Cooling faults appear as output drift weeks before they appear as a shutdown, so manage ducts, filters and coolant on a schedule. On a water-cooled source the chiller loop runs between −10 and 80 °C; when the displayed coolant temperature or the airflow stops matching the setpoint, the optical output follows it down. Inspect on an interval, not on an alarm.
Thermal faults degrade the measurement long before they stop the instrument, which is why they are the most commonly missed cause of drift. A loaded filter cuts airflow, the arc envelope runs hotter, the arc column moves, and irradiance at the reactor falls gradually while every other reading on the bench looks normal. Nothing trips. Your reaction rate simply declines and you look for the cause in the catalyst.
A partly blocked strainer produces no sound and no error, and you can see a dirty window while missing a loop that has lost half its flow. Only a logged temperature or a repeat irradiance check will show it.
A short checklist at fixed intervals catches nearly every cooling fault before it reaches the data. The intervals below assume a laboratory with normal dust load and several hours of daily operation; adjust them for a workshop environment and record the date of every check, because the trend between checks is more informative than the check itself.
| Interval | Item | What to check | Pass criterion | If it fails |
|---|---|---|---|---|
| Weekly | Intake and exhaust filters | Hold the filter against a lamp; look for a dust mat and for lightening at the edges where air bypasses | Even color, no visible dust layer, no gaps at the frame | Replace, do not vacuum and refit |
| Monthly | Duct joints and fan | Run a hand along joints for leakage; listen at the fan housing | No draft at joints, no whine or rumble at the bearing | Reseal the joint; replace the fan |
| Quarterly | Coolant level and appearance | Look at the reservoir against the mark, then at the color through the wall | Level at the mark, coolant clear, no film or sediment | Top up with the specified mix; change if cloudy |
| Every 6 months | Coolant change and flow | Drain, flush, refill; compare flow indicator against the commissioning value | Flow at the recorded value, no bubbles in the return line | Clean the strainer, check for a kinked hose |
| Annually | Hoses, clamps, seals | Flex each hose, inspect clamps and quick-connects | No hardening, no seepage, no cracking at the bend | Replace the hose set as a unit |
Two entries do most of the work. A filter that has gone dark at the edges is bypassing air rather than filtering it, so unfiltered room air cools the housing and dust collects on the reflector. A flow indicator that has drifted below its commissioning value is the earliest reliable sign of a restriction, and it usually moves before the coolant temperature rises.
The air path is only as good as its weakest joint, and most installations have a joint that was never sealed properly. Air takes the path of least resistance, so a gap at a duct collar or a missing blanking plate sends most of the flow around the lamp rather than across it. Seal the run from intake to exhaust, keep the exhaust outside the room or into an extraction system, and never run the source with the filter frame removed "just for today".
Room conditions belong in the schedule too. A source sharing a room with an oven or an open solvent bath pulls in warm, loaded air, and its cooling margin disappears in summer. Record the room temperature beside the coolant temperature so the two can be compared later.
Three measurements describe the loop: the setpoint, the return temperature and the flow. A chiller that holds its setpoint while the return temperature climbs is telling you the loop cannot remove the heat being generated, which points to flow rather than to the chiller. Read both numbers, not just the one on the front panel.
A cooling problem has a signature: output falls over tens of minutes and recovers after a pause, while lamp power stays constant. That pattern separates it from bulb aging, which does not recover, and from alignment drift, which appears suddenly after a mechanical change rather than ramping with run time. Log irradiance at the start and end of every long run and the signature becomes obvious.
For water-cooled LED sources running between −10 and 80 °C, a setpoint the loop can no longer hold is worth investigating before the next experiment, because the source compensates by driving the diodes harder for the same output and ages faster. Thermal reactor stages deserve the same treatment: read the stage temperature trend against the cooling trend, never in isolation.
Keep one spare filter set, one spare hose set and one container of the specified coolant on the shelf; their absence turns a twenty-minute service into a lost week. Record every check with a date, a reading and initials, and keep the log with the instrument rather than with the person who ran the check.
Watch what happens after a pause. Cooling-related drift recovers when the source cools and then falls again once it reaches operating temperature, while bulb-related loss stays where it is. Compare irradiance at a fixed power setting before and after a rest period; a recovery of more than your repeatability band points at the thermal path.
Only as an emergency measure, and only to finish the current run. Water alone changes the freezing point, the corrosion protection and the conductivity of the loop, and repeated top-ups concentrate whatever remains. Drain and refill with the specified mix at the next service interval rather than accumulating a diluted blend.
No. The filter protects the optical compartment as much as the lamp, and dust that reaches the reflector is difficult to remove without touching the coating. If airflow is the problem, find the restriction instead: a collapsed duct, a closed damper or a filter installed the wrong way round are the usual causes.
