What a Time Delay Is Protecting

Why this matters

Every delay in a sequence is somebody's answer to a physical question. Somebody worked out how long a real process takes, added margin, and wrote a number. The number then travels forward through decades of service work as a bare setting with the question stripped off, which is how a delay ends up being shortened by a tech who found it annoying.

The claim: a delay is always buying time for a specific physical process to finish, and naming that process is what tells you whether the delay can be adjusted and what fails if it is shortened. The corollary is the operating rule: a delay whose physical process you cannot name is one you must not change.

Before you touch a timer setting

A delay protecting a combustion, pressurized, or stored-energy function is part of the equipment's listed safety function. Purge timing in particular is set by the appliance listing and the control's approval, not by field preference, and a service technician does not shorten one. If your analysis says the delay is too short for the process, the finding is that the process has degraded and needs restoring, not that the number needs raising.

For the measurements below: readings inside an enclosure mean de-energizing and lockout under 29 CFR 1910.333(b)(2), proved live-dead-live per NFPA 70E-2021, 120.5, and stored energy such as an accumulator, a spring-return actuator or a pressurized line isolated and relieved under 29 CFR 1910.147. Live readings for diagnosis fall under 29 CFR 1910.333(a)(1), which permits them where de-energizing would introduce additional or increased hazards or is infeasible for the equipment design, with a meter rated at or above the system voltage and leads inspected for damage per 29 CFR 1910.334(c). Take airflow readings from outside the swept path of any driven part with guards in place, and if the machine burns fuel, wear a personal CO monitor on your collar before it fires and stay clear of the flue and relief openings.

The five processes a delay is bought for

Every delay you meet in the field is buying time for one of these. Identify which, and everything else about the delay follows.

Volume turnover. Clearing the contents of an enclosed space by pushing a known number of volume changes through it. The governing physics is volume divided by flow, times the number of changes required. This is the delay class most likely to be safety-listed and least likely to be adjustable.

Pressure or level equalization. Letting a differential across a machine bleed away through whatever leak path exists, so the driver is not asked to break away against a load it was never rated for. The time constant is set by the size of that leak path and by the volumes on each side, which is why this number is published rather than derived in the field.

Thermal settling. Letting a mass at temperature move toward a condition the next step assumes. The slowest of the five by a wide margin, frequently minutes to tens of minutes, and the one most often under-served by a timer sized for something else.

Electrical recovery. Arc quench across separating contacts, dielectric recovery in a switching device, and coil dropout. Milliseconds to tens of milliseconds. Short enough that people forget it is a delay at all, and long enough to matter when two devices are supposed to transfer between each other.

Mechanical travel and settling. An actuator reaching its commanded position, a valve stroking, a rotating load coasting to standstill. Seconds to tens of seconds, and the one you can most easily measure yourself with a stopwatch.

A timer is not a permissive, and the difference decides everything

A timer asserts that enough time has passed. A permissive confirms that a condition is true. They are not interchangeable, and where a sequence uses one in place of the other it inherits a specific weakness.

A timer is correct where the process has a bounded, predictable duration under all conditions the machine will see, and where no available sensor reports completion directly. A permissive is correct where the duration varies with conditions, which is exactly the case for thermal settling and often for pressure equalization.

The practical consequence: any timer standing in for a variable process must be sized for the WORST case, which means it is unnecessarily long in every other case. That is where the pressure to shorten it comes from, and it is why the person feeling that pressure is usually looking at a mild day and reasoning from it.

The four timer shapes, and what each implies about intent

Reading the shape tells you the direction of the protection before you know anything else. An off-delay says the danger is in stopping too soon. A minimum-off says the danger is in starting too soon. Techs routinely misread the second as the first.

Deriving the requirement from the process, not from the setting

The useful move is to compute what the process actually needs and compare it against what is set, rather than reading the setting and assuming it was right.

