PracticeChecklist

A — Why this is the first thing you build

You cannot do accurate hand work on a bench that racks, bounces or is the wrong height. Every plane stroke puts a horizontal force into the bench, and a bench that moves absorbs the stroke instead of the wood doing. This is not sentiment about craft — it is the reason your hand-cut joints have been harder than they should be.

Mass is the whole design. Roughly 200–300 lb is the working minimum for a bench you plane on. Everything else — the joinery, the leg position, the stretchers — exists to keep that mass from moving. Schwarz's argument, which is correct, is that a heavy slab on four legs joined with drawbored mortise and tenon is structurally complete and everything beyond it is decoration.

Height. Set it by the work, not by a chart. For hand planing, the classic test is the top at about the height of your pinky knuckle with arms hanging — roughly 32–34″ for most people. Lower is better for planing (you get your weight over the tool), higher is better for joinery and carving. If you will do both, build it at planing height and use a raised work platform for detail work. Do not split the difference; you get a bench that is wrong for both.

Legs flush with the front edge. This is the single most useful detail and most beginners set the legs inboard. Flush legs let you clamp a long board to the front of the bench with a holdfast in the leg, which is how you edge-plane and how you cut tenon cheeks vertically.

Build it out of cheap wood. Southern yellow pine, Douglas fir, ash, or construction lumber laminated up. Spending money on the bench is spending money in the wrong place. Schwarz builds his out of 2×12 SYP.

What this module actually teaches you: milling rough stock to dimension repeatably, machine setup done once in service of a real thing, the biggest mortise and tenon you will ever cut, drawboring, flattening a wide surface by hand, and sharpening everything you will use for the next eleven months.

B — The laminated slab

Why laminate rather than find a slab. A 4″ thick solid slab is expensive, moves enormously, and will check. A slab laminated from 2″ stock turned on edge is more stable, cheaper, and the glue lines are irrelevant on a work surface.

Stock selection. Buy 2×12 or 8/4 stock and let it sit in the shop for at least two weeks, stickered with air on all faces, before you touch it. Construction lumber is often sold at 15–19% MC and will move violently as it equalises. Take a moisture reading at the start and at the end and log both.

Rip out the pith. Any board with the pith (the centre of the tree) running through it will split. Rip it out and use the two halves.

Mill in stages, not in one pass. Joint one face, plane the opposite face, joint one edge, rip to width. Then let it sit overnight and do it again. Wood relieves internal stress as you remove material, and a board milled flat in one session will not be flat tomorrow. This is the lesson: milling is a process over days, not an operation.

Glue-up in sections. Three or four boards at a time, not all twelve. Each section gets flattened before it joins the next. A twelve-board glue-up in one go has more open surface than any glue's open time and it will come out twisted.

Clamp pressure roughly 150–250 psi for hardwood, less for softwood. Cauls across the width to keep the boards from sliding into a staircase. An even small squeeze-out bead all along the joint — a flood means you are starving the joint, not feeding it.

Use a slow glue. Titebond III or a liquid hide glue for the long open time. Regular PVA gives you five minutes and you need twenty.

C — Flattening a slab by hand

This is the skill the bench exists to teach. You will do it again on every wide panel for the rest of your life.

Read the surface first. Winding sticks at both ends, sighted across. Two parallel sticks longer than the width, and you look across their top edges — any twist shows as one stick crossing the other. A straightedge across the width at six positions, and along the length. Mark the high spots in pencil. You are removing high spots, not 'flattening'.

Traverse first, diagonally, with a scrub or a heavily cambered jack. Work across the grain at about 45° in both directions. A cambered iron (an 8–10″ radius on a scrub, 8″ on a jack) takes a thick hollow shaving and does not tear out badly even across grain. This stage is fast and ugly.

Re-check with the winding sticks every few passes. Twist is the enemy; flatness along the length is easy to recover, twist is not.

Then the try plane or jointer (a #7 or #8, lightly cambered, maybe 0.002″ of camber across the iron) along the length. Overlapping passes. You are now removing the scrub's scallops and establishing flat.

The shavings tell you everything. When a full-width, full-length continuous shaving comes off every pass, the surface is flat. Broken and partial shavings mean you are still on the high spots.

Finish with a smoother, or do not finish at all. A workbench top that is too smooth lets work slide. Many benchmakers stop at the try plane deliberately.

Do not use a router sled unless you must. It works, but you will have built a bench without learning the one thing the bench was for.

D — Drawboring: the joint that needs no clamps

A drawbored mortise and tenon is mechanically locked. It pulls the shoulder tight and stays tight for centuries without glue — every timber frame and every old chair is drawbored. It is also the correct joint for a bench, because a clamped-and-glued bench joint will loosen as the wood moves and a drawbored one will not.

How it works. You bore a hole through the mortise cheeks, dry-fit the tenon, mark through the hole onto the tenon, then re-bore the tenon hole offset toward the shoulder. Driving a tapered peg through forces the tenon deeper into the mortise, pulling the shoulder closed.

The offset is the whole trick. 1/16″ for hardwood, up to 3/32″ for softwood on a joint this size. Too little and it does not draw. Too much and you snap the peg or split the tenon — and the split runs to the end grain and the joint is scrap.

Sequence: cut the joint to a good dry fit first. Bore the mortise cheeks with the joint apart. Assemble dry, push a brad-point or the same bit through the mortise hole to mark the tenon — a light tap, just a dimple. Disassemble. Bore the tenon offset toward the shoulder by the amount above. Reassemble.

The peg matters. Riven (split, not sawn) straight-grained oak or hickory, whittled octagonal then rounded, with a long gentle taper on the leading end. A sawn dowel has grain runout and will shear. Make ten for every eight you need.

Chamfer the entry, and drive the peg with a hammer in steady blows, not one heroic swing. You will feel the joint close.

Reference: Schwarz, The Anarchist's Workbench, chapters 6–9 — the joinery argument and the drawboring method, with drawings. Free PDF.

E — Workholding, and the leg vise

Workholding is the actual function of a bench. Schwarz's framing in The Anarchist's Workbench is the useful one: you hold work on its face, on its edge, or on its end, and a bench needs an answer to all three.

Faces — a planing stop and holdfasts. A single square stop at the end of the bench, adjustable by tapping, handles most planing. You are pushing the work against a stop, not clamping it, and that is faster than any vise.

Edges — a face vise plus a leg jack or a deadman to support the far end of a long board.

Ends (for dovetailing, for tenon cheeks) — the face vise, held high, or a twin-screw.

Holdfasts. The most underrated tool in the shop. A holdfast in a 3/4″ hole, struck with a mallet, holds harder than most clamps and releases with one tap. The bench top must be at least 2″ thick and the hole must not be chamfered on the underside — the holdfast grips on the two edges of the hole. Gramercy holdfasts work in thin tops; cheap cast ones often do not hold at all. You can forge your own in module 06.

The leg vise. A single screw through the leg, a chop, and a parallel guide (a slotted bar with a pin, or a scissor/Crisscross mechanism) at the bottom to stop the chop racking. It gives enormous clamping force with a wooden screw or a cheap steel one, and it is the vise to build if you are building one.

Hardware you can make in module 04: the parallel guide, the pin, the garter, and the chop's leather facing. The screw itself is worth buying.

F — Layout: the part that decides whether any joint works

Almost every bad joint is a layout error, not a cutting error. This is the section to read twice.

Face side and face edge. Choose one face and one edge on every single piece, mark them with the traditional face mark (a loop) and edge mark (a caret pointing at the face mark), and measure everything from those two surfaces only. Never measure from a sawn edge or an opposite face. When four pieces all reference their own face side, the assembly closes even if the stock varies in thickness.

Mark, do not measure, wherever possible. A rule read twice gives two answers. Holding the actual mating piece against the work and knifing its position gives one answer and it is correct. This is the single biggest accuracy improvement available to you.

Knife, not pencil, for anything that will be cut to. A knife line severs the fibres, gives the chisel a wall to register against, and is roughly 0.003″ wide against a pencil's 0.02″. Use a pencil only for waste marks and notes.

Mark the waste immediately with a scribble or an X, every time, before you pick up a saw. Cutting the wrong side of the line is the commonest beginner failure and it is entirely preventable.

Cut lists lie; story sticks do not. For anything with repeated dimensions, make a stick marked with every position, and transfer from the stick. No arithmetic, no cumulative error.

Number every joint and every mating face as you cut it. Corner 1, 2, 3, 4. Hand-cut joints are not interchangeable and you will not remember.

Reference: Robert Wearing, The Essential Woodworker, ch. 1–2 — the clearest account of working to a line in print. Lost Art Press publish a free excerpt containing the whole of chapter 2.

G — Mortise and tenon, properly

The structural joint. It resists racking — the shoulder takes the load, the tenon locates and holds. Everything with a leg and a rail is this joint.

Proportions. Tenon thickness = one third of the stock thickness, rounded to a chisel you own. In 3/4″ stock that is 1/4″, which is convenient because you have a 1/4″ mortise chisel. In 1-1/2″ stock, 1/2″. The rule exists because it balances the tenon's strength against the mortise walls' strength — a fatter tenon just means thinner cheeks that split.

Tenon length = as long as the mortise allows, ideally 5× its thickness or more. A short tenon in a deep rail is the commonest weak joint.

Width. A tenon wider than about 5× its thickness should be divided into twin tenons with a gap, because a wide tenon's own seasonal movement will split the mortise. This is why wide rails get twin tenons, not one big one.

The haunch. Where a tenon meets a stile at the top of a frame, you cut away the top of the tenon leaving a short stub — the haunch. It stops the rail twisting, it fills the groove if the frame is grooved for a panel, and it leaves enough stile above the mortise not to blow out. Haunch depth about 1/3 the tenon width.

Cut the mortise first, always. It is far easier to pare a tenon to fit a mortise than the reverse.

Chopping by hand. Mortise chisel the same width as the tenon. Start in the middle, not at the end — work from the centre outward in both directions, levering the chip out each time, then come back and pare the two ends to the knife line last. Starting at the end drives the chisel's bevel into the end wall and you get a blown-out, over-length mortise.

Machine options. Router with a spiral bit and a fence gives a round-ended mortise — either round the tenon's corners with a chisel (faster) or square the mortise ends. Hollow-chisel mortiser: set the chisel 1/64″ clear of the auger, cut both ends first, then overlapping bites between, never a bite supported on one side only.

Tenon cheeks by hand — saw on the waste side, leaving the knife line, with the piece in the vise angled so you saw down two corners then level the cut. Shoulders last, with a knife wall chiselled to give the saw a track.

The fit. Firm hand pressure to enter, needs a mallet tap to seat, comes apart without violence. If you can push it fully home with your thumbs it is too loose; glue does not fill gaps in a structural joint. If you need a hammer, you will split the mortise at glue-up when the glue swells the tenon.

The shoulder is the joint. A perfect tenon with a gappy shoulder is a failed joint — the shoulder carries the load and it is the only part anyone sees. Pare it dead square to the face side with a wide chisel and a router plane if you have one.

Free reference with photos: Paul Sellers' Common Woodworking has complete free courses on the mortise and tenon and the haunched mortise and tenon.

H — Dovetails

The tension joint. It resists a pull in one direction mechanically — a dovetailed drawer front cannot be pulled off. It is also the joint your work will be judged on, fairly or not.

Angle. 1:8 for hardwood, 1:6 for softwood. The reason is short grain: a steeper tail in a soft, weak-in-shear wood has corners that crumble. 1:6 is about 9.5°, 1:8 is about 7°. Buy or make a marking gauge for one and stop thinking about it.

Tails first or pins first? Either works and people are religious about it. Tails first is easier to transfer accurately and is what most hand-tool teaching uses. Pins first is better when you need the pins very fine. Pick one and do it the same way every time; the consistency is worth more than the choice.

Layout. Set the gauge to the mating board's thickness plus a hair and scribe the baseline on both faces of both boards. Lay out tails with the two half-pins at the edges first, then divide the rest by eye or by dividers. Do not lay out with a ruler and arithmetic — dividers walking across the board give an even spacing with no cumulative error and it takes twenty seconds.

Sawing. The whole skill is sawing to a line. Saw on the waste side, leaving the line visible. Start the cut with the saw at a low angle and a backward pull, let the weight of the saw do the work, and watch the line and not the saw. Saw all the left-slope cuts, then all the right-slope cuts — changing the body angle once rather than every cut is worth several thousandths.

Removing waste. Fret saw or coping saw close to the baseline, then chop. Chop from both faces, meeting in the middle — going straight through blows out the back face. Undercut the middle very slightly so the shoulder closes.

Transfer with a knife, holding the boards firmly and squarely, with the tail board's baseline exactly on the pin board's end. This is the step where accuracy is lost. Mark each corner individually and number it.

Fit. Dry assemble once, tap home with a block. If it needs force, find the tight spot — a shaving off the right place, not a general loosening.

Half-blind (drawer fronts) — the pin board is not cut through, so all the chopping happens in a socket and there is no coping saw help. Depth about two-thirds the front's thickness.

Through, at scale — on a 24″ or 30″ case corner the failure mode is cumulative: each tail a thousandth off adds up and the last one will not close. Lay out from one reference edge, transfer each corner individually, and dry assemble the whole case before any glue.

Free reference: Paul Sellers' free dovetail and double-dovetail courses at Common Woodworking; Rob Cosman's YouTube channel for the sawing technique specifically.

I — The rest of the vocabulary, and when each one is right

Butt joint. Resists nothing. End grain to long grain glues at roughly 20% of a long-grain joint's strength, because end grain wicks glue away from the joint like a bundle of straws. Only acceptable with mechanical reinforcement — screws, dowels, a spline, a domino. Sizing the end grain with thinned glue first and letting it tack helps measurably but does not rescue it.

Dowelled butt. Adds shear resistance and alignment. Its weakness is that two dowels define a line, so the joint can still pivot; three or a domino stops that. Dowel holes must be bored perpendicular from a registered face or the joint racks.

Biscuit. Alignment, not strength. Useful for edge-joining panels and for carcase alignment. Do not use it where the joint carries load.

Domino / loose tenon. Structurally a real mortise and tenon with both mortises machined. As strong as an integral tenon for most purposes and vastly faster. The honest position is that it is the right joint for a lot of production work and the wrong one for demonstrating hand skill.

Housing / dado. Resists downward load and racking across the joint. The shelf sits in a trench. Cut the trench to the actual thickness of the shelf, which is never nominal — measure with a caliper, then shim a dado stack or use a router with two passes against a fence. A stopped housing hides the joint at the front edge.

Sliding dovetail. A housing that also resists being pulled out. The joint for a fixed divider in a case, or a breadboard end. It must be cut with a taper (tighter at the back) on anything over about 12″, or it will bind solid halfway in and you will never get it home or out.

Half lap. Two pieces each cut to half thickness, crossing. Resists racking well, weak in tension, dead easy, and completely honest in a shop fixture or a light frame. Depth exactly half — cut both from the same setup and any error cancels.

Bridle joint (open mortise and tenon). A mortise and tenon with one mortise wall removed. Much easier to cut and fit than a blind mortise, almost as strong in racking, and the end grain shows. Correct for shop-made squares, frames, and anything where the visible end grain is fine.

Mitre. Resists nothing on its own — it is two end-grain surfaces glued together. The reason it exists is that it hides end grain and continues a moulding round a corner. Always reinforce: splines across the corner, keys let into a sawn kerf, a loose tenon, or a domino. A keyed mitre with contrasting keys is both stronger and better looking than a hidden fix.

Finger / box joint. Enormous long-grain glue area, mechanically interlocked in one axis, easy to machine on a jig. Structurally excellent, visually mechanical. The right joint for boxes, drawers in a shop context, and anything where you want the repetition to read.

Wedged through tenon. The tenon passes through and is split and wedged from the far side, spreading it in a flared mortise. Mechanically locked, no glue needed, visible, and honest. Flare the mortise about 3–5° on the exit side only, on the two walls the wedges will spread against. Saw the kerf to about two-thirds the tenon's depth and stop it with a small drilled hole so the split does not run.

Tapered round tenon (the staked joint). A round tenon tapered to about 6° driven into a matching reamed socket. This is how chairs, stools and Windsor undercarriages are built. It self-tightens under load and it is far more forgiving of angle errors than a square joint. Schwarz's The Anarchist's Design Book, chapters 3–18, is the complete free treatment of it.

