Reliability test plans are critical to the success of any new electronic medical device or technology. The tests selected must be stressful enough to identify defects but also show a correlation to realistic use environments. The preferred test plan development approach is to use a combination of industry standards and Physics of Failure (PoF) techniques. This results in an optimized test plan that is acceptable to regulatory bodies, management, and users. But even the most streamlined qualification testing can take months or years, and can still result in a design that may not meet reliability requirements.
Significant opportunities exist for medical electronics. There is an increasing public awareness and perception of issues with medical devices due to high profile recalls and adverse events. Furthermore, the perception that one’s life may be dependent upon these products creates a powerful emotional effect, which means that assuring reliability becomes of critical importance.
Once a product has been designed, data files are created that manufacturers use to build the product. Now, software tools exist that can take those very same manufacturing files and build the product “virtually.” The virtual product can then be subjected to the anticipated user environment to determine the reliability and identify any weak areas. No longer are months required to build prototypes of the product and subject them to reliability tests — design changes can now be made and reviewed almost instantly. What-if analyses to optimize a design can be done in a fraction of the time it takes to even get on a typical designer or fabricator’s schedule. Mechanical simulation can be run in minutes, not days, providing the answers needed to build a product better, faster, and at a reduced cost.
The greatest potential for reliability and quality improvement is also when the cost is lowest: before the first prototype is ever built. The most requested qualification tests have been packaged for the medical device designer and include: virtual shock testing, virtual vibration testing, virtual thermal cycling, and virtual CAF (conductive anodic filament) testing.
Medical device designers can use reliability modeling software to get a jump on testing by reducing the risk of failing qualification tests. Using a physics-based automated design analysis (ADA) tool is like testing the design before a single prototype is built. Design weaknesses can be identified early, allowing for iterative design improvements while maintaining an aggressive design schedule. The ADA tool can also identify areas where specific, targeted testing would be beneficial.
Design is always an exercise in tradeoff analysis. Some choices are driven by cost, some by function, some by form, some by risk and safety. The ability to rapidly compare design choices and understand quickly how those choices affect reliability helps ensure that both the likelihood of premature failure and the long-term cost of ownership remains low.
The use of the reliability modeling tool is limited only by the needs of the user. There are a wide range of problems that it can either solve or provide insight on, in regard to designing a reliable product. Some of these uses are identified below:
• Determine thermal cycle test requirements needed to replicate the use environment.
• Determine proper environmental stress screen (ESS) conditions.
• Determine impact of component package modifications or changes to the circuit board laminate.
• Determine impact of changing to Pbfree solder.
• Determine expected failure rates and times for a given set of conditions.
Setting reliability goals and targets at product design kickoff is crucial to success. Desired lifetime and product performance metrics must be identified and documented. The lifetime may be defined by regulatory requirements, company requirements or, most importantly, when the customer needs will be satisfied. These lifetime goals should be actively used in development of the product qualification activities.
For illustrative purposes, two examples of medical device manufacturers using modeling for reliability goals are shown below:
1. Total service life of 6 years where year 1 includes transport and storage and years 2–6 are in service with a patient. The maximum acceptable failure rate is set at 5% after 6 years. This example will be explored further in this article.
2. Goal of 10 years' life with 98% reliability.
For medical devices, it is vital to consider transportation and storage life conditions in the overall test plan. Many times, these environments are significantly more severe than the actual product use environments. Transport and storage are frequently overlooked and underestimated as failure sources. Xerox published results (http://www.ista.org/forms/LEINBERGER_Dimensions06_paper.pdf) from a one-year study of ocean container temperature and humidity conditions for shipments between Asia, Europe, and North America. Highlights included port, sea, rail transport, and storage conditions, as well as a number of unexpected conditions. Fig. 1 shows some of the environmental conditions that can be reached in various transportation modes. For most medical devices, these conditions would be far more severe than the expected use environment.
Product performance metrics include returns during the warranty period, survivability over lifetime at a set confidence level, and mean time between failures (MTBF) or mean time to failure (MTTF) (try to avoid these unless required).
Some companies set reliability goals based on survivability, which is often bounded by confidence levels such as 95% reliability with 90% confidence over 15 years. The advantages of using survivability are that it helps set bounds on test time and sample size, and does not assume a specific failure rate behavior (decreasing, increasing, steady-state). The disadvantages are that it can be reinterpreted through MTTF and MTBF.
