As an Amazon Associate, Akermin earns from qualifying purchases. Product links on this page are affiliate links.
A temperature data logger is a sensor, an analogue to digital converter, a clock, a block of non volatile memory and a battery, run by a microcontroller that spends almost all of its life asleep. On each tick of the clock it wakes, powers the sensor, takes a conversion, applies its stored calibration coefficients, writes the value into memory, updates its minimum and maximum, checks any alarm thresholds and goes back to sleep. You later read the accumulated values out over USB, Bluetooth or a dock. Everything that makes the resulting record trustworthy or useless is decided in how that cycle is configured, not in the sensor.
The five parts inside a logging instrument
Strip the housing away and there are five functional blocks, and each one sets a limit on the record.
- The sensing element. Either a digital sensor that returns a number over a serial bus, or a thermistor, platinum resistance element or thermocouple that needs excitation and conditioning. The element sets the measurement range and the intrinsic tolerance.
- The converter and controller. Turns the analogue signal into counts and applies the stored coefficients that translate counts into degrees. This is where averaging of several conversions happens, and it is why a logger can be quieter than a bare sensor.
- The clock. A low power real time oscillator. Its accuracy sets how well the horizontal axis of your record corresponds to actual time, which matters enormously when you line a logger up against another record.
- Non volatile memory. Flash or EEPROM that survives a battery change. Its size and the number of bytes per reading set how many readings fit, which is the arithmetic in the next section but one.
- The power source. Usually a coin or AA cell. It is the ceiling on deployment length once memory stops being the limit, and it is why interval choice is a power decision as well as a resolution decision.
The pillar page on what an environmental data logger is covers how the same five blocks are arranged for humidity, light, carbon dioxide and multi channel instruments.
What happens during one logging interval
The awake window is typically a few milliseconds to a few tens of milliseconds, and the sequence inside it is always roughly the same. A clock interrupt pulls the controller out of its deepest sleep state. Power is applied to the sensor and a settling delay elapses, because a sensor that has just been energised has not stabilised. One or several conversions are taken and averaged. Stored calibration coefficients convert the raw result into engineering units. The value is written to the next free memory location. Running minimum, maximum and average registers are updated. Alarm thresholds are compared. Then the controller returns to sleep and does nothing at all until the next tick.
Two details in that sequence are worth knowing about. The settling delay is a real contributor to accuracy, and it is the reason a logger that samples far more often than it records is often better than one that does not: the record is an average of several conversions rather than a single one. And nothing in the sequence writes a timestamp, because most designs store only the start time and the interval. The HOBO temperature and humidity USB logger states selectable intervals from one second to eighteen hours and a one year battery life, which brackets everything from a short thermal study to a season of ambient monitoring, and its bullets also state that software is required to work with it, so the record is not simply a file you copy off the device.
Because times are reconstructed rather than stored, the clock’s error accumulates down the record. An oscillator losing two seconds a day has shifted the record by a minute after a month and by six minutes after half a year. On its own that rarely matters. It matters a great deal the moment you try to line one logger up against another, or against a building management record, to work out which of two events came first.
Temperature makes it worse, because the oscillators used in low power designs drift with temperature themselves, and a logger in a freezer or an engine bay is running its clock well outside the conditions where it is best behaved. Two habits defuse this: synchronise the clock from the host immediately before deployment, and create a deliberate marker you can recognise in the data, such as a brief known excursion at the start, so the two records can be aligned on content rather than on trust. When several instruments have to agree, a single multi channel logger avoids the problem entirely by sharing one clock, which is the main argument in multi channel temperature loggers.
Memory arithmetic sets how long you can leave it
Deployment length is a division sum, and it is worth doing before you walk away from an instrument rather than after. The HOBO MX1101 temperature and relative humidity logger publishes the two numbers you need: 128 KB of storage holding up to 84,650 measurements. Because it records temperature and relative humidity, each interval consumes two measurements, which gives 42,325 intervals of capacity. From there the arithmetic is straightforward.
| Logging interval | Intervals available on two channels | Deployment length | Practical limit |
|---|---|---|---|
| 1 minute | 42,325 | About 29 days | Memory |
| 2 minutes | 42,325 | About 59 days | Memory |
| 5 minutes | 42,325 | About 147 days | Memory |
| 10 minutes | 42,325 | About 294 days | Memory and battery together |
| 15 minutes | 42,325 | About 441 days | Battery |
| 30 minutes | 42,325 | About 882 days | Battery |
Read the right hand column as the real constraint. Below ten minutes you run out of memory first and the interval is the lever. Above it you run out of battery first, so lengthening the interval further buys very little and costs you resolution on fast events. The same logger states a measurement span of minus 20 to 70 degrees C with a tolerance of plus or minus 0.21 degrees C in standard conditions, and 1 to 90 percent relative humidity with a typical tolerance of plus or minus 2 percent, which is the other half of the planning decision: there is no point logging every minute to chase a change smaller than the instrument can resolve. The selection walkthrough in choosing a temperature and humidity logger works through matching interval, capacity and tolerance to a job.
