What Each Stage of Filtration Removes

Why this matters

A compressed air filter train is not a set of progressively finer sieves. Each stage removes one class of contaminant and is physically incapable of removing the others, and every stage a shop adds "to be safe" charges pressure drop for the rest of its life whether it catches anything or not. The two expensive mistakes are opposite in shape: expecting a stage to remove something it cannot, and installing a stage nothing downstream needs. Both look like diligence on a quote.

Isolate before an element comes out

Changing an element means opening a pressurised path. Close the isolation valves either side of the housing, vent the housing through its own bleed, confirm zero on the housing gauge with the bleed still open, then lock and tag the isolation under 29 CFR 1910.147, which is the standard covering stored mechanical and pressure energy. A filter bowl that still holds pressure will launch when the last thread lets go, which is a struck-by hazard at head height on a wall-mounted housing.

A used element and its bowl hold oily condensate. Wear chemical-resistant gloves and sealed eye protection for the removal, and do not put that liquid down a floor drain: oil-bearing condensate discharged to a sewer falls under the local pretreatment authority's rules, so route the disposal question to that authority or to a licensed hauler rather than deciding it on the truck. Do not blow an element out to reuse it; besides the fact that a loaded element cannot be recovered, cleaning with compressed air is limited under 29 CFR 1910.242(b) to under 30 psi with effective chip guarding and personal protective equipment.

Four things in the air, and only three of them are filtration

Bulk liquid is water and oil already condensed into slugs and running along the bottom of the pipe. It is removed mechanically by an aftercooler and a separator, and by drains. A filter is not the tool for it, and a filter fed bulk liquid floods and passes it straight through.

Solid particles are pipe scale, rust, desiccant fines and ambient dust that made it through the intake filter.

Liquid aerosol is water and oil suspended as droplets fine enough to travel with the air stream. This is what a coalescing filter exists for: it makes small droplets collide and merge into large ones that drain out under gravity.

Vapour is water and hydrocarbon in the gas phase. Vapour passes through every mechanical filter ever built, because there is no droplet to catch. Water vapour is removed by a dryer. Oil vapour is removed by an adsorbent bed.

Where each stage sits

compressor
  |
aftercooler + separator -- bulk liquid leaves here
  |
receiver -- more liquid drops out as air rests
  |
prefilter ------- solids, protects the next stage
  |
coalescer ------- liquid aerosol, no vapour at all
  |
dryer ----------- water vapour
  |
afterfilter ----- fines carried out of a desiccant bed
  |
carbon bed ------ oil vapour, only where required
  |
header to the tools

The order is not a preference. Each stage is protecting the next one, and any stage moved upstream of its protector has a shortened life. A coalescer placed ahead of a separator drowns. A carbon bed placed ahead of a dryer is destroyed by liquid water. An afterfilter has no job at all unless there is a desiccant bed above it shedding fines.

What each stage cannot do

This is where the money is, so it gets stated as flatly as possible.

A coalescing filter removes no vapour. None. Air leaving a perfect coalescer is fully saturated, and it will condense liquid water the moment it cools by one degree anywhere downstream. A shop that fits a high-grade coalescer to solve water at a tool has bought an excellent solution to a different problem, and the water will still be there. Only a dryer changes the dew point.

A particulate filter removes no meaningful oil aerosol. It catches solids. Oil droplets pass.

An adsorbent carbon bed removes no particles and cannot survive liquid. It handles vapour and odour and nothing else, and a slug of water or oil ruins the bed rather than merely loading it.

A desiccant dryer removes water vapour, not oil vapour. Worse, oil arriving at the bed coats the adsorbent surface and takes capacity out permanently. What that does to a bed is covered in its own article and is not re-derived here.

No filter has an opinion about pressure dew point. Purity of one kind does not imply purity of another, which is why ISO 8573-1:2010 states particles, water and oil as three independent class numbers rather than one grade. That standard binds through a customer specification, a contract, or an equipment listing, in the edition named there, never on its own authority.

Stages that come off the list

Two candidates come off more often than they go on, and each has a condition attached that decides it.

