What a Duct Silencer Costs You in Pressure

Why this matters

A silencer is bought on one number off a submittal, the insertion loss, and it is paid for with two others that nobody reads: the static pressure drop and the self-generated noise. Squeeze the unit down to fit the shaft and you have quietly raised both. The reason this bites in the field rather than on paper is that all three numbers are published against the same variable, face velocity, and they move in opposite directions on the thing you want. This card is one schedule line filled in twice, at two face velocities, so you can see the same product buy 19.8 dB in one column and 14.8 dB in the other while its pressure drop roughly triples.

The three published numbers and the basis attached to each

Dynamic insertion loss, per octave band, in dB: the difference in sound power level downstream with and without the unit in the duct, measured to a laboratory method such as ASTM E477 in the edition the product's test report names, at a stated face velocity and a stated flow direction. Forward flow and reverse flow are different columns for the same product, because the airflow and the sound travel the same way in one and opposite ways in the other.

Static pressure drop, in inches of water gauge, at a stated face velocity, with a stated approach condition. The test has a clean, uniform approach. A silencer bolted directly to a fan discharge does not have one, and both its pressure drop and its self-noise are worse than published by an amount the manufacturer's system-effect guidance owns.

Self-noise, also called regenerated noise, as a sound power level per octave band re 1 picowatt, at a stated face velocity. It is the noise the silencer makes by having air pushed through it, and it exists whether or not the fan is loud.

Any one of those three quoted without its face velocity is not a specification, the same way a level quoted without its weighting and reference is not a measurement.

Face velocity is computed, and its denominator is the trap

Face velocity is volume flow divided by area, and there are two areas. Gross face area is the outside dimensions of the unit. Free area is what is left between the splitters or baffles, which on a common dissipative silencer is roughly half the gross. A velocity computed on gross area is roughly half the velocity computed on free area for the same duty, and the published curves are drawn against one of them.

Read which one the manufacturer's data uses before you take a number off it. Reading a free-area curve at a gross-area velocity understates pressure drop and self-noise by a wide margin, in the flattering direction, on both numbers at once.

The pressure figure re-bases the fan's duty

Before you add a silencer's published drop to a system total, write down what that total already contains.

A design external static pressure calculated during engineering usually already carries a silencer allowance, because the engineer knew a silencer was scheduled. Adding the selected unit's full drop on top of that total charges the system twice for the same component. The correct move is a re-basing: the change to the duty is the selected unit's drop minus the allowance already inside the total, and you say in the same clause which of the two you are doing.

The scaling is the second place this goes wrong. Pressure drop through a fixed geometry varies with roughly the square of velocity, holding the geometry, the approach condition and the air density constant at standard conditions of about 0.075 lb per cubic foot. At altitude, or on a hot exhaust stream, density is lower and the drop at a given volume flow falls with it, so a sea-level table read at 6,000 feet or at elevated temperature needs the manufacturer's density correction rather than the square-law alone.

Self-noise is a floor at the silencer outlet, not at the terminal

The most common misreading of self-noise is to treat it as the lowest level the room can reach. It is not. Self-noise is a sound power level generated at the silencer, and everything downstream of the silencer attenuates it exactly as it attenuates any other duct-borne energy: duct length, lining, bends, the terminal itself.

What self-noise does cap is the silencer's own effective insertion loss. The level leaving the unit is the logarithmic sum of the attenuated source and the self-noise, and once those two are within a few dB of each other, more published insertion loss buys almost nothing, because you are no longer listening to the source.

The schedule line, filled in twice

Duty: 8,000 cfm through a rectangular dissipative silencer. The manufacturer's data is published against free-area velocity, and the unit's free area ratio is 0.50.

Column one, the unit as scheduled: gross face 48 in by 24 in, which is 8.0 square feet, so free area is 4.0 square feet and free-area velocity is 8,000 divided by 4.0, which is 2,000 fpm. Say the published table gives, for the 250 Hz octave, a dynamic insertion loss of 21 dB in forward flow and a self-noise sound power level of 66 dB re 1 pW, with a static pressure drop of 0.30 in wg. Note that the gross-face velocity here is 1,000 fpm, and reading the free-area curve at that figure would have understated everything.

Source sound power at the silencer inlet in the 250 Hz octave: 92 dB re 1 pW, from the fan manufacturer's published data.

• Attenuated source: 92 minus 21, which is 71 dB re 1 pW.
• Combine with self-noise of 66 dB: the two differ by 5 dB, so 1.2 dB is added to the higher, giving 72.2 dB re 1 pW leaving the unit.
• Effective insertion loss: 92 minus 72.2, which is 19.8 dB, not the 21 dB on the submittal.

Column two, the same product squeezed into a shaft: gross face 36 in by 18 in, which is 4.5 square feet, so free area is 2.25 square feet and free-area velocity is 8,000 divided by 2.25, which is 3,556 fpm. Round to the manufacturer's nearest published step of 3,500 fpm and read the table rather than extrapolating, because self-noise rises far too steeply with velocity for any exponent you carry in your head to be safe. Say that step gives an insertion loss still of 21 dB in the 250 Hz octave and a self-noise of 76 dB re 1 pW.

