Undercut
A groove melted into the base metal along the edge of the bead, and never filled back in.
The short answer. Undercut means you melted more base metal than you put filler back into. It is nearly always too much heat, an arc that's too long, travel that's too fast, or a work angle pointing the arc at the plate instead of into the joint. Fix the heat first — it's the most common cause and the easiest to test.
What it looks like
Run your fingernail along the edge of the bead where it meets the plate — the toe. If it catches in a groove rather than sliding smoothly onto the parent metal, that's undercut. In good light you'll see it as a dark, narrow shadow line running parallel to the bead, because the groove is deep enough to hold shade even when the rest of the plate is lit.
It shows up in two places. Toe undercut runs along either edge of a finished bead. Sidewall undercut hides inside a groove weld, on the face of the joint, where it will be buried by the next pass — which is worse, because burying it doesn't remove it.
On a horizontal fillet it characteristically appears on the upper toe only. That asymmetry is a strong clue: gravity is pulling your puddle down and away from the top edge, leaving the metal you melted there unfilled.
What causes it
1. Amperage too high
The most common cause by a wide margin. Excess current melts the plate faster than your filler can fill it, so the edges of the weld pool eat into the base metal and the puddle freezes before it flows back. If you have undercut on both toes and the bead is also wide and flat, start here — turn the machine down and run the same bead again.
2. Arc length too long
A long arc spreads and wanders. Instead of a tight, directed cone of heat going where you point it, you get a broad one that washes the plate either side of the joint. With stick especially, a long arc also means less of your rod ends up in the puddle and more of it ends up as spatter. Tighten up — for most stick work the arc should be roughly the diameter of the rod's core wire.
3. Travel speed too fast
You melt the toes on the way past and move on before the puddle can flow back and fill them. The tell is that the bead is also narrow and possibly ropey — see a ropey bead — and the undercut is shallow but continuous down the whole length.
4. Wrong work angle
If the electrode is tilted so the arc points at one plate rather than into the corner of the joint, that plate gets the heat and the other gets the filler. Undercut on one side only, with the bead visibly leaning away from it, usually means angle rather than amperage. On a horizontal fillet, a small upward tilt puts heat back where gravity keeps stealing it.
5. Weaving without pausing at the edges
A weave that sweeps smoothly from side to side spends its shortest time exactly where it needs to spend the longest — at the toes. The fix is a deliberate, brief hesitation at each end of the weave to let the puddle catch up and wet out onto the plate before you come back across.
How to work out which one is yours
Change one thing at a time, and read the bead in between:
- Undercut on both toes, bead wide and flat — turn the amperage down 10% and run it again.
- Undercut on one toe only — it's your work angle, or gravity on a horizontal joint. Fix the angle before you touch the machine.
- Shallow undercut running the full length, bead narrow and high — you're travelling too fast.
- Undercut plus heavy spatter — arc length. Get closer.
- Undercut only at the ends of a weave — pause at the toes.
One change per bead. If you drop the amps, shorten the arc and slow down all at once and the undercut disappears, you have learned nothing about which of the three was responsible — and you'll be guessing again on the next joint in a different position. Changing one variable per coupon is slower for one bead and much faster over a week.
Why it matters
Undercut reduces the thickness of the base metal right at the edge of the weld, and it leaves a sharp-bottomed groove. Both matter: the section is thinner than the drawing says, and the sharp root of the groove concentrates stress at exactly that thinner point. Under cyclic loading that's a natural place for a crack to start.
That's also why codes treat it seriously and why acceptance limits are usually tighter for undercut running across the direction of stress than along it. What those limits actually are depends on your code, the thickness, and the joint — get the numbers from the standard you're working to and the inspector who'll be checking, not from a web page.
What a photo can and can't tell you
Undercut is one of the few defects a photograph handles honestly, because it's entirely a surface feature — length, depth and continuity are all visible from above given decent light and something in frame for scale. It's the kind of thing an app can genuinely help you see, especially early on when your eye hasn't calibrated yet.
What no photograph can tell you is what's underneath it: whether the root fused, whether there's slag trapped along that same toe, or whether the penetration is anywhere near what the drawing asked for. A bead with no visible undercut at all can still be a bad weld. Surface visual only — internal defects need RT or UT and a qualified inspector.
Related
- Running too hot — undercut's usual travelling companion.
- A ropey bead — when travel speed is the culprit.
- Porosity — the other defect people photograph most.