Carving Hashemi Flowers 26

Decide which strand passes over and which goes under, then undercut the lower one on both faces with a V tool laid flat. Nothing about the wood changes. What changes is the shadow under each crossing, and shadow is the only thing that tells an eye one band went beneath another.

Ever carved a lattice that came out looking like a printed grid? Does your basket pattern read as lines drawn on a board rather than bands passing through each other? The drawing was probably fine. What went missing is the dark line where one strand ducks under the next.

Workshop 26 of Carving Hashemi Flowers is an hour of nothing but that, front and back, on a grid that was already finished enough to leave alone.

What makes a crossing read as woven

• A decision, made before any tool moves, about which band is on top.
• An undercut on the lower band, shallow enough that the strand keeps its support but deep enough to throw a line of shadow.
• The same treatment on the back.
• Thickness taken down until the whole grid stops looking like a sawn plank with holes in it.
• A final pass with a knife along every flowing line, done last, when the wood is thin enough to be easy.

Why does a lattice look flat when it is carved correctly?

Your eye reads depth from shadow before it reads anything else. Two bands crossing at the same height give it no information at all, so it defaults to calling the whole thing a pattern on a surface. Drop one band a fraction and put a dark line under the join, and the same shapes suddenly pass through one another.

The measurement is smaller than most people expect. From the high point of a band down to the low point runs no more than 5 mm, often as little as 3 mm. Anywhere from 3 to 5 mm covers it. Beyond that you are not weaving, you are digging holes.

How do you cut the undercut itself?

Turn the V tool onto one of its two walls and use that wall as a chisel. Held flat against the wood it slides under the edge of the upper band and lifts a sliver away from the lower one, which is exactly the cut wanted, and no gouge does it as cleanly.

Depth stops short on purpose. Go until the wall almost disappears and then leave a hair under a millimetre standing, because that last sliver is what holds the strand up. Take it all and the band is floating on nothing, which looks wonderful for about four seconds.

Step	What this buys	Distance
Settle which band wins	settles the whole design before a tool touches anything	every crossing, marked
Undercut the low band	shadow	stop with 0.8 mm of wall still there
Thin the whole grid	stops the lattice reading as a sawn board with openings cut through it	down to 3 mm
Hollow the hidden faces	makes the edge read far thinner than the band is	until nothing is left at depth except the edge you can actually see from the front

A last practical note for anyone copying this. Mark which band wins on the wood itself in pencil before you start, not just on the drawing, because once chips are flying the two bands look identical and you will guess. Guessing at a crossing is how a weave ends up with two strands that both pass over, which no amount of later undercutting can put right. Five minutes with a pencil saves the panel.

Does every crossing need the same treatment?

Some do not, and knowing which is worth a few minutes of looking. A crossing that will sit against a wall, or one buried behind a leaf, becomes almost invisible once the panel is mounted, and carving it to the same standard as the front is time spent on something nobody sees. I skip several during this session and say so.

Everywhere else gets the full treatment, including places I could have left alone. A difference exists between work that looks nice and work that is right, and on a panel this size the difference costs an afternoon. I chose the afternoon. Anyone in a hurry may choose otherwise and the panel will still look good from the front.

Which band goes on top?

Look at the design and follow each band through the whole panel before deciding. A strand that goes over at one crossing has to go under at the next, all the way along, or the weave falls apart the moment somebody traces it with their eye. I stop repeatedly in this workshop to check which one goes below and which goes above.

Getting it wrong is not fatal and it is very annoying. Nothing about a wrongly woven lattice looks broken, exactly; it simply reads as decoration instead of structure, and a carver who knows will spot it across a room. Real weaving is a rule, not an effect.

Do you have to do the back as well?

On a panel with access from behind, yes, and I say plainly in the session that strictly I did not have to. It would still have looked nice. What I wanted was better than nice, which meant repeating the same cut behind every crossing so the weave holds up from any angle.

Flipping is easy here because a holding plate lives on both faces, so the piece turns over and mounts again without unscrewing anything. That setup has its own workshop, and this is the session where it pays for itself.

The back matters for a second reason. Once the finished panel is applied to a wall or a background it stands proud of whatever is behind it, so the lattice casts real shadow underneath, and every small undercut on the reverse deepens that shadow. Not much on its own, and every little bit of it helps.

How thin should the grid go?

Three millimetres, about an eighth of an inch, with 5 mm as the absolute ceiling anywhere on the panel. One semi flat gouge brings it down in long passes, and I keep saying out loud that I hope not to break it, having already demonstrated earlier in this course that I am perfectly capable of overdoing things.

Thin is what separates carved openwork from a fretsaw panel. Leave a lattice at its sawn thickness and every opening shows a square wall inside, which announces the bandsaw, and no shadow forms anywhere, because nothing is shaped for light to fall across. More on that in the workshop on thin leaves.

