What a Loop Header Does That a Dead-End Main Cannot

Why this matters

Most compressed-air distribution in small shops grew rather than got designed. A main was run to the first machine, then extended to the second, then teed off for the third, and ten years later the far end of that run is where every complaint comes from. Topology is the one thing about a distribution system that cannot be fixed later with a bigger machine, a better dryer, or a higher setpoint, and it is decided on the day someone hangs the first stick of pipe. A loop costs roughly double the pipe of a dead-end run covering the same floor. What it buys is three things the dead-end run cannot buy at any price: a lower loss at the worst drop, a demand event fed from two directions, and the ability to take a section out of service without shutting the bay.

What a dead-end run forces on you

In a dead-end main, every cubic foot going to the last drop has already travelled the full length of the run and passed every drop before it. Friction loss in a round pipe at fully turbulent flow, with the friction factor treated as roughly constant over the range, scales with the square of volumetric flow and inversely with about the fifth power of the bore. Length enters linearly. So the far drop pays the full length at the full flow, and it pays it every time the plant is busy.

The second thing it forces on you is arithmetic you cannot escape by being careful: a demand event at the far drop is fed down one path. There is no second route, so the whole flow deficit has to travel a single restriction, and the local pressure sag is as deep as that single path allows.

The third is operational. Any repair anywhere on the run kills everything downstream of it. That is why so many dead-end systems have never had a single leaking fitting replaced: the only window to do it is a plant shutdown, and shutdown time gets spent on production equipment.

The loop's arithmetic

Take a loop as a closed ring passing through the receiver connection and out around the bay, with the two halves of equal length and equal bore, and take the worst drop as the one directly opposite the feed. By symmetry the two halves each carry half the flow to that drop, over half the ring.

  dead-end main

  receiver >==+==+==+==+==+  closed end
              1  2  3  4  5

  air reaching drop 5 has passed drops 1
  through 4, at full flow, over the whole run

  loop header

              +==1==+==2==+
              |           |
  receiver >==+           +== drop 3, opposite
              |           |   the feed
              +==5==+==4==+

  air reaching drop 3 arrives by two paths,
  each carrying part of the flow

Compare like with like. The far point sits a route distance L from the receiver. A dead-end main covering that route is length L carrying the full flow Q, so its loss goes as Q squared times L. The loop covering the same floor has a perimeter of 2L, and the far point is fed by two paths of length L each carrying Q over 2, so the loss along either path goes as one quarter of Q squared times L. Same bore, same flow, same route distance: about a quarter of the loss, for double the installed length.

Those conditions are load-bearing. Move the dominant load off the point opposite the feed and the two paths no longer split the flow evenly, because the split follows the relative resistance of the two routes, and the improvement shrinks toward nothing as the load approaches the feed. Make one half of the ring a smaller bore than the other and the same thing happens for the same reason. And this is a scaling comparison, not a sizing calculation: hand the actual bore selection to the pipe manufacturer's tables with your measured peak simultaneous flow, not to this ratio.

It is worth setting that quarter against the alternative use of the same money. Going up one bore step on a dead-end main also cuts the loss, and a bore 25 percent larger cuts it to about a third, because 1.25 to the fifth power is 3.05. On the far-drop loss alone the two options are close. The loop wins on everything else.

What the loop does not change

It does not reduce leak flow. Leaks are choked openings venting to atmosphere and their mass flow tracks absolute header pressure, not layout. A loop reduces leak flow only through whatever setpoint reduction its lower loss allows.

It does not add capacity. If the machine cannot make the air, the ring delivers the shortfall more evenly.

It does not make a drop leg unnecessary. Liquid water in a header runs along the bottom of the pipe. Take every drop off the top of the header, run the branch down past the takeoff, and put the shutoff and the filter above a short drop leg with a drain at its bottom. That is true on a loop and a dead-end run alike.

It does not decide the slope for you, and here the loop actually makes the question harder. On a dead-end run you slope with the flow toward a drain at the far end. On a loop the flow leaves the feed in both directions, so there is no single downhill. Slope each half away from the feed and put a drained low point at the far end of each half, or you have built a ring with a puddle in it.