Which to choose, in one line: if it will be racked, mortise and tenon or bridle. If it will be pulled apart, dovetail. If it carries a shelf, housing. If it must come apart, drawbore or wedge, no glue. If it is a corner that must look continuous, keyed mitre. If it is a fixture, half lap and move on.

J — Wood movement, and designing so the joint survives it

This is the knowledge under every joint decision and it is the one thing that makes furniture fail years later.

Wood moves across the grain and effectively not at all along it. A 24″ wide panel in a domestic hardwood will change roughly 1/4″ across a full seasonal cycle in a heated building. The same board will change a few thousandths in its length. Every joinery problem in solid wood follows from this asymmetry.

Tangential movement is about twice radial. Flatsawn boards move about twice as much across their width as quartersawn ones, and flatsawn boards cup away from the heart. If a wide surface must stay flat, use quartersawn, or make it a frame and panel, or accept the cup and design for it.

Calculate it, do not guess. Dimensional change = width × the species' shrinkage coefficient × the change in moisture content. The coefficients are tabulated free in the USDA Wood Handbook, chapter 4, and at wood-database.com. Do this calculation once for the species you use most and you will remember the number forever.

Equilibrium moisture content is where the wood ends up for a given temperature and relative humidity. Interior heated space in a northern winter can drop to 6% EMC; the same room in August can be 11%. That is the range your work has to survive. Build at the middle of it, around 8–9%.

The frame-and-panel solution. A panel floats in a groove, unglued, free to move, while the frame — all long grain — stays put. Finish the panel before assembly or an unfinished line appears when it shrinks. Leave clearance at the panel's edges, more in a dry winter, and use a space ball or a dab of glue at the centre of the top and bottom edges only so it stays centred.

Cross-grain construction is the trap. Gluing a long-grain rail across a wide panel's width will split the panel, guaranteed. Wherever you must fix across the grain — a tabletop to an apron, a case to a steel stand, a breadboard end — use slotted holes running across the grain, with washers wide enough to bridge, and fix solidly at one point only (usually the front centre).

Breadboard ends are the classic cross-grain problem solved: a tongue the full width, pegged through elongated holes in all but the centre peg, and glued only at the centre few inches.

Free references: USDA Forest Products Laboratory, Wood Handbook FPL-GTR-282, chapter 4 (moisture relations and shrinkage coefficients) and chapter 13 (drying, distortion and warp). These are free government PDFs and they are the source everything else paraphrases.

A — Sharpening

The back matters more than the bevel. A chisel is two surfaces meeting; a rounded back means no edge regardless of bevel.

Flatten the first 1″ of the back on the coarse plate until it is uniformly scratched edge to edge. Then 1000, then 8000, until it mirrors. This is a one-time job per tool and takes 20–40 minutes the first time.

For plane irons use the ruler trick: a thin steel rule under the iron’s far end raises a tiny back bevel, so only the last 1/32″ needs polishing.

Bevels. 25° primary ground, 30° secondary honed. A honing guide is not cheating; freehand can come later.

Sequence: 1000 until a wire edge forms along the full width (check by running a thumbnail up the back), 4000, 8000, then strop or one pass on the back to remove the wire.

Tests. Chisel: pare end grain on pine and leave a shiny surface, not a torn one. Plane iron: full-width shaving you can read newsprint through. Scraper: shavings, not dust.

B — Bench plane setup

Check the sole with a straightedge. It needs to be flat at the toe, the mouth, and the heel — the middle can be hollow.

Set the frog so the mouth opening is about twice the shaving thickness. Tight mouth for figured wood, open for rough work.

Chipbreaker set 0.010–0.020″ back from the edge for general work; 0.005″ for difficult grain.

The chipbreaker’s leading edge must contact the iron with no gap, or shavings jam. Lap it if it does not.

Lateral adjust by eye, sighting down the sole against a light.

Depth: back off until it stops cutting, then advance until it cuts, to take up the backlash.

F — Saw sharpening and setting

A saw that will not track is not a skill problem. This is a half-hour job and after it you will stop blaming yourself.

1. Joint. Run a mill file flat along the tooth line until every tooth shows a flat on its tip. You are making them all the same height. The flats are also your reference for the next two steps.

2. Set. A saw set plier bends alternate teeth outward. Set is roughly 10–15% of the plate thickness per side — barely visible. Set only the top third of the tooth; setting down to the gullet cracks teeth off. Less set for hardwood and for accuracy, more for green or wet stock.

3. File. Triangular saw file, held level, one face against the tooth. Rip: file straight across, 0° fleam, rake about 8°. Crosscut: skew the file about 15–25° fleam, rake about 15°. File every other gullet down the saw, turn it around, do the rest. Stop when the jointing flat just disappears — that is how you know each tooth is the same size.

Test cut in scrap, not in work. Drifting to one side means one side has more set — stone that side lightly with a fine stone laid flat against the plate.

Japanese pull saws are not resharpenable (induction-hardened teeth). They are consumable. Know which of yours are which before you start filing.

G — The card scraper

The scraper is the answer to figured wood that tears out under any plane, and to a sanded surface that looks dead.

1. File the edge square with a mill file held in a block, until the old burr is gone and the edge is flat and 90° to the face.

2. Stone it. Face and edge, on a fine stone, until the edge is polished. A burr can only be as good as the edge it is turned from — this is the step people skip and then decide scrapers do not work.

3. Consolidate. Lay the burnisher flat on the face and stroke a few times, then flat along the edge. Steel is being work-hardened, not cut.

4. Turn the burr. Burnisher along the edge at about 5° off square on the first pass, 10–15° on the last, two or three firm strokes. Both corners of both edges — that is four burrs per card.

It is cutting when it makes shavings, not dust. Dust means no burr. Flex it with both thumbs to control the cut and to keep the middle from digging.

Re-turn the burr when it stops; you get several re-turns before it needs filing again.

E — Drilling wood, and the drill press

Square the table to the spindle first. Chuck a length of 1/4″ rod, bend the bottom 2″ out 90° to make an L. Rotate the spindle by hand and check the rod tip against the table at 0° and 180°. Target ≤0.005″ over a 6″ swing. A table out of square bores a hole that is not perpendicular, and every dowel and draw-bore pin after it is wrong.

Speed by diameter and material. Small bits fast, big bits slow. Wood: 1/4″ twist ~3000 RPM · 1/2″ ~1500 · 1″ Forstner ~700 · 2″ Forstner ~350 · 3″ hole saw ~250. Too fast burns; too slow tears.

Bit types. Brad point for clean entry in a flat face. Forstner for flat bottoms, overlapping holes and holes on an edge or angle — the rim cuts before the chip. Twist is a metal bit and wanders in wood; it is what you have, not what you want. Auger pulls itself in and is for hand braces and deep holes.

Backer board every time. It stops blowout on the exit face and it is free. Clamp it — a loose backer is not a backer.

Never hand-hold on a drill press. A bit that binds in a hand-held part spins it into a club. Fence and clamp, every hole.

Peck-drill anything deeper than three diameters. Retract fully, clear the chip, go again. A packed flute overheats, burnishes the wall and can split the stock.

Hollow-chisel mortiser, if you have one. Set the chisel 1/64″ clear of the auger — a slip of paper. Too tight and the auger seizes; too loose and the chisel burns. Cut both ends of the mortise first, then take overlapping bites between them, never a bite with the chisel supported on one side only.

A — Aligning the table saw

Unplug the saw. Raise the blade to full height. Mark one tooth with a marker.

Blade to miter slot. Clamp the dial indicator to a bar that rides the miter slot. Zero the indicator on the marked tooth at the front of the blade. Slide the bar to the rear, rotate the blade so the same marked tooth is at the rear, and read. Target ≤0.002″.

To correct: on a contractor saw, loosen the trunnion bolts underneath the table and tap the trunnion. On a cabinet saw, loosen the four bolts holding the table to the cabinet and tap the table. Re-check after every tap; tighten in a cross pattern.

Fence to blade. Run the indicator along the fence face using the miter slot as the reference. Set the fence parallel, or with 0.002–0.003″ toe-out at the rear (fence trailing edge further from blade). Never toe-in — that traps the offcut and causes kickback.

Arbor runout. Indicator against the arbor flange, rotate by hand. Over 0.002″ means a bent arbor or debris behind the flange. Clean and re-check before blaming the blade.

Stops. Set 90° and 45° against the blade plate, never the teeth. Confirm with a test cut, not with the square alone.

Record all six numbers before and after in the notebook.

B — Blade change and dado shimming

Unplug. Arbor nut loosens toward the blade rotation on most saws (left-hand thread). Use two wrenches, never the blade as a lever.

Clean pitch off the flange seat. A chip under the flange reads as runout.

Reinstall riving knife and guard every time. Set riving knife within 0.003″ of blade-plate alignment and just below tooth height.

Dado. Measure the actual plywood with the caliper first. 3/4″ nominal ply typically measures 0.700–0.725″.

Stack: two outer blades + chippers to just under target. Add shims (0.005″, 0.010″, 0.020″) between chippers, spaced evenly around the arbor, never stacked in one place.

Test cut in scrap of the same thickness. The ply should push in with hand pressure and not fall out when lifted. Adjust by one shim at a time.

C — Building the crosscut sled

Base: 3/4″ ply or MDF, 18″ deep × 27″ wide. Flatten and seal both faces so it does not cup.

Runners: hardwood or UHMW, milled to a slip fit in the miter slots — 0.001″ under slot width, and 0.010″ shy of slot depth so the runner never bottoms out.

Set runners with pennies or shims in the slots, glue and screw the base to them from above, with the base held tight against the fence as a straightedge.

Front fence: 2-1/2″ tall, structural only. Rear fence is the working fence — 3″ tall, dead flat on the reference face, fixed with one screw at the left end plus a second screw at the right end through a slightly oversized hole so it can pivot for adjustment.

Raise the blade through the base. Add a blade guard block over the exit slot.

D — The five-cut method

Take a piece of flat ply about 16″ square. Label the edges 1, 2, 3, 4 clockwise.

Place edge 1 against the sled fence and trim a whisker off. Rotate clockwise, cut edge 2, then 3, then 4.

Fifth cut: rotate once more and cut a strip about 1″ wide off the edge you cut first.

Measure that strip at both ends with the caliper. Call them A and B.

Error = (A − B) ÷ 4 over the strip’s length. The division by four is because each of the four cuts added the same error.

To correct: error per inch × the distance from the fixed screw to the adjustable screw = how far to move the loose end. A tap with a mallet and a dial indicator is the usual method.

Repeat until (A − B) ÷ 4 ≤ 0.001″ across the strip, which is about 0.005″ over 24″.

A — Jointer setup

Outfeed table height is the whole job. It must sit exactly at top dead centre of the knife arc.

Test: lay a straightedge on the outfeed table so it overhangs the cutterhead. Rotate the head by hand. The knife should just kiss the straightedge and carry it forward 1/8″ to 3/16″, then release.

Too low: the board dips at the end (snipe on the trailing end). Too high: the board stops feeding or tapers.

Knife setting. Use a magnetic jig or a straightedge with the drag test above, one knife at a time, both ends of each knife. Torque gib screws from the centre outward.

Table coplanarity. Straightedge across both tables with the cutterhead lowered. Feeler gauge for any gap. Over 0.003″ and the machine needs its ways adjusted, not its knives.

Fence square. Set with an engineer’s square against the outfeed table, not the infeed.

Face jointing technique. Cup down. Light pressure on the infeed only until 6″ of board has passed the head, then transfer all downward pressure to the outfeed table and keep it there. Never push down on the unsupported infeed end — that flattens the cup and it springs back.

Two passes maximum before re-checking. Stop when the whole face is cut, not when it looks flat.

B — Planer

Set the bed rollers barely above the bed for rough stock, flush for smooth stock.

Snipe is the head lifting as the board leaves one roller. Fixes, in order: support infeed and outfeed exactly level with the bed; lock the head if the machine has a lock; feed boards end-to-end so the next board holds the last one down.

Deliberately reproduce snipe once, then eliminate it, so you can diagnose it on a student’s board.

Tearout: read grain direction before feeding. If it tears, reverse the board end-for-end, reduce depth of cut to 1/32″, or slow the feed rate.

C — Bandsaw setup, in order

1. Blade on, guides and thrust bearings backed fully off.

2. Tension. Ignore the scale. Tension until a 1/2″ blade deflects about 1/4″ under firm side pressure at mid-span with the guides raised.

3. Tracking. Rotate by hand, adjust the upper wheel tilt until the blade runs centred on the crown (or with the gullets just past the crown on a flat tyre).

4. Guides. Side guides set 0.004″ off the blade (a dollar bill is about 0.004″), positioned just behind the gullets so they never touch the teeth.

5. Thrust bearings set 0.015″ behind the blade — they should not spin until you feed wood.

6. Table square to blade with an engineer’s square. Reset the stop.

7. Wheel coplanarity. Straightedge across both wheel rims. Out of plane means the blade will never track or tension correctly.

Drift test. Pencil a line down a 2″ wide board, freehand cut along it, stop mid-cut and clamp a bevel gauge to the table along the board edge. That angle is your drift. Set the resaw fence to it.

A — Why a chair, and which chair

A chair is the standard test because it is the hardest common object: every joint is loaded in use, nothing is square, the angles are compound, and a person's whole weight goes through it repeatedly for decades. It also happens to hit four of your weakest audit blocks at once — the wood lathe, bending, shaping, and structural thinking.

Build a stick chair or a Windsor, not a joined chair. The reason is technical, not stylistic. A joined chair (square mortise and tenon in square stock) requires you to cut compound-angled mortises accurately, and an error of one degree shows. A staked or Windsor chair uses tapered round tenons in reamed sockets, which are drilled at an angle, self-centre, self-tighten under load, and are forgiving of small errors. You will finish it and it will be strong.

Two free complete courses exist. Curtis Buchanan's twenty-part video series builds a democratic chair start to finish and his plans for it are free. Christopher Schwarz's The Stick Chair Book is 568 pages, free as a PDF, and covers seats, leg joinery, stretchers, arms, sticks and assembly with five sets of plans.

Buy one book: Peter Galbert, Chairmaker's Notebook. It is the most technically complete woodworking book written in the last twenty years — 500 of his own drawings, and the appendices on sightlines and drill-bit grinding are worth the price alone.

You do not need green wood or a shaving horse to start, although both help. A stick chair can be built from dry hardwood with a bandsaw, a lathe, a drill and hand tools.

B — Green wood, dry wood, and the joint that tightens itself

This is the deepest lesson in the chair and it is invisible if nobody tells you.

A tenon that is drier than its mortise gets tighter as they equalise. The tenon absorbs moisture from the mortise and swells; the mortise loses moisture and shrinks onto it. Traditional chairmaking exploits this deliberately: kiln-dry or oven-dry the tenon to roughly 5–6% MC, cut the mortise in stock at 10–12%, assemble, and the joint locks itself over the following days.

Do it backwards and the joint loosens forever. A wet tenon in a dry mortise shrinks as it dries and the chair racks within a year. This is why chairs bought cheap come apart.

Drying a tenon. A toaster oven or a warm box at about 110–140°F for a few hours. Check with a moisture meter if you have one; otherwise weigh the part, dry it, weigh it again, and stop when the weight stops falling.

Assemble quickly once dry — a dried tenon starts reabsorbing moisture from the room within the hour.

Grain direction in a leg or a stick is everything. Riven (split) stock follows the grain and is far stronger than sawn stock of the same species, because sawing cuts across fibres and creates short grain that snaps. For legs and spindles, split the blank from a log with a froe and a mallet, then shave it round. If you must use sawn stock, choose boards with grain running the full length and reject any with runout.

Reference: Jennie Alexander, Make a Chair from a Tree, 3rd ed. — the whole book is built around this moisture-differential joint. Galbert, Chairmaker's Notebook, covers it for Windsors.

C — Sightlines and resultants: drilling a compound angle

Every leg and every stick in a chair goes in at a compound angle — splayed to the side and raked front-to-back at the same time. Trying to hold two angles at once freehand does not work. The traditional method reduces two angles to one, and it is elegant.

The sightline is a line drawn on the seat, in plan, along which you stand to sight the drill. If a leg splays 10° to the side and rakes 14° to the back, there is one direction from which those two angles appear as a single angle.

