Flow, metering and water loss

How much straight pipe does a flow meter need?

• Measured in diameters, not metres. Ten diameters on DN100 is a metre; on DN500 it is five.
• Upstream matters most. Disturbance downstream propagates back, but less severely.
• Not all fittings are equal. Two bends in different planes create swirl that persists far longer.
• Short of space, favour repeatability. A consistent bias still detects change reliably.

Why the meter cares about the profile

An ultrasonic meter does not measure flow directly. It measures the time taken for sound to travel diagonally across the pipe with the flow and against it, and derives velocity from the difference. It then multiplies by the cross sectional area to give volumetric flow.

That multiplication is where the assumption lives. The meter samples velocity along one or more acoustic paths, not across the entire cross section, and converts what it measured into an average using a profile correction factor. That factor assumes fully developed flow: symmetrical, with the familiar faster centre and slower walls, and no rotation.

Disturb that profile and the assumption fails. The meter still measures its paths perfectly accurately, then applies a correction derived from a profile that is not present, and produces a confidently wrong number. This is why the error from poor installation is systematic rather than noisy, and why it will not average out over time.

Why some fittings are much worse than others

The headline figure of ten diameters is a general guide. What actually matters is what kind of disturbance the upstream fitting creates.

A single bend produces an asymmetric profile: flow is pushed to the outside of the bend and recovers over a reasonable distance. This is the case the common guidance addresses.

Two bends in different planes, one after the other, produce something worse: swirl. The whole body of fluid rotates as it travels, and rotation decays far more slowly than asymmetry. A swirling flow can persist for many times the distance an asymmetric profile needs, which is why this configuration deserves considerably more straight run than the standard figure.

A partially open valve upstream is worst of all, because it is a variable disturbance. The profile depends on the valve position, so any error changes when somebody adjusts it. A throttling valve should always be downstream of the meter, never upstream. Pumps, reducers and tees each bring their own signature, which is why an installation manual gives different figures per fitting rather than a single universal number.

What to do when the pipe does not allow it

Retrofit work rarely offers a textbook straight run, and the honest answer is that you make an informed compromise rather than abandoning the measurement.

Give the upstream side the space, at the expense of downstream, if you have to choose. Upstream disturbance dominates.

Be explicit about what accuracy the compromise buys, and match it to the purpose. If the meter exists to detect change, a stable installation with a consistent bias remains genuinely useful, because minimum overnight flow analysis compares the pipe against itself rather than against an absolute standard. If the number is going on an invoice or a regulatory return, the same compromise is not acceptable and the pipework may need modifying.

Where nothing suitable exists, a flow conditioner upstream can restore a usable profile in a shorter distance, and relocating the measurement point to a different part of the system is often cheaper than trying to fix the pipework at the original one.