A xenon bulb is near end of life when it stops reproducing its commissioning baseline at the same power setting. On a PLS-SXE300E, a 300 W setting that no longer yields the 50 W of total optical power recorded at commissioning, together with slow ignition and a drifting arc, is a replacement signal. Judge the trend, not the hour counter.
An hour meter measures time, not the thing that ruins experiments, which is output loss. Xenon bulbs rarely fail open; they decay. The arc wanders inside the envelope, the envelope darkens, ignition needs more energy, and the same electrical input yields less light at the reactor. A bulb partway through its rated hours can already sit well below the baseline your method was validated against, while a bulb near the end of its rated hours can still be inside tolerance.
The counter also ignores ignition count. Every start stresses the electrodes more than steady operation does, so two lamps with identical hours are not equivalent if one is switched on daily and the other runs continuously.
Signs arrive in a fairly consistent order, so one unusual observation proves little while a pattern across several runs proves a lot. The table lists what each sign looks like, how to confirm it without opening the housing, and what the observation justifies.
| Sign | First appearance in the data | Confirmation method | What it justifies |
|---|---|---|---|
| Slow ignition | The lamp strikes after a longer delay than at commissioning, sometimes on the second attempt | Time the interval from trigger to a stable arc on three consecutive starts | Schedule replacement; stop running unattended starts |
| Slower output ramp | Irradiance takes longer to reach its plateau and keeps creeping through warm-up | Log irradiance at fixed intervals from a cold start | Extend warm-up before calibration; plan the swap |
| Spot drift | The bright core no longer sits on the reactor window center and the edge falloff changes | Re-check the spot on a target at the working distance | Re-align first, replace if drift returns |
| Color shift | The beam looks warmer or duller and a filter cut-off no longer matches earlier runs | Compare against a stored reference image and a spectral scan | Re-qualify the filter set or replace the bulb |
| Lower output at the same setting | The supply reads the same power but the radiometer reads less | Measure at fixed distance and geometry against the commissioning log | Replace the bulb |
Read the table as a sequence rather than a checklist. Slow ignition and a sluggish ramp typically precede measurable output loss by weeks, which is exactly the planning window you want. Spot drift is the exception: it can appear on its own after a mechanical change, a bumped housing or a lamp that was not fully seated, so confirm alignment before condemning the bulb.
A comparison is only valid when geometry, power setting and warm-up state are identical. Fix the sensor at the same distance and angle, let the lamp reach thermal equilibrium at the same power setting, and read with the same meter on the same range. A casual check at a different distance is worse than no check, because it hands you a number you will later compare against a real baseline.
A defensible trigger is one you wrote down before the bulb aged, not one you invent once results look odd. Record the baseline at your working power setting and distance during commissioning, decide how much departure you will accept, and name who authorizes a swap. A trigger based on a fixed number of hours is a guess; a trigger based on measured departure from baseline is a decision you can defend to a reviewer.
Once the trigger is crossed the bulb is on borrowed time. Loss accelerates and the risk changes from slightly lower irradiance to an unstable arc in the middle of a run, which costs an entire series rather than a single point. Treat the crossing as a scheduling event, not a warning to monitor.
Plan the swap around the run schedule instead of reacting to a failure. Keep a spare on the shelf, run the long experiments while the lamp is fresh, and do not start a multi-day campaign on a bulb that has already failed two of the five checks. Replacement itself is mechanical and repeatable; the post-swap re-verification matters more than speed.
Do not assume a new bulb reproduces the old baseline. Re-measure irradiance, re-check the spot, and run a reference sample before new data joins the old series. A source change is a method change until you have shown otherwise.
Only if it still meets its commissioning baseline at your working power setting. Reliable ignition says the arc starts, and nothing about how much light reaches the reactor. A bulb that strikes cleanly but delivers visibly less irradiance biases every rate you calculate, and the bias looks like a change in catalyst performance rather than a change in hardware.
At the start of every experimental campaign, and whenever two consecutive runs disagree by more than your established repeatability band. Frequency matters more than a fixed calendar, because a bulb that sits idle ages differently from one that runs continuously; the only useful reference is your last good measurement at the same geometry.
No. A misaligned housing, a hazed window, a loaded cooling filter or a partly closed aperture all reduce irradiance at the reactor with a healthy bulb installed. Check the optical path and the cooling first, then repeat the measurement at a fixed power setting. Only a loss that survives those checks belongs to the bulb.
