equal measure, and the equality is between the three axes rather than between the drawing and the world.

Measured from the projection: a unit along x images to 0.8165 of the drawing’s unit, and so do y and z. The three agree to twelve decimal places, which is where the name comes from. None of them is 1.

The other system with three equal scales #

The figure above is worth reading carefully, because it refuses the obvious summary of the name. Two rows come back with all three axis scales equal, not one. Isometric is the expected one. The other is cavalier, whose three scales are 1, 1 and 1 exactly.

Cavalier gets there by a different route and the difference is the whole point. It is an oblique construction: the front face is drawn true, and the depth axis is drawn off at 45° at its full length, by decree rather than by projection. Nothing is foreshortened because nothing is projected — there is no direction in space from which a cube looks like a cavalier drawing of one. That is why cavalier drawings look distended, and it is why cabinet exists, which is the identical construction with the depth axis halved to 0.5 for no reason except that it looks better.

So equality of the three scales is not, on its own, what isometric means. Equality under an orthographic projection is, and the two conditions pick out different sets. Stated the loose way — three equal scales — the name describes a drawing that is not a projection at all alongside the one that is. Stated the exact way, isometric is the unique viewing direction, up to symmetry, that foreshortens all three axes equally, and 0.8165 is what that costs.

The figure asserts both halves now: that every system whose three scales coincide is either isometric or an oblique one, and that isometric is the only orthographic member of that set. The second assertion is what the caption had been claiming without it; the figure beneath it had been labelling two rows “all three equal” throughout.

Where 0.8165 comes from #

The number is √(2/3), and the derivation is two lines.

Isometric projection views the object along the direction (1, 1, 1) — the body diagonal of a cube — so that the three axes make equal angles with the line of sight. Project the unit x-vector orthographically onto the plane perpendicular to (1, 1, 1): the component along the view direction is 1/√3, so the length remaining in the picture plane is

$$\sqrt{1 - \tfrac{1}{3}} = \sqrt{\tfrac{2}{3}} = 0.8165$$

By symmetry the same holds for y and z. The equality of the three is forced by the symmetry of the viewing direction, and the value follows from it.

There is a satisfying consequence: the diagonal of the cube being viewed along projects to a point, so an isometric drawing of a cube is a regular hexagon with the near and far corners coincident at its centre.

True isometric and the isometric everyone uses #

Here is the thing that produces most of the confusion.

True isometric projection foreshortens everything to 0.8165. A drawing made this way is a correct orthographic projection, and a ruler applied to it gives 0.8165 of the real dimension.

Isometric drawing, which is what technical drawing means by the term and what every isometric grid provides, ignores the foreshortening and draws every axis at full scale. It is the same shape enlarged by 1/0.8165 = 1.2247.

The second is not a projection of anything, strictly. It is a projection scaled up by a factor. But the scaling is uniform in all three directions, so all the useful properties survive: parallels stay parallel, ratios are preserved, and the drawing is measurable — with the convenience that a ruler now reads true dimensions directly, which is the entire reason for the convention.

The cost is that an isometric drawing of an object is 22.5% larger than an isometric projection of it, and the two are frequently compared without anyone noticing the factor. In a context where drawings are scaled to fit a sheet nobody sees it; in a context where a projection is being checked against a photograph, it matters.

Why isometric and not dimetric #

Isometric’s dominance in technical illustration is not obvious from the geometry, since dimetric and trimetric have a real advantage: they avoid the degenerate views.

The isometric direction is symmetric in all three axes, which means every face of a box is drawn at exactly the same angle and the same foreshortening. That is elegant and it is why the drawing feels neutral — no face is privileged.

It is also why edges coincide. A cube in isometric hides three of its twelve edges exactly behind the other three; the near and far body-diagonal corners project to the same point; and any two features aligned along the view direction overlap perfectly. In a complicated assembly that ambiguity is a real hazard.

Dimetric — two scales equal, the third half — breaks the symmetry deliberately and eliminates most of the coincidences. Its scales measure 0.943, 0.943 and 0.471, with the depth axis exactly half the other two, which is the reason for choosing those particular rotation angles.

So why isometric anyway? Because of the grid. The isometric axes are at 30° to the horizontal, 30° is constructible with a standard set square, and isometric graph paper exists. Dimetric requires angles of 7° and 41°, needs a protractor, and has no paper. The convention that won is the one that could be drawn quickly with the tools on the drawing board, which is the same reason the involute won over the cycloid in a different subject: tooling, not merit.

