Referring to ANSI-A 108, limits for deflection are expressed in AN-2.3 as “floor areas over which tile is directly bonded to subfloor shall not have a deflection greater than L/360 of the span when tested per ASTM C627.”
Here’s how this works: L refers to the length of the span in inches. L is then divided by 360. For example, if the area in question is a 12-foot span, convert feet to inches (12 feet equals 144 inches) and divide by 360 to get the maximum allowable defection.
144 inches divided by 360 = 0.4 inches is the maximum allowable deflection.
Remember, allowances must be made for a live load (people) and impact as well as dead loads (equipment, furniture, tile weight, setting bed). The Marble Institute of America requires a deflection equation of L/720, as dimensional stone requires a more stable substrate.
ASTM C627 is the test method for evaluating floor tile systems using a Robinson Floor Test. The test is designed solely for the purpose of evaluating complete ceramic tile installations for failure under loads. It tests for deflection on various bases, such as mortar for Portland cement installations, concrete for thin-bed installations, and plywood and composition for installations (plywood with a composition board or other sheeting material).
There is some concern in the industry that, with increasing use of large format tile and the expanded use of engineered wood systems (EWS) with longer spans, the Robinson Floor Test may not provide results with suitable accuracy. However, I believe it is perfectly adequate for most applications.
When confronted with unusual floor joist/truss spacing applications, the Tile Council of America (TCA) offers the recently published detail F-147-01, “Wood Subfloor, 24” o.c. Joist Spacing, with Uncoupling System.” I believe this also helps when applied to long joists. F-147-01 limits tiles to no smaller than 4-inch-by-4-inch. It also presents a deflection limit for the span of no greater than L/360 when measured under a 300-pound concentrated load, per ASTM C627. Note that the detail refers to an uncoupling system rather than a cleavage membrane, leading to a proprietary product.
ANSI A-108-99 Section A 2.4.3 cautions that, when dealing with wood backing materials, many panels and non-veneer panels such as wafer board, oriented strand board (OSB), and lauan plywood can expand and contract due to moisture. However, plywood manufactured with fully waterproof adhesive, and with an exposure durability rating of Exposure 1 may be used on residential horizontal surfaces when installed according to paragraph AN-3.4.
The emphasis on stiffer framing is accelerating. The Marble Institute’s L/720 is also the standard used by the Canadian Tile and Marble groups. The APA/EWS has voluntarily raised its standard to L/480, 33 percent higher than L/360.
I am a great believer in isolating the tile and setting bed from any unstable substrate. In many cases, doing so will help avoid problems resulting from deflection, shrinkage, substrate expansion and racking.
I would like to thank David deBear, CTC for excerpts from his research report for the Ceramic Tile Institute of America, as well as ANSI and the TCA for the use of their details and standards.