What a Relief Valve Does in a Hydraulic Circuit and What It Costs

Why this matters

A relief valve has one job, which is to give pump flow somewhere to go before pressure destroys something. What almost nobody reads off the dial is that the pressure a relief valve actually holds depends on how much flow is going through it at the time. The gap between the pressure where it first opens and the pressure it needs to pass the pump's whole output is called override, and it is a property of the valve's construction rather than of your adjustment. Override is what decides the real peak pressure every hose and fitting in that circuit sees, and it is why a machine can chew through hose assemblies while its gauge reads exactly what the drawing says.

The valve's actual job, and what it is not

It is protection. It is set above the highest pressure the circuit needs to do its work, and in normal operation it should pass nothing. Every gallon that crosses it is a gallon that did no work and became heat in the oil, which the card on what a hydraulic system is trading away works through.

It is not a control. Turning a relief up does not make a machine push harder against a load that only needs 900 psi, because pressure is set by the load, not by the valve. That is the sibling card on what pressure and flow each decide, and it is worth reading first if the distinction is not already automatic.

Cracking pressure, full-flow pressure, and the number in between

Cracking pressure is where the valve first begins to pass oil. Full-flow pressure is the pressure it reaches when the entire pump output is going through it. Between them the valve is partly open, passing part of the flow.

Override is the difference, and it is quoted here as a percentage of cracking pressure. It is not an error band and it is not a spread; it is a deterministic point on the valve's own flow-versus-pressure curve, so the same valve at the same flow gives the same number every time.

Direct-acting relief valves put a spring straight on the poppet. To pass more flow the poppet must lift further, which compresses the spring further, which raises the pressure. Override is large. Illustrative for a valve of the size used below: cracks at 2,000 psi, needs 2,600 psi to pass 8 gpm, an override of 600 psi or 30 percent of crack.

Pilot-operated relief valves use a small pilot valve to control a large main stage, so the main poppet can open a long way without the pilot spring moving much. Override is much flatter. Illustrative for the same duty: cracks at 1,950 psi, passes 8 gpm at 2,080 psi, an override of 130 psi or under 7 percent. The trade is a slower response to a fast spike, and a tendency to chatter at very low bypass flow.

Both sets of figures are illustrative. The valve's own published flow-versus-pressure curve owns the real ones, and it is the document to ask for.

Three costs, and the order they bite

The pressure everything sees. The circuit's genuine peak is full-flow pressure, not the dial. If the relief is set by cracking pressure and the machine ever dead-heads, every component is exposed to crack plus override.

Flow lost below the dial. A direct-acting valve begins passing oil at cracking pressure, so a circuit whose working pressure creeps up toward the setting starts bleeding a slice of flow over the relief continuously, with no dramatic symptom. The machine gets slower and the oil gets warmer and the gauge looks normal. This is the cost that is hardest to catch, and it is the argument for setting the relief with real headroom above the highest genuine working pressure rather than just above it.

Heat. Flow across a relief does no work, so the whole of it converts to heat right at the valve. The line downstream of a relief that is passing runs hot, which is a free field test: on a circuit that should never touch its relief, a relief return line at oil temperature is right and a warm one is a finding.

Setting one without lying to yourself

Set it at flow, not at crack. The number you want to control is full-flow pressure, so the valve is adjusted while the circuit is actually dead-heading the pump through it, and the gauge is read then. A relief set on a bench at low flow, or set by listening for the crack, has set the wrong end of the curve.

Never set a relief above the lowest pressure rating in the circuit. Walk the rating of every hose assembly, fitting, cylinder and accessory in the loop and set below the smallest one. The related card on why you never test a relief valve to prove a point covers the other half of that discipline.

Adjust with the machine running only where the valve is designed for it, with the locknut backed off and the cap left in place. Do not remove an adjustment cap or a spring cover on a pressurized valve; the spring is under load and the port behind it is at system pressure. Stand out of line of the adjuster, with eye protection on, and make small changes.

The case: hoses at the pump outlet, every four months

The signal. One clamp machine has been eating hose assemblies at the pump outlet on roughly a four-month cycle. Two identical machines in the same shop, same duty, same hose lot, run for years on the same assemblies. The failures are ruptures along the length, not abrasion at a bracket.

Hypothesis one, a bad hose batch. Killed immediately: the assemblies came from the same lot as the two machines that are fine, and were made up on the same crimper on the same day.

Hypothesis two, routing and abrasion. Killed on inspection: the failed assemblies carry no chafe marks or flattening, the routing is identical to the healthy machines, and the bend radii are the same.