MTTF / MTBF
Defined as mean time to failure (MTTF) or mean time between failures (MTBF), MTTF applies to non-repairable items while MTBF applies to repairable items. MTBF is typically calculated through a parts count method. Every part in the design is assigned a failure rate. This failure rate may change with temperature or electrical stress, but not with time. Failure rates are summed and then inverted to provide MTBF. Most calculations assume single point of failure while some calculations take into consideration parallel paths.
MTBF/MTTF calculations tend to assume that failures are random in nature and provide no motivation for failure avoidance. It is very easy to manipulate these numbers to reach a desired MTBF simply by modifying quality factors for each component. These calculations are also frequently misinterpreted. For example: A 50K hour MTBF does not mean there will be no failures in 50K hours (but rather that nearly half the products will fail in this time period). Basically, these calculations are a better fit toward logistics and procurement and not failure avoidance activities.
MTBF calculations do not take into account wear-out mechanisms such as solder joint failures. Common MTBF calculation assumptions include:
• Perfect Design
• All stresses/use data known
• Failures are random
• Any part failure causes a system failure
• Parts models are up-to-date and accurate
Reliability Test Conditions
The appropriate test conditions can be determined by first generating a solder joint fatigue model based on the expected field conditions of the product. The percent failure at the required life can then be determined through ADA modeling of the product. The model is then rerun using the desired thermal cycle test conditions (say 10 to 50 °C). The number of cycles required to generate the same percent failure shown in the previous model is how many cycles are required (with no great percent failure). Naturally, the number of cycles would be increased if the sample size is reduced.
Early in the design phase of a product is the best time to run various what-if scenarios for the design. These might include experimenting to determine where the mount point locations should be in order to reduce strain on sensitive components. One may also run thermal cycle modeling using the various package options available for critical integrated circuits. The impact of material choices such as the circuit board laminate type or the solder alloy type (SnPb vs. Pb-free) can also be evaluated.
Example of Potential Environment Model Conditions for an Implanted Medical Device
A potential implantable medical device with 85 parts and 274 holes/vias was selected for modeling and analysis (see Fig. 2). Component types modeled included 0402 capacitors and resistors, QFPs, QFNs, TSSOPs SOICS, and SOTs. A total service life of 6 years was selected where year 1 includes transport and storage and years 2–6 are in service with a patient. The maximum acceptable failure rate was set at 5%.
Two profiles were established for thermal analysis. One was for year 1 transport and storage conditions and one was for the use conditions in years 2–6. A minimum temperature of 12.5 °C. a maximum temperature of 47.5 °C, minimum dwell time of 720 minutes, and 60 cycles were used as transport and storage conditions. For use conditions, a minimum of 32 °C, a maximum of 35 °C, a minimum dwell time of 720 minutes, and 305 cycles were entered. An above average PCB manufacturer setting was used along with an assumption of supplier based qualification (no additional bare PCB qualification performed like Interconnect Stress Test /IST). Based on these conditions, the product will be successful at meeting its reliability targets as shown in Fig. 2. As one might expect, thermal cycle failure is not a concern in an implantable. In this product type, component failure is the main contributor to the failure rate. The combined results of the all the modeling tests performed are less than 5% at the end of 6 years. In terms of prediction, the 5% probability of failure would not be exceeded until sometime beyond year 12.
However, the conductive anodic filament (CAF) analysis did turn up areas of concern (see Fig. 3). The CAF module identified 13 pairs of holes which were closer than the desired 20mil separation. Redesign for further investigation could be performed to reduce the likelihood of CAF formation.
For this example analysis shown, tin lead solder was used. The analyses were also performed using Pb-free alloy SAC 305, but the solder did not significantly impact any reliability results. Now that a model of the device has been created, it is simple to change reliability goals and environments and assess impact. Fig. 4 shows the impact of changing life from 6 to 15 years. Now, the product will not be capable of achieving its reliability goals since the combined failure rate exceeds 5% before year 15.
In conclusion, designing in reliability up front pays off immensely over the life of the product. To date, there has not been a quick, easy-to-use method of estimating the wear-out life of an electronic product. Automated design analysis (ADA) software is designed to fill this need and does so by allowing a rapid assessment of electronic systems reliability utilizing Physics of Failure (PoF). ADA software modeling is a powerful reliability tool that can be used by the entire engineering design and management team. It allows the reliability group to get involved in the design process as well. They can now better quantify tradeoffs before the product is ready for testing and optimize testing plans based on model predictions of potential issues. This is the future of reliability: the integration of design rules, best practices, and use of a physics-based understanding of product reliability.
This article was written by Cheryl Tulkoff and Randy Schueller, Ph.D., for DfR Solutions, College Park, MD. For more information, Click Here .