Most teams trust the O₂ and H₂O numbers on the main display. They should not.
We placed four sensors in one 1500 mm nitrogen box: plenum return, floor center, 40 cm above the tray, and taped to a steel fixture just transferred from the antechamber. Ambient was 23 °C, 47 % RH. Target was O₂ < 5 ppm and H₂O < 5 ppm.
The dashboard looked perfect. The sample did not agree.
Plenum lies the least and matters the least
After a normal transfer, the plenum probe read 4 ppm O₂. The floor probe read 17 ppm. The probe near the warm fixture read 29 ppm.
Why? Transfer gas rolls along the floor. Fans mix the bulk, but the first 90 seconds are layers, not mixture. If your sensitive powder sits low, it breathes the worst air in the box.
Sensor refresh rate hides spikes
We compared two loggers:
- 60-second refresh: 0 transfer spikes flagged
- 5-second logging: 34 of 40 transfers showed O₂ > 25 ppm for 15–70 seconds
A material that reacts in 30 seconds does not care about your average. The dashboard averaged the damage away.
Probe height changed our batch yield
We ran 20 sulfide-electrolyte batches under the old rule: “open inner door when display < 5 ppm.”
Failures: 6 of 20.
Then we changed one rule: wait until the sample-height probe reads O₂ < 10 ppm and H₂O < 10 ppm.
Failures: 1 of 20.
Same gas. Same purifier. Same box. Different measurement point.
Glove-finger sweep is free performance
Butyl gloves hold lab moisture. After 200 insertions, glove-wall H₂O near the fingers stayed around 26 ppm even when the chamber read 4 ppm.
We added a 3-minute, 15 L/min N₂ sweep through the glove fingers before sensitive work. Near-glove H₂O dropped from 26 ppm to 6 ppm in 12 minutes.
No new hardware. One new habit.
The takeaway
Do not manage a glove box by the panel. Manage it by the sample.
Use these rules:
- Put at least one sensor at sample height.
- Log at 5 seconds, not 60.
- Ignore plenum-only “clean” numbers.
- Sweep glove fingers before air-sensitive work.
- Define success as ppm at the sample, not ppm at the fan.
The box was never lying. The instrument was standing in the wrong room.