How to Set Up a Logging Run That Answers a Question
Why this matters
Most logging runs fail the same way. Somebody hangs a logger, leaves it a week, comes back, downloads a file with thousands of rows, and finds that the file does not contain the answer. Not because the logger malfunctioned. Because the run was never designed to distinguish between the two things it was supposed to distinguish between, so the data is consistent with both and settles nothing. That is a week of customer patience and two trips spent on a file nobody can use.
A logging run is a piece of test equipment you build on site, and like any test it is designed backwards from the decision it has to support. The question comes first, the observation that would separate the candidate answers comes second, and the channel list, the interval and the duration fall out of those. Getting that order right is the whole job. Choosing a logger first and hoping the data speaks is the failure mode.
Build the run as a written plan
Write the plan before you touch the equipment. Eight fields, and every one of them is a decision you would otherwise make badly under time pressure on the roof.
Field 1: the decision. One sentence, in the form "if X, we do A; if Y, we do B." Not "find out why it trips." If you cannot name two actions that depend on the outcome, you are not running a test, you are collecting data, and you will collect the wrong data.
Field 2: the discriminating observation. What would you have to see to choose A over B? This is the field the whole run is built around. State it as a value, a duration and a coincidence: "a supply voltage below the equipment's stated minimum, lasting longer than the control's stated ride-through, in the same second as the trip."
Field 3: the channels. Everything the discriminating observation names, plus the one channel that proves the run was valid at all. That last one is usually the event itself, or a channel that tells you the machine was running. A logger that records perfect data through a week the equipment never operated is a null result nobody can interpret.
Field 4: the placement. Where each sensing element goes, in words a second tech could follow. Placement is where most logging runs quietly die: a temperature element in the airstream instead of on the pipe, a current clamp above a tap that feeds something else, a pressure port teed into a line that also serves a control.
Field 5: the interval. Set by the shortest event you must be able to see, never by how long you want the batteries to last. A companion article covers exactly what an interval hides; the short version is that an event shorter than the interval does not appear as a small event, it appears as nothing or as something that never happened, and the record cannot be repaired afterward.
Field 6: the duration and the stop rule. Long enough to capture the event at least three times, so a single occurrence cannot be a coincidence. Write the stop rule as a condition, not a date: "three trips captured, or fourteen days, whichever comes first."
Field 7: power and retrieval. How the logger stays powered for the full duration, how you get the data off, and what happens if the site loses power. A run that ends silently on day four and is discovered on day fourteen has cost you the ten days you could have been re-running it.
Field 8: the disqualifiers. What would make the run invalid, decided in advance. Somebody changed a setting. The equipment was off for four of the seven days. A sensing element moved. Deciding this before you see the data is the only defense against deciding it after you see the data, which is how a run gets talked into an answer it does not contain.
Safety belongs in the plan, not after it
Fitting a logger is an installation, and it creates hazards the reading itself does not.
If installing sensing elements means opening an energized enclosure, 29 CFR 1910.333(a)(1) requires the parts to be deenergized first unless the employer can demonstrate that deenergizing introduces additional or increased hazards or is infeasible due to equipment design or operational limitations. Where the run genuinely requires the machine energized, the approach boundary and the PPE come from the employer's electrical safety program, which most shops build on NFPA 70E in the edition their employer or the site has adopted (numbering here follows NFPA 70E-2021), binding through that program or the site contract rather than on its own. Inspect leads, cables, probes and connectors for external defects and damage before use, as 29 CFR 1910.334(c)(2) requires.
Breaking into a pressurized line to fit a gauge or a transducer is a stored-energy task under 29 CFR 1910.147: isolate, relieve the pressure through a bleed or vent to a controlled point, and verify zero at a gauge before the fitting comes loose, because a line that reads isolated at a valve can still be holding. For a hot line, let it cool or fit at a cooled point rather than at service temperature, and keep hands and face out of the plane of the joint while it breaks.
Then the hazard your own plan creates: a logger left inside an enclosure for two weeks becomes part of that enclosure. Route leads so the door cannot pinch them, keep the body and the leads clear of any rotating or reciprocating part and clear of combustion and flue surfaces, and confirm the logger's own environmental and location rating covers where you are leaving it. Tag it, on the outside, with the date, your shop and a callback number, so the next person to open that panel knows what it is and does not pull it.
