Temperature & Humidity / Validation
Temperature Mapping Procedure for Stability Chambers with 9 15 27 Probe Points
A sensor grid that proves the whole volume, not the control point alone
Temperature mapping places a grid of calibrated sensors through a chamber’s working space and logs them together for a day or more, to prove that every corner of the volume holds the setpoint, not just the spot where the control probe happens to sit. The map locates the hot and cold pockets, fixes where the monitoring sensor belongs, then marks any zone too far out of band to store product in.
Why one sensor is not enough
The chamber controls itself from a single probe, usually sitting in the return airflow where the conditions are steadiest. That probe reads a clean number. It says nothing about the corner by the door, the shelf above the cooling coil, the dead spot a loaded rack creates downstream of the fan.
Air inside an insulated box does not hold one temperature everywhere. It stratifies, it short-circuits back to the return before reaching a far corner, it pools cold where the supply dumps in. A gradient of a degree or two across the working space is ordinary. On a stability study run to a two-degree tolerance, a degree or two is the whole budget.
Mapping is the test that finds those differences before a regulator or a ruined batch does. It reads the chamber as a volume to be sampled point by point.
One trusted number is not a map.
The protocol, written first
A mapping study that means anything is planned on paper before a single sensor goes in. The plan is approved before the run starts.
The protocol fixes the things that would otherwise be chosen to flatter the result: how many sensors, where each one sits, how long the run lasts, what counts as a pass. It names the calibrated loggers by serial number, draws the grid as a diagram, sets the duration, then writes the acceptance criteria as numbers the data will be judged against. None of it is decided after the fact.
This order is what makes the map evidence under good practice. A criterion set before the run cannot be bent to fit a borderline result; a placement drawn in advance cannot quietly skip the corner that was going to fail. The protocol, approved and signed, is the promise that the study was run to find the truth and not to pass.
When the run is done, the data is checked against that protocol line by line. Any sensor or moment outside the band becomes a deviation, logged and investigated. The discipline is the one the whole stability programme runs on: decide the rule first, then live by what the data says.
Nine, fifteen, twenty-seven
The probe count is a sampling decision. The common steps are nine, fifteen and twenty-seven.
Nine sensors set the floor for a modest chamber: one at each of the eight corners of the working space, one at the centre, the classic corner-and-core cube. A larger box needs a finer net, so the count climbs to fifteen, then to twenty-seven, the full three-by-three-by-three grid with a sensor at every corner, every edge midpoint, every face centre, the middle. The reasoning is plain. Temperature is uneven inside a box driven by one airflow, so a warm pocket can hide in the gap between two sensors. Widen the spacing and a gradient slips through unseen; tighten it and the map catches the pocket near the door seal or above the coil before product ever sits there. The count rises with volume because the grid has to stay fine enough, in real distance, to resolve the spots that matter. A reach-in might map at nine, a double-door at fifteen, a walk-in at twenty-seven, with the largest rooms adding a sensor for every few cubic metres beyond that corner-and-centre minimum, so the finest grid always follows the biggest box and no pocket of it goes unsampled. Below that minimum a map gambles that the unsampled gaps behave like their neighbours, a gamble a regulated study will not take, so the grid is sized to leave no gap wide enough to hide a pocket that matters; the cost of an extra logger counts for nothing against a cold corner found a year into a study, with the whole run thrown into doubt behind it.
The figures are not arbitrary ritual. Guidance from the WHO technical reports and the ISPE good-practice guides, alongside the chamber-performance method in IEC 60068-3-5, all point the same way: enough sensors, spread through the real volume, to map the distribution across the whole space.
One sensor among the grid stays beside the chamber’s own control probe, so the map can say how far the number the chamber trusts sits from the worst place in the box.
Choosing the sensor count
Nine, fifteen and twenty-seven are the common counts, with the guidance behind them tying the number to the volume, sized up as the box grows.
A small chamber of a few hundred litres is well served by nine, the eight corners and the centre catching the spread a box that size develops. A larger reach-in or a double-door earns fifteen, the extra sensors filling the long faces a bigger volume stretches out. A walk-in moves to twenty-seven, the full grid; a true room keeps adding from there, a common rule putting roughly one sensor in every cubic metre or two on top of the corner-and-centre frame. The published guidance, from the WHO technical reports and the ISPE practice guides, frames the counts this way so that two labs mapping the same chamber reach for the same grid. A regulator reading two studies can then weigh them on equal terms, sensor for sensor.
