News

Rework of a defective or incorrectly placed component occurs when it fails visual inspection or electrical test. The problem can be attributed to a number of causes related to the material or manufacturing process. The most common methods for repairing defective components are by using a solder tip that fits the component (for small devices) and hot air (for large multi-leaded fine-pitch packages).

Small through-hole components are typically solder-sucked with a special tip designed for this purpose. Multi-pin through-hole connectors and components are typically taken to a solder pot for rework. (Solder pot rework is not covered here.)

The aim is a fast and reliable rework process with minimal disruption to the manufacturing flow. Pb-free (LF) rework makes this more difficult due to the increased melting point of the solder. The solder melting point can range from 34°C higher for tin-silver-copper alloys to 44°C higher for tin-copper alloys.

A variety of components were evaluated with good results. This paper covers three years of LF rework development on yield and reliability. Table 1 provides LF components and board test vehicles. Table 2 shows equipment and materials used.

Network interface card. The first Pb-free rework was done on a functional network card. A few selected locations, shown by red circles in Figure 1, were reworked to get some initial reliability data. Unfortunately, the component surface finish information was unavailable. Most likely the surface finishes were tin. Two manufacturing technicians did all the rework on 24 boards with no issues. All rework was done with 700°F soldering tips. Reliability results showed no failures and all boards passed functional tests. Reliability tests were 1000 cycles of temperature cycling from -45°C to 85°C, 1000 hours of bake at 125°C, mechanical shock and vibration. Cross-sectional analysis of a few selected components showed no cracks or abnormalities.

Component surface finish. A test vehicle became available which permitted a more in-depth study of different component surface finishes and a rework matrix was designed to cover a variety of components. Figure 2 shows components selected for rework circled in red. Passive and daisy-chained components could be electrically tested but active components could not be tested. Active components included all the oscillators, crystals and polyswitches. A total of 14 of 26 components could be electrically tested using a digital multimeter.

Two manufacturing technicians did all the hand-solder rework with one of them being new to the Pb-free rework process. All hand-solder rework was done with 700°F solder tips. Large SMT components and SMT connectors were removed with a hot-air rework machine. There was some initial trouble using the hot-air rework machine as the technicians became accustomed to Pb-free solder temperatures. After increasing heater temperatures and extending the airflow time, the technicians were able to remove the large SMT components and connectors with no major issues.

The rework results initially were not as good as the previous Pb-free rework done on the NIC card. Some oscillator leads did not get solder to the pads, while some through-hole pads experienced lifting and board damage. Figure 3 shows a couple of different type of oscillator leads that failed to solder to pads. Figure 4 shows through-hole damage that occurred during desoldering and re-soldering.

Many of these rework issues occurred because 800°F solder tips were not available to the manufacturing technicians at the time of the study. If hotter tips had been available, failures would have been decreased. The technicians were familiar with the SnPb solder that permitted them to go at a faster pace than could be achieved with the Pb-free solder alloy.

A total of 23 boards were put through reliability tests. Eight boards went through 1000 to 1500 cycles of temperature cycling from -45° to 85°C, four boards went through 1000 hours of static bake at 125°C, three boards went through 1000 hrs. of humidity at 85°C/85%RH, four boards went through mechanical shock and four boards went through vibration. All components that could be tested passed. Cross-sectional analysis was done on many of the components and the only cross-sections that showed signs of stress were the temperature cycle solder joints.

One crystal with SnBi showed total separation from the bottom of the board pad and also cracks started to propagate into the bulk of the solder joint after 1500 cycles of temperature cycling (Figure 5). Resistor-packs also showed many cracks starting at 1000 cycles of temperature cycling. This was not unexpected since SnPb resistor-packs also exhibit the same cracking characteristics at similar temperature cycles. Figure 6 shows the typical resistor-pack solder joint cracking on a tin surface finish.