Wrap or stop when the memory fills
When the last free location is used, a logger does one of two things, and you choose which at setup. Stop on full freezes the record: everything from the start is intact and everything after the memory filled is simply absent. Wrap around keeps writing, overwriting the oldest readings, so you always have the most recent window and never the beginning.
Neither is the safe default. If you are documenting a shipment and the question is whether the cold chain held throughout, stopping on full is a disaster because the end of the journey is missing. If you are monitoring a running process and the question is what it is doing now, wrapping is correct and the first week is of no interest. The failure mode is choosing without thinking, discovering six weeks later that the mode did not match the question, and having no way to recover the readings that were overwritten.
One more wrinkle: a wrapping logger with an implied timestamp still needs its start time and interval intact to reconstruct the time axis, so a battery change or a reset partway through can leave you with values and no reliable times. Checking the record immediately after retrieval, rather than months later, is the cheap insurance here. The HOBO data logger page walks through what retrieval looks like on one of these instruments in practice.
Reading the data out of the device
Retrieval is where otherwise good records get damaged, because every path out of the device can resample or reformat on the way.
- Direct USB. Either a mass storage device presenting a file, or a proprietary link requiring the manufacturer’s software. The first is self sufficient; the second ties the record to an application you have to keep working.
- Wireless to a phone. Convenient, and worth checking whether the app shows a decimated chart or exports the full record. A chart image is not data.
- Probe and mounting considerations. The Elitech GSP-6 dual probe logger states a separable probe that removes the need to precool the device after each move, and a built in magnet for mounting directly on a refrigerator. Both are thermal lag features: they let the sensing point sit in the controlled space while the body sits outside it, which changes what the record is about.
- Record integrity requirements. The same listing states that it is designed to meet CDC and VFC requirements and FDA 21 CFR Part 11 compliance. That is about audit trails and controls on altering stored data, not about sensor quality, and it is a separate axis from tolerance.
- A rolling chart is not an export. The Autopilot desktop monitor and data logger states a digital memory chart over an adjustable 24 hour or 7 day period with minute, hour, day and week views. That is a display window rather than a retrievable record, and an export path is not stated in the listing bullets. For a monitoring job that is fine; for a record somebody else will read, it is the wrong instrument.
If you want retrieval without visiting the instrument at all, Bluetooth temperature loggers trade a little battery life for being readable from the doorway, and the same questions about what the app exports apply with extra force.
Setting one up so the record holds up
Six decisions, made before the logger goes anywhere, determine whether the record answers the question later.
- Pick the interval from the fastest change you need to see, then check the memory arithmetic and the battery life against the intended deployment length.
- Choose wrap or stop deliberately, based on whether the event you care about is at the start or the end.
- Synchronise the clock from the host at setup, every time.
- Place the sensing point where the question is, not where the device is convenient, and use a remote or separable probe when those differ.
- Start logging after placement, or plan to discard the opening part of the record while the instrument equilibrates.
- Retrieve and open the record once early in the deployment, so that a configuration mistake costs you a day rather than a season.
Frequently Asked Questions
Does a logger store a timestamp with every reading?
Usually not. Most store a start time and a fixed interval, then reconstruct each reading time as the start plus the number of intervals elapsed. That is why changing the interval mid deployment is often blocked, and why a clock that gains or loses a few seconds a day shifts the later part of the record.
What happens when the memory fills up?
One of two things, and you choose which at setup. Stop on full preserves the beginning of the record and loses everything after the memory ran out. Wrap around keeps the most recent window and overwrites the oldest readings. Neither is safer in general; it depends on whether the event you care about is at the start or the end.
How does interval choice affect battery life?
A logger spends almost all its time asleep, so average current is dominated by sleep current plus a small awake cost per sample. Halving the interval roughly doubles the sampling contribution but leaves the sleep contribution unchanged, so at long intervals battery life is set by the sleep current and barely responds to interval at all.
Why are the first readings in a record always suspect?
Because the logger reports the temperature of its own sensor, and that sensor has to reach the temperature of its surroundings first. Thermal mass and housing insulation mean the opening part of any record describes the logger equilibrating rather than the environment, which is why you discard it or start logging after placement.
Is a rolling trend chart the same as a data logger?
No. A rolling chart keeps a fixed recent window for display and discards what falls off the back. A logger keeps a record you can retrieve as data. If you need to hand the record to somebody else, the question to ask of a listing is whether an export exists and in what format.
What does compliance mean on a logger specification?
It refers to record integrity rather than sensor quality, covering audit trails, electronic signature handling and controls on altering stored data. A logger can be accurate and not compliant, or compliant and badly placed, so it is a separate axis from measurement performance.