Why a Bag Is the Filter and a Cyclone Is Not

This page contains affiliate links. If you buy through one we may earn a commission, at no cost to you.

Not for the reason it sounds like. The complaint reads as being about emptying a bin over a kitchen caddy, and that is a real annoyance, but it is not the mechanism. The mechanism is that a bag is not a container with a filter attached. The bag is the filter, and it has about five times the media area of the pleated cartridge that replaces it.

Take the bag out and three things happen in order. The fine dust that used to land on 0.19 square metres of paper now lands on 0.04 square metres of pleat. The cyclone put in front of it to help cuts off at somewhere between 0.6 and 2 micrometres, which is above the particle size that same filter is specified at. And the dust cup, which is sold as the convenience, decides how much of that fine dust accumulates before anybody intervenes.

A cyclone has a size below which it does nothing

Start with the separator, because it is the part people assume is doing the work. A cyclone does not filter. It spins the air and lets inertia throw the heavy things outward, so whether a given particle gets caught depends on whether its momentum can carry it across the streamlines before the air leaves.

That gives a cut diameter, the size at which it catches half of what arrives. The standard expression puts it at the square root of nine times the air viscosity times the inlet width, divided by two pi times the number of effective turns times the inlet velocity times the particle density.

Run it on three plausible geometries, with air viscosity at 1.81 times ten to the minus five, household dust at 2,000 kilograms a cubic metre and five turns. A handheld with an 8 millimetre inlet at 20 metres a second cuts at 1.02 micrometres. A full size upright with a 25 millimetre inlet at 15 metres a second cuts at 2.08. A small multi-cyclone cone of 4 millimetres at 25 metres a second cuts at 0.644.

Now the number that matters. A Eureka PowerSpeed H13 vacuum filter set publishes exactly the figure the comparison needs, stating that "H13 media captures 99.97 percent of particles as small as 0.3 microns". Three tenths of a micrometre. Every cut diameter above sits between 2.1 and 6.9 times higher than that.

Which settles what the cyclone is for. It is not reducing the filter's hardest duty. It is removing the bulk that would otherwise bury the filter, which is genuinely valuable and is a completely different job. At the one size the filter is specified at, the cyclone is transparent.

There is also a direction in that expression worth noticing, because it is the opposite of the intuition. The cut diameter rises with the square root of the inlet width, so a bigger cyclone is worse at fine dust, not better. That is why machines in this category use arrays of many small cones rather than one large one, and why our roundup of cyclone dust collectors is full of multi-cone arrangements rather than single large drums.

The area ratio, which is the actual answer

A four litre paper bag measuring roughly 300 by 200 by 70 millimetres presents about 0.19 square metres of filtering surface, and every square centimetre of it is working, because the bag is the filter. A handheld pleated cartridge 60 millimetres across and 60 tall, with thirty pleats ten millimetres deep, presents about 0.04 square metres.

A ratio of 4.75. At a representative airflow of 0.02 cubic metres a second, the face velocity through the bag is 0.105 metres a second and through the cartridge 0.50.

Both halves of that hurt, and they compound. Cake accumulates in proportion to the dust arriving per unit of area, so nearly five times faster. And pressure drop through a cake rises with the velocity you are pushing through it, so each layer of cake that does form costs more. A bagged machine spreads the same dust over five times the paper at a fifth of the speed, and then throws the whole thing away rather than asking anybody to rinse it.

Our guide to choosing the right replacement vacuum filter is the page a buyer lands on at this point, and it is about fit and type, as it should be. The quantity nobody publishes is the area. The Eureka record is honest enough to name the gap in its own drawbacks: "No dimensions given for the filter pieces". On a product that is nothing but filter, the area is the specification, and it is absent.

The dust cup sets the clock, and bigger is worse

Here is where two published figures in this catalogue produce a result neither listing mentions.

A Runful handheld car vacuum publishes a 100 millilitre dust cup and sells it as roomy: its record notes the cup "holds 100ml, which the maker notes is larger than many competitors". A VACDRIVE DW688 handheld cordless car vacuum publishes 680 millilitres and sells it harder, as "up to three times larger than typical mini handhelds", with the reasoning spelled out: "On a car interior full of sand and crumbs, cup size decides whether you finish the job or stop halfway to empty it."