The logic is resolution. Two sensors a metre apart cannot see a pocket that forms between them, so as the box grows the grid has to grow with it to keep the spacing fine enough to catch a real gradient. A study that mapped a walk-in on a small chamber’s nine sensors would pass it on a grid too coarse to find the cold corner that ruins a batch. The number is a floor to build on, never a ceiling. Where a chamber is known to be awkward, an extra sensor goes wherever judgment says a pocket might hide. A careful study adds where in doubt, since a sensor costs little against a missed gradient that fails a study a year in.
What reads the temperature
The grid is only as good as the sensors hung in it. Mapping usually uses small self-contained data loggers, each a probe with its own memory, dropped into place and collected at the end; the alternative wires thermocouples or platinum-resistance probes out to a central recorder. Wired resistance probes give the tightest accuracy, while wireless loggers give a clean install with no cable bundle forced through the door seal.
Whatever the type, a sensor’s accuracy has to be a fraction of the tolerance it judges. A two-degree band is no place for a probe good only to a degree, so mapping sensors are picked to resolve a tenth or two, fine enough that the spread they report is the chamber’s own.
That accuracy traces back to a national measurement standard through an unbroken chain, the link that lets one site’s map be compared to another’s, or to last year’s. A logger reading a tenth of a degree high across the board would lift the whole map and bury a real cold spot, so the chain is checked, documented, then carried into the report beside the data.
Where the probes go
The grid is not scattered at random. Sensors go where temperature is known to misbehave, with the corner-and-centre cube as the skeleton and the suspect spots filled in around it.
Corners and walls come first. Close to the insulated skin, far from the supply airflow, the extremes of the box live there.
The door plane gets its own sensor. Every opening floods it with room air. It is the slowest part of the chamber to recover afterward.
So does the region right off the cooling coil or the heater, where the supply air is at its coldest or hottest before it has mixed into the room. A shelf parked there sees a harsher condition than the setpoint reads.
The last one shadows the control and monitoring probe, the reference that ties the whole map back to the single point the chamber runs from day to day.
The worst spot wins
Once the map is read, the monitoring sensor for daily use is moved to the hottest or coldest spot the mapping found. The chamber is then watched at its weakest point, so the day the worst corner drifts out of band is the day the alarm sounds.
Where mapping sits in qualification
Mapping is one step in a larger qualification; knowing where it fits keeps it in proportion.
A chamber is qualified in stages: that it was installed right, that it operates across its range, that it performs in real use. Mapping is the evidence behind the operating and performance stages, the data that proves the box holds its condition through the working volume before any product is trusted to it. The formal naming of those stages is its own discipline; mapping is the measurement that feeds them.
It runs after the chamber is installed and calibrated, before product goes in. An empty map proves the equipment; a loaded map, often run as part of the performance stage, proves it under the real conditions of use. Only when both pass does the chamber carry product on a study a regulator will read. So the map is not paperwork for its own sake. It is the one piece of the qualification that directly measures the air the product will sit in. Every later claim about the chamber stands on it.
Empty, then loaded
An empty chamber maps the equipment. With nothing inside, the airflow is unobstructed and the result describes what the box can do on its own, which is the right baseline for comparing one chamber to another or one service interval to the next.
A loaded map describes real use. Product and racks block airflow, add thermal mass, then create the dead spots an empty box never shows, so a chamber that passes empty can still hide a cold pocket once it is full.
Serious qualification runs both. The empty map proves the equipment; the loaded map proves the way the site fills it in service.
Loading for the worst case
A loaded map is only as honest as the load it runs, so the load is chosen to be the hardest the chamber will ever face.
The worst case for airflow is a full, dense load, racks packed close so the air struggles to reach the middle, the configuration that opens the widest gradient. Some studies also map a minimum load, a near-empty box, because a sparse load behaves differently again, the air short-circuiting with little mass to even it out. Mapping both ends brackets the real operating range. A study that proves the two extremes covers everything that lies between them.
The load is not boxes alone. It mimics the product: the same containers, the same packing density, the same thermal mass, since a tray of dense glass vials loads the air differently from a few light cartons. A map run on an unrealistic load proves little about the real one.
This is why the loaded map carries more weight than the empty one for daily use. The empty map describes the equipment; the loaded map describes the chamber doing the job it was bought for, with the product fighting the airflow the way it will every day the study runs, which is the condition the product itself will face on the shelf. A study that mapped only an empty box, then filled the chamber to the brim in service, would have proven the easy case and shipped the hard one untested.