For volume turnover the derivation is direct: required time equals the number of volume changes, times the enclosed volume, divided by the flow actually established during that step.

Three conditions attach to that formula and each one has bitten somebody:

• The flow term is the flow during THAT step, which is frequently not the run-condition flow. Where a damper sits at a purge position rather than fully open, the flow during purge is lower, sometimes considerably, and using the run figure overstates how fast the volume clears.
• The volume term is the whole space being cleared, including sections a casual look does not count.
• The multiple volume changes are not conservatism for its own sake. Real enclosures do not displace their contents cleanly: there are dead pockets and short-circuit paths, so some of the delivered flow leaves without having swept anything. The multiplier is the allowance for that imperfect mixing, which is precisely why it is specified rather than left to a single change.

Worked example: a purge that no longer clears what it was sized to clear

A machine whose documentation specifies four volume changes of a 60 cubic foot enclosed section, and whose purge timer is set at 36 seconds.

Reverse the setting to find the flow it assumed. Four changes of 60 cubic feet is 240 cubic feet to move. Moving 240 cubic feet in 36 seconds, which is 0.6 minutes, requires 240 / 0.6 = 400 cubic feet per minute. So the 36-second setting was sized against 400 cubic feet per minute of purge-condition flow.

Measured on this machine, with the damper at its purge position and the machine in its purge step: 300 cubic feet per minute.

Recompute the requirement at the actual flow: 4 x 60 / 300 = 0.8 minutes, which is 48 seconds.

The machine needs 48 seconds and is set at 36. The shortfall is 12 seconds against a 48-second requirement, which is 25 percent short of the specified clearance. Put the other way, the 36-second purge is delivering 36 / 48 = 0.75 of the required volume changes, so three changes where four were specified.

What that means and what it does not. It does not mean the timer should be changed to 48 seconds. The setting was correct for the machine as designed; the flow is what moved. Raising the timer to compensate for degraded flow hides a real fault behind a number, and it leaves the machine with no margin at all if the flow drops further, which it will if the cause is progressive. The finding is a 25 percent flow deficit at the purge position, and the repair is whatever caused it: a damper not reaching its purge position, added resistance downstream, a degraded air mover.

What would change the answer. If the measured flow had come back at or above 400 cubic feet per minute, the purge is meeting its specification and the complaint that brought you there is somewhere else entirely. If the volume figure in the documentation excludes a section that was added later, the requirement is higher than 48 seconds and the deficit is worse than computed, so confirm the volume corresponds to the machine as it exists now rather than as it shipped. And if the measurement was taken at run-condition damper position rather than purge position, the number is not comparable to the derivation at all and has to be retaken.

The failure mode. Measuring flow in the run condition, getting a healthy figure, and concluding the purge is fine. Run-condition and purge-condition flow are two different numbers on the same machine, and only one of them is in the derivation.

What a delay is not

It is not a fault detector. A timer that expires does not know why the process did not finish, only that time ran out, which is why a fault raised by an expired timer names the timer's own step and not the cause.

It is not a substitute for a permissive on a variable process, as above.

And it is not an accumulator of margin. Three techs each shortening a minimum-off time slightly, over years, each by an amount that felt harmless, consume the same margin as one large change, and nothing on the machine displays how much is left. That is the same failure that eats the separation between a control and a limit, and the sibling article on that separation carries the full stack.

References

• 29 CFR 1910.333(a)(1) and 1910.333(b)(2) for energized troubleshooting conditions and electrical de-energizing, 1910.334(c) for test instrument inspection
• 29 CFR 1910.147 for isolation of stored energy including pressurized and spring-loaded elements
• NFPA 70E-2021, 120.5 for live-dead-live verification
• Manufacturer and listing documentation for specified purge volume changes, enclosed volume, minimum off time, and whether a given delay is field-adjustable at all
• See related: Why Short Cycling Protection Looks Like a Fault; The Difference Between a Limit and a Control; What Proving Actually Means in a Sequence