The resultant is that single angle — the true angle of the hole measured in the plane of the sightline.

The arithmetic: the sightline angle from the centre line is arctan(splay ÷ rake), and the resultant angle from vertical is arctan(√(tan²splay + tan²rake)). You do not have to derive these — Galbert tabulates them, and the free plans give you the numbers directly.

In practice: draw the sightline on the seat in pencil. Stand so you are looking straight down it. Set a bevel gauge to the resultant and stand it next to the bit. Drill with the bit leaning in the plane of the sightline, matched to the bevel. Have someone sight you from the side, or clamp a mirror — you cannot judge your own lean.

The forgiveness: because the tenon is tapered and reamed, a degree of error changes the leg's position slightly but does not weaken the joint. This is why the staked system is the right one to learn on.

Ream after drilling. A tapered reamer opens the through-hole to match the tenon's taper. The classic taper is about 6° included. Ream from the top face, check the fit repeatedly with the actual leg, and stop when the leg seats with about 1/4″ of tenon still proud — you want room to drive it further at assembly.

Wedge across the grain of the seat, never with it. A wedge driven parallel to the seat's grain splits the seat. This is the mistake that ruins a finished chair at the last step.

D — Carving the seat

A saddled seat is the difference between a chair you sit in and a chair you sit on. It is also your shaping practice, which the audit says you have almost none of.

Map it first. Draw the contour lines on the blank — the free Buchanan plan includes a seat contour map. You are cutting to depths, not carving by feel.

Drill depth gauges. Bore a grid of shallow holes to the finished depth at each point with a depth stop. Carve until the bottoms of the holes just disappear. This single trick turns an intimidating sculptural operation into a bounded one and it is how professionals do it.

Tool sequence: adze or scorp for bulk removal, working across the grain from the high areas toward the low. Travisher to fair the hollow — a curved-sole spokeshave that rides the surface and leaves a clean concave cut. Scraper and a curved card scraper to finish. Sandpaper only at the end, and some chairmakers skip it entirely.

Work from the pommel outward. Keep the raised ridge between the legs and the rear edge crisp; those two lines are what make the seat read as carved rather than dished.

Grain direction reverses across a saddled seat. You will be cutting downhill on one side and uphill on the other of every hollow. Turn the work, not the tool.

Sharpen constantly. A travisher that is not sharp tears end grain and you will blame yourself. Every twenty minutes is not too often.

Reference: Galbert, Chairmaker's Notebook, the seat chapters and the appendix on making a travisher. Buchanan's free video series shows the whole operation in real time.

E — Steam bending the bow

Your audit has bending at 0.86/4 and this is where you fix it, on a part that must succeed because the chair depends on it.

Species and moisture decide the outcome before technique does. White oak, red oak, ash, elm, hickory and beech bend. Maple is poor. Cherry and walnut are marginal. And critically: air-dried stock at 12–16% MC bends; kiln-dried stock tears. Kiln drying permanently sets the lignin and the fibres will not slide. Most failed steam bends are this, not technique.

Riven stock, not sawn, for a bow. Grain runout across the bend is where it breaks.

Build the steambox first. 3/4″ exterior plywood (interior ply delaminates in steam), about 7″ square inside, dowel racks so steam circulates all round the part, a hinged door, and a drain. Heat with a propane burner and a few gallons of water. It must vent — never seal it, or you have built a pressure vessel. Michael Fortune's steambox article is free on Fine Woodworking and is the one to follow.

Target 200–212°F. One hour per inch of thickness. Oversteaming is a real failure mode — it wrinkles the compression face. Do not leave it in 'to be safe'.

The compression strap is not optional on a tight radius. Wood is strong in compression and weak in tension; the strap with end stops forces the outer face into compression rather than letting it stretch and tear. Without one you get the long splinters everyone blames on bad luck.

You have about 60 seconds from opening the box to clamped on the form. Rehearse the whole sequence cold, with gloves on, with every clamp laid out. Then do it.

You get about 10 minutes of working time once it is on the form to fine-adjust, so do not panic, but do not dawdle either.

Leave it on the form at least a week, ideally until it reads 7–8% MC. Taking it off early gives you a bow that slowly opens.

Springback is 5–15% and unpredictable. Overbend the form. Bend two, expect to lose one.

The physical explanation for all of the above is in the USDA Wood Handbook chapter 19, free, which covers plasticising, steaming times by thickness, and why the fixing period matters.

F — Turning the legs and spindles

You have not touched a lathe in four years and you scored 0.14/4 on it. Twelve to sixteen chair parts is exactly the right amount of turning to fix that, because it is enough repetition to build the motion and it is all in service of something real.

Speed by diameter. Keep diameter in inches × RPM under roughly 6,000–9,000. A 2″ leg blank: 1,500–2,000 RPM. Start slower than feels right.

Knock the corners off square blanks on the bandsaw first. It halves the roughing and removes the worst of the interrupted cut, which is where the fear comes from.

The one idea: the bevel rides the wood and the edge is brought into the cut by raising the handle. Rest the bevel on the spinning work with the edge not cutting, raise the handle slowly until a shaving starts, then move along. Every catch you will ever have is the edge getting ahead of the bevel.

Tool rest 1/8″ from the work, just above centre, and rotate the work by hand one full turn after every single rest adjustment. That habit is what stops you hitting the rest at speed.

The four tools: roughing gouge (square to cylinder only — never on a faceplate, its tang will break), spindle gouge (beads and coves, always cutting downhill from high point to low), skew (the planing cut, using the middle third of the edge only — leaves a surface needing no sanding, and is worth 50 inches of scrap practice before you touch a leg), parting tool (sizing cuts against calipers).

Sharpen far more often than feels necessary. An 8″ slow-speed grinder, white or blue aluminium-oxide wheel or CBN, and a Wolverine-style jig with a Vari-Grind arm. Freehand grinding a fingernail profile takes years; the jig takes an afternoon. Angles: roughing gouge 45°, spindle gouge 35–40°, skew 20–25° per side, parting tool 30° per side.

Make a story stick with every diameter and position marked, and a set of calipers preset. Sixteen parts that must match is a repeatability problem, not a turning problem.

Face shield, not glasses. Stand out of the firing line for the first ten seconds.

Free reference: Richard Raffan's YouTube channel — unedited, unscripted, including his mistakes and recoveries, which is what you actually need to see. The AAW Safety Guidebook for Woodturners is a free 26-page PDF and is the document to adapt into your shop's posted lathe policy.

H — Sharpening turning tools

Turning tools are sharpened far more often than bench tools — every few minutes on abrasive timber. If you are not sharpening often you are not turning, you are struggling.

Grinder, not stones. An 8″ slow-speed grinder (1725 RPM) with a white or blue aluminium-oxide wheel, 60 or 80 grit. Grey wheels burn. CBN wheels are better and do not need dressing.

Dress the wheel with a diamond dresser whenever it glazes. A glazed wheel burns the edge, and burnt steel loses its temper and never holds an edge again.

Use a jig. Wolverine-style with a Vari-Grind arm for gouges. Freehand grinding of a fingernail grind takes years; the jig takes an afternoon and gives you a repeatable edge you can return to in ten seconds.

Angles: roughing gouge 45° · spindle gouge 35–40° with the fingernail sweep · bowl gouge 45–55° (60° or more for the inside of a bowl) · skew 20–25° per side, and keep it slightly curved rather than dead straight · parting tool 30° per side.

Light touch, and keep it moving. You are removing a few thousandths. Blueing the edge means you pressed too hard.

No honing needed for gouges — straight off the wheel with the grinder burr is the working edge. The skew is the exception: hone it, because its whole value is the surface it leaves.

E — Bent lamination, steam bending, kerf bending

Three ways to make wood curve. They fail differently and you should know all three by feel, not by reading.

BENT LAMINATION — the reliable one.

Saw laminae on the bandsaw, 1/8″ or thinner for a tight radius. Rule of thumb: a lamina will bend to a radius about 100–150× its thickness without steam. Resaw, then run each through a thickness sander or plane — a bandsawn face glues badly.

Keep the stack in order and mark a triangle across the edge before you separate them, so the grain lines up again and the finished edge reads as one board.

Two-part form, both halves. A single form with clamps applies point loads and starves the glue between them. Cut both halves on the bandsaw from the same layout line, allowing for the stack thickness, and face them with cork or thin rubber.

Glue matters more here than anywhere. PVA creeps — the curve slowly opens over years. Use urea-formaldehyde (Unibond 800, Cascamite) or epoxy for laminations that must hold shape. Read the open time before you spread any of it; a six-lamina glue-up is not fast.

Springback is roughly 2–5% of the bend with UF glue, more with PVA, more with fewer and thicker laminae. Measure it on this piece and write the number on the form — then make the form that much over-bent next time. That number is the whole point of this exercise.

STEAM BENDING — the fast one that fails.

Species matters absolutely. White oak, ash, elm, beech, hickory bend. Maple is poor. Cherry and walnut are marginal. Kiln-dried anything is much worse than air-dried. Most failed steam bends are a species and moisture problem, not a technique problem.

One hour of steam per inch of thickness at boiling, in a box that is not airtight (it needs to vent or it is a pressure vessel).

You have about 60 seconds from opening the box to having it clamped on the form. Have the form, the strap and every clamp laid out and rehearsed dry, with gloves on. Rehearse it once cold.

Use a compression strap with end stops for anything past a gentle curve. Wood is strong in compression and weak in tension; the strap forces the outer face into compression rather than letting it pull apart. Without one, the outside fibres tear and you get the long splinters everyone blames on bad luck.

Leave it on the form until dry — days, not hours. Springback is 5–15% and unpredictable, which is exactly why lamination is used for anything that must be repeatable.

KERF BENDING — the cheap one, for hidden or veneered work.

Saw kerfs across the back, stopping 1/8″ short of the show face. Spacing sets the radius.

Find the spacing empirically, once: cut one kerf in a test piece, clamp one end flat, lift the other until the kerf just closes, and measure how far the free end rose over a known distance. That gives the angle one kerf absorbs. Divide the total angle of your curve by it to get the number of kerfs, then divide the arc length by that to get the spacing.

It is weak and it telegraphs. Fill the kerfs with glue or epoxy while bent, and put it where the back never shows.

C — Mortise and tenon

Rule of thumb: tenon thickness = 1/3 of stock thickness. In 3/4″ stock, a 1/4″ tenon.

Mortise first, tenon to fit it. It is far easier to trim a tenon than to widen a mortise.

Layout. Set a mortise gauge to the chisel width, then mark from the face side of every piece, always. Knife lines, not pencil.

Hand chopping. Start 1/8″ in from each end, chisel vertical, bevel toward the waste. Work to the middle, lever out, then pare the ends to the knife line last.

Tenon by hand. Saw the cheeks on the waste side of the line, then the shoulders. Pare with a router plane or chisel to a consistent thickness.

Fit. It should go together with hand pressure and hold itself when lifted. If you need a mallet it is too tight and will split the stile when glue swells it.

Haunch. On a grooved stile the tenon needs a haunch at the top — a short stub that fills the groove where it runs out at the end of the stile. Length = groove depth.

A — What an instrument forces you to learn

You raised this yourself and it is a good instinct. An acoustic instrument is the densest woodworking object there is: it is built to roughly ±0.005″ in places, out of stock thinned to 0.090″, bent in two planes, with a structure that has to be simultaneously strong enough to resist about 160–180 lb of string tension and light enough to vibrate.

What it teaches that nothing else on this plan does: working to thickness by feel and by tap tone rather than by number; bending thin stock over a hot pipe; precision layout at the scale of a thousandth; inlay and binding; a structural joint (the neck) that must be both rigid and removable; and French polish, which is the most controlled finishing technique in woodworking.

Be honest about the time. A first steel-string takes most people 150–250 hours. This plan budgets around 130, which means you should either build a smaller instrument (a parlour guitar, a tenor ukulele, a lap dulcimer) or accept that it runs into the following month. A tenor ukulele or a parlour guitar teaches ninety percent of the same lessons in half the time and is the sensible first build.

The sculptural option. If a conventional guitar does not interest you, the same technical content sits inside an invented instrument — a monochord, a bowed sound sculpture, a set of struck metal and wood resonators. You keep the bending, the thicknessing, the acoustics and the precision, and you get an object that reads as your work. Say which you want and the module adapts; the procedures below apply either way.

Buy one book and read one free correction page. Cumpiano & Natelson, Guitarmaking: Tradition and Technology, is the standard. Cumpiano publishes a free updates-and-corrections page that you must read before building from the book — it replaces the book's single-action truss rod and its pinned neck joint, and corrects several dimensions.

B — Bending sides on a hot pipe

Different from steam bending and easier to control. You are plasticising a thin piece with heat and a little moisture and bending it by hand, continuously, feeling it go.

Thickness the sides to 0.070–0.090″ first, evenly, checked with a caliper across the whole surface. Thickness variation is the main cause of a crack at the waist.

Pipe temperature 300–350°F. A steel or aluminium pipe over a heat gun or a purpose-made element. Test with a drop of water — it should dance and evaporate in about a second, not instantly.

Wet the wood lightly or wrap it in a damp cloth or a thin metal slat. You want steam generated at the surface, not a soaked board.

Keep it moving, always. Constant slow motion back and forth with firm even pressure. A pause scorches, and a scorch is a weak line that cracks later.

Bend past the target and let it relax back. Check against a plan drawn full size or against the mould, repeatedly.

Figured wood, and especially highly figured maple, cracks. Go slower, wetter, and cooler, and expect to lose one. Order enough for three sides.

Into the mould immediately and leave it to cool and set for at least several hours, ideally overnight.

Free reference: Robert O'Brien's Luthier Tips du Jour YouTube series covers side bending, and StewMac's free resource library has around 765 articles and videos including bending, binding and fretwork.

C — Thicknessing, bracing and voicing

This is where instrument making stops being woodworking and becomes something closer to engineering by ear, and it is the part worth learning even if you never build a second one.

Soundboard thickness is a starting number, not a target. Spruce tops for a steel string run roughly 0.100–0.120″, thinner at the edges. But the number depends on the stiffness of the particular piece, which varies by a factor of two between two boards of the same species.

Deflection testing is how you replace guesswork. Support the plate on two rails a fixed distance apart, hang a known weight at the centre, measure the deflection with a dial indicator. Do it across the grain and along it. Now you are thicknessing to a stiffness, not to a dimension, and this is the single professional technique that separates a good instrument from a lucky one.

Tap tone. Hold the plate at a node and tap. You are listening for a clear ringing note with sustain rather than a dead thud. It is real information, it is learnable, and it is also not a substitute for measurement — use both.

Bracing does two jobs at once: it resists the roughly 160–180 lb of string tension trying to fold the top, and it tunes the plate's stiffness distribution. Every brace you shave makes the top livelier and weaker. The craft is knowing where the structural minimum is.

Shave braces in small increments, re-tapping as you go, and stop early on your first build. A slightly over-braced first guitar is quiet; an under-braced one belly-ups and is firewood.

The engineering-literate treatment is Trevor Gore & Gerard Gilet, Contemporary Acoustic Guitar Design and Build — the Design volume explains the acoustics and mechanics properly and relates measured stiffness to sound. Buy it if this module interests you; skip it if you are building one and moving on.

Free and peer-reviewed: the Savart Journal is open access and is where the serious instrument-acoustics research is published.

D — Fret layout and the neck joint

Fret spacing is not a craft decision, it is arithmetic, and being wrong by 0.010″ makes the instrument audibly out of tune.

The rule of 18, properly: each fret sits at the remaining scale length divided by 17.817. Equivalently, the distance from the nut to fret n is L − L / 2^(n/12), where L is the scale length. Work from the nut to each fret with a single cumulative calculation — never measure fret to fret, or the errors stack.

Buy a fret slotting template or a pre-slotted board for your first build. The saving in frustration is enormous and nobody will think less of you.

Compensation. A fretted string goes slightly sharp when pressed, because pressing stretches it. The saddle is therefore set slightly further from the nut than the theoretical scale length — roughly 1/16″ for the treble side, 1/8″ for the bass, and angled between. Get this wrong and the instrument plays in tune at the nut and sharp at the twelfth fret.

The neck angle determines the action, and it is set at the neck joint, not at the saddle. Aim the neck plane so that a straightedge lying on the frets arrives at the top of the bridge (or a hair above). Too low and you have no saddle height left to adjust; too high and the action is unplayably tall.