• Attenuated source: still 92 minus 21, which is 71 dB re 1 pW.
• Combine with self-noise of 76 dB: the two differ by 5 dB, so 1.2 dB is added to the higher, giving 77.2 dB re 1 pW.
• Effective insertion loss: 92 minus 77.2, which is 14.8 dB. Five decibels of the purchase disappeared into the unit's own noise, with no change to the printed insertion loss.

Pressure, re-based rather than added. The published drop at 2,000 fpm is 0.30 in wg. Scaling by the square of the velocity ratio, at constant geometry and standard air, 3,500 over 2,000 is 1.75, squared is 3.06, so the drop at the squeezed selection is about 0.92 in wg. The design external static pressure on the schedule is 2.40 in wg, and that figure already contains a 0.30 in wg silencer allowance from engineering. So the change to the fan's duty is 0.92 minus 0.30, which is 0.62 in wg, and the new total is 3.02 in wg. It is 0.62, not 0.92, because the 2.40 already had a silencer inside it. Adding the full 0.92 would have charged that allowance twice.

What the pressure change costs. Air power varies with the product of volume flow and total pressure, so at constant flow and constant fan efficiency, 3.02 against 2.40 in wg is a 25.8 percent increase in air power. Restoring the original flow also means more fan speed, and a given fan's sound power rises steeply with speed at constant fan size and a geometrically similar operating point, so part of the pressure penalty comes back as additional source noise upstream of the silencer you just bought. The size of that return belongs to the fan's published sound data at the new speed, not to a rule of thumb.

Sibling-rule check. Every acoustic figure above is a sound power level re 1 picowatt, unweighted, in the stated 250 Hz octave band, so the quantity, reference and bandwidth stay consistent and a power level is never mixed with a pressure level. The logarithmic combination uses the addition convention the decibel-addition card states rather than an arithmetic sum. The velocity basis is named as free area in every reading. The pressure correction is stated as a re-basing, with the amount already inside the baseline written down first. Both roundings are checked for direction, and they do not run the same way. Reading self-noise at the published 3,500 fpm step rather than at the computed 3,556 fpm understates it slightly, and the square-law pressure figure is likewise about 3 percent low, so the 0.92 in wg is a floor rather than a value and the real duty change sits a little above 0.62 in wg.

Working on a silencer installation without adding a hazard

Reading static pressure across a running unit means drilling test ports in live duct and standing beside a running fan: use a step drill with the duct supported, keep every fan and drive guard in place as 29 CFR 1910.212 requires, and where your own exposure at that position reaches the 29 CFR 1910.95 action level for the time you will be there, wear the protection the program specifies. Pulling a silencer module is a rigging job, not a two-hands job; the units are heavy and their lifting points are on the manufacturer's drawing. Where the media is fibrous and has degraded, cutting or removing it releases airborne fibre, so use the respiratory control the product's safety data sheet specifies under a 29 CFR 1910.134 program rather than a nuisance dust mask.

One placement question is a code question, not an engineering preference. A media-filled silencer is not permitted in a grease-bearing commercial cooking exhaust duct, and what is permitted there is governed by NFPA 96 in the edition your authority having jurisdiction has adopted and amended, which binds the facility and reaches your work through the permit and the inspection. Ask before you schedule one, not after.

What would change the selection

Low-frequency content. Dissipative silencers work by absorption and their insertion loss falls off in the low octaves, which is exactly where fan noise and breakout live. Where the governing bands are 63 Hz and 125 Hz, a longer dissipative unit is a poor trade and the conversation moves to reactive or packless designs, to duct shape, or to the fan selection itself.

Reverse flow. An exhaust or return application reads a different published column, and a selection made from the supply column is the wrong number.

No room for the drop. Where the fan has no margin left, the pressure penalty is not a cost, it is a hard stop, and the answer is a larger cross-section or a different path.

How to verify after installation

Measure static pressure across the unit at design flow and compare it with the published figure at your computed free-area velocity, not at the tested one. A measured drop well above published usually means the approach is non-uniform rather than that the unit is wrong, and the fix is straight duct ahead of it. Then measure at a fixed downstream position before and after, with the same quantity, weighting, bandwidth, time basis and position on both sets. If the improvement lands near the effective insertion loss rather than the published one, the selection is behaving; if it lands well below both, look for a breakout path around the silencer, which the unit was never in a position to treat.

References

• ASTM E477, laboratory measurement of duct silencer insertion loss, pressure drop and self-noise, in the edition the product's test report names
• Silencer manufacturer's published data at your face velocity and flow direction, and their system-effect guidance for a non-uniform approach
• Fan manufacturer's published sound power data at the operating point, which owns the source spectrum and the speed effect
• NFPA 96, in the edition adopted and amended by the authority having jurisdiction, which binds the facility and reaches the work through the permit, for anything installed in a commercial cooking exhaust duct
• See related: Why Duct-Borne Noise and Duct Breakout Are Different Faults; Why Decibels Do Not Add the Way Numbers Do; Why Sound Power and Sound Pressure Are Not the Same Number