Headers are usually described by their primaries: bore, length, material, finish. The collector gets a sentence at the end, as if it were plumbing that gathers four pipes into one. It is the most sensitive part of the assembly. Every primary terminates in it, every pulse passes through it, and the geometry there decides how much of each pulse is allowed to leave and how much is sent back up the pipe it came from.
Treat it as a junction and the behaviour starts to make sense. A pressure wave arriving at a change of area is partly transmitted and partly reflected, and the proportions depend on the size of the change and how abruptly it happens. The collector is a large area change in a short distance, repeated every time a cylinder fires, and it is doing that work whether or not anyone designed it deliberately.
Each primary delivers a pulse with a shape: a steep front as the valve opens, a plateau while the cylinder empties, and a trailing edge as the valve closes. By the time that pulse reaches the collector it has been distorted by the pipe it travelled through, cooled, and slowed. It arrives as a moving slug of high-pressure gas with a low-pressure region behind it.
That shape matters more than the peak. A pulse with a steep front carries its energy in a short window and reflects a sharp signal; a smeared pulse carries the same energy over a longer window and reflects something blunt. Primary length is what shapes the pulse before it ever reaches the merge, which is why collector changes and primary changes are never independent of each other.
Two pipes meeting at a shallow angle encourage the arriving gas to keep going in one direction. The pulse crosses the junction, the other primaries see a pressure rise but not a violent one, and the flow leaves with most of its momentum. Two pipes meeting at a sharp angle behave differently: the arriving pulse hits a wall of gas, splits, and a significant part of it turns back up the pipe it came from.
That reflection is not automatically bad. A controlled reflection returning to the port at the right moment during overlap helps empty the cylinder. It becomes bad when it returns at the wrong moment, or when its strength varies cylinder to cylinder because the merge geometry favours one primary over another. In a badly shaped collector the four cylinders do not get the same exhaust event, and no amount of tuning elsewhere fixes that.
Where primaries are joined before the final merge, the pairing has to follow the firing order. Two cylinders that fire close together should not share a junction if it can be avoided, because the second pulse arrives into gas the first one has just pushed through, and the two interfere rather than help. Two cylinders spaced evenly around the cycle share a junction comfortably, and the pulse spacing becomes regular enough to be useful.
This is the real content of the argument between a single merge and a staged one. It is not simply that one design favours low speed and the other high speed. The staged design lets the pairing be chosen, and the pairing decides which cylinders cooperate and at what engine speed the cooperation lands.
Change the cam, the firing order or the crank layout and the conclusion changes with it. A pairing copied from an engine with a different firing sequence is not a design decision. It is a coincidence that happened to work once.
Downstream of the merge the gas is fast and the area is small. Most of the kinetic energy in that fast moving column can be turned back into pressure if the area opens gradually, which is what the cone is for. The word gradually is doing the work. A diffuser that opens too quickly separates: the flow leaves the wall, the effective area stops growing, and the recovery that was supposed to happen does not.
The same principle governs the shape at the merge itself. A collected transition that eases the four streams into one keeps the flow attached and keeps the pressure loss low. A flat plate with four holes punched in it and a cone welded over the top has the right silhouette and almost none of the behaviour, because the flow separates at every step and never reattaches before the next one.
Where space forces primaries of different lengths, the pulses arrive at different times. Each cylinder then sees its reflection at a slightly different point in its cycle. The result is not a single tuned peak but a wider, shallower region in which each cylinder is doing something slightly different, and the engine feels smooth without feeling sharp.
That is a legitimate compromise and often the only one available in a real engine bay. It should be a decision, though, rather than something discovered after the fact. A fabricated set with one primary coiled to fill a gap is a packaging solution that has exhaust consequences, and the honest way to describe it is as a trade rather than as a tuned design.
Feed a turbocharger and the collector no longer terminates in open space. It terminates in a housing that swallows the pulse almost completely, because the turbine extracts energy from pressure rather than reflecting it back. The tuning argument changes shape: the objective is no longer to time a reflection at the port but to deliver the highest possible total pressure to the wheel, with as little loss as possible on the way.
That favours short, direct primaries with generous radii and a merge that does not fight the flow, and it makes the merge angle less about wave timing than about keeping the streams from colliding head on. A design copied from a naturally aspirated header onto a turbo manifold usually keeps the length and loses the point.
All of the above describes wave behaviour in a pipe of constant section with a clean junction. Real headers have bends, welds, wall thickness changes, supports and a hot gas that is not uniform. Each of those damps or scatters the wave, and the sharper a predicted effect is, the more of it is lost to that damping.
The reasoning also stops being reliable at the extremes of the rev range, where the pulse spacing becomes so short that pulses overlap and the individual events stop being individual. Predicting what happens there needs simulation or a test, not an argument from first principles over a bench.
Most of the value in a header is decided in the first conversation, before any tube is cut.
We fabricate headers and the systems they feed, and the parts are on our titanium exhaust headers pages. The rest of the path is under stainless steel exhaust system, heat control around it under exhaust heat blanket and exhaust heat shield, and the charge side under intercooler with filtration under air filters and prefilters.