Use a bolt-on neck joint on a first build. Cumpiano's own corrections page recommends replacing the book's pinned mortise and tenon with a hardware-based joint, precisely because it is adjustable and repairable. A glued dovetail neck joint is traditional, beautiful, and a serious problem when the neck needs resetting in twenty years.

Double-acting truss rod, again per Cumpiano's corrections — a single-action rod can only correct relief in one direction.

Free reference: StewMac's Trade Secrets library, 354 free items, is the best source for fretwork, nut and saddle setup and neck adjustment. Their fret calculator is free.

E — French polish

The finish worth learning, because it is the most controlled application technique in woodworking and it teaches a feel that transfers to every other finish.

What it is: shellac, applied in hundreds of thin layers with a cloth pad (a 'muñeca' or rubber) charged with shellac and lubricated with a trace of oil. Each pass melts slightly into the last, so there are no witness lines and the film ends up remarkably thin — which matters on an instrument because a thick film damps the top.

Shellac cut. Mix from flakes, fresh. A 1 to 1.5 pound cut for bodying, thinner for the final spiriting off. Old shellac never fully hardens — buy flakes, not a can that has sat for two years, and mix what you need.

Pore filling first on open-pored woods like rosewood or mahogany: pumice worked in with the pad, or a proprietary filler. A finish over unfilled pores will sink and look patchy no matter how many coats you apply.

The pad. A wool or cotton core inside a lint-free linen outer, charged from the inside, never dripping. It should feel almost dry.

Never stop the pad on the surface. Land it moving and lift it moving, in a continuous figure-eight or circular motion. A pad set down stationary sticks and pulls the film.

A trace of oil — a single drop of mineral or walnut oil on the pad face — is the lubricant that makes the motion possible. Too much and the finish stays soft and cloudy.

Bodying, then spiriting off. Many sessions of building the film, with rest periods of hours between, then final passes with a nearly dry pad and a little alcohol to remove the oil and level the surface.

It is repairable, which is the other reason to know it: a damaged shellac finish can be re-amalgamated with alcohol rather than stripped.

Reference: Robert O'Brien's online course covers his French polish method in detail; Flexner, Understanding Wood Finishing, has the chemistry of why it behaves as it does.

C — The router, handheld and in the table

The router is the most dangerous tool in a wood shop that nobody is afraid of. 20,000 RPM, no guard worth the name, and it climbs if you feed the wrong way.

Feed direction. Handheld, the bit spins clockwise seen from above, so you move the router left to right along an outside edge and right to left inside a frame. In the table the bit is upside down, so you feed right to left, against the rotation, workpiece between you and the fence. Getting this backwards is a climb cut — it snatches the work out of your hands.

Climb cutting on purpose is a real technique for a very light final pass on tearout-prone grain, handheld only, very light, never on the table with a small piece.

Bit depth in passes. Never more than about 1/8″ of cut per pass in hardwood on a table, less for a large profile. A full-depth pass in one go burns the bit, burns the wood and loads the motor.

Bit shank fully seated, then backed out 1/16″ before tightening, so the collet is not gripping on the shank radius. A bit bottomed hard in the collet works loose. Clean the collet — pitch in a collet is the usual cause of a bit creeping down mid-cut.

Speed by bit diameter. Up to 1″ full speed · 1–2″ ~18,000 · 2–2.5″ ~16,000 · over 3″ (panel raisers) 12,000 max. A big bit at full speed has a rim speed that will explode it.

Table fence is a split fence — the outfeed half supports the work after the cut. For a full-width cut (jointing on a router table), shim the outfeed half out by the depth of cut, exactly like a jointer.

Starting pin for any bearing-guided curved work done without a fence. The work contacts the pin first, pivots into the bit. Starting a curved piece straight into a spinning bearing bit with no pin is how the work gets thrown.

Featherboards before the bit and above the work, never after the bit — a featherboard past the cut traps the offcut.

D — Template routing

This is how you get identical parts, and it is the technique that makes curved and repeated work possible at all.

Make the template out of 1/2″ MDF, not ply — MDF takes a fair edge off a spokeshave and a sanding block and holds it. The template edge is the finished edge; every flat spot and bump transfers to every part. Spend the time here.

Bandsaw the part to within 1/16″ of the line first. A flush-trim bit is for taking off a sixteenth, not a quarter inch. Asking it to remove too much is how you get burn, tearout and a grabbed workpiece.

Bit choice. Flush-trim has the bearing at the tip and the template goes underneath. Pattern bit has the bearing at the shank and the template goes on top. Having both lets you flip the work to cut downhill on grain that wants to tear.

Grain direction changes around a curve. On an S-curve you will reach a point where you are cutting uphill and it tears. That is the moment to stop, flip the work over, swap to the other bit style, and come back at it from the other direction.

Fix the work to the template with double-sided tape (a lot of it, pressed hard) or toggle clamps on a purpose-built carrier. Screws only where the holes will be hidden.

Spiral bits over straight bits for this. Up-cut clears chips and pulls the work down onto the table; down-cut leaves the cleanest top edge but packs chips. Compression bits do both and are worth it on veneered ply.

F — Veneer and glue-up

Substrate. MDF or high-grade ply. Never solid wood under veneer across the grain — the substrate moves and the veneer cannot.

Balance every panel. Veneer on one face only will cup, without exception. Put a backer veneer of similar thickness and similar species on the other face, grain running the same direction, glued at the same time.

Jointing veneer edges for a book-match: tape the two leaves together, clamp between two straight battens with the edges just proud, and plane or shoot the pair in one pass so any error is mirrored and closes. Tape the show face with veneer tape (water-based, removable), not masking tape.

Vacuum bag is the method that works: even pressure everywhere, no starved spots. 10–14 inHg is enough for veneer; more is not better and can crush softwood cores.

Hammer veneering with hot hide glue is the method that needs no press, and it is worth learning because it also lets you repair, re-lay a bubble with an iron, and work on a curve. Glue both surfaces, let them tack, then squeeze the glue out from the centre with the veneer hammer, reheating with an iron as you go.

Cauls and a platen if you have neither: a flat platen, the panel, waxed paper, and clamps working from the centre outward. Crown the cauls slightly so the middle gets pressure first.

GLUE-UP, generally. Open time is how long the glue stays workable after spreading; clamp time is how long it must stay clamped; cure time is when it is at full strength and can be machined. Standard PVA: ~5 min open, 30–60 min clamp, 24 h cure. These are three different numbers and people confuse the last two.

Always dry-fit with every clamp, timed. If the dry run took eight minutes and your glue has five minutes of open time, you need a slower glue or a staged glue-up — not more speed.

Clamp pressure: hardwood wants roughly 150–250 psi, softwood less. Clamps at 45° spread from each clamp face, so spacing should be about twice the stock thickness from the edge and close enough that the cones overlap. Squeeze-out should be an even small bead, not a flood — a flood means a starved joint.

Cold weather kills PVA. Below about 50°F it chalks and never reaches strength. Check the shop temperature before a winter glue-up.

E — Wood finishing

The one distinction that organises everything: film finishes sit on top, penetrating finishes go in. Film finishes (shellac, lacquer, varnish, polyurethane, water-based) protect better and can be rubbed out to any sheen; penetrating finishes (oil, oil-varnish blends, wax) look and feel better, repair easily and protect poorly. Everything else is detail.

Surface prep decides the finish, not the finish. Sand through the grits with no skips — 80, 120, 150, 180, 220 — and each grit must remove the previous grit’s scratches entirely. Stop at 180–220 for film finishes; going finer burnishes the surface and can stop it absorbing stain evenly.

Raise the grain before the first coat of any water-based finish or water stain: wipe with a damp rag, let it dry, knock the fuzz back with 220 or 320. Skip it and the first coat raises it for you, inside the finish, where you cannot fix it.

Light a raking light across the surface before finishing. Every scratch, glue spot and planer ripple you cannot see now becomes permanent and obvious under the finish. This single habit separates work that looks finished from work that looks made.

SHELLAC. Dissolved lac resin in alcohol. Dries in minutes, every coat melts into the last so there are no witness lines, non-toxic when cured, and it sticks to everything and everything sticks to it — which makes dewaxed shellac the universal barrier and sealer coat. Weak against water and alcohol. Sold as flakes measured in pounds cut (lb of flake per gallon of alcohol); a 1–2 lb cut is what you actually want to brush or pad. Mix it fresh — it has a shelf life of months and old shellac never hardens.

OIL AND OIL-VARNISH. Pure tung or boiled linseed penetrates and cures slowly; ‘Danish oil’ and most wiping finishes are oil plus varnish plus thinner, which is why they build a little. Flood, let it sit, wipe off every bit of excess after 15 minutes. Oil left to dry on the surface stays gummy forever. Oily rags spontaneously combust — linseed oxidises exothermically in a wadded rag. Spread them flat or drown them in water. This has burned down real shops.

VARNISH AND POLYURETHANE. The toughest brushable film, best water and heat resistance. Slow drying means dust nibs — thin the first coat, sand between coats with 320, and work in a clean space. Oil-based ambers with age; water-based stays clear but looks colder on warm woods.

WATER-BASED. Low odour, fast, clear, and it is now genuinely good. It raises grain, it does not amber (so cherry and walnut look flat under it unless you seal with shellac or dewaxed shellac first), and it is fussy about temperature and humidity.

LACQUER. Sprays beautifully, dries in minutes, burns into itself, and is the production shop standard. Needs real ventilation and an explosion-proof setup — check what your shop is actually rated for before spraying it. Pre-catalysed lacquer is more durable and still sprayable.

WAX. Not a finish; a topcoat for feel and sheen over something else. Wax under anything prevents adhesion permanently — this is why silicone and wax contamination is the great enemy in a finishing room.

Blotch-prone woods — cherry, birch, maple, pine — take stain unevenly because of variable density. Seal first with a washcoat of 1 lb cut shellac, or use a gel stain that does not penetrate, or dye rather than pigment stain.

Dye vs pigment: dye is molecular, colours the wood itself, stays transparent and fades in UV. Pigment is ground solids that lodge in pores and scratches, so it emphasises grain and open pores, and is lightfast. Most ‘stain’ on a shelf is both.

Rubbing out a cured film: 600 then 1000 then 1200 wet with soapy water or mineral spirits, then 4/0 steel wool and wax for satin, or pumice and rottenstone or automotive polish for gloss. The finish must be fully cured first — days to weeks, not hours — or you drag it.

A — Four-jaw indicating

Mount the work roughly centred, jaws snug but not tight.

Indicator on the work, rotate by hand, find the high spot.

Loosen the jaw opposite the high spot by half the indicated error, tighten the jaw at the high spot by the same amount. Work in opposing pairs.

Repeat, halving each time. Target ≤0.001″ TIR in five minutes once you have done it twenty times.

B — Tool height

Wrong height is the cause of most chatter, poor finish and parting disasters.

Rule test: hold a 6″ steel rule vertically between the tool tip and the work, lightly pinched. If the rule stands vertical, the tool is on centre. Leaning with the top away from you = tool above centre; toward you = below.

Facing test: face a bar. A tiny nub left at the centre means the tool is off centre. On centre leaves nothing.

For turning, on centre. For parting, exactly on centre or fractionally above — never below.

C — Turning to a dimension, and speeds

RPM = (SFM × 3.82) ÷ diameter in inches.

Surface speeds to start from: mild steel 90 SFM HSS / 350 SFM carbide · aluminium 300 HSS / 800 carbide · brass 200 HSS / 500 carbide · stainless 304 60 HSS / 200 carbide.

Example: 1.000″ 1018 with carbide → (350 × 3.82) ÷ 1.0 = 1337 RPM.

Feed: 0.002–0.005″ per revolution for roughing, 0.001–0.002″ for finishing.

Always leave 0.010″ for a finish pass, then take exactly that 0.010″ at finishing feed. Sneaking up in 0.001″ steps gives a worse finish, not a better one, because the tool rubs instead of cutting.

Remember the cross slide removes twice its reading from the diameter on most lathes. Check which yours does before your first part.

Measure with a micrometer, not a caliper, for anything under 0.002″ tolerance. Let the part cool first — hot steel reads large.

D — Parting off

The operation that scares everyone. It is a rigidity problem, not a courage problem.

Blade exactly on centre. Absolute minimum overhang from the holder. Blade square to the work in both planes.

Half the turning speed. Steady, continuous feed — if you hesitate, the tool rubs, work-hardens the material and then digs in.

Part close to the chuck. Never part a workpiece held between centres.

Flood coolant or cutting oil. Withdraw and clear chips on deep cuts.

If it chatters: shorten the overhang, raise the tool 0.005″ above centre, slow the speed, increase the feed.

E — Threading

Die. Chamfer the end 45° to just past the thread minor diameter. Turn the blank to the nominal size minus 0.002″. Start the die square using the tailstock chuck as a guide. One full turn forward, half a turn back, repeat. Cutting oil throughout.

Tap. Drill size = major diameter minus thread pitch for 75% thread. Start square, same forward-back rhythm.

Single point. Compound set to 29°, feed on the compound only, 0.005–0.010″ per pass roughing, 0.002″ finishing. Never disengage the half-nut until the thread is finished. Test with a nut, not by eye.

A — Tramming

Indicator on an arm in the spindle, sweeping a 6″ radius on a clean table or a ground plate.

Read front, rear, left, right. Adjust the head nod and tilt until all four read within 0.001″ over that 12″ sweep.

Tighten the head bolts and re-sweep — tightening always moves it.

Out-of-tram shows up as a faceted, stepped finish on a fly-cut or face-milled surface. If you see that, stop and re-tram.

B — Squaring a block, six sides

Face 1: clamp on parallels, mill the largest face flat. This is now the reference face.

Face 2: reference face against the fixed jaw, with a round rod between the work and the moving jaw so the rod forces the reference face flat against the fixed jaw. Mill.

Face 3: face 2 down on parallels, face 1 against the fixed jaw. Mill.

Face 4: mill to thickness, referencing face 3.

Faces 5 and 6: stand on end against a square, or use a 123 block. Mill to length.

Check every pair with a known square and the caliper. Target ≤0.002″ on all six.

The rod trick is the whole technique. Without it the vise pulls the part out of square and you never find out why.

C — Fits

Free running: 0.003–0.005″ clearance. Anything that must turn under load.

Sliding / location: 0.001–0.002″ clearance. Hinge pins, locating pins.

Transition: 0.000–0.0005″. Needs a light tap.

Press: 0.0005–0.0015″ interference per inch of diameter. Needs an arbor press. Freeze the pin or heat the housing.

Measure both parts with a micrometer, not a caliper. Subtract. The number is the fit.

D — Hand-fitting a square

Mill both parts close, then finish with files and stones. Machining alone will not get you to 0.001″.

The flip test. Scribe a line along the blade against a straight reference edge. Flip the square over and scribe again from the same point. Any error doubles between the two lines. A 0.001″ error shows as a 0.002″ gap at the far end.

Correct by filing the stock face, not the blade. Take a few strokes, retest. This takes an hour and it is the hour that teaches you what square means.

Blue the mating surfaces to find the high spots if you cannot see them.

E — Cutting metal stock

Horizontal bandsaw is the shop workhorse. Blade pitch by wall thickness: at least three teeth in the cut at all times. Thin wall (tube, angle) wants 14–18 TPI; solid bar wants 6–10. Too coarse a blade strips teeth on tube; too fine a blade packs and burns on solid.

Speed: mild steel ~150–250 FPM · stainless ~80–120 · aluminium 500+ FPM. Running stainless at mild-steel speed work-hardens it under the tooth and the blade stops cutting within a minute.

Feed by weight, and let the saw cut. Riding the head down is how blades break. Coolant or cutting oil on anything but aluminium and cast iron.

Cold saw for a square, clean, burr-free cut — the best cut you can get and the one to use on anything that gets welded or machined after. Carbide or HSS blade, slow RPM, flood coolant. Clamp tight and close to the cut.

Abrasive chop saw is fast, loud, throws sparks, and leaves a cut that is hot, burred and rarely square. Fine for rough stock, not for a mitre you intend to weld without dressing. Never cut aluminium on one — aluminium loads the wheel and the wheel can burst.