Can you make it look thinner than it really is?

You can. This is the part worth stealing. Bring the grid to 3 mm, then hollow the inside faces away leaving just the outer edge at full depth. Anyone looking at the panel sees an edge that reads like a sixteenth, while the band behind it stays at full depth and keeps all its strength.

Structure and appearance come apart at that point, and that is the whole trick. The wood you left is where the load is. What you removed was only ever going to be seen from an angle nobody looks from, and taking it away buys you an apparent delicacy that a genuinely thin band would pay for with breakages.

Old carvers worked this way constantly. Look closely at openwork foliage by Grinling Gibbons in the V and A and the leaf edges appear impossibly fine, yet the pieces have survived three hundred years of being moved about. Something behind those edges is carrying the load, because a genuinely paper thin band would have gone long ago. Exactly the same reasoning turns up again in the Gibbons class, where the leaves get thinner still.

Which V tool suits this job best?

Honestly, any of them will do the work. Mine is a number 16 at 6 mm for most of the work and a 3 mm version where the crossings tighten up, and I say on camera that any V tool will do the same thing. The angle matters more than the width, since a narrow V leaves a tighter shadow line.

Thinning falls to a wide, shallow gouge, and the one I reach for is stamped number 4 while looking a great deal more like a number 3. Makers are careless about marks. After a year with a tool you stop reading the stamp and go by how the edge sits in the wood, the only measurement that ever mattered.

Two more links for anyone building this vocabulary. Interlace of exactly this kind runs through Celtic knotwork and through Islamic geometry, and both traditions solved the same problem centuries before anyone wrote it down. Our panel borrows the logic without the geometry. To practise the principle on something smaller before risking a whole panel, a rosette is the right size, and a narrow V tool is the only thing you must own.

What about the grain?

Against it, often, and there is no rotating your way out. Some crossings simply cannot be reached from the friendly side, so the tool goes over on its side and works against the grain deliberately, survivable only because the tool is genuinely sharp and the pass is tiny.

End grain turns up constantly at the crossings too, since a band curving through a lattice presents its fibre ends somewhere in every arc. Sharpness is not optional there. End grain has its own session and the rule is the same: less bite, keener edge.

A knife covers what the V tool cannot reach. Mine is a number 12 whose tip snapped at some point and got reground shorter, and the shorter version turns out to work better for this, an accident I ended up grateful for. Regrinding a broken tool is worth learning before you replace one.

Speed is the surprise of this stage. Thinning and undercutting a whole grid goes quickly compared with everything that came before, and the danger sits exactly there. Quick work invites pressure, pressure breaks a lattice, and I catch myself hurrying on camera and say so out loud. Nothing on this panel punishes impatience faster.

Choppy is allowed on the way through, too. Partway along the back the lines look ragged and I say not to worry, because a ragged line at this stage is still fixable and a broken band is not. Carving forgives ugly far more readily than it forgives force.

One question always follows this workshop: how did anyone carve a weave before photographs, when a carver could not simply look one up. They had pattern books and they had each other, and interlace is old enough that a Roman floor laid in tesserae shows the same logic in stone. Draw one on paper first. Follow every band with a pencil until the alternation holds all the way round, and the carving afterwards is only the easy half of the job. Getting that drawing right is design work, and it is where a weave is won or lost.

Light is the other thing to plan for. A panel meant for a north wall gets softer shadows than one facing a window, so the same undercut reads differently depending on where the piece will hang, and knowing the wall beforehand changes how deep I go. Carvers in the Baroque were ruthless about this. They carved for the room it would hang in, never for the bench it sat on.

What is the last thing you do?

Draw the final line in pencil, then follow it with the knife. By that stage the grid is thin, the weaving reads properly, and all that remains is making every band flow naturally rather than in the slightly choppy way it looks after a session of undercutting. Only a few spots need it.

Thinness makes that last pass easy, which is why it comes last. Trying to correct a flowing line while the band is still at full thickness means fighting a great deal of wood for a small change, and the same correction on a thin band takes one stroke. Order matters more than skill in a lot of carving.

Want the whole sequence rather than the summary? Every workshop of this course sits inside tuition at the school, and this page is where people who have never carved begin. Bring basswood, since Tilia holds a band this thin where most timbers split.

One honest note to finish on. Rushing is tempting during this step because it goes quickly and looks like tidying up, and I catch myself trying to hurry on camera and say so. Pressure is what breaks a lattice, and a lattice breaks at the one crossing you were not thinking about.

The wood is doing you a favour throughout. A band of lime at 3 mm survives being undercut from both faces, and the same section in oak or ash would split along a ray while you watched. That is why nearly every historic piece of pierced acanthus in Europe is carved in a soft, even timber, and why reading the grain matters more here than anywhere else in the course.

Shall we carve?