233 Lancaster Ave · Kitchen nook
Two runs, four pine treads, eight legs. About four hours. Work top to bottom; nothing here needs a decision.
Everything is H − 1″
H is your stool’s seat height. The treads are 1″ thick, so the top of the ledger and every leg are H − 1″. At a 17¼″ stool that is 16¼″. Get this wrong and all eight legs are wrong.
Tape and pencil, before any saw
H — top of a stool. Write it on the wall.
West run — cabinet face to the north wall. Should read 78½″. You will cut 78¼″.
Snap the ledger line at H − 1″, both walls, level along the whole length.
Mark every stud, both walls, ABOVE the line so the ledger does not cover your marks.
| Cut | To | |
|---|---|---|
| Tread 1 | leave at 48″ | north seat, no cut |
| Tread 2 | 39⅛″ | west piece A |
| Tread 3 | 39⅛″ | west piece B |
| 2×4 | 78¼″ | west ledger |
| 2×4 | 48″ | north ledger |
| 1×2 | 78¼″ and 48″ | the two rails |
| Poplar 2×2 | ONE leg at H − 1″ | stop here — step 3 |
Tread 4 stays whole. Keep the 2×4 offcuts.
Stand the single leg under a tread beside the table and sit on it. Height is the one thing you cannot fix afterwards. Happy? Cut the other seven — two per poplar board, four boards.
Both west pieces mark identically
Measure from the outer end of each west piece — the end that is not the joint — and the two are the same layout:
West A and West B, each: 6⅝, 11⅝, 16⅝, 21⅝, 26⅝, 31⅝, 36⅝
North tread: 6½, 11½, 16½, 21½, 26½, 31½, 36½, 41½ — from either end, it is symmetric
All on a line 3½″ from the BACK edge — the flat one, not the nosing.
Optional: piece B’s first three (6⅝, 11⅝, 16⅝) sit past the end of the radiator and vent nothing. Skipping them saves six minutes and leaves the line stopping short of the end. Everything else is load-bearing.
Three things that halve the misery
Clamp scrap underneath every hole. No blowout on the underside, and the plug stays put instead of spinning.
Slow speed, both hands. A 3″ saw grabs. Let the pilot seat before the teeth bite, keep the drill square, back off if it smokes.
Clear the plug through the saw’s side slot every single hole. A jammed plug turns a two-minute hole into five.
Then ease every rim, top and bottom, with sandpaper rolled into a tube.
Face down on a flat surface
Lay both west pieces face down, ends butted tight. Run the 78¼″ rail across the joint, set back about 1¼″ from the front edge so it tucks behind the nosing.
Glue and screw: glue the whole length, then 1¼″ screws every 12″. The glue is what makes the rail count.
Clamp or weight both halves down onto the rail while it sets. That is what forces the joint flush — a step you can feel is the only real failure here.
Glue a 6″ block across the joint underneath while you are there.
Same for the north tread: 48″ rail, glued and screwed, no joint to worry about.
One coat of the Varathane. Top face, nosing, both ends, and inside every bore — the bores are end grain and thirsty. Stir, do not shake. Long strokes, one pass, never go back into an area that has started to set. It goes on milky and dries clear.
Flip and do the underside once the clamps come off. That is the only free shot at it. Then stop at one coat — more can wait, and one coat is the easiest thing to paint over if you go black later.
This is the part that has to be right
West ledger first, 78¼″, running into the corner. Face-flat against the wall — 3½″ tall, projecting 1½″. Top edge on your line.
Clearance-drill the ledger before it goes up. If the threads bite the ledger the screw jacks it off the wall instead of pulling it tight, and you will not see it until the seat sits proud.
Three 3½″ screws into every stud. Then the north ledger, 48″, butted to it, starting 11½″ off the west wall.
Check level along the whole length before you move on. The seat copies the ledger exactly.
West piece A into the corner first, then B against it. Then the north tread butted to the west run.
1¾″ screws every 12″, on a line 3/4″ off the wall, countersunk. That line sits behind everyone and between the wall and the hole line.
The leg top must touch the tread
Bearing directly on the underside is 700–900 lb. Hanging off screws is 240. Push each leg up until it touches, then fasten. The rail only locates it.
West, from the corner: 0, 19⅝, 39⅛, 58⅝, 78¼ — the middle one lands under the joint.
North, from the west end: 0, 24, 48.
