Why Raising System Pressure to Fix a Symptom Costs More Than the Symptom

Why this matters

Turning up the compressor is the fastest fix in the building. It takes a minute, it usually makes the complaint go away for a while, and it leaves no record. It is also the only fix in the building that keeps charging you after it stops working, because a setpoint is paid for on every hour the machine runs, not once. Most shops with a chronically high setpoint cannot tell you what it was raised to fix. Somebody raised it, the reason was resolved or forgotten, and the plant has been carrying it ever since.

The point of this article is not that raising pressure is always wrong. It is that the same symptom, low pressure at a machine, has two completely different causes, and the setpoint is the wrong answer to one of them and an expensive stopgap for the other. One measurement separates them.

The two prices a psi is paid at

Before the gate, the arithmetic, because you cannot weigh the fix without it.

On the compression side. Compression work rises with the logarithm of the pressure ratio, not with pressure itself, so the marginal cost of one psi depends on where you already are. At a discharge of about 100 psig from an atmospheric intake, one more psi is roughly four tenths of a percent of input power for the same air made. That coefficient was derived at about 100 psig and does not travel: at a lower discharge the same psi is a larger percentage because it is divided by a smaller absolute pressure, and at a higher discharge it is smaller.

On the demand side. Any opening venting to atmosphere from a header above roughly 13 psig is choked, meaning the air leaves the throat at the local speed of sound and the mass flowing out is proportional to the absolute upstream pressure. Every leak, every unregulated blow-off, every open drain and every device fed above the pressure it needs passes more air when you raise the header, in direct proportion. In an unmanaged system that population commonly totals a fifth to a third of compressor capacity, so it is a large term to apply a percentage to.

Those two stack. You pay more per unit of air and you make more units.

And there is a ceiling that is not negotiable. The receiver's maximum allowable working pressure is stamped on its nameplate, under the ASME Boiler and Pressure Vessel Code Section VIII in the edition your state's boiler and pressure vessel law has adopted, which binds the vessel manufacturer and reaches you through that stamp and your jurisdiction's inspection requirement. A setpoint may never be raised toward the relief valve setting: a spring-loaded safety valve, required on air receivers by 29 CFR 1910.169, is a last-resort protective device, not a pressure controller, and a system that runs it open is running without its protection. Check the weakest rated component too, which is often a hose, a plastic filter bowl or a section of non-metallic pipe rather than the receiver.

The gate: is the pressure missing at the header, or between the header and the tool

One measurement, taken during the complaint, with the same gauge moved between two ports. Using one instrument matters: a bourdon gauge's dominant error at a given point on the dial is a fixed systematic offset, and a fixed offset from one instrument cancels in a difference, leaving only that error's contribution to the difference itself. Read the two ends on two different gauges and the offsets do not cancel. If all you know about each is a worst-case bound, those bounds add linearly and you have doubled the uncertainty on the number the whole decision turns on.

To fit a gauge into a live line, isolate the section at its upstream valve, open a bleed and confirm the gauge reads zero before breaking the joint, and hold the isolation under lock and tag per 29 CFR 1910.147, which governs stored mechanical energy of this kind. A line opened under pressure ejects the fitting and whips the hose.

Read the header at the receiver outlet and read the point of use at the tool inlet, both at the moment the complaint happens.

• Header good, point of use low: the pressure is being lost in the drop chain. Go to case one.
• Header low too, and every other gauge in the plant is low at the same moment: the pressure is missing before distribution ever gets it. Go to case two.

Case one: the grinder that stalls

A shop's die grinder bogs under load at one station. During the complaint the receiver outlet reads 98 psig and the tool inlet reads 62 psig. The compressor band is 98 psig load, 108 psig unload, so 98 is the bottom of a normal band and the header is fine.

Thirty-six psi is being lost between the header and the tool. Raise the band by 10 psi, to 108 load and 118 unload, and here is what happens. The tool inlet rises by less than 10 psi, because the higher upstream pressure drives more flow through the same restriction and loss through a fixed restriction rises with the square of flow. The 36 psi restriction is untouched. Meanwhile the header absolute goes from 112.7 psia to 122.7 psia, a ratio of 1.089, so every leak and unregulated opening in the plant passes about 9 percent more air; and the compression side costs about 4 percent more input power, at four tenths of a percent per psi around 100 psig.