Shear for sheet: clean, no heat, no distortion, no dust. Limits are on the machine plate and they are not suggestions — shearing over capacity bends the blade bed. It leaves a slight edge camber and a burr on the underside; know which side of your cut is the good side.

Burrs come off before the next operation, always. A burr on a face means the part sits wrong in the vice and everything you machine after is out.

F — Drilling and tapping steel

Speed. RPM = (SFM × 3.82) ÷ diameter — the same formula as the lathe. SFM for HSS: mild steel 80–100 · stainless 40–60 · aluminium 250–300 · cast iron 60–80 · brass 150–200. So a 1/2″ drill in mild steel: (90 × 3.82) ÷ 0.5 ≈ 690 RPM. Most people run twice that and wonder why bits die.

Cutting fluid on everything except cast iron and brass, which cut dry. Dark sulphurised oil for steel, kerosene or a wax stick for aluminium.

Peck and clear. Deeper than 3× diameter, retract fully every peck.

Centre drill or spot drill first on anything that matters. A twist drill walks; a spot drill does not.

Pilot then full size for anything over about 1/2″: pilot to roughly the width of the big drill’s web.

Tap drill size = major diameter − pitch (metric), which gives about 75% thread. Imperial: use a chart, or major − (1 ÷ TPI). Examples: M6×1 → 5.0 mm · M8×1.25 → 6.8 mm · 1/4-20 → #7 (0.201″) · 5/16-18 → F (0.257″) · 3/8-16 → 5/16″.

75% thread is the target, not 100%. Going from 75% to 100% adds about 5% strength and roughly doubles the tapping torque — that is the trade that breaks taps. In a blind hole in stainless, drop to 65%.

Taper, plug, bottoming — that is the sequence in a blind hole, in that order. Taper alone will not thread to the bottom.

Start square. Use the drill press or mill with the spindle unpowered and the tap in the chuck, or a tap guide block. A tap started crooked is a broken tap.

Quarter turn forward, back off until you feel the chip snap, forward again. That is the rhythm. In a blind hole, back the tap out completely and clear the chips two or three times.

Extracting a broken tap: tap extractor with fingers down the flutes if enough is protruding; otherwise a carbide burr in a die grinder, or an EDM if the part is worth it. Alum solution dissolves HSS taps in steel slowly and does not touch the steel — genuinely useful and almost no one knows it.

Thread-forming (roll) taps in aluminium and mild steel: no chips, stronger thread, much harder to break. Different, larger drill size — check the chart, do not use the cutting-tap number.

A — What the dials mean

Voltage sets arc length, and therefore bead width and flatness. More volts = wider, flatter, shallower bead. Too high = spatter and undercut; too low = a ropey, humped bead that sits on top.

Wire feed speed is amperage in MIG. The machine is constant voltage — it holds voltage steady and lets current follow the wire. More wire = more amps = more penetration and more deposit.

Constant voltage (CV) is for wire processes. Constant current (CC) is for stick and TIG, where the arc length is controlled by your hand, so the machine has to hold the amps steady instead.

DCEP (electrode positive, ‘reverse polarity’) puts about 2/3 of the heat at the work: deeper penetration. Standard for solid-wire MIG and most stick rods.

DCEN (electrode negative) puts more heat at the electrode: less penetration, more deposition. Used for self-shielded flux core.

Get polarity wrong and the arc is erratic, spattery, and does not fuse. On a wire machine polarity is changed by swapping the leads inside the drive compartment, not by a switch.

Gas. C25 (75% argon / 25% CO₂) is the default for short-circuit MIG on steel: stable arc, low spatter. 90/10 or 98/2 for spray transfer. Pure argon on steel gives a narrow, erratic arc and terrible wetting — it is for aluminium only. Straight CO₂ penetrates deepest and spatters most.

Stickout 3/8″. Longer stickout = less current at the same wire speed = less penetration.

Wire stubs into the plate when wire speed is too high for the voltage. Burnback to the tip when wire speed is too low. Both are a voltage-to-wire-speed mismatch, not a machine fault.

Push vs drag. Push (forehand) gives a flatter, wider bead with less penetration — good for thin material. Drag (backhand) gives more penetration and a narrower bead. Flux core is always dragged, so the slag follows.

Duty cycle at 40% means six minutes of arc time in any ten. Exceed it and the machine thermally trips; do it repeatedly and you cook it.

B — The parameter ladder

Cut 25 coupons of identical 1/4″ plate. Set up a T-joint fillet on each.

Pick five voltages across your machine’s useful range and five wire speeds. Weld every combination.

Stamp or scribe the settings into each coupon before you weld it. Not after — you will mix them up.

Line them up and photograph. You now own a settings chart made on your machine, with your gas, in your shop air.

This is two hours of work and it replaces years of guessing.

C — Sectioning and etching

Cut the weld across the bead with a bandsaw or cut-off wheel.

Grind flat, then sand to 400 grit. The surface must be scratch-free or the etch shows nothing.

Swab with ferric chloride or 10% nital for 10–60 seconds. Rinse, dry, oil.

The weld metal, the heat-affected zone and the parent metal will show as three distinct tones.

Look at the root. A gap or a dark line at the root of the joint is lack of fusion — the defect you cannot see from outside and the one that fails.

Photograph every section next to its settings coupon.

D — Producing defects deliberately

Undercut: too much voltage, too much travel speed, wrong gun angle. A groove melted into the parent metal at the toe of the weld.

Porosity: pull the gas line, or weld over mill scale, paint or moisture. Gas trapped as the weld freezes.

Cold lap: too little heat, too much travel. Weld metal sits on unfused parent metal — looks fine, holds nothing.

Lack of fusion: wrong gun angle into a tight joint, or insufficient root opening. The bead bridges instead of penetrating.

Photograph each, section each, and write the cause. This becomes a teaching set no book gives you.

E — Distortion control

Weld metal shrinks as it cools and pulls the joint toward the weld.

Tack every 4–6″ before welding anything long. Check the assembly is square after tacking and before welding.

Back-step: work in short runs, each one laid in the opposite direction to the overall progression.

Balance welds side to side — weld a bit on one side, then the matching bit on the other.

Stitch rather than run continuous on light section. Continuous beads on 11 ga tube will banana it.

Clamp to a flat table and leave it clamped until fully cool.

Measure your test frame before and after. Log the number for each method.

F — The angle grinder hour

Discs: cut-off (1–3 mm, never side-load it), grinding (6 mm, for heavy stock removal), flap (blending and finishing), wire cup (scale and rust), strip disc (paint).

Never fit a cut-off wheel to a grinder with the guard removed. Never use a cut-off wheel for side grinding — it shatters.

Guard position: between the wheel and your body, always. Rotate it for the cut you are making.

Kickback goes in the direction opposite the wheel’s rotation at the point of contact. Grip so that if it kicks, it kicks away from you and not into your leg.

Let the wheel reach full speed before contact. Ease in; do not plunge.

Check every disc for cracks and check the expiry date on bonded abrasives. They do expire.

G — Why the brace goes that way

A four-sided frame is a parallelogram and it will rack. A triangle cannot.

The diagonal must run from the bottom hinge corner up to the top latch corner. In that orientation the brace is in compression, and steel in compression over a short brace does not stretch or sag.

Run it the other way and the brace is in tension — which works for a cable but is the wrong use of a rigid member, and any weld creep shows as sag.

Fit the brace tight. A brace with a gap does nothing until the gate has already sagged.

H — TIG

TIG is the process you do not have at all (M3 scored 0/4 on every item), and it is the one a sculpture department asks about, because it welds aluminium, stainless and thin material, and because it is what a student wants for visible welds.

What the letters mean. GTAW. A non-consumable tungsten electrode makes the arc; you add filler by hand with the other hand; argon shields it. Nothing is consumed but the filler and the gas, so there is no spatter and no slag.

Polarity is the whole thing. DCEN (electrode negative) for steel and stainless — about 70% of the heat goes into the work, deep penetration, tungsten stays cool. AC for aluminium — it alternates, and the electrode-positive half of each cycle blasts the aluminium oxide layer off the surface. Aluminium oxide melts at about 3,700°F while the aluminium under it melts at 1,200°F; without that cleaning action you are trying to weld under a ceramic skin. DCEP is almost never used — it cooks the tungsten.

Amperage: roughly 1 amp per 0.001″ of steel thickness. 16 ga (0.060″) → ~60 A. 1/8″ (0.125″) → ~125 A. Aluminium wants about 25–30% more because it dumps heat away so fast — and then wants less as the part heats up, which is what the foot pedal is for.

The pedal is an amperage control you use continuously, not an on/off switch. Come in hot to start the puddle, back off as the part heats, taper out at the end to fill the crater. A crater crack at the end of a bead is a pedal problem.

Tungsten. 2% lanthanated (blue) does everything, AC and DC — buy that and stop thinking about it. 2% thoriated (red) is the old standard for DC and is mildly radioactive when ground; ceriated (grey) is good at low amps. Pure tungsten (green) is obsolete on modern inverter machines.

Grinding the tungsten: grind along the axis, never around it. Grinding marks running around the tip make the arc wander. Dedicated wheel or a belt sander, never the wheel you grind steel on. Taper 2–2.5× the diameter for DC, with a tiny flat on the tip; blunter for AC and high amps.

Touch the tungsten to the puddle or the filler and it is contaminated — the arc goes fuzzy and wide instantly. Stop, break the end off, regrind. Everyone does this constantly at first. Keep three ready.

Gas. 100% argon for both steel and aluminium. 15–20 CFH through a #6–#8 cup. More gas is not more protection — too high a flow goes turbulent and pulls air in. Post-flow 1 second per 10 amps, so the tungsten and the hot weld are still shielded as they cool. A gas lens doubles the useful coverage and costs almost nothing.

Filler. ER70S-2 for mild steel · ER308L for 304 stainless · ER4043 for general aluminium (more fluid, less crack-prone) or ER5356 (stronger, better colour match if anodising). Filler diameter roughly the material thickness.

Cleanliness is not a suggestion in TIG. Stainless brush kept only for stainless, acetone wipe, no grinding dust, no oil, no mill scale. On aluminium, wire-brush the oxide off with a stainless brush immediately before welding — the oxide re-forms in minutes.

Torch angle 10–15° from vertical, filler in from about 15° at the front. Arc length equal to the tungsten diameter, and that is much closer than feels right. Dip the filler into the leading edge of the puddle and withdraw — never hold it in the arc.

Technique order: run puddles with no filler until you can walk a consistent puddle the length of a coupon. Then add filler. Then do joints. Skipping the first stage is why people flounder for months.

I — Stick and flux-core

You already stick weld reasonably. What follows is the part you said you do not have — why the numbers are what they are.

SMAW. The flux coating does three jobs: it creates a shielding gas as it burns, it forms slag that slows cooling and shapes the bead, and it carries alloying elements into the weld. That is why the coating being damp ruins the weld — hydrogen from moisture goes into the steel and cracks it days later.

Reading an electrode number, E7018: E = electrode. 70 = 70,000 psi tensile. 1 = all positions (2 = flat and horizontal only). 8 = the coating and current type — here iron powder, low hydrogen, AC or DCEP.

E6010 — deep digging penetration, cuts through rust and paint, cellulose coating, DCEP, and it is the standard root pass on pipe. Whip and pause technique.

E6011 — the same idea but runs on AC, so it is the buzz-box electrode.

E7018 — low hydrogen, the structural electrode. Beautiful bead, easy slag, but it is hygroscopic: out of a sealed can it picks up moisture in hours and must be kept in a rod oven at 250°F. A 7018 left on the bench overnight is a hydrogen-cracking problem, and this is the single most common shop failure with it. Drag technique, tight arc, no whipping.

E6013 — easy, shallow, for thin material and for teaching. What a beginner should start on.

Amperage roughly = electrode diameter in thousandths. A 1/8″ (0.125″) rod runs around 90–130 A. Start in the middle of the range printed on the box and adjust by sound.

Arc length equals the electrode core diameter. Too long and you get spatter, undercut and porosity; too short and it sticks.

FCAW. Flux-cored wire. Self-shielded (gasless, E71T-11) runs outdoors and in wind and is DCEN — opposite polarity to MIG, so the machine must be re-wired at the terminals. Dual-shield (E71T-1) uses gas as well, runs hot and fast and is what structural shops use on heavy plate.

Chip and wire-brush every pass in both. Slag trapped between passes is the classic inclusion, and it is what shows up when you section the coupon.

J — Oxy-fuel: cutting, heating, welding

The most versatile and the most dangerous thing in the shop. It cuts, heats, brazes, welds and bends steel, and it is two tanks of compressed gas.

Setup, in order. Chain the cylinders. Crack each valve briefly to blow out dust, pointed away. Regulators on, adjusting screws backed out before opening the cylinder valve. Open oxygen fully (it is a double-seated valve and must be all the way open); open acetylene only 1/2 to 1 turn, so it can be shut fast, and leave the key on the valve.

Never exceed 15 psi on acetylene. Above that it becomes unstable and can decompose explosively in the hose, with no oxygen involved at all. This is the hard number in the whole process.

Flashback arrestors and check valves at the torch and at the regulator. Not optional.

Lighting and the three flames. Acetylene first with a striker, never a lighter. Open it until the smoke stops. Then add oxygen. Carburising (excess acetylene) has a feathery white inner cone — it adds carbon, used for hardfacing. Neutral is when that feather just disappears into a sharp blue inner cone, about 5,600°F — this is what you use for almost everything. Oxidising (excess oxygen) has a shorter, sharper, hissing cone — it burns the steel and is used only for brass.

How cutting actually works — and this is worth knowing because it explains the limits. The preheat flames bring the steel to about 1,600°F, the bright orange kindling temperature. Then the cutting oxygen lever sends a jet of pure oxygen that burns the iron into iron oxide, and the reaction is exothermic so it sustains itself. The jet blows the molten oxide out as slag. You are not melting the steel; you are oxidising it.

Which is why it only works on carbon steel. Stainless and aluminium form refractory oxides that will not blow out. This is the answer to the question every student asks.

Cutting technique: preheat the edge to bright orange, squeeze the lever, move only when the sparks blow through to the underside. Too fast and the kerf trails and does not cut through; too slow and it widens and gouges. Drag tip resting on the plate for straight cuts — steadier than free-handing.

Heating and bending. A rosebud tip heats a broad area. Heat a narrow band across a bar to dull red and quench one side, and the bar pulls toward the heated side as it cools — flame straightening. This is how distortion from welding gets pulled back out, and it is a genuinely skilled operation worth practising on scrap.

Shutdown: oxygen off at the torch first, then acetylene — then close both cylinder valves, bleed both lines, back the adjusting screws out. Always the same order.

Never oil or grease on any oxygen fitting. Oxygen and hydrocarbon under pressure ignite spontaneously. Not a slow hazard — immediate.

K — Plasma cutting

What it is: compressed air is forced through a small orifice and an arc ionises it into plasma at roughly 25,000°F. It melts the metal and the air jet blows it away. It is melting, not burning — so unlike oxy-fuel it cuts any conductive metal: stainless, aluminium, copper, thin sheet.

Clean, dry air is everything. Moisture and oil in the line destroy consumables within minutes and give a ragged, double-arcing cut. A dedicated dryer or at minimum a filter at the machine, drained every session. If cut quality has suddenly gone bad, check the air before anything else.

Air pressure to the machine’s spec, usually 60–90 psi, measured while the torch is firing, not at rest. A regulator that reads fine at rest can sag under flow.

Amperage by thickness, from the machine chart. The published ‘severance’ thickness is the ugly maximum; the quality cut thickness is about half of it and that is the number to design around.

Travel speed. Fast enough that the sparks blow out the underside at roughly a 15–20° trailing angle. Sparks blowing back up at you means you are too fast or underpowered and not cutting through.

Drag tips let the nozzle rest on the work for thin material. Standoff tips need a consistent 1/8″ — use a standoff guide or a roller, because holding it by hand gives a wandering cut.

Pierce at an angle, then rotate upright once through, so the blowback goes sideways and not straight into the nozzle. Piercing flat is what kills tips.

Consumables are a set. Electrode, swirl ring, nozzle, shield — change them together. A worn electrode (a pit deeper than about 0.040″ in the hafnium insert) with a new nozzle will destroy the new nozzle.

Template cutting: a metal template with a standoff roller running against it. Non-conductive templates burn. Thin sheet cuts far better clamped down on a sacrificial slat table than floating.

