Target and Spec Limits
Target values or specification limits for the critical parameters might have been developed as part of the QFD/first House of Quality effort, as discussed in Chapter 7. The specification limits are involved in the calculation of the P/T ratio in the measurement system analysis, the design capability analysis, and the tolerance allocation topics discussed in later sections of this chapter.
If the critical parameter is a lower-is-better type parameter, then it will generally just have one specification limit, the maximum. Examples of such one-sided parameters include leakage currents, defects or defect densities, costs, weight, delay times, and power consumption. The target in this situation could be half of the specification limit or maximum, or perhaps an achievable low value that would represent a value considered desirable for the customers.
Similarly, if the critical parameter is a higher-is-better type parameter, then it will have one specification limit corresponding to the minimum. Examples include battery life, drops-to-failure, mean-time-to-failure (MTTF), efficiency, and resolution. Some of these examples are bounded on both sides by the nature of the metric or by physics; for example, percent efficiency is bounded by 0 and 100 percent, even though it is considered a higher-is-better type parameter. The target in this situation could be twice the lower specification limit, or an achievable high value that would be considered desirable for the customers.
If the critical parameter is a target-is-best type parameter, then it will have both an upper and a lower specification limit. Examples could include total radiated power (TRP) for a transmitted signal and some timing requirements in a clocked system constrained by issues such as race conditions. Generally, for two-sided limits, the target will be midway between the upper and lower specification limits; however, there will be exceptions to this, such as situations where the critical parameter is believed to follow a lognormal distribution, in which case the target might be the geometric average of the upper and lower specification limits (that is, the square root of the product of the upper and lower specification limits). Alternatively, the target is an achievable value that would be considered desirable by most of the customers; ideally, if the manufacturer could produce all parts with exactly that value, the customers should be satisfied (if not downright ecstatic).
Companies are rife with examples of problems with measurement systems analysis (MSA), capability indices, SPC, and customer issues that trace back to specification limits set arbitrarily, such as to some target ±10 percent. The specification limits should be based on what is needed to meet the customers' expectations—and, subject to that consideration, the spec limits should be as wide apart or as generous as reasonable for the design team. This enables the design team to have the best chance of success in meeting the specifications with high confidence, and creates a high likelihood that the customers will be satisfied by the result of the design teams' innovation, optimization, and robust design of the product.
In some instances, appropriate specification limits may be hard to pin down. One possible cause for this fuzziness might be that different customers or sets of customers may have different expectations. There are at least three alternative approaches that can be used to deal with this issue: the best and widest-spaced compromise can be selected to satisfy the largest groups of key customers, the product can become multiple products each tuned to the expectations of different customers, or the characteristic can be designed to be tunable, programmable, or selectable by the customers.