Put the two together with leaks at a quarter of output, which is inside the fifth-to-a-third band above. Total output rises about 2 percent, specific power rises about 4 percent, so input power rises about 6 percent, permanently, and the grinder still bogs when the next operator leans on it.

The actual fix is in the drop chain, and it is found by measuring the same difference in pieces: read across the filter element under flow and compare against the manufacturer's loaded differential rather than its clean rating, then read across the coupler, then read the hose. The fifth-power dependence on bore means the smallest bore in that chain dominates, and the smallest bore is almost always a quick-coupler or a long undersized hose, not the pipe.

Verdict: do not raise it. There is no version of this where the setpoint is the cheapest path.

Case two: the Monday-morning sag

A different plant. On the first start of the week the header falls from 98 psig to 74 psig for about 90 seconds, then recovers and behaves for the rest of the day. Every gauge in the building moves together. Nothing is being lost in a drop chain; the plant is briefly asking for more than the supply side can deliver, and the storage is not deep enough to bridge it.

The shop's own tolerance, from the worst point of use plus its measured loss, is 85 psig. So the sag goes 11 psi below what the plant can work at.

Now the honest version of what a setpoint increase buys here. Ride-through you can count on is measured from the load setpoint, not the unload setpoint, because that is where the header may be sitting when the event starts. Today that guaranteed band is 98 minus 85, which is 13 psi. Raise the band by 10 and it becomes 108 minus 85, which is 23 psi, a 77 percent increase in usable stored air. Unlike case one, that is a real effect on the real symptom: the same mass deficit now starts 10 psi higher and bottoms out at 84 psig instead of 74.

It also still misses, by one psi, and it costs the same roughly 6 percent of input power every hour of every day to solve a 90-second event once a week.

Verdict: raise it as a contained stopgap with a written end date, while receiver volume or a controlled storage arrangement is sized and installed, then walk it back and confirm the sag is gone with the storage doing the work. Storage is bought once; a setpoint is bought every hour. The one thing that must not happen is what usually happens, which is that the stopgap becomes the permanent setting because nobody wrote down that it was temporary.

Walking back a setpoint you inherited

Most shops meet this problem from the other end: the setpoint is already high and nobody knows why. Do not simply drop it to a number that sounds right, because if there is a genuine consumer of that headroom you will find out through a production stoppage.

Adjusting a mechanical pressure switch or a controller setpoint means opening an enclosure with energized conductors in it. De-energize the starter, lock and tag it, and prove the circuit dead with a meter before working inside, per 29 CFR 1910.333(b)(2), which is the electrical standard for this work; 1910.147 expressly excludes exposure to electrical hazards from work on conductors and equipment in electric utilization installations at (a)(1)(ii)(C). On a construction site the electrical counterpart is 29 CFR 1926.417.

Then walk it down in steps of about 2 psi, holding each step for a full production cycle of the plant, typically a week for a shop whose work mix varies by day. Two psi is small enough that a genuine constraint announces itself as a single complaint at a single machine rather than as a plant-wide stoppage, and a full cycle is long enough that you have actually run the work that needs the pressure. Record the step, the date, and the complaint if one comes. When a step produces a complaint, stop, go back up one step, and go measure the header and the point of use at that machine, which is the gate above. That measurement, not the setpoint, tells you what to do next.

The record is the deliverable. A setpoint with a written reason beside it stops being raised by the next person who wants a quick fix, and that is worth more over ten years than the psi you recovered this week.

References

• 29 CFR 1910.169, air receivers, which requires an indicating pressure gauge and one or more spring-loaded safety valves on the receiver
• 29 CFR 1910.147, control of hazardous energy, for isolating and verifying depressurisation before a pressurised joint is broken, and 1910.333(b)(2) for the electrical work of changing a setpoint inside an enclosure (1926.417 is the construction electrical counterpart)
• ASME Boiler and Pressure Vessel Code Section VIII, in the edition adopted by your state's boiler and pressure vessel law, which binds the vessel manufacturer and reaches you through the stamped nameplate
• See related: Why Pipe Size Decides More Than Compressor Size; What Artificial Demand Is and Why Nobody Sees It; How to Put a Number on What Leaks Are Costing