Density decides whether a panel works or just decorates the wall

Sound dampening panels are layered assemblies that convert acoustic energy into heat through friction inside a porous core, and core density controls how far down the frequency scale that conversion keeps working. The definition sounds dry until you stand inside a room where it went wrong. Cheap sound reducing panels built on 24 kg/m³ glass wool feel soft to the touch and test beautifully above 1 kHz, then do close to nothing against a 200 Hz compressor hum. The lab file we keep for our 80 kg/m³ rock wool build lists an absorption coefficient of 0.62 at 250 Hz; the 48 kg/m³ version of the same panel tests 0.41. Same thickness. Same facing. The cheaper panel is out of the running one page into its own data sheet.

My working floor for machinery rooms is 80 kg/m³ rock wool, full stop. Glass wool at 32 to 40 kg/m³ has its place in HVAC duct linings where weight matters and the spectrum sits mid-band, but diesel generators and screw compressors put their energy exactly where light cores give up. Below roughly 60 kg/m³, a mineral wool core stops absorbing meaningfully at 250 Hz and under, whatever the catalogue prints. That sentence has settled more procurement arguments than any presentation I could build.

If you read quotes the way most purchasing departments do, price per sheet first and data sheet last, this is the spot where you get burned. Two panels can carry the same printed NRC while one weighs half of the other, because the average hides the low band. Which brings us to the number everyone quotes and almost nobody reads correctly.

NRC 0.95 hides more than it shows

NRC is the arithmetic average of absorption coefficients at 250, 500, 1000 and 2000 Hz, rounded to the nearest 0.05, measured in a reverberation room per GB/T 20247-2006, the Chinese twin of ISO 354:2003. Four frequencies, one number, no low band. A compressor spectrum peaks between 63 and 500 Hz, so most of the energy in a machinery room sits at or below the bottom edge of that average. The printed NRC tells you almost nothing about the figure you will measure at acceptance.

Mounting type changes the story further. Type A means the panel sits flat against a solid backing; Type E-200 means it floats on a 200 mm air cavity. The same 50 mm panel can post 0.55 at 125 Hz on a Type A mount and 0.85 on Type E-200, and both figures are legitimate because each is printed beside its mount in eight-point type on page six. Brochures quote the better one. Search any platform for sounds dampening panels and the first page mixes foam, curtains and real panels, every listing printing an NRC with no mount named. You are holding a marketing number, not a test result.

Demand the third-octave absorption curve from 100 to 4000 Hz instead. One page, lab stamp, mount stated, and it lets you read absorption exactly where your spectrum peaks. A CMA-qualified lab issues it with every report. A vendor who cannot produce the curve within a day has never read it either.

How thick do the panels actually need to be?

Thickness buys the low end. Twenty-five millimetres handles speech in offices and conference rooms, where nearly all energy lives above 500 Hz. Fifty millimetres covers machinery mid-band down to roughly 200 Hz. Dominant tones at 125 Hz and below call for 100 mm of core, or 50 mm spaced off the wall on a 100 mm cavity, which reaches similar low-frequency numbers for less material cost, because the air gap extends the porous layer downward in frequency while the panel face keeps taking the mechanical abuse.

Honestly, I would not put 25 mm polyester anywhere near a compressor. I have replaced enough of it after failed acceptance tests to know how that story runs: the room sounds a shade less harsh on speech, the meter moves 2 dB, and nobody can hold a conversation beside the machine. Sound dampening boards under 40 mm total thickness belong in offices, dialogue studios and corridors. Machine rooms start at 50 mm and stay there.

Match the table to your octave-band measurement before you so much as glance at the budget column. Budget follows function in this trade; the moment it runs the other way, the acceptance test collects the difference with interest.

Fire class, the number that kills cheap boards at inspection

GB 8624-2012 grades burning behaviour from A1 down through B1 to B2, and for occupied or machinery spaces B1 is the floor, not the target. The core needs oxygen index ≥32% per GB/T 2406.2 and the fabric facing must hold the same class, since a B1 core behind B2 cloth reads as a B2 wall at inspection. Our compressor room file carries one combustion report from a CMA-qualified lab covering core and facing together, matched to batch numbers. That sheet passed fire inspection on first submission and moved handover forward by two weeks.

I once watched a school gymnasium in South China strip an entire ceiling of budget sound dampening boards because the facing failed the burn check at final acceptance. The panels had been up four months. Removal, re-purchase and re-installation cost more than triple the original saving. These are building materials first and acoustic products second, so treat the fire file as part of the product itself.

(Ask the supplier to staple the batch combustion report to every delivery note. Public project inspectors match batch numbers panel by panel, and one missing sheet can stall handover for weeks while a retest gets arranged.)

Wall panels and ceiling panels are not interchangeable

Sound dampening wall panels take hits in working plants: trolleys, pallets, forklift masts. That means perforated steel facing of 0.8 to 1.2 mm over the mineral wool core, with a fiberglass tissue layer between steel and wool so fibres cannot drift through the perforations. Keep the perforation rate at 20% or higher, otherwise the facing starts reflecting the very band you paid to absorb. Fabric-wrapped sound reducing wall panels belong where hands and furniture are gentler: offices, lobbies, meeting rooms, broadcast studios.