The video-game convention, which is a third thing #

The projection used in a great many games and in pixel art is called isometric and is not.

True isometric puts the axes at 30° to the horizontal, which on a pixel grid means a slope of tan 30° = 0.577 — an irrational number, so the lines alias badly.

The convention used instead is a slope of exactly 1/2: two pixels across for one pixel down. That gives an axis angle of 26.57° rather than 30°, and it is a dimetric projection: two of the three axis scales are equal and the third is not.

The reason is entirely about pixels. A 2:1 slope draws as a clean repeating stair pattern with no aliasing, tiles seamlessly, and lets a sprite be reflected to make the opposite direction. A 30° line does none of that.

So there are three distinct things all called isometric: the true projection at 0.8165, the technical drawing at full scale, and the 2:1 dimetric of games. They look similar and they are three different maps, and the only way to tell which is in use is to measure the axis angle or the scale — which is the reason this essay’s figures print the numbers rather than naming the system.

What the equal scales are worth #

The practical value of all three scales being equal is that one ruler works everywhere.

In a dimetric or trimetric drawing, a measurement has to be taken with a different scale depending on which axis it lies along, and a measurement along a face diagonal is not simply available at all. In isometric, one scale serves, and a length along any axis reads directly.

That is why isometric survives for hand drawing and for anything that has to be read by someone with a ruler. It is also why it matters that the scale is 0.8165 in the true projection: a person measuring a true isometric projection with a plain ruler gets numbers 18% too small, consistently, and the error is invisible because it is uniform.

The technical-drawing convention of scaling up to full size exists exactly to remove that trap, and it works. It is worth knowing about only when a drawing has to be compared with something that is a genuine projection — a photograph, a render, or a perspective view approaching its parallel limit.

Reading a drawing for its system #

Given an unlabelled parallel drawing, the system can be identified with a ruler and no other information.

Measure the drawn length of a unit along each of the three axes. Three equal is isometric or an isometric drawing, distinguished by whether the value is 0.8165 or 1 relative to the stated scale. Two equal is dimetric — check whether the odd one is half the others, which is the standard choice. None equal is trimetric.

If the front face is undistorted with true right angles, it is oblique rather than axonometric, and the depth scale says which: 1 is cavalier, 0.5 cabinet, anything else is somebody’s own convention.

That check takes a minute and settles arguments, which is the general recommendation this site keeps arriving at. A drawing system is identified by measuring the drawing, not by what the drawing is captioned.

The hexagon, and what it hides #

An isometric cube draws as a regular hexagon divided into three rhombi, and that is worth knowing both as a construction shortcut and as a warning.

The construction: the outline of a cube in isometric is a regular hexagon whose side equals the cube’s edge times 0.8165 in true projection, or times 1 in the technical-drawing convention. The three visible faces are three identical 60°–120° rhombi meeting at the centre.

The warning: the near corner of the cube and the far corner both project to the centre of the hexagon. They are the same point in the drawing. So an isometric view of a cube cannot distinguish the near corner from the far one, and the drawing is genuinely ambiguous — this is the Necker cube, and its bistability is a property of the projection rather than a curiosity of perception.

The same coincidence happens throughout an isometric drawing of anything: features aligned along the (1,1,1) direction overlap exactly. In a simple object that is a pleasant symmetry. In an assembly it hides parts behind other parts with no visual cue, and it is the principal reason dimetric is preferred for complex mechanical illustration where the ambiguity would be costly.

Isometric circles #

A circle in a face of an isometric cube draws as an ellipse, and the numbers are fixed and worth having.

In true isometric projection the ellipse has an axis ratio of 1:√3, which is 0.577, and its major axis lies along the face’s long diagonal. In the full-scale technical-drawing convention the same ratio holds — the uniform 1.2247 scaling does not change ratios.

That single ratio covers all three faces, which is another consequence of the three axes being symmetric. In dimetric there are two ellipse ratios and in trimetric three, which is one more reason isometric templates are the ones that exist.

The centre of the drawn ellipse is the image of the circle’s centre, exactly, because parallel projection preserves ratios along a line — which is the one convenience the perspective case does not offer, where the two points differ by 3 or 4%.