Butt each leg behind the rail and drive two 2″ screws horizontally through the rail into it. Pilot the leg first — 1¼″ of screw into a 1½″ leg splits it otherwise.
| Ledger top, leg length | H − 1″ |
| Ledger | 2×4 face-flat — 3½″ tall, 1½″ out |
| Seat depth | 11½″, nosing to the room |
| Hole line | 3½″ from the back edge |
| Rail setback | about 1¼″, behind the nosing |
| Ledger → stud | 3½″, three per stud |
| Seat → ledger | 1¾″, every 12″, 3/4″ off the wall |
| Rail → seat | 1¼″, every 12″ |
| Rail → leg | 2″, two per leg, pilot first |
Not now, and none of it makes you undo anything.
The short list
Counter wrap — tread 4 at 25″, the leftover 1×2 for its rail, two legs from the 2×4 offcuts.
Table anchor — rubber pads under the pedestal base tonight; the block in the crook once the benches settle.
Finish — two more coats of the Varathane, or scuff and B-I-N if you go black.
Screen the holes — hardware cloth, not fly screen.
USB — cables up through two holes; a tamper-resistant USB-C receptacle rather than a brick in the cavity.
Window sill.
The bean is built and in — 42¼ × 54¼, mounted on the old pedestal through a plywood puck, in at 1:40pm on 30 August against a 3:35 estimate. This is v2 — and it is second. Build the benches, sit on them, then measure where the table edge wants to be; that number sets the arms, and everything on this page follows from them.
Arms
46¼ × 58¼ in, from the corner
Crook corner
4 in radius, was 8½
Counter end
16⅛ in flat, then 45°
Tabletop
2151 sq in · +20.5%
One 4×8 sheet, water-based poly rather than shellac, and the puck re-drilled to a new balance point. The arms below are the best numbers before the benches exist — treat them as the current answer, not a commitment.
What the benches will tell you, and what it changes
Sit on the finished bench and find where the table edge wants to be, measured out from the wall. You want it landing 4–6″ over the seat, so it is on your lap rather than at your knees. Call that number D, and the bean’s right-angle corner sits D from each wall.
D does not change the shape at all — the arms, the 4″ corner and the flat are the table’s own dimensions. What D moves is where the table sits: the walkway number, and how far the flat overhangs the wrap. So nothing on this page is wasted by waiting.
What could change is the arm lengths, if sitting there says the table should run further along a bench than 46¼ and 58¼. The east arm is capped near 47″ by the sheet’s 48″ width; the west has room. And if they do change, the flat follows a formula rather than needing me:
flat = √(east² + west²) − west
That is the free maximum for any pair of arms, exactly — it is the point where the 45° line lands on the fattest point, which is what makes it free. The fattest point itself is at east²/√(east²+west²) along the east arm and west²/√(east²+west²) down the west.
| East arm | West arm | Free flat | Fattest point | Tabletop |
|---|---|---|---|---|
| 42¼ | 54¼ | 14½ | 25-15/16 / 42-13/16 | 1829 |
| 44 | 56 | 15-3/16 | 27-3/16 / 44-1/16 | 1967 |
| 46¼ | 58¼ | 16⅛ | 28¾ / 45⅝ | 2151 |
| 47 | 58¼ | 16⅝ | 29½ / 45-5/16 | 2188 |
| 46¼ | 62 | 15⅜ | 27⅝ / 49-11/16 | 2285 |
Note the last row: a longer west arm gives a shorter flat. The two trade against each other, because both are spending the same diagonal. Worth knowing before you decide the table should reach further down the bench.
The shape
Both instincts are right, and one of them is more exactly right than you probably realised.
| v1 | v2 | ||
|---|---|---|---|
| East arm | 42¼″ | 46¼″ | |
| West arm | 54¼″ | 58¼″ | |
| Crook corner | 8½″ | 4″ | 1.9″ more reach |
| Tabletop | 1785 sq in | 2112 sq in | +328, or +18% |
| Balance point | 17.9 / 23.0 | 19-9/16 / 25¼ | new puck position |
| With the 16⅛″ flat | — | 2151 sq in | +20.5% — the spec |
“+4 on each axis” really is a 4″ offset
Worth confirming because it did not have to be true: growing both semi-axes by 4″ moves the curve 3.97 to 4.00″ from the old one at every point along it. So you get a genuine parallel push-out, and you can still lay it out with the same batten-and-offsets method rather than anything clever.
Still one 4×8 sheet — 46¼ leaves 1¾″ on the width, so ask for a single rip at 46½″ and keep the factory edge for the long straight side.