Fume. Plasma on galvanised or painted steel produces genuinely dangerous fume — zinc oxide gives metal fume fever. Grind the coating off the cut line or use forced extraction. Never plasma-cut a drum or vessel that held anything.

L — Brazing and soldering

The definition, which is the part people get wrong: in welding, the base metal melts. In brazing and soldering the base metal never melts — a filler with a lower melting point is drawn into a close-fitting joint by capillary action. The line between them is 840°F: above it is brazing, below it is soldering. That is the whole distinction.

Why it is worth having: it joins dissimilar metals, it joins thin material without burning through, it joins castings that would crack if welded, and it distorts far less. For sculpture it also gives you a visible brass line as a deliberate mark.

Joint gap is the critical variable. Capillary action needs 0.002–0.005″ for a proper brazed joint. Too tight and the filler cannot enter; too loose and there is no capillary draw and the joint is just a weak fillet. A brazed joint with the right gap is stronger than the filler alone because the thin layer is mechanically constrained.

Clean and flux. Bright metal, then flux, every time — flux dissolves oxide and lets the filler wet the surface. Black flux for high-temp silver brazing on stainless, white for general silver, borax-based for brass on steel.

Heat the work, not the filler. This is the technique error that defines a bad braze. Bring the joint up to temperature with the flame moving, and touch the rod to the joint — if it melts on contact with the metal and runs into the gap, the temperature is right. If you have to melt the rod in the flame and drip it on, the joint is cold and it will not have penetrated.

Fillers. Silver braze (BAg series, 1,145–1,400°F) — strongest, flows into the finest gap, best for stainless and for bandsaw blades. Brazing brass / bronze (RBCuZn, ~1,600°F) — the general steel and cast-iron filler, done with oxy-fuel. Aluminium braze exists and is difficult: the working range sits close to the aluminium’s own melting point, so the part collapses if you overshoot by very little.

Soldering. Soft solder for copper and electrical: lead-free 95/5 tin-antimony for plumbing, 60/40 tin-lead for electronics (and it is legal for that, just not for potable water). Rosin flux for electronics; acid flux for plumbing and sheet metal, and it must be washed off afterwards or it corrodes.

Silver soldering in jewellery is really brazing — hard, medium and easy grades melt at progressively lower temperatures so you can make a second joint without the first one flowing. Plan the order of joints before lighting the torch.

Never braze or solder galvanised without removing the zinc first. Same fume hazard as plasma cutting it.

B — Grit progression

Each grit exists to remove the scratches of the previous grit. Nothing else.

Never skip more than one grit. 80 → 120 → 180 → 220 → 320. Jumping 80 → 220 means you sand for an hour and the 80 grit scratches are still there under the finish.

Change direction 90° with each grit. When the previous grit’s scratches are gone, that grit is done.

Raking light is the only honest test. Hold a light nearly parallel to the surface. Scratches you cannot see under a lamp will appear under finish.

On wood, stop at 180–220 for oil and film finishes. Beyond 320 you burnish and the finish stops penetrating.

Keep a witness strip: mask 1″ of the board at each stage so the finished board shows the whole ladder side by side.

D — HVLP spraying

HVLP means high volume, low pressure — under 10 psi at the air cap. That is what gives the high transfer efficiency; a conventional gun at 40 psi wastes most of the material.

Tip size: 1.3–1.4 mm for lacquer and thin finishes, 1.7–1.8 mm for primers and heavy-bodied coatings.

Thin to the manufacturer’s spec, or test with a viscosity cup. Too thick gives orange peel; too thin gives runs and poor build.

Distance 6–8″, gun kept perpendicular and moving parallel — arcing your wrist gives a heavy centre and thin edges.

50% overlap on every pass. Start the pass before the panel and finish after it.

Faults: orange peel = too thick, too far, or too little air. Runs = too close, too slow, too much material. Dry spray = too far, too much air, or too fast a solvent.

Clean the gun immediately. A gun left ten minutes with lacquer in it is a two-hour job.

F — Hot work: the forge

Forging is not a separate craft to be intimidated by — it is another way to move metal, and for sculpture it gives you shapes no machine will make.

Heat colours, and they matter absolutely. Dull red ~1,200°F · cherry red ~1,400 · orange ~1,700 · yellow ~1,900 · white ~2,200. Forge mild steel at orange to yellow. Below red it work-hardens and cracks. Above white it burns — the sparks come off and the steel is destroyed, not recoverable.

Work in a darkened area or you cannot judge the colour. This is a real reason forges are dim.

The four operations. Drawing out — making it longer and thinner; work over the far edge of the anvil or with the peen, then planish flat. Upsetting — making it shorter and thicker by striking the end; heat only the section you want to upset or it buckles somewhere else. Bending — over the horn, in the hardy hole, or in a bending fork. Punching and drifting — punch most of the way from one face, flip, punch through over the pritchel hole, then drift to size and shape. Punching makes a stronger hole than drilling because the grain flows around it instead of being cut.

Hit while it is hot and stop when it darkens. A beginner’s instinct is to keep hitting a cooling bar. Back in the fire. Many short heats beat one long one, because a long soak scales the surface badly and can burn the edges.

Scale is the flake of oxide that forms in the fire. Brush it off before striking or you hammer it into the surface and it stays there as a pit.

Forge welding — the hard one, and it is fine to fail at it this month. Both pieces to bright yellow-white (~2,300°F), borax flux to keep oxygen off the mating faces, scarf both ends so the faces meet in the middle first and squeeze air outward, then set the weld with several fast light blows before heavy ones. It fails from: not hot enough, oxide in the joint, or hesitating between fire and anvil.

Heat treatment, since the forge is right there. Annealing — heat to non-magnetic and cool as slowly as possible (in vermiculite or ash) for maximum softness. Normalising — heat to non-magnetic and air cool; refines the grain and relieves stress, and you should do it two or three times after forging any tool. Hardening — heat to non-magnetic (about 1,420°F for plain carbon steel; the Curie point is a free thermometer, test it with a magnet), then quench — oil for O1 and most tool steels, water for plain high-carbon, air for A2. Tempering — immediately after hardening, reheat to draw back brittleness: light straw ~420°F for a scraper or a knife, brown ~480°F for a chisel, purple ~540°F for a punch, blue ~570°F for a spring. Polish the surface first or you cannot see the colour run.

A hardened tool not tempered will shatter, and it will shatter at the worst moment. Temper within the hour.

Tongs must fit the stock. Loose tongs throw hot steel. Make or dress a pair to fit before forging anything small.

E — Cold forming: bending, rolling, the press brake

What happens in a bend. The outside goes into tension, the inside into compression, and somewhere between them is the neutral axis, which does not change length. That line is not at the centre — it sits at roughly 0.4 × thickness from the inside face for typical bends. Everything about bend allowance follows from that.

Bend allowance (the length of material consumed in the bend) = angle(rad) × (inside radius + K × thickness), with K about 0.33 for tight bends and 0.45 for generous ones. Cut your flat blank to finished length minus the allowance. Guessing this is why brackets come out wrong and everyone blames the machine.

Minimum bend radius. Mild steel will take an inside radius of about 1× thickness cold; aluminium depends entirely on alloy and temper — 6061-T6 cracks at anything under about 3–4× thickness and 5052 will go to 1×. If a bend is cracking on the outside, the radius is too tight for the material, not a machine problem.

Bend across the grain. Rolled sheet has a grain direction. Bending along it splits; across it holds. On a large sheet this decides how you lay out the blank.

Springback is always there. Mild steel about 1–3°, stainless and high-strength aluminium considerably more. Overbend by the measured amount — bend a test piece in the same material and thickness, measure what you actually got, and set the machine to overshoot by the difference. Every setup gets a test piece.

Air bending vs bottoming. Air bending (the punch does not force the material fully into the die) gives any angle from one die set but more springback variation. Bottoming gives repeatability and needs far more tonnage. Know which your brake is doing.

Tube and pipe bending: a rotary draw bender with the right die for the tube OD and the radius, a mandrel inside, and a wiper die behind. Without a mandrel, thin-wall tube collapses on the inside of the bend and ovalises. Minimum centreline radius of about 3× the tube diameter without a mandrel; tighter than that needs one. Filling with sand and capping the ends is the low-tech mandrel and it works.

Slip rolls. Three rolls; the adjustment is the position of the third. Multiple light passes, not one heavy one — creep the radius down and check against a template. The ends will always have a flat where the material was not yet gripped by all three rolls: either pre-bend the ends in a press, or roll it long and cut the flats off.

Shrinker / stretcher for curving flat bar on edge and for shaping sheet, and it is the tool nobody in an art shop knows about. Shrinking gathers material and curves the bar toward the shrunk edge; stretching does the opposite.

F — Knowing the metal you have

A rack of unlabelled steel is a real problem: it decides whether something welds, hardens, machines or rusts, and a wrong guess wastes a week.

The four-digit system. 10xx is plain carbon steel and the last two digits are the carbon content in hundredths of a percent. 1018 = 0.18% carbon, the general mild steel bar and the one you weld and machine freely. 1045 = 0.45%, medium carbon, stronger, will harden, needs preheat to weld without cracking. 1095 = 0.95%, spring and blade steel. 41xx adds chromium and molybdenum — 4130 chromoly is the tube-frame steel, strong, weldable with care. 12L14 is leaded free-machining steel: it machines beautifully and welds terribly and it is what a lot of cheap bar stock turns out to be.

A36 is not a chemistry spec but a structural strength spec (36 ksi yield) — it is what hot-rolled structural plate and angle is, and its composition varies between mills.

Hot-rolled vs cold-rolled. HR has black mill scale, rounded edges, loose tolerance and internal stress — it moves when you machine one side of it. CR is bright, accurate, stronger at the surface, and also stressed, so it will bow when you cut it unevenly. Neither is stable; if a part must be accurate, rough it, let it sit, then finish it.

Stainless. 304 is the general austenitic one: non-magnetic, excellent corrosion resistance, work-hardens fiercely, weld with 308L. 316 adds molybdenum for salt and chemical resistance. 303 is the free-machining version and should not be welded. 400 series (410, 416, 430) is magnetic, hardenable, less corrosion resistant — and the fact that it is magnetic is how people wrongly conclude a stainless part is mild steel.

Aluminium. 6061-T6 — the structural, weldable, machinable standard, but it loses roughly half its strength in the heat-affected zone when welded, which matters for anything load-bearing. 5052 — sheet alloy, bends tightly, does not machine well. 2024 and 7075 — aerospace strength, not weldable, and 2024 corrodes. Cast tooling plate (MIC-6) — flat and stable, the right choice for a jig base.

The spark test is fast, free and surprisingly reliable once calibrated. Grind a known sample first. Mild steel: long straw-yellow streaks, few forks near the end. High carbon: shorter, whiter, bursting into many repeated star-like forks along the whole streak — more carbon, more bursts. Cast iron: short, dull red, breaking into small fine sprigs close to the wheel. Stainless (300 series): short, dull orange-red, almost no forking. Nickel: short orange streaks with no bursts at all.

The file test for hardness: a sharp file skates on hardened steel, bites on annealed. Backed up with a magnet (300-series stainless is non-magnetic; 400 series and carbon steel are magnetic) it separates most of a rack in ten minutes.

Then label the rack. Paint-marker the end of every bar. It is an hour’s work and it is the single most useful thing you can do for a shop you run.

A — What you actually do not know yet

You have poured bronze and aluminium. So this module is not about learning to pour — it is about learning the controls, so that when a casting fails you can name the cause instead of guessing, and so that you can run a shell room rather than assist in one.

The four things that decide whether a shell casting succeeds, in order of how often they are the culprit: the shell (slurry condition and drying), the dewax (thermal shock), the gating (whether metal reaches everywhere before it freezes), and the pour itself (temperature and speed). Most people only think about the last one.

Ceramic shell versus block investment, honestly. Shell is thinner, lighter (one comparative study measured a shell mould at 2.5 lb against 120 lb for the equivalent investment block), burns out in under an hour rather than three days, reproduces detail better, and leaves less to chase. It is also more fragile before firing and unforgiving about drying.

Read the three free defect atlases before you start, not after you fail. The Investment Casting Institute publishes an Atlas of Shell Defects, an Atlas of Casting Defects and an Atlas of Wax Pattern Defects as free PDFs. Each entry is a photograph, the appearance, the probable causes and the cure. They are the single most useful documents in this module.

Buy one book: Ron Young, Methods for Modern Sculptors. It is the ceramic-shell-for-sculptors text — slurry formulation and handling, waxes, sprues, dewax, melting, pouring, chasing, and building your own kiln and furnace.

B — The wax, and why the wax decides everything

Every defect in the wax is faithfully reproduced in bronze, then costs ten times as long to fix in metal. An hour spent perfecting wax saves a day of chasing.

Wax types. Microcrystalline for modelling and building up; sheet wax for surfaces and shells; sticky wax and a hot tool for joining; water-soluble waxes for cores. Foundry waxes such as the SculptaCast range are formulated for low ash and low expansion — ordinary candle wax leaves residue and will ruin a burnout.

Surface finish. Work the wax to the finish you want in metal. A solvent wipe (mineral spirits on a soft brush, very lightly) or a low flame passed quickly will level tool marks. Fingerprints in wax become fingerprints in bronze.

Wall thickness for hollow work: 3/16″ to 1/4″ is the normal range for bronze. Thinner risks misrun; much thicker wastes metal, adds shrinkage porosity and adds weight.

Check wall thickness everywhere before spruing — a needle probe or a caliper through a small hole. Thin spots are where you get non-fill.

Chase every seam, bubble and flaw now. The ICI Atlas of Wax Pattern Defects catalogues what to look for: non-fill, distortion, flow and knit lines, graining, flash, sink and cavitation, cracking, air bubbles.

Store wax flat and cool. A wax left leaning against a wall in a warm shop distorts, and you will not notice until the bronze comes out wrong.

C — Gating and spruing for art bronze

The gating system exists to get metal to every part of the cavity before any of it freezes, without turbulence, and then to feed the shrinkage as it solidifies. Those are two different jobs and they need two different features.

Turbulence is the enemy, not speed. Molten bronze that tumbles entrains air and oxide skin, and that oxide ends up as an inclusion inside the casting. The critical (non-turbulent) flow velocity is around 500 mm/s for copper alloys. Above it you are folding the surface film into the metal.

Taper the sprue downward. Metal accelerates as it falls, so a parallel sprue lets the stream separate from the wall and aspirate air. A sprue tapered narrower at the bottom keeps it full. This is straight Bernoulli and it is the most commonly ignored rule in art foundries.

Gate into thick sections, not thin ones, and gate at the bottom or the side wherever possible rather than dropping metal from the top.

Feeders (risers) feed shrinkage. Bronze shrinks roughly 1.5–2% on solidification. The feeder must freeze after the section it feeds, which means it must be bulkier. The modulus rule: casting : neck : feeder = 1 : 1.1 : 1.2, where modulus is volume divided by cooling surface area. In practice: a feeder must be fatter than what it feeds.

Gate thickness roughly half the local wall thickness at a hot spot, so the gate freezes first and does not itself become a shrinkage path.

Vents from every high point and every trapped pocket. Air has to leave or metal cannot enter. On a figure, vent the head, the raised hand, the top of every undercut.

Draw the whole system on paper before you touch the wax, and keep the drawing. When the casting comes out, cut one open across the heaviest section and check whether the feeder actually fed it. That one sectioned casting will teach you more than five successful pours.

Free reference with the actual formulas: A.K.M.B. Rashid, Design of Gating and Feeding Systems — choke area, sprue taper, feeder modulus, critical velocities, gating ratios. It is a university course PDF and it is free.

D — The slurry, and the numbers that control it

This is the part that separates someone who dips from someone who runs the room. A slurry is a live system and it drifts.

What it is. A colloidal silica binder carrying refractory flour in suspension, kept moving by a mixer so it does not settle. Primary slurry (fine, zircon or fine fused silica) captures detail. Backup slurry (coarser) builds strength.

The controls, with targets: binder solids 25.5–27.5% by specific gravity · slurry temperature within ±4°F of ambient and never above 76°F · viscosity measured with a Zahn cup and logged every session · pH · refractory solids (around 64% is a typical target).