The second coalescing stage stacked behind the first. Two coalescers in series are commonly sold as insurance. Once the first stage is operating at its rated efficiency and is fed a separator-protected stream, the second one has very little left to remove, and it charges its pressure drop continuously for the whole of its life. The condition that keeps it: a stated downstream requirement that the first stage's rated carryover does not meet on its own, in which case the two-stage arrangement is a specification, not insurance.

A carbon bed on a general shop-air header. A carbon bed has a finite adsorption capacity, no differential pressure signature that reports saturation, and a saturated bed releases what it was holding rather than simply stopping work. On a header feeding impact tools, blow guns and cylinders, it is a drop charge with no downstream claim on it. The condition that keeps it, and it is not a small one: any breathing air, food contact, pharmaceutical, or process requirement written into a specification. Where that requirement exists, the bed stays and gets a scheduled change interval from the manufacturer, because you cannot measure your way to its end of life from the pipe.

Reading a rating without being sold by it

Filter grades are quoted as a retained efficiency at a stated particle size, or, for coalescers, as a remaining oil carryover concentration. Both are measured under stated conditions, and the conditions travel with the number.

Coalescer carryover is quoted at a stated inlet oil concentration, a stated flow and a stated air temperature. Efficiency falls as temperature rises, because warm oil is less viscous and drains differently through the media, so a coalescer rated at a cool test condition and installed immediately downstream of a hot aftercooler is not delivering its published figure. If the rating on a data sheet does not name its test conditions, treat it as a marketing number.

Pressure drop is quoted the same way: at rated flow, clean. A clean element commonly starts in the 1 to 2 psi range at its rated flow, and manufacturers typically set a change point in the 2 to 3 psi range; read both off the element's own data sheet. The conversion from psi of drop to compressor power belongs to the distribution article, which uses the rule that roughly 1 percent of specific power tracks each 2 psi of discharge pressure on a lubricated rotary screw at the 100 psig class.

Worked: a six-stage train, measured

A machine shop with recurring water at one station had a train built up over years, one housing at a time. Measured differential across each stage at peak plant flow, using a single gauge moved from tap to tap so that the gauge's own fixed offset is common to every reading and cancels in the differences:

The first coalescer at 2.1 psi is at the top of its manufacturer's stated change range and is due. That is a maintenance finding, not a design one.

The second coalescer and the carbon bed together account for 2.7 psi, about 26 percent of the train's total drop. This header feeds impact wrenches, blow guns and air cylinders, with no breathing air, food contact or process specification anywhere on it, and the shop could produce no document requiring either stage. Both came out. Total train drop falls to 7.7 psi, and by the rule stated above, 2.7 psi is a bit over 1 percent of the compressor's specific power handed back permanently.

Note what did not change: the water at the station. Removing two stages did not fix it and was never going to, because neither stage removes vapour. The water was a dew point problem, measured and resolved by the method in the dryer article. The filter finding and the water finding were two separate items on the same report, and keeping them separate is the point. A tech who bundles them writes "cleaned up the filter train and resolved the water," and the next person to see water at that station has no idea which of two changes to look at.

The failure mode on the other side is worth naming because it is the one that gets a tech in trouble: pulling a carbon bed off a line that turns out to feed a paint booth or a food line, on the strength of nobody in the room remembering a requirement. Absence of memory is not absence of a specification. Get the requirement in writing, or leave the stage in and charge for the element.

References

• 29 CFR 1910.147 for isolating and verifying depressurisation of a filter housing before an element is changed, and 29 CFR 1910.242(b), which limits compressed air used for cleaning to under 30 psi with effective chip guarding and personal protective equipment
• Local sewer authority pretreatment rules for disposal of oil-bearing compressed air condensate, which vary by jurisdiction and are decided there rather than on site
• ISO 8573-1:2010 purity classes for particles, water and oil, which bind through a customer specification, a contract or an equipment listing in the edition named there; filter manufacturer data sheets for efficiency, carryover and clean pressure drop with their test conditions
• See related: What Oil Carryover Does Downstream; Why a Filter That Is Not Changed Costs More Than One That Is