Hypothesis three, a shift spike. Plausible on a machine that shifts a directional valve under pressure. Set aside for the moment because it predicts an occasional transient, and this machine's exposure turns out to be continuous.

What nobody had looked at. The cycle is 5.5 seconds of extend, 40 seconds of clamp hold with the directional valve centered on a closed center, and 4.1 seconds of retract, so about 50 seconds and 72 cycles an hour. During those 40 seconds the pump has no path through the valve, so its whole 8 gpm crosses the relief. Everyone who had put a gauge on this machine had watched it during the stroke, when it reads what the load needs, and walked away before the hold.

The reading during the hold. 2,600 psi, on a direct-acting relief whose dial had been set to 2,000 psi during a bench check at low flow. That is the override showing up: 2,000 psi of cracking pressure plus 600 psi to pass 8 gpm.

What the gauge could and could not prove. The gauge was a 0 to 5,000 psi bourdon-tube dial with an accuracy class quoted as plus or minus 2 percent of full scale. That is a percent-of-span basis, not percent of reading, so it is a fixed band of about 100 psi anywhere on the dial: at 2,600 psi that is under 4 percent of the reading, at 500 psi it would be 20 percent. Within one gauge over one session that band behaves as a systematic offset, so the 600 psi gap between the low-flow setting and the full-flow reading is solid even though neither absolute value is better than about plus or minus 100 psi.

That matters here, because the conclusion rests on an absolute. The hose assemblies are marked 2,500 psi maximum working pressure, a rating that comes from the hose construction standard the manufacturer builds to and reaches you through the assembly's own marking and your purchase specification rather than through any regulation. A reading of 2,600 psi plus or minus 100 psi cannot cleanly resolve whether the machine is above 2,500 psi or exactly at it. Either answer is a finding: hose ratings are a ceiling, and this machine was sitting at or above that ceiling for 40 seconds out of every 50, which is 80 percent of its running life.

The heat, corrected. The card on what a hydraulic system trades away computes this machine's hold heat at the nominal 2,000 psi setting and gets about 19,000 BTU per hour from the hold alone. Override raises it in the same proportion as the pressure: 2,600 divided by 2,000 is 1.30, so the real figure is about 24,700 BTU per hour. Which is also why this machine had been given a cooler two summers earlier and the other two had not.

The fix, and the better fix. Replacing the direct-acting relief with a pilot-operated one of the same rating brings full-flow pressure to about 2,080 psi, comfortably under the hose marking, and drops the hold heat to about 9.7 hydraulic horsepower from 12.1. That is a real improvement and it is still paying 9.7 horsepower to hold a clamp.

The better fix is the one the two healthy machines already had, and it explains why they never showed the problem: their directional valves use a tandem center, which routes pump flow back to tank at low pressure during the hold. Their relief valves have never passed full flow in service, so their override has never been visible and never mattered. That change is covered in the card on what a directional control valve is doing in each position.

Why hypothesis three was worth keeping. A shift spike is real on this machine and probably contributed, but it could not be the primary cause, because the two healthy machines shift the same way and the failing one differs only in what it does between shifts. When one machine in a set of identical machines fails, look for what is different about that one, not for what is wrong with the design they share.

How to verify override on a machine in front of you

• Read the gauge during the condition where the pump has nowhere to go, not during the stroke. On many machines that is the hold, on some it is end of travel, on a few it is a deliberate stall test the manufacturer permits.
• Compare that reading to the dial setting. A gap of a few percent says pilot-operated. A gap of tens of percent says direct-acting, and the number you must design around is the higher one.
• Feel the relief return line during normal operation with the back of a hand, briefly, and only where it is not hot enough to burn. At oil temperature, the valve is not passing. Warmer than the rest of the return, it is, and the circuit is losing flow you have not accounted for.
• Use a gauge whose full scale is roughly twice the pressure you expect to read. A percent-of-span accuracy class is a fixed band in psi, so a gauge running at the bottom of its dial carries an error that is a large fraction of the reading.

References

• Hose assembly manufacturer documentation and the assembly's own marking for maximum working pressure and the construction standard it is built to, which bind through that marking and your purchase specification rather than through regulation
• Valve manufacturer published flow-versus-pressure curve, which owns cracking pressure, full-flow pressure and override for the specific valve
• Pressure gauge accuracy class as marked on the instrument, which states whether the specification is percent of span or percent of reading
• See related: What Pressure and Flow Each Decide in a Hydraulic Circuit; What a Hydraulic System Is Actually Trading Away; What a Directional Control Valve Is Doing in Each Position; Why You Never Test a Relief Valve to Prove a Point