An Investigation into the Failure Characteristics of External PCB Traces with Different Angle Bends

4.1

Maximal Current Testing

Table 2 illustrates the summary statistics, in particular the mean, median and the standard deviation, of the maximal currents for each of straight traces, traces with 45° bends and traces with 90° bends. The median is well known to reduce the effect of outliers in the measurements, which justifies its inclusion. At first glance, the maximal currents for the three trace types appear similar, perhaps indicating that the presence of bends does not clearly change the maximum current rating of a trace despite any over-etching that may occur at bend sites during manufacture. The standard deviation values demonstrate that the maximal currents have larger variability for traces with bends than for the straight traces. This could possibly indicate that the manufacturing tolerances are larger for traces with bends. However, we note from Table 1 that the straight traces, with length 85 mm, are shorter than the traces with 45° and 90° bends, which both have length 145 mm. Therefore, only the results for the traces with 45° and 90° bends are directly comparable with each other.

.

Table 2 Summary Statistics for Maximal Currents

Full size table

It is also beneficial to examine the distribution of the individual maximal currents, in particular to understand whether they have a Normal/Gaussian distribution. Figure 2 compares the histograms of maximal currents for traces with 45° bends and traces with 90° bends, each of which are formed from 60 measurements as illustrated in Table 2. In order to explain the histogram bin labelling, for an example bin of 2.7 A, the measurement in this bin had a maximal current of greater than or equal to 2.7 A, but less than 2.8 A. The distribution of maximal currents for the traces with 90° bends appears somewhat Gaussian, but is left skewed rather than symmetric. The distribution of maximal currents for the traces with 45° bends is clearly multimodal (i.e. having multiple local maxima) rather than Gaussian. For example, looking at the histogram for traces with 45° bends, there are local maxima at 2.4, 2.8 and 3.1 A.

Fig. 2

Histogram of Maximal Currents for Traces with Bends

Full size image

A possible reason for these observations is differences between the PCBs produced for a given trace type. The 60 samples tested for each of the traces with 45° bends and traces with 90° bends were distributed across 4 PCBs, with 15 traces per PCB. Figure 3 compares the histograms of maximal currents for traces with 90° bends as a function of the PCB on which the trace was located. Table 3 illustrates the summary statistics, in particular the mean and the median, of the maximal currents for each such PCB. The distribution of maximal currents for each PCB considered individually is basically unimodal (i.e. has a single local maximum), but the individual distributions exhibit clear differences both in terms of their mean values and variance. PCB 3 has the “peakiest” distribution (i.e. lowest variance in maximal current), whereas some of the other PCBs (especially PCB1 and PCB4) exhibit a relatively large variance in the maximal currents. This demonstrates that PCBs ostensibly made according to the same specification (and in the same manufacturing run) can exhibit significantly different electrical characteristics. From the perspective of characterising the electrical properties of traces with different bend angles, it suggests we not only need to consider a large number of such traces overall, but also a large number of PCBs accommodating those traces, in order to gain visibility of not just trace variations, but also PCB variations.

Fig. 3

Histogram of Maximal Currents for Traces with 90° Bends

Full size image

Table 3 Summary Statistics for Maximal Currents Across PCBs For Traces with 90° Bends

Full size table

4.2

Failure Location Testing for Traces with Bends

Figure 4 compares the histograms of failure locations for traces with 45° bends and traces with 90° bends, each of which are formed from 60 measurements. There are significant differences between the results for the two types of traces. In particular, traces with 45° bends are more likely to fail at the bend than traces with 90° bends. Conversely, traces with 90° bends are more likely to fail on the straight segments of a trace than traces with 45° bends.

Fig. 4

Histogram of Failure Locations for Traces with Bends

Full size image

Figure 5(a), (b) and (c) show thermal images of straight traces, traces with 45° bends and traces with 90° bends respectively shortly before failure occurs due to overcurrent. These illustrate that the temperature at the hotspot or point of failure generally exceeds the maximum measurement temperature of 270 ˚C for the thermal camera in use when the failure occurs. Note that the failure location was not determined based upon these thermal images because the temperature distribution resolution along the trace is too low; instead, the methodology described in Sect. 3.4 was used to determine the failure location.

Fig. 5

Thermal Images of Traces Shortly Before Failure a Straight traces, b Traces with 45° bends, and c Traces with 90° bends

Full size image

4.3

Time to Failure Testing

Table 4 illustrates the summary statistics, in particular the mean, median and the standard deviation, of the time to failure for each of straight traces, traces with 45° bends and traces with 90° bends. It is immediately clear that straight traces take much longer to fail on average than traces with bends, which suggests that straight traces are more robust than traces with bends. The time to failure measurements for straight traces were significantly more variable than those for traces with bends, as indicated by a larger standard deviation. However, the minimum measured time to failure for straight traces was 42 s, and this is greater than the maximum measured time to failure for traces with bends of 40 s, so it is clear that straight traces always take longer to fail.

Table 4 Summary Statistics for Time to Failure

Full size table