A full day, at least
A map runs for 24 hours as a floor, often longer. The point is to catch the chamber through a complete span of its own behaviour: the compressor cycling on and off, an automatic defrost dumping a brief warm pulse into the space, the slow drift between a cool night and a warm afternoon in the room outside.
A snapshot would miss all of it. Only a record long enough to cover every cycle shows whether the worst excursion of the day still lands inside the tolerance, which is the number that decides a pass.
Open the door, cut the power
Two deliberate upsets usually ride along with the steady-state map. The door is held open for a set time to measure how far the conditions fall and how long they take to recover, the test that mirrors a real loading.
Power is then cut, to see how long the loaded chamber holds inside its limits with no help from the system. That recovery and that hold time tell a site how much of an excursion a real outage would cost. They show how long the site has to react before the product slips its band and the run is at risk. A heavy, well-insulated chamber can hold its band for many minutes after the power drops; a small one with little thermal mass leaves the band in a fraction of that. The figure that comes back is exactly what a contingency plan is built on.
When the map fails
A map does not always pass, and what happens then is part of the procedure.
If a corner reads out of band, the first move is to understand why. A blocked airflow, a worn door seal, a fan losing speed, a load packed too tight: each has a fix, with a re-map after it to prove the fix worked. The aim is to bring the whole working space inside the tolerance.
Where the cold corner cannot be cured, the volume is derated. The failing zone, the shelf right at the door or hard against the coil, is fenced off and marked as not for product; the qualified volume shrinks to the part that holds the band. A chamber can ship with a usable working space smaller than its physical one, the difference written into the report. What a study cannot do is pass a chamber by ignoring the spot that failed. The map records every sensor, the failing one included. A zone outside the band is either fixed, fenced off, or reason to call the chamber unfit. There is no fourth option that keeps the data honest.
How often to re-map
A map is a snapshot of a chamber on the days it ran. A chamber changes, so the map has a shelf life of its own.
The usual rhythm is a re-map on a set period, often yearly, plus a re-map on change: after a repair to the refrigeration or the airflow, after the chamber is moved, after a shift in how it is loaded. A compressor swapped, a fan replaced, a new shelving layout each can move the pockets the first map found.
Between full re-maps, the daily monitoring at the worst spot stands watch. That single sensor, placed where the mapping found the chamber weakest, catches a drift as it starts; a creeping trend is itself a signal that a fresh map is due before the schedule calls for one. The periodic re-map keeps the qualified volume honest over the years. A chamber that mapped clean at install can develop a cold corner as its door seal ages or its coil fouls. Only a repeat study, not a hopeful assumption, proves it still holds. The cost of a re-map is small against the cost of finding, a year into a study, that the qualified volume quietly stopped being qualified somewhere along the way.
Humidity mapping too
A stability chamber controls humidity as well as temperature, so the map covers both.
Humidity sensors join the grid, reading relative humidity at the same points the temperature probes do, the working space judged against the humidity tolerance, commonly five percent. The two are linked, since relative humidity depends on temperature, so a cold corner reads damp and a warm one dry even when the water in the air is even.
Humidity is the harder of the two to hold across a volume. The moisture has to be generated, carried by the air, then kept from condensing on a cold surface, so the damp corners and the spots near the humidifier or the coil are watched with extra care. A chamber tight on temperature can still wander on humidity in a far corner.
For a study run at 30 degrees and 75 percent, the humidity map is as much the qualification as the temperature one. A product sold on a moisture claim was tested at a humidity the chamber has to prove it held, evenly, everywhere a sample sat.
Mapping a warehouse, a truck, a fridge
The same logic reaches well past a stability chamber. Anywhere a regulated product is held or moved at a controlled temperature gets mapped the same way.
A warehouse holding finished drug is mapped with sensors spread through its volume, more of them and over a longer span than a chamber needs, because a building carries bigger gradients: sun on one wall, a loading door open to the weather, warm air stratified under the roof. The map runs across days to catch the daily and seasonal swing. It finds the hot aisle near the dock or the cold corner under a vent.
A refrigerated truck or a shipping container is mapped in motion, with the doors opening at stops, to prove the cold chain holds from the warehouse to the pharmacy. A pharmacy fridge, small as it is, gets a lighter version of the same study, since a vaccine ruined by a warm shelf is ruined whether the shelf was in a chamber or a clinic.
