From a photo or a tape measure to a panel layout and a Fanuc top-rail program, one step at a time. Every point on the top edge is over from the left and up from the bottom; every segment between two points is its own line or circular arc, defined by its rise — nothing is inferred from a neighbour.
Opening
Inches. Type fractions as 24 1/2, 24-1/2 or 24.5. Points snap to 1/16".
Top-edge points
| # | Over | Up | To next | Rise | R |
|---|
First and last points are anchored to the left and right edges (⚓) — set their height only. Click the plot to add a point, drag a point to move it. To next = straight line or arc to the following point. With Auto-smooth on, each arc's rise is chosen so the chain reads as one fair edge; a rise you type or drag is kept. Rise = how far the template bows away from a straightedge laid across the two points, measured at the middle (+ out, − in); drag the small green handle on the plot to set it by eye. R is the resulting radius.
Start from
Template photo
Align by 3 clicks (starts on its own when a photo loads): click the left and right ends of the bottom line (the sill, or a line you drew on the template) and type its width, then click one point at the top and type how high it is above that line. The photo is rotated, scaled and placed from those three points — width and height are scaled separately, which also takes out most of the tilt in a phone photo. Then trace. Photos stay in this tab only.
Outline cut — machine
Generic Fanuc-style post, inches, absolute, arcs as G2/G3 with I/J. Program origin = the template's bottom-left corner, Z0 = top of stock. The cut runs the outline clockwise: up the left edge, over the top, down the right, back along the bottom — which puts the part on the right of travel, so with the tool outside the part G41 (cutter left) is the normal choice; G42 would cut inside the line. Dry-run above the stock before the first real cut.
Layout — the questions
Stiles are 2" wherever a vertical stile can be used. Margin is the gap panel-to-frame and panel-to-T-post. The top rail follows the curve, measured square to it, and is two boards glued face to face so its louver holes can be cut as half-round grooves with a ball nose, one half at a time. Louvers run on the pitch all the way to the top of the rail: the ones that can swing half their size are working louvers, the ones above are fixed — glued in, blank cut to match the inside of the top rail. "First hole" defaults to half the pitch above the bottom rail.
Top rails — blanks & program
Router table 145" × 48", zone lines every 13.1" in X. Each half of a laminated rail is its own blank, sitting at Y0 on the next free zone line; a blank wider than one zone takes two.
Outline — left to right along the top edge
| Seg | From → To | Type | Chord | Rise | Radius | Center (over, up) | Sweep | Length |
|---|
Each segment stands alone: chord between its two points, rise at the middle, and the circle those three points define. Center and radius are what the next layer (panel rules, then G2/G3 toolpaths) consumes.
Layout files
| Part | Qty | Size | Notes |
|---|
| Panel | # | Up from panel bottom | Pin span | Left end in | Right end in | Function |
|---|
Tools
| T# | Ø | Description | Feed | RPM | Use |
|---|
One tool is the profile cutter (outer edge, inner edge, end cuts) and one is the groove ball nose (half-round louver-pin grooves, one half per board). Other tools ride along in the header for the operator.
Program
Arc audit — every G2/G3 in the program
| Blank | Pass | Dir | Start | End | R | I | J | Sweep | Chord / 2R | R-word risk |
|---|
Chord / 2R is how close the arc is to a semicircle. Past ~0.98 the R word can no longer place the centre reliably; at 1.0 or above no arc of that radius fits the two points and the control alarms. I and J are the centre measured from the start point.
Arc words — what R and I J mean, and why it matters
A G2 (clockwise) or G3 (counter-clockwise) block tells the control to move from where the tool is to the X Y end point along a circle. Two points on their own do not define a circle — infinitely many circles pass through them — so the block has to carry one more piece of information. Fanuc gives you two ways to supply it, and they behave very differently.
G2 X5.1674 Y1.2188 R16.5313 says "end here, on a circle of radius 16.5313". The control must work out where the centre is: it drops a perpendicular from the middle of the chord and goes out √(R² − (chord/2)²). Two circles of that radius pass through the two points (left figure), so a positive R means the short way round (≤ 180°) and a negative R the long way (> 180°). That is the whole convention, and it has four consequences.
1. The centre is computed from rounded numbers. Start, end and R are all written to four decimals. Near a semicircle the square root above is taking the difference of two nearly equal numbers, so a rounding of 0.0001" in R moves the centre by far more than 0.0001". The Chord / 2R column in the audit shows how close each arc sits to that cliff.
2. No arc can be longer than the chord allows. If the two points end up further apart than 2R — which rounding alone can cause when the arc is close to a semicircle — no circle of that radius fits and the control stops with a radius alarm. A full circle (start = end) cannot be written with R at all, because every circle of radius R through one point qualifies.
3. The sign carries the geometry. Any arc that sweeps past 180° must be written with a negative R. Write it positive and the control cuts the other arc — same two points, same radius, bulging the other way — without complaint. That is the classic R-word failure: a program that looks right, runs without an alarm, and leaves a part whose curve is wrong.
4. Nothing is checked. The control always starts where the tool is and always lands on the commanded end point. Whatever radius makes that possible is the radius that gets cut. If the end point is slightly off the intended circle, the error is absorbed silently into the shape.
G2 X5.1674 Y1.2188 I-16.4894 J-1.1786 says "the centre is I in X and J in Y from the start point, go clockwise to the end". On Fanuc, I and J are always incremental from the start point, whatever G90/G91 is in force (the K word does the same for Z). There is exactly one circle, any sweep from a sliver to a full 360°, and the control now has enough information to verify the block: it measures the radius at the start and at the end, and if they differ by more than the tolerance in parameter 3410 it raises PS0020 "OVER TOLERANCE OF RADIUS". A geometry mistake shows up at proof-out on the screen, not in the wood.
Mirrored blanks (A-2 against A-1) flip the sign of I and swap G2 for G3 while J stays the same; the studio derives all three from the mapped points rather than mirroring the words. The two arcs of a rail edge that share a point are each written on their own, so no tangent is assumed between them — a compound arch stays a chain of independent circles exactly as it was measured. The R5 on Fanuc's G05.1 Q1 R5 line in the header is unrelated: that is the AI-contour smoothing level, not a radius.