That is true and it is only half the consequence. At a bulk density of 200 kilograms a cubic metre those cups hold 20 grams and 136 grams of debris. If two percent of the collected mass is finer than the cyclone's cut diameter, then 0.40 grams and 2.72 grams reach the filter between emptyings. Divided by the same 0.04 square metres of media that is 10 grams a square metre against 68.

A ratio of 6.8, which is exactly the ratio of the cups, because the filter area did not grow with the cup. The bigger cup does not reduce the filter's work. It postpones the intervention that would have cleaned it, and it collects 6.8 times as much fine dust before that happens.

Which reframes a detail on the VACDRIVE listing. It ships two washable filters, and its record gives the stated reason: "you rotate them, so while one dries the other is in use. That is a small design decision with a real consequence, because the usual failure pattern with washable filters is putting a damp one back in and losing suction." That reasoning is sound. The arithmetic adds the part the listing does not say, which is that a 680 millilitre cup needs two filters more than a 100 millilitre one does, and shipping the second is the right answer to a problem the cup created.

One of our records publishes a speed that cannot make its own pressure

The Runful record contains the clearest arithmetic error we have found in this catalogue, and it is in a single sentence of its design text: "A 24 volt maximum power battery paired with a 1000rpm motor reaches a stated 25000 pascals of suction."

A centrifugal fan's pressure rise is set by the square of its impeller tip speed, times the air density, times a coefficient of order one. Invert it. To make 25,000 pascals in air at 1.2 kilograms a cubic metre, the tip speed has to be the square root of 25,000 over 1.2, which is 144.34 metres a second. On a 50 millimetre impeller that is 5,773 radians a second, which is 55,129 revolutions a minute.

At 1,000 revolutions a minute the tip speed is 2.618 metres a second and the pressure rise is 8.22 pascals. Twenty five thousand divided by 8.22 is 3,041. The published speed is short of the published pressure by a factor of three thousand in pressure, or fifty five times in speed.

And the cross-check is on the shelf next to it. The VACDRIVE publishes "an 80,000 RPM brushless motor". Eighty thousand revolutions a minute on the same 50 millimetre impeller gives a tip speed of 209.44 metres a second and an ideal rise of 52,638 pascals, which lands on 25,000 at a perfectly ordinary pressure coefficient of 0.475. So 80,000 is the right order of magnitude for a 25 kilopascal handheld and 1,000 is not, by a factor of eighty.

The two figures on the Runful cannot both be right, and the speed is the one that fails. Everything else on that record is consistent and useful: 24 volts, 4000mAh, about 32 minutes, a 100 millilitre cup, a washable filter, a two year warranty, and an unusually candid note that "Powerful car vacuums tend to be noisy" and that it is not built "for very long continuous sessions". Which is itself a clue, because noise in a handheld comes from tip speed. A machine that is genuinely loud is a machine whose impeller is turning fast, and the record says it is loud two paragraphs after saying it turns at 1,000 revolutions a minute.

The gap is symmetrical, and it is the whole category

The VACDRIVE's failing is the mirror image, and its own record states it better than we could: "What the listing never tells you is how strong that suction actually is. There is no figure in pascals or kilopascals anywhere, only words like relentless and hurricane. That is the central gap in an otherwise well-documented product, and it means you cannot compare this directly against a rival that does publish a number."

That is exactly right, and the record repeats it in the drawbacks as "No suction figure published in pascals or any other unit". It is the most useful thing on the listing.

So one machine publishes a pressure and a speed that contradict each other, and the other publishes a speed and no pressure. Neither can be compared with the other, and there is a third figure missing from both and from every vacuum in this catalogue: the airflow.

That matters more than the missing suction, because pressure and flow are the two ends of one curve and a vacuum works somewhere in the middle of it. A sealed suction figure is measured with the hose blocked, at zero flow, which is the one operating point at which the machine is doing no cleaning whatsoever. Our page on what kPa means on a vacuum sealer establishes the conversion that makes the scale readable, that a suction figure is a fraction of one fixed atmosphere of about 101 kilopascals, so the Runful's 25,000 pascals is 24.7 percent of everything available. What it cannot tell you, and what this page cannot compute either, is where on the curve the machine sits when air is actually moving, because that needs the flow and nobody states it.

