Zeolite holds nitrogen by preference and lets oxygen pass - and it does so by equilibrium rather than by rate.

A carbon molecular sieve in a nitrogen generator works kinetically: oxygen diffuses into its pores faster than nitrogen, so the separation is a race and the cycle must be timed to stop before nitrogen catches up.

Zeolite works differently. Its structure and surface charge attract the nitrogen molecule more strongly than the oxygen molecule, so nitrogen is preferentially held at equilibrium. Oxygen passes through and leaves as product.

The practical consequences are that the cycle timing is less knife-edge, and that anything else the zeolite holds strongly is a problem. Water is the significant case: zeolite is powerfully hygroscopic, adsorbs moisture in preference to nitrogen, and does not release it during normal depressurisation.

So the two machines look similar and share a principle, but their sensitivities differ - and the oxygen machine's sensitivity is water.

A genuinely dry supply, to the manufacturer's stated dewpoint - and this is the requirement most often underestimated.

Zeolite adsorbs water far more strongly than it adsorbs nitrogen. Water molecules reaching the bed occupy adsorption sites and are not driven off by the pressure swing that regenerates the sieve for nitrogen. They accumulate.

The result is a bed that gradually loses capacity: purity falls slowly, flow falls, and eventually the sieve must be replaced. The decline is gradual enough that it is often attributed to age rather than to a drying failure.

So the dryer ahead of the generator is not a general air-quality nicety but a protective component, and its performance should be verified rather than assumed - a dryer that is undersized, fouled or has lost its regeneration will still deliver air that looks dry at the outlet while being far above the required dewpoint.

Monitor dewpoint continuously where the generator is significant to the operation.

Purity in the low-to-mid nineties, with flow set by machine size and falling as purity is pushed toward the ceiling.

The argon limit means there is no purity above roughly ninety-five per cent to aim at, so the specification question is different from nitrogen: rather than choosing a purity from a wide range, you are generally working near the ceiling and sizing for flow.

Within that, the same trade applies - drawing product more aggressively from a given bed reduces purity, and backing off increases it at the cost of flow.

Performance also falls as feed air temperature rises, which matters more than people expect. A machine rated at a moderate inlet temperature will not meet its figures in a hot plant room or an unventilated container in summer.

When comparing quotations, confirm the purity and flow are stated together, at a defined inlet air temperature and pressure, and check that those conditions match the intended installation.

Everything the oxygen touches must be oxygen-compatible and oxygen-clean, starting at the generator outlet.

An oxygen-enriched stream makes ordinary materials burn readily. Hydrocarbons are the acute hazard: oil or grease in contact with pressurised oxygen can ignite without any external ignition source, and the energy released can ignite the metal around it.

So pipework, valves, fittings, gaskets and instruments in the oxygen line must be selected from oxygen-compatible materials and supplied cleaned for oxygen service. No ordinary lubricants may be used on threads or seals - only oxygen-compatible products.

This applies to maintenance as much as to installation. A fitter using standard thread sealant or wiping a component with an oily rag has created the hazard, and it will not be visible.

Ventilation matters too: leaked oxygen enriches the surrounding air, making clothing and ordinary materials dangerously flammable, so plant rooms need designed ventilation and enrichment monitoring.

Cheaper for continuous industrial duty, and free of boil-off and delivery logistics - but capped in purity.

Liquid oxygen delivered to a site tank is very pure and can supply high flows, but it evaporates continuously in storage, so a site pays for product it never uses. It also requires a cryogenic tank, its own safety regime, and regular deliveries.

A generator has no boil-off, no deliveries and no cryogenic hazard. Its costs are capital, maintenance and the electricity to compress its feed air.

For aquaculture, wastewater treatment and ozone systems running continuously, particularly at sites where deliveries are difficult, the generator usually wins clearly - and remote sites often value the independence more than the arithmetic.

Where purity above the argon ceiling is genuinely required, or where demand is very high and intermittent, delivered oxygen remains appropriate. Some sites run a generator for base load and keep delivered gas for peaks or high-purity use.

Noticeably - performance falls as feed air gets warmer, and installations in hot rooms routinely underperform their data sheets.

Adsorption capacity decreases as temperature rises, so a zeolite bed fed with warm air holds less nitrogen per cycle. The machine therefore delivers less oxygen, at lower purity, than the same machine fed cooler air.

Ratings are quoted at a stated inlet temperature. A generator in an unventilated plant room, inside a container, or downstream of a compressor with poor aftercooling can be receiving air far warmer than that, and the shortfall is then blamed on the generator.

The practical measures are adequate aftercooling on the compressed air, a ventilated location for the generator, and checking the actual inlet temperature at commissioning rather than assuming it.

Where ambient conditions are genuinely hot, tell the supplier at the quotation stage so the machine is sized for the real conditions rather than for a reference figure.

Years in a well-protected installation, and much less where moisture reaches it - the difference is entirely down to the drying upstream.

Zeolite is not consumed by normal cycling. What degrades it is water, which it adsorbs preferentially and does not release during the pressure swing that regenerates it for nitrogen. Moisture accumulates in the bed, occupying sites permanently and reducing capacity.

The symptom is a slow decline in both purity and flow over months, with no mechanical change - and it is frequently attributed to the sieve simply ageing when the real cause is a dryer that is undersized, fouled or has lost its regeneration.

So the sieve's life is effectively a measure of how well the dryer has worked. Verify dewpoint continuously where the installation matters, rather than trusting the dryer to be doing its job.

When replacing sieve, establish why the previous charge failed first, or the new one will follow it.