In every case the question is the one the chamber map asks: does every part of the space hold the condition the product needs, or is there a pocket where the label’s promise quietly fails? Good distribution practice rests on the answer.
Reading the map
The analysis splits the logged data two ways. Uniformity is the spread across the sensors at one instant, the hottest minus the coldest, the spatial picture. Stability is the wobble at a single sensor over time, the temporal picture. A chamber can be tight on one and loose on the other, so both get reported against the tolerance.
Every sensor carries a calibration. Each logger is checked against a traceable reference before the run and again after; a probe that has drifted between the two has its data weighted or thrown out, since a map is only as honest as the sensors that drew it.
A pass is every sensor staying inside the band for the whole run, defrost pulses and door tests included. What comes out is a qualified working volume, the part of the box certified for product, with the door plane or the coil shelf fenced off if they could not hold the line. That map is what the daily monitoring, the loading plan, the next requalification all build on, a single document that fixes where the chamber may be trusted and where it may not for as long as it stays in service.
The report the map becomes
The output of all this is a document, and the document is what a chamber is finally sold and audited on.
The report gathers the protocol, the calibration records, the placement diagram, the logged traces from every sensor, the analysis of uniformity and stability against the tolerance. It states the qualified working volume, marks any fenced-off zone, names the spot chosen for daily monitoring, then records the open-door and power-loss results. It is signed, dated, and kept for as long as the chamber is in service and well after. Years on, when a batch made today is questioned, that report is what proves the chamber it was stored in held its condition, so the record outlives the run that made it by a wide margin and has to stay legible the whole time.
An auditor reading it should be able to reconstruct the whole study without having been there: which sensors, calibrated when and against what, placed where, reading what, judged how. A map whose report cannot be followed that way is a map that did not happen, whatever the chamber did on the day.
That is the plain fact of mapping. The measurement matters as much as anything in the qualification, and the record of the measurement matters as much, because a condition no one can later prove was held is, to a regulator, a condition that was not.
Questions on temperature mapping
How many probes does a mapping study need?
Common practice steps through nine, fifteen and twenty-seven: nine for a small chamber as eight corners plus the centre, twenty-seven for the full three-by-three-by-three grid of a large one, fifteen in between. The largest walk-ins add a sensor for every few cubic metres above that minimum. The count rises with volume so the grid stays fine enough to catch a local pocket.
What is the difference between uniformity and stability?
Uniformity is the spatial spread, the hottest sensor minus the coldest at one moment across the working space. Stability is the temporal fluctuation at a single point over time. Mapping reports both against the chamber’s tolerance, since a chamber can be even in space but unsteady in time, or the reverse.
How long should a mapping run last?
At least 24 hours, often longer, to capture a full span of the chamber’s behaviour: compressor cycling, any automatic defrost, and the day-to-night swing of the surrounding room. A shorter snapshot can miss the worst excursion of the cycle, which is the value that decides whether the run passes.
Should the chamber be mapped empty or loaded?
Both, in a thorough qualification. An empty map characterises the equipment as a baseline; a loaded map shows what happens once product and racks block the airflow and add thermal mass. A chamber that passes empty can still hide a cold pocket when it is full, so the loaded map reflects real use.
How often should a chamber be re-mapped?
On a set period, often yearly, and on change: after a repair to the refrigeration or airflow, after the chamber is moved, or after a shift in loading. Between full re-maps the daily monitor at the worst spot stands watch, and a creeping trend there can call for a fresh map before the schedule does.
What happens if a mapping study fails?
The cause is found and fixed, then the chamber is re-mapped to prove the fix. Where a corner cannot be cured, that zone is fenced off and the qualified working volume shrinks to the part that holds the band. A failing zone is fixed, fenced off, or reason to reject the chamber, never ignored.
Where is the monitoring probe placed after mapping?
At the hottest or coldest spot the map found, so the chamber is watched at its weakest point. Monitoring the worst location means an alarm fires when that spot drifts out of band, well before the problem reaches the calmer place where the control probe sits.
What standards guide chamber mapping?
There is no single global standard, but WHO technical reports, ISPE good-practice guides, EU good-distribution-practice guidance, and the chamber-performance method in IEC 60068-3-5 all inform it. They converge on enough calibrated sensors, spread through the real volume, held long enough to describe the distribution across the whole space.