Why the equality is worth having at all #

The deeper reason isometric is useful is that it makes a drawing measurable with a single scale in three directions at once, which is the property technical drawing is organised around.

An orthographic multi-view drawing — plan, front elevation, side elevation — is also measurable, more precisely, and it is what dimensioned engineering drawings actually use. What it does not do is show the object as a whole; the reader has to assemble the three views mentally, which is a learned skill.

An isometric view shows the whole object in one picture and stays measurable. It gives up some precision — a face diagonal is not directly measurable, and neither is anything at an angle to the axes — in exchange for immediate legibility. That is why isometric is the standard for assembly instructions, exploded views, patent drawings and anything read by someone who is not a draughtsman.

The equal scales are what make that trade work. Without them the reader needs three rulers and has to know which axis they are on, and the whole advantage disappears.

Where the name misleads in practice #

Three concrete situations where “isometric” being read as “same as life” causes trouble.

Reading a true isometric projection with a plain ruler. The dimensions come out 18.35% short, uniformly. Because the error is uniform it looks like a scale factor rather than a mistake, and it will be attributed to the drawing having been reduced.

Mixing a projection and a drawing on one sheet. A true isometric projection of one part beside an isometric drawing of another puts the two at different scales with nothing to indicate it. This is the practical reason technical drawing standardised on the full-scale convention and stopped using the true projection.

Comparing an isometric view with a render. A modelling program’s orthographic camera produces a true projection, at 0.8165, and a hand isometric drawing of the same object is 1.2247 times larger. Overlaying them for comparison requires knowing which is which.

The general defence is the one this site keeps recommending: measure the drawing rather than trusting its caption. Three axis scales, taken with a ruler, identify the system unambiguously and take a minute.

What the word should have been #

Axonometric — measurable along axes — is the honest name for the family, and it is the name the family actually has. Isometric, dimetric and trimetric are its three cases, distinguished by how many axis scales coincide.

Read that way the vocabulary is exact: isometric means the three measures are equal to each other, which is a statement about the projection’s symmetry and says nothing about their relation to the object. The confusion comes entirely from reading the prefix as a comparison with the world rather than among the axes.

It is worth the paragraph because the same misreading recurs. Conformal does not mean a map preserves shape; it means it preserves angles locally, which preserves the shapes of very small things and not of large ones. In both cases a precise term describing a relationship among the projection’s own quantities gets read as a promise about fidelity to the world, and in both cases the promise is one no projection can make.

The other misread name #

One more piece of vocabulary in this family gets read as a promise it does not make, and it is worth pairing with the isometric case.

Orthographic means the projecting rays are orthogonal to the picture plane. It says nothing about the object being shown at true size, and an orthographic projection foreshortens every direction that is not parallel to the picture plane — which is why isometric, itself orthographic, foreshortens all three axes.

The only directions an orthographic projection shows at true length are those lying in the picture plane. That is why a multi-view engineering drawing has three views: each one shows two of the three axes at true length and the third not at all, and the reader recovers the object by combining them.

So a single orthographic view is never fully measurable, and a set of three is. That is the trade the drawing office made, and the isometric pictorial view exists alongside it precisely because a single view that is nearly measurable and immediately legible is worth having next to a set of three that are exactly measurable and are not.

The one number to remember #

If a single figure survives from this essay it should be 0.8165, together with what it is a ratio of: a unit along a world axis to the length it draws at, under true isometric projection.

Everything else follows. Three axes equal is where the name comes from. 1.2247 is its reciprocal and is the factor by which technical drawing scales the projection up so a ruler reads true. √(2/3) is where it comes from, and the constraint that the three axis scales’ squares sum to 2 is why no orthographic projection can draw all three axes at full size — which is the reason cavalier oblique exists and why it is not orthographic.

The number is also the quickest test of which of the three things called isometric a drawing uses. Measure a unit along one axis against the drawing’s stated scale: 0.82 is a true projection, 1.00 is a technical drawing, and an axis angle of 26.6° rather than 30° is the games convention, which is dimetric and not isometric at all.

The same number, from the other side #

√(2/3) has a companion. The equal-angle direction that foreshortens all three axes equally is also the direction that gives a figure with three perpendicular aspects as much of each as any single view can — and that share is √(1/3), which is the other half of the same identity.