Snapping to grid

[Ozzyoz]

I am using Xara Extreme 4, and am trying to produce an isometric drawing as I have previously done in CAD software. I have found that the lines drawn do not exactly "snap" to the grid points and some innaccuracy is evident if the drawing is magnified. Is this a known "bug" or am I doing something wrong?

[Soquili]

Hello Ozzyoz,

Welcome to Talkgraphics.

Check your snap radii settings to see if they have been changed from the defaults. See attached image.

[Albacore]

I am not sure that the grid is an accurate isometric grid but I could be proved wrong by Bill's setting.

[Xhris]

Hi Ozzyoz, there's an online resource of Xara tutorials available at http://xhris.digitalred.net. Specifically, a tutorial covering everything to know about the grid and snapping is available here (online version not done yet though).

[jclements]

Also, the Isometric Grid of Xara is, by definition, not a true isometric grid.

[JET]

Also, the Isometric Grid of Xara is, by definition, not a true isometric grid.

The so-called "isometric" grid in Xtreme is not true isometric.

It uses diagonal axes at 26.565° off horizontal. The reason is because that is the sine of one-half. Why one-half? Because monitors, being arrays of rectangular pixels, cannot draw diagonal lines; they "fake" diagonals by "stair-stepping" pixels.

The stair-stepping is least distracting if it is small and uniform. Obvously, the least distracting diagonal, assuming square pixels, is 45°, because the stair-steps are one pixel of rise for every one pixel of run.

True isometric, by definition, requires 30° diagonals. But using square pixels, you cannot get a clean, uniform stair-stepping pattern using only whole pixels at 30°.

Back in the days of low-res monitors, and low bit-depth raster imaging without antialiasing, that was a problem. So the desire for a small and consistent stair-step pattern of whole pixels drove the decision to use the aspect ratio of one pixel rise for every two pixels of run--in other words, an aspect ratio of 1:2--in other words, the arctangent of 1/2--in other words, the angle of 26.565°.

Using this angle is an archaic throwback to the days before antialising, high bit-depth, transparency, fast antialiasing, and fast processing of animated vector graphics. It is an intentional compromise of proportion, which should NOT be called "isometric", but frequently is.

Nor can it properly be called "dimetric", which is another common falacy. The angles of dimetric drawing are no more arbitrary than those of isometric. Merely having the two diagonals at equal angles from horizontal is not all there is to dimetric. A true dimetric unit cube using 26.565° as the angle for the diagonals would require a vertical axis of 86.6 % of true length. Why? Because as explained above, the "rise" of 26.565 is one-half the "run". The arcsine of .5 is 30°. This means that the top of the unit cube is "tilted" upward toward you at 30°. That requires that the vertical edge of the unit cube is "tilted" away from the picture plane at an angle of 30°. The length of its orthographic projection would be 86.6, because .8660 is the cosine of 30°.

But in the fake so-called "isometric" grid of Xtreme, the vertical edge length is 50%. Again, this is merely an accommodation to the bitmap pattern of square pixels using a 1:2 rise:run aspect ratio, so as to allow the resulting grid to be used like an isometric grid. That is, in an isometric grid, the diagonals of the two visible sides of the unit cube are on the same line as the opposing (piercing) axis.

Again, this convention is just a throwback to archaic bygone limitations. Especially given that Xtreme's claim-to-fame is lightning fast use of transparency and raster effects applied to live vector paths; and especially given today's capability of employing real scaleable vector graphics in animation; and especially given today's ubiquity of on-the-fly antialising of everything rendered to the screen--there is little need for the severely distorted compromises explained above.

It would be fine for Xtreme to continue to include the fake "isometric" grid. But a truly modern Xtreme should provide a fully user-definable correct axonometric snap-grid. (See Inkscape for one example.)

JET

[Stealth]

The so-called "isometric" grid in Xtreme is not true isometric.

It uses diagonal axes at 26.565° off horizontal. The reason is because that is the sine of one-half. Why one-half? Because monitors, being arrays of rectangular pixels, cannot draw diagonal lines; they "fake" diagonals by "stair-stepping" pixels.

The stair-stepping is least distracting if it is small and uniform. Obvously, the least distracting diagonal, assuming square pixels, is 45°, because the stair-steps are one pixel of rise for every one pixel of run.

True isometric, by definition, requires 30° diagonals. But using square pixels, you cannot get a clean, uniform stair-stepping pattern using only whole pixels at 30°.

Back in the days of low-res monitors, and low bit-depth raster imaging without antialiasing, that was a problem. So the desire for a small and consistent stair-step pattern of whole pixels drove the decision to use the aspect ratio of one pixel rise for every two pixels of run--in other words, an aspect ratio of 1:2--in other words, the arctangent of 1/2--in other words, the angle of 26.565°.

Using this angle is an archaic throwback to the days before antialising, high bit-depth, transparency, fast antialiasing, and fast processing of animated vector graphics. It is an intentional compromise of proportion, which should NOT be called "isometric", but frequently is.

Nor can it properly be called "dimetric", which is another common falacy. The angles of dimetric drawing are no more arbitrary than those of isometric. Merely having the two diagonals at equal angles from horizontal is not all there is to dimetric. A true dimetric unit cube using 26.565° as the angle for the diagonals would require a vertical axis of 86.6 % of true length. Why? Because as explained above, the "rise" of 26.565 is one-half the "run". The arcsine of .5 is 30°. This means that the top of the unit cube is "tilted" upward toward you at 30°. That requires that the vertical edge of the unit cube is "tilted" away from the picture plane at an angle of 30°. The length of its orthographic projection would be 86.6, because .8660 is the cosine of 30°.

But in the fake so-called "isometric" grid of Xtreme, the vertical edge length is 50%. Again, this is merely an accommodation to the bitmap pattern of square pixels using a 1:2 rise:run aspect ratio, so as to allow the resulting grid to be used like an isometric grid. That is, in an isometric grid, the diagonals of the two visible sides of the unit cube are on the same line as the opposing (piercing) axis.

Again, this convention is just a throwback to archaic bygone limitations. Especially given that Xtreme's claim-to-fame is lightning fast use of transparency and raster effects applied to live vector paths; and especially given today's capability of employing real scaleable vector graphics in animation; and especially given today's ubiquity of on-the-fly antialising of everything rendered to the screen--there is little need for the severely distorted compromises explained above.

It would be fine for Xtreme to continue to include the fake "isometric" grid. But a truly modern Xtreme should provide a fully user-definable correct axonometric snap-grid. (See Inkscape for one example.)

JET

ic

[handrawn]

Hi Stealth

regarding xara's 26 degree 'isometric' grid..

I have no real idea whether Jet's argument has any merit or not, but I would agree with Covoxer's explanation at post #22 here:

http://www.talkgraphics.com/showthre...648#post352648

you should be able to google for true isometric grids that you can import...

[Stealth]

does isometric, axometric use curves or just circles, lines, and squares. one of the things i dont get yet is what exactly true isometric art is. im refering to vector art in a non pixel/cad way. anybody got any isometric xara art they like to share? doesnt have to be by you. but vector. or am i missing the point yet again.

[handrawn]

if you are not going to be doing architecture/engineering/scientific plans and drawings, you may not need to worry about isometric at all

I'm no expert - I tend to think only in terms of vanishing point[s] when I need to put depth into a drawing