In our never-ending search for truth in specifications, we often lose sight of reality. We're inundated with advertising, product data, test reports, and white papers; where once specifiers complained about a lack of information, we now struggle to keep up with what we receive. We can't know what we don't know, and we have no time to evaluate what we have seen. As if that weren't bad enough, we often find that what we thought we knew isn't true.
As an example, consider building insulation. The way it works and its value have been understood since antiquity, and until recently we have felt comfortable with evaluating, specifying, and detailing various types of insulation. And then everything began to unravel.
Several years ago, the accepted nominal R-value of polyisocyanurate insulation was reduced. Until then, manufacturers conditioned insulation for six months prior to testing and those properties were published, even though it was known that the R-value continued to decline after six months. Architects typically assumed the published values to be constant and gave no more thought to the issue.
When plotted, the insulating value of foam insulations is seen to follow an asymptotic curve, always decreasing, but at a rate that also decreases. The result is a curve that always decreases toward a limit, but never reaches it. The term LTTR (long-term thermal resistance) became part of our vocabulary, a method to calculate a nominal R-value closer that more closely represents the properties of insulation. The reason for the reduction is the off-gassing of the blowing agents, which slows down as the concentration in the cells decreases.
I recall the discussion surrounding the introduction of LTTR. Maybe my memory is faulty, but as I recall it was focused on polyisocyanurate insulation. As it happens, extruded polystyrene also loses R-value. More important is the definition of LTTR, “the time-weighted average of thermal resistance over 15 years.” In other words, it does not show the R-value of insulation that remains in place longer than 15 years. What does that mean for buildings designed for a service life of 50 or 100 years? How will the operating costs change? The nature of the curve, which has been shown to be reasonably accurate, means there won't be much change, but it's a question worth asking.
The reported performance of insulation is itself suspect. In Info-502: Temperature Dependence of R-values in Polyisocyanurate Roof Insulation, Building Science Corporation shows the results of testing 15 samples of polyisocyanurate insulation. The range of R-values is about 1.5 points at 25 degrees mean temperature, a 30 percent difference. Yet all manufacturers, with one exception, claim the same, ASTM-standard R-value for their products. The exception is a company that reports a 50-year R-value for its XPS. As you might expect, it's not the usual 5.0 we all use, but 4.2. Doesn't it make sense to calculate energy use and mechanical system performance based on the lower value?
One of the problems with all testing is confidence in the results. ASTM and other standards-producing bodies often refer to two aspects of measuring this confidence - repeatability and reproducibility. In theory, a given test method would always produce the same results for the same material. In practice, that rarely occurs due to a number of influences. Because humans are involved, results can vary even when tests are performed by equipment. There may be slight differences in the type of equipment used, and even when equipment is regularly recalibrated, it may not give the same readings.
Briefly, repeatability is a measure of consistency of a test when performed by the same person in the same laboratory using the same equipment, within a short period of time. In contrast, reproducibility measures the differences caused when a test is performed by different people in different locations using the same equipment. I looked at several ASTMs and found both repeatability and reproducibility reported as 95%, which means that the accuracy of test results will be consistent.
The point of all this is that we seem to be caught in a Spock syndrome, citing performance values of unsupportable accuracy. If, for example, you specify an R-value of 5.0, that is the number you expect after rounding. In practice, of course, you would accept a higher R-value regardless of rounding, but the minimum would be 4.95. But when test results show a variation of 30 percent, and manufacturers simply claim a nominal value stated in a standard, what are you really specifying, and what are you getting? Sprayed products present even more problems. I wasn't completely comfortable with material that went on at 60 mils; I cringe when I see product data that claims only 6 mils is required.
I understand the problems involved in manufacturing, application, testing, and so on, and we need some way to compare materials, but I question the validity of Spock-like precision.