A filled-in plan: nuisance trips on a commercial exhaust fan
The customer reports the fan tripping "a couple of times a week, mostly afternoons," always resettable. Two candidate causes are live: a supply voltage sag during a large motor start elsewhere in the building, which would be a fast event, or a slow thermal problem in the motor or the overload relay, which would build over tens of minutes.
The decision: if the trip coincides with a voltage sag, the finding goes back to the building's electrical service and the fan is not the problem. If the trip follows a slow current or temperature climb, the fan's mechanical load and the relay come under investigation.
The discriminating observation: a supply voltage excursion below the control's stated minimum, lasting longer than its stated ride-through, in the same second as the trip - versus a current rise over tens of minutes ahead of the trip with no voltage excursion.
The channels: supply voltage, motor current, and a contact channel on the starter that tells you when the fan is actually running. Three channels, and the third one is the validity channel.
The interval, and where the plan changes. The sag candidate is a sub-second event. To see it as an event rather than as a single stray sample you need several samples inside it, which puts the interval on the order of a tenth of a second. At 0.1 second, one week is 6,048,000 records per channel. At 1 minute it is 10,080. No general-purpose logger holds the first for a week, and the second cannot see the sag at all.
So the run splits, and this is the part that has to be decided in the plan rather than improvised: one instrument with a fast peak-and-excursion capture watching supply voltage, and a second slow channel at 1 minute carrying current and the run contact for the thermal candidate. Two instruments, one clock. Set them from the same time source before they go in, and write both start times down, because a two-minute offset between them is enough to break the coincidence test that the whole run exists to make.
The duration and the stop rule: at roughly two trips a week, seven days yields about two occurrences and fourteen days about four. Two is not enough to separate a pattern from a coincidence, so the stop rule reads "three trips captured, or fourteen days, whichever comes first."
Power and retrieval: both instruments on their own batteries rather than on the panel that may be the subject of the investigation, with a mid-run check at day seven. The day-seven visit is not optional. It is what catches a dead battery on day four while there is still time.
Disqualifiers: any change to the starter, the overload setting or the load; the fan not running for more than one of the days; a sensing element found moved at retrieval.
What the run returned. The day-seven check found two trips recorded, one short of the stop rule, so the instruments went back in and were retrieved on day eleven with four trips in the file. That is how a count-based stop rule actually works in the field: you cannot count occurrences without downloading, so the rule sets when you come back, and the final count is whatever it is when the rule is satisfied. In three of the four trips the fast instrument captured a voltage excursion within the same second. In the fourth, no excursion was captured, and the slow channel showed no current climb ahead of it either. That fourth event is not tidy, and reporting it as a fourth confirmation would be a lie: three of four is the finding, and the fourth is an unexplained event that either fell in a gap in the fast instrument's capture or has a different cause. It goes in the report as exactly that.
How to verify the run before you leave it
Do not walk away from a logger you have not seen produce a record. Three checks, on site, in about a quarter of an hour.
- Take one live reading with a handheld at each sensing point and compare it to what the logger is reporting at that moment. This catches a reversed clamp, a channel wired to the wrong input, and a scaling setting left from the last job. It is the only check that tests the whole chain from the sensing point through to the stored value.
- Force the validity channel to change state and confirm the record shows it. Cycle the equipment if the site permits it. A channel that never changes during the run is indistinguishable from a channel that was never connected.
- Download the first few minutes of data before you leave. Not to read it, to prove the file exists, the timestamps are the times you think they are, and the interval is the one you set rather than a default the instrument reverted to.
References
- 29 CFR 1910.333(a)(1), OSHA general industry, deenergizing before work on or near live parts and the demonstrations that permit energized work
- 29 CFR 1910.334(c)(2), OSHA general industry, visual inspection of test instruments, leads, probes and connectors before use
- 29 CFR 1910.147, OSHA general industry, control of hazardous energy for isolation and relief of stored pressure
- Manufacturer documentation for the logger's sampling limits, storage capacity, and environmental and location ratings
- See related: What a Sampling Interval Hides; What a Data Log Can Tell You That a Single Visit Can't