Ceilings answer to weight and access instead of impact. Sound reduction ceiling panels over a live plant usually mean 600 x 600 mm mineral wool tiles at NRC 0.55 to 0.60, or vertical baffles where ducts and cranes make a full ceiling impossible. Baffles are the underused option in my experience, since both faces of each blade see the room, so a given tonnage of material absorbs more than a flat layer of the same footprint and every service overhead stays reachable. In the compressor hall we ran perforated steel on the walls and a baffle grid above at 600 mm centers, and that combination pulled mid-band reverberation from 1.4 s down to 0.68 s.

Coverage follows a plain rule: treat at least 60% of total room surface in machine rooms, spread across two non-parallel planes, or the numbers will not close. One treated wall in a boxy room does next to nothing, because the bare parallel faces keep trading energy back and forth.

What actually went into the room that passed

The job is a 320 m² compressor hall, four screw compressors, dominant tones at 125 and 250 Hz, acceptance criteria written against the indoor noise limits in GB 55016-2021. Pre-treatment reverberation measured 2.3 s at mid-band and the operator platform sat at 84 dB(A). The specification was supplied by Sino Acoustic Panels (Export), factory-direct, with their engineers on site for measurement and layout before any quote number existed, and every figure below matched the CMA lab reports line for line when the goods landed.

\\\\n
• Walls: 50 mm perforated steel sound dampening wall panels, 0.8 mm steel, 22% perforation, 80 kg/m³ rock wool core, class B1 per GB 8624-2012
\\\\n
• Ceiling: 100 mm vertical baffles at 600 mm centers, same core class, oxygen index 34%
\\\\n
• Coverage: 68% of combined room surface across walls and upper volume
\\\\n
• Result: 0.68 s at 500 Hz to 1 kHz, operator platform 78 dB(A), acceptance signed at first inspection
\\\\n

The acceptance measurement ran at seven microphone positions per the acceptance sheet, and every third-octave band from 125 Hz upward landed inside the GB 55016-2021 limits with margin to spare. Paper like that ends up framed in the workshop.

Six decibels at the operator platform dropped the room out of the mandatory hearing protection zone, and the plant bought fewer square metres than their first contractor proposed, because that first quote padded thin panels everywhere while this layout put density and thickness only where the spectrum demanded them. Spec hard where the tones live, standard-issue everywhere else. That is the whole discipline, and it is why the paperwork cleared in one visit.

One check that settles any panel quote

Before signing anything, ask the vendor for one document set: the third-octave absorption curve from 100 to 4000 Hz per GB/T 20247-2006 with the mount stated, the batch combustion report per GB 8624-2012, and the formaldehyde emission result per GB 18580-2017 showing E1 at ≤0.124 mg/m³. Three sheets of paper. A factory-direct supplier with in-house testing sends them inside a day; a reseller starts explaining why the general catalogue page is good enough. One email, and you know which type of vendor is on the other end.

Then read the curve against your own spectrum. Find your dominant band on the horizontal axis, run a finger up to the curve, and if the absorption coefficient there sits below 0.60 for a machine room, that panel will not close your acceptance no matter how friendly the price looks. This is the judgement call I make on every quote that crosses my desk, and it has a cleaner hit rate than any rule of thumb I was taught as an apprentice.

If the vendor answers a document request with a discount instead of a PDF, you have your answer. Price concessions do not survive an acceptance test; test curves do.

How much do sound dampening panels cost per square meter?

Factory-direct, 25 mm PET panels run USD 8 to 15 per m², fabric-wrapped 50 mm rock wool panels USD 18 to 30, and perforated steel sound dampening panels with an 80 kg/m³ core USD 25 to 45 depending on steel thickness and perforation pattern. Buying through a distributor adds 15 to 25% on top of those figures.

What NRC rating does a machine room actually need?

Target NRC 0.90 or higher on a 50 mm minimum core, then judge the 125 to 250 Hz values on the third-octave curve. A panel printing NRC 0.95 with 0.45 at 125 Hz will fail a compressor room acceptance. NRC alone never settles a machinery spec.

Do sound dampening panels block noise from the next room?

No. Sound dampening panels cut reverberation inside a room, typically 4 to 8 dB(A) at distance from the source. Noise passing through a wall needs mass and decoupling: mass-loaded barriers, double stud frames, sealed penetrations. Absorption and insulation are different physics, and confusing the two is the classic buying error.

How long does installation take for a 300 m² machine room?

With a four-person crew, a 300 m² machine room takes 6 to 9 working days for walls plus baffles, including scaffolding and keel work. Perforated steel arrives prefabricated to grid, which keeps the job under two weeks. A fabric-panel office of the same area finishes in 3 to 4 days.

What fire rating do panels need in public buildings?

B1 per GB 8624-2012 is the minimum for occupied spaces, with oxygen index ≥32% per GB/T 2406.2 on core and facing. Metro and hospital projects increasingly specify A2. Keep batch combustion reports on site, because inspectors match batch numbers at handover.

Can I install sound dampening wall panels myself?

PET and fabric-wrapped panels, yes: construction adhesive plus impaling clips, a laser level and one helper. Perforated steel needs a keel system, lifting gear for full-height sheets and torque-controlled fastening, so leave that to a crew. Open seams and unsealed edges leak the low band you paid for.

Send your floor plan or octave-band spectrum to the Sino Acoustic Panels (Export) engineering desk through the contact page or the WhatsApp number listed on their site, and the free measurement, layout proposal and factory-direct quote with the full test file attached will land in your inbox within two working days.