Viscosity is the one you will actually use daily. Too thin and the coat is too light and you get shell that cracks; too thick and it bridges across detail and traps air. Measure it with a Zahn cup and a stopwatch, write the number and the date on a board by the tank, and adjust with binder (thinner) or flour (thicker) in small increments.

Wetting agent for the primary coat, so the slurry wets the wax rather than beading off it. Wax is hydrophobic; without a wetting agent or a pre-etch you get skip and bubbles all over the first coat.

Antifoam — entrained air in the slurry becomes bubbles on the pattern becomes positive bumps on the casting.

Biocide. A slurry tank is a warm wet organic-friendly environment. Bacterial contamination degrades the binder and it smells. Grotan or equivalent, dosed on a schedule.

Keep the mixer running continuously, even overnight. A slurry allowed to settle may not fully recover.

Free reference: Berta & Duncan (Ransom & Randolph), Slurry Analysis & Shell Material Testing Methods, hosted free by the Investment Casting Institute — every test above with the method and the target range.

E — The dip schedule

Typical schedule: 8 to 12 coats over about a week, 6–12 hours drying between coats, 2–3 stucco grits, finishing at roughly 3/8″ total shell thickness. Those are the numbers to start from and then adjust to your own climate and slurry.

Primary coat. Dip slowly, rotate to wet everything, drain deliberately, and watch for bubbles — break any you see with a brush or a puff of air immediately. Then stucco with the finest sand, by rainfall or fluidised bed, not by throwing it. The primary coat is the casting surface; everything about detail is decided here.

Backup coats. Progressively coarser stucco. The coarser grits build thickness fast and mechanically key the layers together.

Drying is the most misunderstood step. You are removing water until the binder gels. Roughly 70% of the water must leave a 20%-silica slurry to reach gel point, only about 50% for a 30%-silica slurry. Water removal to strength is non-linear with a plateau — past a point, more drying buys nothing and over-drying causes cracking.

Control humidity and airflow, and keep them the same every time. Too dry or too much air gives drying cracks and rat-tailing; too damp and the shell never gels and delaminates in the autoclave. A fan on a timer in a humidity-controlled corner is worth building.

Reinforce where you must: stainless wire or ceramic fibre in the slurry for long thin sections and around the cup.

The cup. Build a proper pouring cup with a flat rim and enough capacity to stay full through the pour. A cup that empties mid-pour aspirates air into the sprue.

Notch the shell before burnout where you intend to break it off later, if the form allows.

Log every dip: date, time, coat number, slurry viscosity, stucco grit, room temperature and humidity. When a shell fails you will have the data to say why. This log is also the evidence that you can run a facility.

Free reference: ICI Atlas of Shell Defects — organised chronologically through exactly these steps, with probable cause and remedy for bubble entrapment, bridging, metal penetration, shell cracking and finning, drying cracks, rat-tailing, prime coat lift, spalling, delamination and bulging.

F — Dewax and burnout

Two separate operations and people conflate them. Dewax removes the bulk of the wax fast, before it can expand and crack the shell. Burnout fires the shell and destroys the residue.

The thermal-shock problem. Wax expands more than ceramic when heated. If the shell warms slowly, the wax at the shell wall expands against a cold rigid shell and splits it. The whole design of dewaxing is about melting the wax at the interface before the bulk heats up.

Steam autoclave: about 320°F steam, and pressurisation must reach roughly 90 psi in under 5 seconds. A slow pressure rise cracks shells. This is why a purpose-built autoclave beats an improvised one.

Flash fire: the shell goes straight into a furnace at 1400–1600°F, cup down. The inner wax melts and runs out almost immediately. Wax is normally out within 10 minutes and burnout complete within 20.

Cup down, always, over a catch tray. Wax must have somewhere to go.

If the pattern is 3D-printed resin rather than wax, everything changes. Resin needs a minimum of 1600°F and often 2000°F+ for complete destruction, and burning resin can pressurise the shell from inside and crack it. Given that you build your own 3D printers, this is directly relevant — plan for it, use a burnout-grade or castable resin, and vent generously.

Burnout fires the shell as well as clearing it. The ceramic needs to reach temperature to develop strength and to drive off chemically bound water.

Preheat before pouring. A common working figure is a kiln at around 1450°F, with the mould poured hot. A cold mould chills the metal at the wall and gives misruns and cold shuts on thin sections; too hot and you get excessive shrinkage and a rough surface.

Inspect every shell after burnout, in good light. Cracks, thin spots, spalling. A shell you are unsure about is repaired with refractory cement or scrapped — a shell that fails during a pour is a serious injury, not a lost casting.

G — The pour

Silicon bronze (Everdur, C87300) is the standard art alloy: roughly 95% copper, 4% silicon, 1% manganese. Melting range 1550–1780°F. Pour temperature 1900–2250°F, and small foundries often cannot reach the top of that range. Aluminium pours far lower, around 1300–1400°F.

Measure the temperature; do not judge by colour. An immersion pyrometer or a reliable optical one. Colour judgement in a bright shop is unreliable by hundreds of degrees.

Overheating is a real fault, not just a waste of gas. It increases gas absorption, oxidation and shrinkage. Get to temperature and pour.

Skim the dross off the melt surface immediately before pouring, and hold the skimmer back as you pour. Dross in the stream becomes an inclusion.

Degas if the alloy and the furnace call for it — a tablet or an argon lance. Dissolved hydrogen comes out of solution as the metal freezes and gives you porosity you cannot chase out.

Pour fast enough to stay ahead of freezing, steady enough not to entrain air, and keep the cup full throughout. A cup that runs empty sucks air down the sprue. One continuous pour, no hesitation.

The crew has roles and they are rehearsed: pourer, skimmer, shanker, shovel and a designated person who touches nothing and watches for problems. Walk the sequence cold before the metal is up.

PPE is not negotiable: full leathers — jacket or cape, chaps, spats over boots — a face shield over safety glasses, and a hard hat with a screen. Cotton or wool under the leathers, never synthetics, which melt onto skin.

Anything wet explodes. Damp tools, damp floor, damp metal charge, a damp shell. Preheat every tool that will touch metal. Never charge metal by dropping it into a crucible of melt.

Afterwards: devest, cut the gates, chase, weld up defects with matching filler (silicon bronze is readily gas-shielded arc weldable, which is why it is the art alloy), then blast or wire-wheel — no solvent cleaners — before patina.

Name the cause of every defect you get and write it beside the photograph. The ICI Atlas of Casting Defects gives you the vocabulary: cold shut, misrun, gas, gas entrapment, slag, hot tear, ceramic inclusion, shell failure, rat-tailing, shrink in the gate area, surface shrinkage.

H — Write the shell room procedure

This is the deliverable that turns the module from a personal project into a professional credential. A department hiring a technician wants to see that you can write the document that lets other people work safely.

Model it on a real one. The University of Lethbridge Fine Arts Metal Casting — Lost Wax Process: Safe Work Practice Manual is public and is structured exactly the way yours should be: studio by studio, operation by operation, with who is permitted to do what.

Sections it needs: wax working (tools, ventilation, burn hazards) · slurry room (skin and respiratory, silica, biocide, mixer guarding, tank access) · dipping and drying (handling, storage, drying room) · dewax and burnout (technician-only, hot surfaces, wax fire procedure) · melt and pour (crew roles, PPE, exclusion zone, what to do if metal spills) · devesting (respirator, silica dust, eye protection) · chasing and finishing (grinding, blasting) · patination (acids, ventilation, neutralisation, waste).

Name who may do each operation. Student, student under supervision, technician only. The Lamar University foundry policy document is a good short model for the pour-day rules specifically.

The silica hazard deserves its own paragraph. Respirable crystalline silica from dry slurry, stucco, and devesting is the long-term health risk in a shell room and it is invisible. Wet methods, HEPA, fitted respirators, and a written exposure-control section.

Put your own measured numbers in it — your slurry targets, your dip schedule, your burnout and preheat temperatures, your log format. A procedure carrying real numbers from a real season is evidence of competence; a generic one is evidence of a download.

Then hand it to someone and watch them use it without you. Fix whatever they got stuck on. That is the test.

A — Blasting

Never silica sand. Free silica causes silicosis; it is banned for abrasive blasting in most jurisdictions and there is no safe way to use it.

Aluminium oxide 60–120 grit: aggressive, sharp, leaves an angular profile for coating adhesion.

Glass bead 60–100: peens rather than cuts. Satin, bright, no profile — wrong under paint, right for a finished surface.

Garnet 30–80: moderate cut, low dust, reusable a few times.

Walnut shell / corn cob: soft media for wood, castings, delicate work.

Pressure: 80–90 psi for AlOx on steel. 40–60 psi for glass bead. 20–40 psi for soft media. Higher pressure buys speed, not quality, and will warp sheet.

Standoff 6–10″, nozzle at 45–60°, never square to the surface — square blasting rebounds media into the stream and cuts nothing.

PPE: supplied-air or a proper blast hood, not a dust mask. Hearing protection.

For your matrix, stamp each coupon with media and pressure before blasting.

C — Patina

A patina is controlled corrosion. Heat, concentration and dwell time are the three variables.

Ferric nitrate — browns and reds on bronze. Apply in light coats to metal heated to about 200–250°F. It builds; do not flood it.

Liver of sulfur (potassium sulfide) — browns to black on copper alloys and silver. Cold, dilute, repeated. Strong solutions give a sooty layer that rubs off.

Cupric nitrate — blues and greens. Heat applied, built in layers.

Surface prep decides everything: blast or abrade to a uniform surface, degrease completely, and do not touch it with bare hands after.

Seal or it keeps moving. Renaissance wax for indoor and reversible; Incralac for outdoor bronze; lacquer for a hard barrier.

For the ladder: same metal, same prep, one chemistry, five dilutions, half heated and half cold, half sealed and half not. Date each. The unsealed ones will tell you in three months what the sealed ones prevented.

D — Concrete

Water-cement ratio 0.35–0.45 by weight. Lower is stronger. Use a superplasticiser for workability, not more water — every extra litre of water is lost strength.

Mix design for a cast element: 1 part cement : 2 parts sharp sand : 2 parts 3/8″ aggregate, by weight, plus AR glass fibre for crack control in thin sections.

Vibrate to release entrapped air. A vibrating table, or an orbital sander held against the form. Do not over-vibrate — aggregate segregates and the surface goes to laitance.

Cure damp for at least 7 days. Concrete does not dry, it hydrates. A piece left to dry in the air reaches a fraction of its potential strength.

Reinforce anything spanning. Rebar or fibre, with at least 3/4″ of cover to any face.

Demould at 24–48 h, keep wrapped and damp for the rest of the week.

A — Diagnostic sequence, any machine

Reproduce the fault before you open anything. Most 'broken' machines are set up wrong.

Then work outward from the cut: workpiece → cutter → workholding → setup → alignment → drive → bearings → structure.

Isolate one variable at a time. Change one thing, re-cut, measure. Changing two things teaches you nothing.

Sound, heat and vibration are your three instruments. A bearing tells you it is failing an hour before it tells anyone else — run the machine with the cutter off and listen.

Write the fault, the tests, and the fix in the log. In August this log becomes the maintenance schedule.

B — Bearings and belts

Identify by the number on the race, not by measuring. A 6203-2RS is a standard part and costs a few dollars.

Never hammer on the outer race to drive a bearing onto a shaft, or you brinell the balls and it fails in weeks. Press on the race that is being fitted — inner race for a shaft, outer race for a housing.

Use a puller, not a screwdriver and a hammer. Heat the housing or freeze the shaft if it is tight.

Seals face out. Shielded bearings can be repacked; sealed ones cannot.

Belts: match by the code on the belt. Check pulley alignment with a straightedge across both pulley faces — misalignment eats belts and bearings both.

Tension to about 1/64″ deflection per inch of span. Re-tension after the first hour of running.

C — Electrical

Read the nameplate first: voltage, phase, full load amps, RPM, frame size, service factor. Everything else follows from it.

Single phase motors need a start capacitor and often a centrifugal switch — a motor that hums but will not start is almost always one of those two, not the motor.

Three phase motors are simpler, more reliable and reverse by swapping any two leads. A VFD lets you run a three-phase motor from single-phase supply and gives variable speed — size it by motor amps, not horsepower, and set the motor nameplate values in the parameters before first start.

Magnetic starters drop out on power loss, so the machine does not restart by itself when power returns. Any machine a person stands at should have one.

Cord repair: correct gauge for the amps and the run length, strain relief at both ends, ground bonded to the frame. Test continuity and ground before plugging in.

If you are not certain, stop. Electrical work in a university shop has rules, and doing it yourself without authorisation is a different kind of problem from doing it badly.

D — Pneumatics

Moisture is the enemy. Drain the tank every session; an undrained tank rusts from the inside and eventually bursts.

Filter-regulator-lubricator in that order, at the drop, not at the compressor.

Air tools want 90 psi at the tool, which means more at the regulator once you account for hose loss. Undersized hose is why your tool feels weak.

Rebuilding a tool: kits are cheap. Lay the parts out in order as they come off, photograph each stage, and replace every o-ring in the kit rather than only the failed one.

Compressor maintenance: oil level, intake filter, belt, safety valve tested by pulling the ring.

E — The maintenance schedule

One row per machine per interval. Columns: machine, task, interval, tolerance, who, last done.

Daily chips, guards, blade condition, dust bin. Weekly belt tension, table wax, fence and stops, blow out motors. Monthly alignment checks, knife setting, tracking, lubrication, dust gates. Quarterly bearings, full alignment sweep, tram, level. Annual electrical inspection, motor service, filters, document review.

Use the numbers you measured this year. A schedule carrying your real tolerances is evidence you did the work; a generic one is evidence you downloaded something.

G — Dust collection

Required airflow: table saw 350–400 CFM · 6″ jointer 350 · 12″ planer 400 · bandsaw 350 · drum or wide-belt sander 550–800 · router table 300.

Branch velocity must stay near 4000 FPM or wood dust drops out and the duct blocks.

CFM = velocity × duct area. A 4″ duct is 0.0873 ft² → ~350 CFM at 4000 FPM. A 6″ duct is 0.196 ft² → ~785 CFM. That is why 4″ is the floor for one machine and useless for a sander.

Smooth-wall duct, not flex — flex has roughly three times the resistance per foot. Keep flex under 5 ft. Long-radius elbows only.

One or two blast gates open at a time, never more.

The health issue is the fraction under 10 microns, which a bag filter passes straight through. A 0.5 micron cartridge or a vented-outside system protects lungs; a chip collector only protects floors.

F — What a checkout must contain

1. Hazards specific to this machine, with the injury each one causes. 2. PPE and where it is kept. 3. Pre-use inspection. 4. Startup and shutdown sequence. 5. Permitted operations and, explicitly, prohibited ones. 6. Material restrictions. 7. The demonstration the student must perform to pass. 8. What to do when it goes wrong — e-stop, who to tell, incident form. 9. Signature and date, retained.

The test of the document is that another technician can run the checkout with it and without you. That is why you hand it to someone and watch.

I — Running a critique

You scored 2.29/4 here, the highest block in the teaching section, so this is about making an instinct explicit enough to hand to someone else — which is exactly what a search committee asks you to demonstrate.

Say the format out loud before it starts. Students perform badly in critique mostly because they do not know what is being asked of them. Naming the structure removes half the anxiety in the room.

The order that works: 1. The maker says nothing at first. 2. Description only — the room says what is physically there, materials, scale, how it is made, where it sits. No judgement yet. This is the step everyone skips and it is the one that makes the rest possible, because it forces looking. 3. Interpretation — what it seems to be doing, what it recalls, how the material reads. 4. The maker speaks — intention, what they were after, what they are stuck on. 5. Evaluation against their own stated intention, not against what you would have made. 6. Concrete next moves — two or three things to try.

Ask questions rather than deliver verdicts. ‘What happens if the base is gone?’ does more than ‘the base is wrong’, because the student has to do the thinking and keeps authorship.

Separate the technical from the conceptual explicitly, and say which you are talking about. ‘This weld will fail’ and ‘this weld reads as careless’ are different sentences with different fixes. Confusing them is the commonest way a critique goes wrong in a sculpture department.

Manage the room. Call on quiet people directly and by name. Cut off the student who talks every time by thanking them and moving on. Do not let the group pile on — when three people say the same criticism, name it as one point and move.