What it buys in seats: fewer than you would hope
You get 3 easy / maybe 4 on the west and 2 easy / maybe 3 on the east. In kid-widths that is 3.88 and 3.02. v2 takes them to 4.16 and 3.30.
So it does not add a new place anywhere — it turns both maybes into yeses. Four along the west and three along the east stop being a squeeze and become the normal arrangement. That is worth having, but it is a different claim from “seats another person”, and it is the honest one.
The corner is the sleeper win. Dropping 8½″ to 4″ pulls the table back only 1.66″ from the crook instead of 3.52″, so whoever sits in the corner gets 1.9″ closer to their plate — and you are right that nobody needs the wide radius to get in, because people arrive along the bench, not through the corner.
The check to run once the benches are in
v2 pushes the curve 4″ further into the room, and the walkway is measured to that curve. Taking the benches to 12″ hard against the wall bought 5.7″; v2 spends 4″. The bench change more than pays for the bigger table — the path lands near 22¼″, about 1½″ better than the nook has today, not worse. That turns this from a trade into a free upgrade, so the check below is a sanity test rather than a decision.
Unless your chair argument holds — and it is checkable. Push a chair in as far as it will go, then measure from the table edge out to the back of the chair. If the chair back sticks out past the edge, the chairs already define the pinch, v2 costs you nothing on a normal day, and you should build it. If the chair tucks inside the edge, the tabletop is the pinch and v2 costs the full 4″.
Do it sitting on the finished bench, at the same time you find D. Both numbers come off the same five minutes, and neither is knowable before then.
Your photo settles the part I could not see: the gap between the table and the counter is not how anyone gets to the benches, so widening the table there costs nothing to get in and out. And it turns out there is an exact number where it stops being free — and it is slightly more than you asked for.
16⅛″ of flat, and you pay nothing at all
Run the edge straight off the west arm’s tip for 16⅛″, then a straight run at exactly 45°, and that run touches the old curve precisely at its fattest point — 28¾″ along the east arm, 45⅝″ down the west arm. From there the curve is completely unchanged.
That is the whole trick. Not one point of the original curve moves. The walkway at the crook, the clearance at the corner, the reach past the east arm — all identical, to the thousandth. You are only filling in a wedge the ellipse was leaving empty against a cabinet.
And it is self-checking: if your straight run leaves the flat at 45°, you are exactly at the limit. A speed square is the whole verification.
| Flat | Table | vs v1 | Straight edge to sit at | Walkway cost |
|---|---|---|---|---|
| none | 2116 | +18.5% | 0 | — |
| 12″ | 2131 | +19.4% | 24¾″ | none |
| 14″ | 2140 | +19.9% | 29¼″ | none |
| 16⅛″ | 2151 | +20.5% | 34″ | none — the limit |
| 18″ | 2165 | +21.3% | 38¼″ | 1.9″ |
| 20″ | 2181 | +22.2% | 42⅞″ | 3.9″ |
| 24″ | 2224 | +24.6% | 52¼″ | 7.9″ |
16⅛″ is the spec. You asked for 12–14 and the free allowance is bigger, so there is no reason to leave any of it. And note which column is worth the most: not the 35 square inches, but 34″ of continuous straight edge at the seat you use every day, where today the curve turns away the instant it leaves the arm.
Marking it out
Three marks and a straightedge, before any batten comes out:
1. West arm 58¼″ from the corner. 2. Square across from that tip, 16⅛″ — call it A. 3. The fattest point: 28¾″ along the east arm, 45⅝″ down the west arm. Join A to it with a straightedge and check the angle — it must read 45°. If it does not, one of the three marks is off, and you have found it before cutting anything.
Then plot the rest from the offsets below and spring the batten from the fattest point round to the east tip. The batten never touches the flat or the 45° run — those are straightedge work, which is faster and more accurate than a curve anyway.
| Along the east arm | Down the west arm | |
|---|---|---|
| 28¾ | 45⅝ | the 45° run lands here |
| 30 | 44-5/16 | |
| 32 | 42-1/16 | |
| 34 | 39½ | |
| 36 | 36-9/16 | |
| 38 | 33-3/16 | |
| 40 | 29¼ | |
| 42 | 24⅜ | |
| 44 | 17-15/16 | it drops fast from here |
| 45 | 13-7/16 | |
| 46 | 6-1/16 | |
| 46¼ | 0 | 90° tip, same as v1 |
Why the flat is worth more than its area
35 square inches is a rounding error on a 2,150″ table. The reason to do this is that the bean’s arm tips are 90° corners — you found that out cutting v1 — and an ellipse leaves a 90° corner heading straight out, perpendicular to the arm. So at the exact spot you sit, today, the edge turns away from you immediately.