Through-hole cross-sections showed minor cracking at the top and bottom of the board pads. There were no internal connections on this board for the through-hole parts. Figure 7 shows a typical crack pattern after 1500 cycles of temperature cycling.

Desktop board. A fully functional Pb-free desktop board was next up for additional rework development. This is a four-layer board and could be tested for short and opens using an in-circuit test (ICT) fixture. Full functional test could also be run after passing ICT. Again, a variety of SMT and through-hole components were selected for rework (in red circles in Figure 8).

Four manufacturing technicians performed all the hand-solder rework with two of them being new to Pb-free rework. All hand solder rework was performed with 700°F solder tips. Large SMT parts were removed with a hot-air rework machine. There were again problems on the rework as the new technicians had trouble with fine-pitch components. There were 15 locations that had solder shorts after rework (Figure 9). There were also some broken and lifted pads after rework that required repair. All boards were fixed and passed shorts and opens test. The boards then went through a full functional test and passed.

A total of 20 boards were put through reliability tests. Eight boards went through 1000 cycles of temperature cycling from -45° to 85°C, four boards went through 1000 hours of static bake at 125°C, four boards went through 1000 hrs. of humidity at 85°C/85%RH and four boards were sequentially test through vibration and then mechanical shock. The 20 boards passed functional testing after completing all reliability tests. Cross-sectional analysis showed some minor cracking on a resistor-pack with tin finish (Figure 10).

There was also some pad lifting on the through-hole reworked components but this did not cause any reliability failures. Figure 11 shows typical damage that occurred to the board after rework was completed.

Server test. A 0.093", 10-layer test board was designed for expanding the Pb-free envelope into the server space. This new board would provide additional reliability data on the Pb-free hand solder rework process. A few locations, circled in red (Figure 12), were selected for rework.

One experienced manufacturing technician did all the hand solder rework. A total of nine boards were reworked and all rework was done with 800°F solder tips. The large fine-pitch component was removed with a hot-air rework machine. No rework issues occurred. All the components were tested with a multimeter and passed. The large fine pitch component was daisy-chained. The nine boards were put through 870 to 1200 cycles of temperature cycling with no failures. Cross-sectional analysis did not show any cracks in any of the SMT components. Some small solder cracking did occur on the through-hole capacitor.²

Unfortunately, it was discovered that the through-hole capacitor locations were not tied to any internal copper layers. This forced another revision of the test board to tie the capacitor locations to two, three and four copper layers. Also, with this revision some of the capacitor locations would have thermal relief and directly connected pads (Figure 13). With the new revision of the board, work began on achieving good capacitor hole-fill after rework. Initial rework showed that the average hole-fill was approximately 26%. This was not acceptable, so preheating the board was tried next.

A series of experiments and lots of cross-sections showed that a good preheat temperature for the board and component is 125°C. At this temperature, a 95% average hole-fill was achieved with two, three and four copper layers combined. Figure 14 shows the copper layers of the server test board.

Individual copper plane layers were examined. This showed that two layer direct attached and thermal relief pads did much better in hole-fill than the three and four layer connections. The 3 layer locations (two, three and nine) are direct attached and the four layer locations (two, three, eight and nine) are thermally relieved (Figure 15). The four layer locations also demonstrated a hole-fill average 8% greater than the three layer locations because of the thermally relieved copper connections.

With the hole-fill data suggesting that thermally relieved pads preformed better than the direct attached pads, statistical analysis was performed on the data to see if this was true. Figure 16 showed that this is a correct observation. Even though the circles slightly overlap there is a statistical difference between them. With a t-Test number at 0.043 which is less than an alpha number of 0.05, making this statistically different.

A total of 63 capacitors were reworked for temperature cycling analysis on the various copper layer locations. Thirty four locations were cross-sectioned after 1000 cycles to see if any failures occurred. Some minor solder cracking occurred at the top of the solder joint fillet but none at the bottom (Figure 17).