A gas lens cup held almost touching the work, the arc small and close. Behind it, a formed titanium shell with a shallow shoulder, and on its face a small attachment sitting against the metal with two or three bright gold tacks already in place. Beside the attachment there is an opening in the shell, and the heat glow is still running out of the joint.
Read the frame as a moment rather than a picture. The shell is not a flat plate and it is not a tube off the saw. It has already been formed, trimmed and fitted. The attachment is not a pipe joint. It is a bracket, a tab, a boss, one of the small features that a real exhaust system needs and that catalogues never show.
A bracket looks like an accessory. In service it is a load path. It carries the weight of a section, resists the movement of a hot system, and takes every vibration the engine sends down the pipe. The tacks in this frame are perhaps the smallest welds on the part, and they are the ones with the least material to argue with.
That is the shape of the problem. A structural joint is usually designed with a generous section, a defined throat and room to add material. An attachment welded onto a thin formed wall has none of those. The wall is thin because the shell has to be light and because it had to be formed in the first place. You cannot thicken it now. The tacks either sit in that wall properly or the bracket is a fatigue crack waiting for a date.
This is why shops that are competent at pipe work still get brackets wrong. Pipe welding rewards consistency along a seam. Bracket welding rewards judgement about heat, sequence and what is happening on the far side of a wall you cannot see.
Run a full seam around a bracket on a thin wall and the heat goes in as a closed loop. The loop contracts as it cools, the wall inside the loop is trapped, and the result is a shallow cone with the bracket at its centre. On a flat panel you would see it immediately. On a curved shell it hides in the curvature until the part is bolted up and the mating face no longer sits flat.
Tacks break that loop. Heat goes in at points rather than around a perimeter, and the material between the tacks stays cool enough to hold its shape. What you trade for that is concentration: the same load now passes through small sections, so the tacks must be placed where the load actually goes and must be sound rather than merely present.
The tacks in the frame are also not decoration holding a part still while someone else does the real work. On an attachment this small, the tacks are the weld. That is why they are sized deliberately and not simply made as small as the welder can manage.
Every tack pulls the metal toward itself as it cools and shrinks. That is not a defect, it is the mechanism. It also means the sequence is a control input, not a habit.
The working rule is to place tacks in opposing pairs and to check the fit between them as you go. Tack one side, then the side across from it, then measure. If you work around the bracket in a circle, the first tacks close the gap on their own side while opening it on the far side, and by the time you close the circle the bracket is sitting in a distortion you created and cannot undo without cutting.
On a flat wall the movement is easier to predict. On a formed shell the movement depends on the local curvature, the thickness and what the wall is doing a few centimetres away. So the welder reads the gap rather than trusting a pattern, and the pattern exists to keep the reading possible.
Titanium will take oxygen, nitrogen and hydrogen into the metal while it is hot, and it does not forgive any of them. The front of the weld is protected by the cup and the trailing gas. The back is protected by whatever the shop arranged before the torch was lit.
On a pipe this is routine: purge the bore, weld, done. On a formed shell you are welding onto a closed volume. There is no open end to push gas through at the moment you need it. Either the shell was purged through a port while it was still accessible, or the attachment was welded at a stage when the interior was still open, or the joint is designed so the back can be reached.
That decision was made hours earlier, by someone planning the build order, and it is invisible in the photograph. A welder can produce a perfect-looking front face on a joint whose back is contaminated. The part will pass inspection, look right, and fail from the inside months later.
A tack cools in seconds, which is exactly the window where shielding is most fragile. The cup is close but the tack is short, and the gas that covers a steady run has less time to establish itself over a brief one.
A contaminated tack is worse than a contaminated seam because the seam can be dressed and re-welded, while the tack gets buried. You tack for alignment, then you weld over the tack, and the compromised metal is now inside the joint. Nothing on the outside records it.
Two things give it away. One is colour: on titanium the surface oxide changes with temperature and atmosphere, and a tack that finishes grey or chalky white was not protected, whatever it looks like from arm's length. The other is the back, if you can see it. This is why the colour on a titanium part deserves to be read rather than admired; it is the only part of the weld that reports on itself.
On a flat bracket a pull can be corrected by shrinking, planishing or a press. On a finished shell, none of those are available. The shell is the shape it is, and any correction risks changing a surface that was formed and finished before welding started.
So the practical answer is to stop the pull rather than fix it. Clamp the bracket flush before the first tack, so the first tack has nothing to pull down. Place that first tack where the gap is already closed, not where the gap is widest. Keep the run of tacks short enough that the wall between them never gets hot enough to move. If a bracket must sit on a surface that cannot stay flat, design it to float, with a slot or a shim, instead of forcing it with welding.
You do not need a laboratory to judge this joint. Look at the spacing of the tacks: even spacing suggests a sequence, uneven spacing suggests someone worked around a problem. Look at the colour: straw and gold are normal on titanium, grey and white are not. Look at the wall around each tack for a dent or a dished area, which means heat went in faster than the wall could take it. Check the bracket face with a straight edge; a bracket welded into a pull will rock or show a gap on one side.
Then ask the question that separates a workshop from a welder with a torch: where was this part purged from, and at what stage was the bracket welded on. A shop that can answer that has a build order. A shop that cannot has a habit.