Time it and say the time. Fifteen minutes each, timer visible. Uneven critique time is read as favouritism, accurately.

Close every critique with something specific the maker is going to do next. Ending on open discussion leaves them with nothing to hold.

Write it up as a one-page document with the steps, the timing and three example questions per step. Bring it to an interview. Almost no candidate has one.

J — Mentorship and the studio visit

The studio visit is the core teaching form in a graduate programme and the one you will be asked to demonstrate for any teaching job above adjunct.

Look first, for longer than is comfortable. Walk the space. Look at what is on the walls, what is in the bin, what is half-finished in the corner. The unfinished work and the discarded work tell you more than the finished piece on the stand.

Open with description, again. ‘Here is what I am seeing.’ It proves you looked and it gives the artist something to correct, which is a much easier way into a conversation than ‘tell me about your work’.

Four questions that do real work: ‘What is the one you keep coming back to?’ · ‘What would you do if there were no deadline?’ · ‘What is the part you are avoiding?’ · ‘Which of these would you not show?’ The last one surfaces their own standard, which is what you are actually there to help them see.

Name what is strong before what is weak, and be specific about both. Vague praise is worse than none because it teaches nothing.

Offer references, materially. Two artists, one book, one technical resource, written down before you leave. An unwritten recommendation is forgotten in an hour.

Know the line between mentorship and therapy. Studio conversations go to difficult places. Be warm, be direct about the work, and know your institution’s referral route for a student in real distress — find out what it is at Syracuse before you need it.

Follow up once, in writing, within a week. Three sentences. This is the thing almost nobody does and it is the difference between a visit and a relationship.

L — Tooling and consumables: what to stock and what it costs

This is the block (S3, 1.0/4) that separates a technician from a user. A shop runs on knowing what wears out and having it before it does.

Build a real list, with part numbers, supplier, unit price, and lead time. Columns: item · spec · where it is used · consumption rate · reorder point · supplier · price.

Woodshop consumables: saw blades and sharpening service · jointer and planer knives or inserts · bandsaw blades by width and TPI · router bits · sanding abrasives by grit and format · glue · finish · dust bags and filters.

Metal shop: MIG wire by diameter, contact tips, nozzles, liners · TIG tungsten, collets, cups, gas lenses · grinding, cutting and flap discs · bandsaw blades · drill bits and taps · cutting fluid · gas cylinder rental and fill.

Know the wear items that actually stop work: MIG contact tips, plasma consumable sets, bandsaw blades, sanding belts. Those get stocked deep. Everything else can be ordered.

Carbide vs HSS: carbide costs three to ten times more, lasts ten to fifty times longer in abrasive material, and chips if the setup is not rigid. In a teaching shop where students crash things, HSS on the hand tools and carbide on the fixed machines is usually the right split.

Sharpening is a service and a budget line. Find the local sharpener, get their price list and turnaround, and build it into the schedule so blades go out in rotation rather than when everything is dull at once.

Suppliers worth having accounts with: an industrial supply (McMaster-Carr for speed, MSC or Grainger for accounts), a metal supplier with a will-call counter, a hardwood dealer, an abrasives supplier, and your gas supplier. Know which of these can deliver tomorrow — that is what makes you useful in a crisis.

Write down the price of an hour of the shop: consumables plus power plus your time. Departments ask for this and almost no technician can produce it.

M — The professional apparatus

P3 scored 1.44/4. These are the documents that get you shortlisted, and they are all writeable in a week.

Shop budget. Annual, split: consumables · tooling replacement · machine maintenance and repair · capital requests · safety and PPE. Each line with last year’s actual and next year’s ask, and a one-sentence justification. A capital request needs: what it is, what it costs installed, what it lets students do that they cannot do now, what it replaces, and its service life.

Purchasing. Learn your institution’s actual process — requisition, PO, P-card limits, sole-source justification, and who signs. At Syracuse, find this out from the department administrator and write it down. Knowing it is a large part of the job.

Teaching materials to have ready: syllabus · one full assignment sheet with a rubric · three machine checkouts · a demo plan · a safety orientation · a one-page critique format.

The job application set: CV · artist statement · teaching philosophy · diversity / inclusive-teaching statement (asked for by most US institutions now and the one people leave to the last day) · 20 images with a numbered image list giving title, date, medium, dimensions · student work samples for a teaching job · three references who have actually agreed.

Teaching philosophy, concretely. Two pages maximum. What you think a shop education is for · how you teach a skill (demo, supervised repetition, independent application, is the honest answer) · how you handle a mixed-ability room · how you assess · and one specific example of a class that went wrong and what you changed. The specific example is what makes it readable.

Grants and residencies. A project budget that adds up, a timeline, a work sample, and a project description that states what will exist at the end. Keep a folder with the current version of all of it so an application is an afternoon and not a fortnight.

Documentation is professional infrastructure, not an afterthought. Every piece photographed properly the week it is finished, with the caption written then, while you still remember the dimensions.

E — CV rewrite

Skills line: remove ‘Foundry’. Add moldmaking, CNC, machining. Describe welding honestly — what this year actually proved, with the positions.

Split Exhibitions into Solo / Two-person / Group.

Cut the 2014 school shows and the 2012–15 animation certificates.

Separate Teaching (three adjunct appointments) from Assistantships.

Add: drawing and painting, ceramics and kiln, kinetics, concrete, machine repair.

Fix ‘Art Manger’ → Manager, and the two spellings of Mouradian/Moradian between CV and thesis.

F — Thesis bibliography

Add the five in-text citations with no bibliography entry: Greenwald, Deleuze, Howard & Hussain, Morozov, Benjamin.

Remove or actually cite Flusser, Debord, Rosa, Williams — currently listed and never mentioned.

Fix the Deleuze misquote: disciplinary societies, not ‘dispersed’.

Attribute ‘hyperobject’ to Timothy Morton.

Reconcile the two Heidegger editions — footnote cites Lovitt 1977, bibliography lists Krell 1993.

Page number on the Schiffman quotation.

Add a list of works with medium and dimensions for all eleven figures.

A — The brief

One piece that uses every process on this plan. Your idea, and it is the right one — a year of separate objects reads as coursework; one object that could only have been made by someone who can do all of it reads as a body of work resolved.

The proposal: a freestanding inhabitable element. Something at the scale of a reading station, a threshold, a bench with a shelter, a gate with a seat — architectural in reference, sculptural in intent, and demanding every skill at once. You name it and you set the content; what follows is the technical spine it has to carry.

It must contain, and the checkpoint is that it does:

— Welded steel structure, load path calculated on paper before cutting, distortion controlled, out of wind.

— At least one TIG-welded element where the weld is visible and not ground away.

— A forged component you made hot — a bracket, a handle, a latch, a foot — with at least one punched rather than drilled hole.

— A cast element in bronze or aluminium, ceramic shell, poured by you, chased and patinated.

— Solid wood joinery carrying real load, cut by hand, with the seasonal movement calculated and allowed for.

— A bent or turned wood element, springback measured.

— A concrete or stone footing, mix designed rather than bagged.

— Every surface deliberately finished — patina, paint, wood finish — each one chosen and each one sealed.

— Rigged, moved and installed by you, with the loads calculated first and a written rigging plan.

— Documented to publication standard, with the drawings, the calculations and the failures included.

Build it so it can be taken apart. Knockdown joinery is not a compromise, it is what lets the piece be shipped, shown, stored and reinstalled — which is the difference between a sculpture that has a life and one that lives in your studio. Schwarz's Campaign Furniture, free, is the best single source on demountable joinery and let-in hardware.

B — Joining materials that behave differently

This is the engineering problem the capstone exists to teach, and it is where most mixed-material sculpture quietly fails over a few years.

Wood moves, metal does not. A solid wood member 18″ wide will change about 3/16″ across a seasonal cycle. Bolt it rigidly to a steel frame across that width and it will split — not maybe, will. Fix solidly at one point only, and use slotted holes running across the grain everywhere else, with washers large enough to bridge the slots. Calculate the movement first and slot for at least that much, in both directions from the fixed point.

Metals move too, just predictably. Coefficients of thermal expansion, per °F: steel about 6.5 × 10⁻⁶, stainless about 9.6, aluminium about 13, bronze about 10. An outdoor aluminium member 8 ft long moves roughly 1/8″ across a 100°F swing. Bolt aluminium rigidly between steel and something yields.

Galvanic corrosion. Two dissimilar metals in electrical contact with any moisture form a cell and the less noble one corrodes. The gap in the galvanic series is what matters. Aluminium against stainless or bronze is a bad pair outdoors; steel against bronze is also a pair. Isolate with a nylon or phenolic washer and sleeve, a gasket, or paint film. Stainless fasteners into aluminium need isolation, and this is the single most common outdoor sculpture failure.

Wood against steel. Oak, chestnut and cedar are acidic and will corrode steel and stain themselves blue-black around the fastener. Isolate with tape, sealant or paint, and use stainless or hot-dip galvanised fasteners, not plain steel.

Concrete against metal. Fresh concrete is strongly alkaline and attacks aluminium and zinc. Coat aluminium that will be cast into or bear on concrete.

Water must be able to leave. Every horizontal surface sheds or drains; every tube is either sealed completely or drilled to drain at its lowest point. A sealed tube that is not actually sealed fills with water and bursts in the first freeze. Drill the drain hole.

C — Sizing it on paper before you cut

Know the weight before you build. Volume × density. Steel 490 lb/ft³ · aluminium 169 · bronze about 550 · concrete 145 · hardwood 45–55. Add it up, member by member, and write the total on the drawing.

Depth beats width, cubed. For a rectangular section, I = bd³/12. Doubling the width doubles the stiffness; doubling the depth multiplies it by eight. If something sags, it is almost always too shallow rather than too thin.

Deflection scales with the span to the fourth power. Doubling a span makes it sixteen times as floppy for the same section. This is why a design that worked at 4 ft fails at 8 ft.

A rectangle racks; a triangle does not. Any four-sided frame needs a diagonal, a moment-resisting corner, or a rigid panel. A bolted rectangle with no diagonal is a mechanism, not a structure.

For anyone seatable, design for a person who is heavier than you and who will sit down hard. Use 300 lb as a static design load for a seat and multiply by a dynamic factor of 2 for someone dropping onto it.

Outdoors, wind is a real load. Pressure in psf is roughly (wind speed in mph)² ÷ 390. At 60 mph that is about 9 psf across the projected face; a 4 ft × 8 ft panel sees roughly 290 lb of push. Size the footing and the anchors against overturning, not just against weight.

Factor of safety. Use 4 on anything structural, and more where a failure would hurt someone.

Graphic statics is the right tool for you. Allen & Zalewski, Form and Forces, teaches you to solve a truss by drawing force polygons with a ruler rather than by algebra — which suits how you already think and is enough for anything at this scale. Gordon's Structures gives you the intuition underneath it.

D — Rigging and installing it

Your audit scored 0.5/4 here, this is the part of sculpture that injures people, and it is exactly what a department needs from a technician on install week.

Working Load Limit is not breaking strength. WLL is breaking strength divided by a design factor — normally 5:1 for rigging hardware, 7:1 for anything suspended over people. Every sling, shackle and hoist ring has a WLL stamped on it. Unmarked hardware has no rating and is not rigging hardware.

Sling angle multiplies the load, and this is the number nobody knows. Two legs at 60° from horizontal each carry 58% of the load; at 45°, 71% each; at 30°, 100% each — each leg carries the whole weight. Below 30° the tension runs away. Never rig below 30°.

The load must hang below its centre of gravity or it rotates the moment it leaves the ground. Lift one inch, stop, look, set it down and re-rig if it is not level. That test lift is the single habit that prevents most rigging accidents.

Shackles take load on the pin broadside and the sling in the body. Never side-load a shackle; never substitute a bolt for a shackle pin; mouse or safety-pin anything left hanging.

Eye bolts. A plain shouldered eye bolt loses about 75% of its rating at a 45° side pull, and an unshouldered one is rated vertical only. Use a swivel hoist ring for anything angled.

Synthetic slings cut instantly on a sharp edge — corner protectors every time. Any sling with a cut, a burn, chemical damage or an unreadable tag is scrap, by rule.

Never use a ratchet strap to lift. It is a tie-down and it is not rated for lifting.

Nobody under a suspended load. Tag lines to control rotation; hands off the load.

Anchoring into a building. Know what the wall is before you drill — stud, solid masonry, hollow CMU, or drywall with nothing behind it. Wedge anchors for solid concrete, sleeve or toggle for hollow CMU, a ledger across studs for anything heavy on a stud wall. Get written permission from facilities before you anchor into any wall in an institution, because this is the argument that follows every install.

Free reference: OSHA's Guidance on Safe Sling Use and its sling tables and figures give rated loads for every sling type at 30°, 45°, 60° and 90°, plus angle-of-choke derating. Print the tables and put them on the wall.

E — Documenting the year

The work is not finished until it is photographed and captioned, and a body of work that is badly documented is invisible to every committee that will ever consider you.

Setup. Grey or white seamless, swept so there is no horizon line. Two lights at 45°, one about a stop brighter than the other, both diffused, plus a reflector on the shadow side.

Camera. On a tripod, at the object's mid-height, f/8–f/11, lowest native ISO, shot raw. Put a colour checker card in one frame per setup and then remove it.

Four frames minimum per object: straight on, three-quarter, a detail of the joint or the surface, and one in context or in use.

Caption every image as you shoot it: title, date, materials, dimensions as height × width × depth, and process. You will not remember the dimensions in March.

Photograph the process, not only the result. The sectioned casting, the failed forge weld, the springback measurement written on the form, the log sheets. For a PhD application and for a teaching job, evidence of method is worth as much as the finished object, and almost nobody has it.

Then shoot all the year's objects together as one image. That photograph is the argument.

Reference for reading a bronze backwards from object to process — useful for writing about your own work with precision: the Getty's Guidelines for the Technical Examination of Bronze Sculpture, 558 pages, free and open access.

A — Weight from volume

Weight = volume × density. Do this on paper before you build, every time.

Densities: steel 0.284 lb/in³ · aluminium 0.098 · bronze 0.32 · brass 0.31 · concrete 0.087 · ash / oak ~0.024 · pine ~0.015.

Example: a 1″ × 2″ × 11 ga steel tube, 8 ft long. Wall 0.120″. Cross-section area ≈ (1×2) − (0.76×1.76) = 2.00 − 1.34 = 0.66 in². Volume = 0.66 × 96 = 63 in³. Weight = 63 × 0.284 = 18 lb.

Do five, then weigh them. Target: within 10%. Once you can do this you will never again be surprised by whether two people can lift something.

B — Why depth beats width

For a rectangular section: stiffness ∝ b·d³ and strength ∝ b·d², where d is the depth in the direction of the load.

A 2×6 is actually 1.5″ × 5.5″. On edge: I = 1.5 × 5.5³ / 12 = 20.8 in⁴. Flat: I = 5.5 × 1.5³ / 12 = 1.55 in⁴.

That is 13× stiffer on edge. Section modulus gives 7.56 vs 2.06 — 3.7× stronger.

Break both and log the actual ratio. The number you measure will stay with you far longer than the formula.

This single relationship is why beams are deep, why a sheet of steel folded into an angle is stiff, and why your cantilevered sculpture needs depth at the wall, not width.

C — Centre of gravity and tipping

Suspend the object from one point and drop a plumb line. Mark the line. Suspend from a second point, drop again. The intersection is the CG.

It tips when the CG passes outside the footprint. That is the whole rule.

For a freestanding piece: base width should put the CG well inside the footprint at the worst expected push. A rule of thumb for public work is that it survives a 50 lb horizontal push at 5 ft without moving.

Ballast low is worth ten times ballast high.

D — Case to stand, and where the wood moves

The case is wood and moves; the stand is steel and does not. Bolting a 30″ solid case rigidly to a rigid steel frame across its width will split the case within a year.

Fix solidly at the front centre. Everywhere else, use slotted holes running across the grain so the case can grow and shrink over the steel.

Threaded inserts in the case bottom, bolts up through the stand’s top rails, with washers large enough to bridge the slots.

Isolate wood from steel with tape, sealant or paint film — ash and oak are acidic and will stain and corrode.

Calculate the movement first: an 18″ depth in a domestic hardwood will move roughly 3/16″ across a full seasonal cycle. Slot for at least that, in both directions from centre.