Parts get described as restrictive as though the description belonged to the part. It does not. Restriction is the pressure a component costs at a given flow, and the same component is negligible at one flow and dominant at another. A filter that is fine at idle can be the largest single loss in the system at full load, not because it changed but because the flow through it did.
That has a practical consequence. Any statement about which part is the bottleneck is incomplete without the flow rate it refers to. The useful question is not which part is restrictive but which part costs the most pressure at the condition the engine actually operates in, and how that ranking shifts across the range.
Two kinds of loss dominate in an air path. Friction along a duct rises with the square of velocity and with length, and it is spread evenly along the run. Loss at a transition, a bend, a step or an orifice also rises with the square of velocity, but it is concentrated at the feature and multiplied by a coefficient that depends on how abrupt the geometry is.
Because both scale the same way with flow, the ranking between segments stays roughly stable as flow rises, which is convenient. What does not stay stable is the velocity at each feature. A transition where the area changes has a local velocity that can be far higher than the mean in the duct, and that local velocity is what the loss coefficient multiplies. The narrowest point in the path, not the longest, usually decides the total.
Finding the bottleneck is mostly a matter of choosing where to measure. A tap before the filter, one after it, one at the compressor inlet, one at the throttle body and one in the manifold gives a series of pressures, and the difference between consecutive taps is the loss in the segment between them. That is the whole method.
The discipline is in the taps themselves. Each one has to be flush with the wall, square to the flow, and free of burrs, because a tap that protrudes into the stream or sits in the shadow of a weld reads a pressure that belongs to the tap rather than to the duct. Sealing the tap matters as much: a leaking tap on a suction line reads atmospheric and quietly ruins the comparison.
Static pressure is only meaningful where the flow is behaving. Put a tap immediately after a sharp bend and the flow on that wall may be separated, so the reading reflects a recirculating pocket rather than the pressure driving the system. Put a tap on the inside of a bend and it reads differently from one on the outside, because the velocity profile there is not symmetric.
The practical rule is to tap in a straight run, a few diameters away from anything that disturbs the flow, on a wall where the surface is smooth. Where that is impossible, the reading still has value as a comparison against itself over time, but it should not be treated as an absolute number to be compared with a reading taken somewhere else.
Some restrictions are not visible on a bench at all. A flexible hose with a soft wall and insufficient reinforcement will flatten under suction at high flow. On a stationary engine at idle the same hose is perfectly round and flows freely. Under load, at the moment the compressor is pulling hardest, the wall moves inward, the area drops, and the loss rises far faster than the flow does.
This failure is easy to misdiagnose because it appears as a loss of top-end performance with no obvious cause, and because the hose looks fine every time it is inspected with the engine off. The clue is usually in the shape of the curve rather than in any single reading: a system that behaves normally up to a point and then falls away sharply is describing a component that is changing its geometry, not one that is simply too small.
Restriction measurements taken during a free rev describe almost nothing. The engine accelerates through the range with very little air mass flowing, the throttle closes at the top, and the compressor spends part of the event in a condition it never sees on the road. The numbers that come out are reproducible and irrelevant.
What is needed is a steady state: a load held while the engine sits at a fixed speed, long enough for temperatures to stabilise and for the readings to stop drifting. That is what a dyno with a load cell provides, and it is also why road data logged during a long pull in a high gear is more useful than a series of blips in a workshop.
Once the segment losses are known, the ranking is usually obvious and often surprising. The filter is rarely the largest loss on a well-designed system. Transitions are. A step where a pipe meets a hose, an abrupt enlargement into a plenum, a bend with a tight centreline radius, a mouth that is partly blanked by a panel: those are the features where pressure disappears, and they are also the features nobody quotes a specification for.
This is why improving the part with a published flow figure so often changes nothing. The published figure was measured across the part, in isolation, with clean approaching flow. In the car, the part sits between two transitions that cost more than it does, and the total barely moves.
The additive model treats the path as a series of independent losses, which is fair while the flow is attached and fully developed at each feature. Where features sit close together, the disturbance from one has not decayed before the next begins, and the losses stop adding up in a predictable way. In a tightly packaged engine bay that is the normal condition rather than the exception.
It also stops being reliable across a compressor. The inlet side and the charge side are separate systems with separate references, and a pressure drop upstream of the compressor costs more than the same drop downstream, because the compressor has to work through a lower inlet density. Comparing a suction-side loss with a boost-side loss as though they were equivalent is a common and expensive error.
Measuring is cheaper than replacing, and it settles arguments that specifications cannot.
The cores and pipework we build are on our intercooler pressure testing pages, the rest of the charge path under intercooler and intercooler airflow. On the inlet side the equivalent work is under air intakes and air intake hoses, and the exhaust side, where the same series logic applies to a hot gas, under stainless steel exhaust system.