The fields a distribution spec has to carry

A distribution layout is worth writing down as a spec rather than a drawing, because the fields are what get argued about later. These are the ones that decide whether the system works:

• Topology, and the reason for it in one line.
• Peak simultaneous flow the header was sized at. Not the compressor's rating. The compressor rating is what the plant could draw if everything ran at once, which it does not.
• Allowable loss from the receiver outlet to the worst drop, at that flow, stated in psi.
• Header bore, from the pipe manufacturer's tables at the two figures above.
• Sectionalising valves: how many, where, and what each one takes out of service.
• Takeoff detail: top-of-header takeoff, drop leg, drain.
• Slope, with the direction and the low points named.
• Point-of-use chain: filter, regulator, coupler bore, hose bore and length. This is where the fifth-power scaling does its worst damage, and it is the part usually left to whoever is holding the tool.
• Spare capped takeoffs, so the next machine does not get a tee cut into a live main.
• Test ports, at the receiver outlet and at the worst drop, so loss can be measured with a gauge instead of estimated.

Worked example: the filled-in spec for one bay

A 40 by 80 foot fabrication bay, twelve drops, compressor and receiver in a room off one corner.

• Topology: loop. Chosen for sectional isolation as much as for loss; the shop cannot take a production shutdown to replace a fitting.
• Peak simultaneous flow: 110 cfm, measured rather than assumed, by running the bay's normal busiest hour and reading the compressor's loaded fraction.
• Allowable loss to the worst drop at 110 cfm: 3 psi.
• Header: 2 inch nominal, loop perimeter 240 feet, two halves of 120 feet.
• Loss check: the drop opposite the feed sees each half carrying 55 cfm over 120 feet. Against a dead-end main of 120 feet carrying the full 110 cfm at the same bore, that is 55 over 110 squared, which is a quarter, at the same length. The dead-end version of this bay would have needed a materially larger bore to meet the same 3 psi.
• Sectionalising valves: four, one at each corner. Closing the two valves either side of a corner segment takes that segment out and leaves the other three segments fed, though from one direction only while the segment is out, so the loss at the remaining drops is temporarily worse. That trade is the point of the valve, and it is worth writing down so the next person understands why the pressure moves during a repair.
• Drops per segment: three, twelve across four segments.
• Takeoffs: top of header, each with a drop leg and a manual drain at the bottom.
• Slope: each half slopes away from the feed at a consistent fall to a drained low point at the far corner.
• Point of use: filter and regulator on each drop, coupler and hose bore recorded per drop so a starved tool can be traced to its own chain before anyone touches the header.
• Spares: two capped takeoffs per segment, eight in total.
• Test ports: two, one at the receiver outlet and one at the drop opposite the feed.

Two notes on executing it. Cutting any of those valves or ports into an existing live header means isolating that section at its upstream valve, opening a bleed and confirming a gauge reads zero before the joint is broken, with the isolation under lock and tag per 29 CFR 1910.147, which covers stored mechanical energy of this kind. An air line opened under pressure ejects the fitting and whips the hose, which is a struck-by injury from a system everyone in the building treats as harmless. And do not blow the cuttings out of the pipe with shop air: 29 CFR 1910.242(b) permits compressed air for cleaning only where reduced to less than 30 psi and then only with effective chip guarding and personal protective equipment, and pipe scale driven at header pressure will not stop at an eye.

The failure mode on this spec is the one field people leave blank. Skip the peak simultaneous flow and someone sizes the header off the compressor nameplate, which is a larger number, so the pipe is generous and nobody notices; then the next bay gets sized off the same habit, the number is wrong in the other direction, and the loss lands at the drop farthest from anyone who would recognise it.

References

• 29 CFR 1910.147, control of hazardous energy, for isolating and verifying depressurisation before cutting into or breaking into a pressurised air line
• 29 CFR 1910.242(b), which permits compressed air for cleaning only when reduced to less than 30 psi and used with effective chip guarding and personal protective equipment
• Pipe manufacturer's published inside diameters, and equivalent-length data for the valves and fittings in the intended layout
• See related: Why Pipe Size Decides More Than Compressor Size; How Pressure and Flow Relate in a Real System