Parallel paths add, they do not average
Sound reaching the far room does not care which route it took. Every path is a fraction of the incident energy getting through, those fractions add, and the total is converted back to a level difference. A rating of 50 passes one part in a hundred thousand; a rating of 60 passes one part in a million. Put those two in parallel and the answer is 49.6, not 55. The good path contributes almost nothing to the total, which is the general shape of every result on this page.
Two equal paths sum to exactly 3 dB worse than either one alone, and that is the single most useful sanity check on the whole subject. If you build a second route as good as the first, you lose three decibels. If you build one that is ten decibels worse, you lose almost all of it.
The plenum, worked properly
A partition that stops at the ceiling grid leaves a route over the top: up through the tile in one room, across the shared plenum, down through the tile in the other. The transmission loss of that route is the ceiling rating counted twice, plus whatever the plenum itself takes out, minus a correction for the fact that a whole ceiling collects the sound while the comparison is against one partition.
The defaults show it. A 30 dB ceiling gives 60 dB through two of them, and with 200 square feet of ceiling against a 128 square foot partition the area correction takes 1.9 dB back, leaving 58.1 for that route. Against a 50 dB wall it barely registers — the plenum is carrying under a tenth of the total. Now drop the ceiling to 20 dB and the same route comes out at 38.1, which is worse than the wall by twelve decibels, and the pair together behave like 37.8. The wall was rated 50. The rooms get 37.8, and no amount of work on the wall changes it.
The area correction is worth understanding rather than accepting. Double the room while keeping the same tile and the same partition and the plenum route gets about three decibels worse, because twice as much ceiling is feeding the plenum. This is why the same detail behaves differently in a small meeting room and an open floor, and why a specification written as a ceiling rating with no area attached is incomplete.
The upgrade that goes nowhere
Run the defaults and the assembly comes out at 47.9 against a wall rated 50 — two decibels lost, which sounds tolerable. Now ask the page what a ten decibel upgrade to the wall buys. It returns 51.4, a gain of 3.5 dB for ten decibels of work, and 35 percent of what was spent. The rest went into paths that did not change.
That is the failure mode this page exists for. Somebody measures a disappointing wall, specifies a better wall, builds it, measures again, and finds almost nothing moved. The wall was never the problem. The order of operations is to list every route, rank them by the share of energy each carries, and fix the top row — then recalculate, because the ranking usually flips once and the next dollar belongs somewhere else entirely.
What is not in here
Structure borne flanking, mostly. A floor slab running continuously under a partition carries vibration from one room to the next in a way that has nothing to do with air, and how much depends on the junction detail, the mass of the slab either side, and whether there is a break in it. The same goes for a stud that runs past the wall junction, a continuous facade mullion, and a rigid duct fixed to both structures. Those figures come from measurement or from a consultant who has done the junction analysis, and putting a guess in the table produces a confident wrong answer rather than a cautious right one.
Nothing here is a measurement or a verdict. Ratings come from test reports for specific assemblies, field performance is measured with an instrument, and whether any result is adequate is a judgement for an acoustical consultant and, where the code touches it, for the authority having jurisdiction.
Questions people ask
Why did my new wall not make any difference?
Almost always because the path that was deciding the answer was not the wall. Run the defaults through the page: a 50 dB partition with a floor path, a duct path and a ceiling plenum path behaves like 47.9, and upgrading the wall by ten decibels moves it to 51.4 — a gain of 3.5 for ten spent. Until the top row of the table changes, the wall specification is close to irrelevant.
How do I stop the plenum path?
The route is up through the ceiling and back down, so the two things that touch it are the ceiling system and whatever sits in the plenum. Carrying the partition to the structure above and sealing it at every penetration removes the route rather than reducing it. A barrier above the ceiling line does something and how much depends on how far it goes and how well it is sealed, which is why the field on the form asks what you credit it with rather than assuming a figure.
Where do I get the number for a flanking path?
From a field measurement, from a consultant who has analysed the junction, or from an estimate you can defend. There is no table that gives it, because it depends on the detail, the mass either side, and how the two structures are connected. A guessed figure produces a confident answer that is wrong, which is worse than admitting the path is unquantified.
Does an area correction really matter that much?
It is worth a few decibels and it moves with room size, so yes for the plenum route. A ceiling gathers sound over its whole plan area and delivers it to one plenum, while the comparison is against a partition of much smaller area. Doubling the ceiling area open to the plenum makes that path about three decibels worse with no change to the tile at all.
Can I add my paths and get a rating for the assembly?
No. What comes out is an apparent transmission loss for the arrangement you described, using figures you supplied. A rating is a laboratory measurement of one assembly under controlled conditions. Field performance is measured on site with an instrument, and no calculation substitutes for it where the outcome matters.