The flat replaces that with a proper straight edge, and the 45° run continues it: 34″ of straight in front of the one seat you use every day. That is a laptop and a coffee with room to put an elbow down, instead of a receding curve.
Shellac softens around 150°F and a bowl out of the microwave is 160–180°F. The film softens, traps moisture, and you get a white ring. The useful rule: if you can’t hold it comfortably, it will mark the table. That is not a defect you can finish your way out of with shellac — it is the deal shellac makes, and the other half of that deal is that the rings come out in fifteen minutes.
First fix: re-coat the whole top
One thin coat, ragged on. The fresh alcohol re-dissolves the top of the old film, the rings level out and vanish, and you get the third coat you skipped. Lowest risk of anything here — it treats every ring at once and you cannot really overdo it. Fifteen minutes, no sanding, no visible seam.
Second fix: pad the stubborn ones
Fold a lint-free cotton cloth into a pad, dampen with isopropyl — 91% if you have it, since 70% carries the water you are trying to drive out — then blot it until it is barely damp. Wipe across the ring in long continuous strokes, landing and taking off like a plane, and never stop moving on the surface. The film goes tacky, re-flows, the ring goes.
The risk is real: on a two-coat film a wet pad takes you to bare wood. Practice on a scrap first — shellac an offcut, burn a ring into it with a hot mug, fix that one before you touch the table.
Don’t use the iron trick. That is the standard advice for lacquer and poly; on shellac, more heat is the last thing you want. Alcohol is the direct tool.
And it settles the v2 question
This is data, not a nuisance. If the bean becomes the permanent table, shellac alone is not the finish — a family table meets hot bowls forever and you would be re-padding monthly. That pushes v2 to the stack below: dewaxed SealCoat as a barrier, then three coats of satin water-based poly. You lose the invisible-repair trick and gain a top that does not care about a hot bowl.
Check the can first, because it decides everything. Zinsser Bulls Eye Amber contains natural wax; Clear is dewaxed. Yours came out amber, so assume waxed — and polyurethane does not reliably stick to waxed shellac. That is the classic adhesion failure: fine for a month, then it peels in sheets.
The table you have now — one more coat of shellac, nothing on top
Fifteen minutes and it looks better than it does now. Shellac is uniquely repairable: new shellac melts into old, so a ring or a scratch wipes with alcohol and re-coats invisibly, which no other finish allows. For a family table that is a real strategy rather than a consolation — the trade is that you will be doing it, because water rings, alcohol marks and heat marks are all exactly where shellac is weakest.
v2 — water-based poly, satin, three coats
Stays clear instead of ambering further, recoats in 2–4 hours, low odour indoors. Oil-based is tougher but pushes the colour noticeably more orange and is an overnight-per-coat, smelly job.
The sequence, given the wax: one coat of dewaxed shellac (SealCoat) as a barrier — it sticks to waxed shellac and poly sticks to it, and that one step is the whole fix. Then scuff lightly with the 200 sponge, dry cloth only (no isopropyl, it softens what is underneath), then three coats of satin, hitting the edge hardest since that is still where water finds plywood. Give the underside a coat too so the faces stay balanced.
Test the stack on a scrap first. Shellac an offcut, add SealCoat and poly, wait 24 hours, scratch a cross-hatch into it and press packing tape down hard, then rip it off. Nothing lifts and you are good — that answers the wax question better than any label.
The thing people get wrong about poly: dry in hours, but about a week to fully cure. Wipe spills fast and keep hot plates off it for that first week or you will press marks into a finish that would have been fine.
So, two finishes for two tables
The current top: one thin coat of shellac. Fifteen minutes, the rings go, and it carries the table until v2. Do not put poly on it — the wax is why.
v2: SealCoat barrier, then three coats of satin water-based poly, both faces. That is the finish that stops this being a monthly job.
The number the puck does not change
The puck fixes stiffness; it does nothing for tipping. Stone off, roughly 23 lb on the far tip lifts the base — and v2 puts its tip 4″ further out, so that number gets smaller, not bigger. The bolted mount is what handles it, which is why the top has to be through-bolted to the flange and not just resting on it.