Which is the honest answer to how a loaded filter hurts. A filter's pressure drop is a flowing quantity and a sealed suction figure is a zero-flow quantity, so they cannot simply be subtracted. What can be said is the direction and the shape: every pascal the filter takes is a pascal the nozzle does not get, the filter's share rises as the cake builds, and the cake builds 4.75 times faster per unit area on a cartridge than on a bag. The size of the penalty needs two numbers that no listing in this category publishes.

The honest product in the group is about a different problem

Worth naming the record that gets its own scope exactly right. An EICEM green vacuum dust detection light attachment clips to a vacuum head and rakes a low green beam across the floor so fine debris casts a shadow. Its first drawback is four words long: "Adds visibility, not suction."

That is the correct disclaimer and it points at a genuinely different failure. What a raking light reveals is dust the nozzle did not pick up, which is an agitation and sealing problem at the floor rather than a filtration problem inside the machine. A bagless vacuum and a bagged one with the same head will leave the same debris behind, and no amount of cyclone or filter changes that. Its record also notes that "Effectiveness also drops in bright rooms where the beam contrast fades", which is the right caveat for a device that works by shadow.

Two different complaints get tangled together in threads about bagless machines. One is that it does not pick up, which is the head. One is that suction fades over a few weeks, which is the filter area. The light diagnoses the first and this page is about the second.

How to read the one figure the filter does publish

The 99.97 percent at 0.3 micrometres deserves a note, because it is a strong claim stated in a format that invites misreading, and our own record already flags the right caution: "performance claims such as the 99.97 percent capture figure are maker-stated, not independently measured here."

The deeper point is about what the number is a statement of. Our page on why two 5 micron filters are not the same filter works through the general problem properly, that a single micron number can describe a nominal average or an absolute ceiling and the two differ by an order of magnitude in what gets through. It is written about water and the argument is identical in air: a rating is a pair of numbers, an efficiency and a size, and quoting one without the other is not a specification.

The Eureka figure does quote both, which puts it ahead of most, and 0.3 micrometres is chosen because it is close to the hardest size for a fibrous filter to catch, so efficiency at 0.3 is a genuine worst case rather than a flattering one. What the record still does not give is the area it applies over, which is where this page started.

The honest limit, and the thing that actually decides it

What is exact here is the structure of the cut diameter expression and every ratio inside it. The cut diameter goes as the square root of the inlet width and inversely as the square root of the velocity and the turns, so the three points in the first panel are exactly ordered whatever the inputs are, and a bigger cyclone is exactly and always worse at fine dust. The 6.8 ratio in the third panel is also exact, because it is two published cup volumes divided by one shared filter area.

What is representative is nearly everything geometric, and the list is long because the category publishes so little. No vacuum we carry publishes a filter area, an airflow, an inlet width, an impeller diameter or a clean filter pressure drop. So the 8, 25 and 4 millimetre inlets, the 15 to 25 metres a second, the five effective turns, the 0.19 and 0.04 square metres of media, the 0.02 cubic metres a second of airflow, the 200 kilogram bulk density and the 2 percent fines fraction are all assumptions.

They move the answer in a convenient direction, which is worth saying. Halve the assumed turns and every cut diameter rises by 1.41 times together, pushing all three further above the filter's 0.3 micrometre rating rather than toward it. Change the fines fraction and the third panel scales as a block without the 6.8 moving. Narrow the inlet to one millimetre, which no cyclone is actually built with, and the handheld's cut still only falls to 0.36 micrometres, which is still above 0.3. There is no plausible set of inputs on which a cyclone reaches down to the size its filter is specified at, and that is the one conclusion this page would defend against any objection.

What it deliberately does not compute is how much suction any specific machine loses as its filter loads, and the refusal is the finding rather than a gap in the method. That calculation needs a filter area and an airflow, two numbers absent from every listing in the category, and a fan curve, which is absent from every listing in every category. Our roundups of cordless vacuums and dust extractors run into the same silence from the other direction.

Which leaves the practical version, and it follows straight from the area ratio rather than from any product. The reason bagless machines draw complaints is that they hand the owner a maintenance job the bagged design did away with, and the job is on a fifth of the area so it comes due five times as often. If you own one, the useful habit is not emptying the cup more often, which only changes how much dust is in the cup. It is washing the filter on the cup's schedule rather than when the suction has already gone, and owning the second filter so the first one can dry properly. One of the two handhelds here ships that second filter. The other tells you to dry the only one you have, which is the same advice with a gap in the middle of it.