|This is a computer-aided-design (CAD) image of the reflow pallet.
Contract manufacturers (CMs) producing electronic printed circuit board (PCBs) often must juggle multiple ovens for properly managing reflow processes. Although multiple ovens can bring added capability, they can also bring a degree of difficulty to the validation process, and most CMs would rather simplify their processes for optimum manufacturing effectiveness. By profiling four industrial ovens, a greater understanding was gained in modeling and managing those ovens for CM applications, in particular for applying reflow soldering.
The four ovens were modeled using typical paste parameters, including maximum temperature, time above temperature, and slope to evaluate oven repeatability and variations. In a production environment especially, no two ovens are the same, so creating the profile of one oven for a model does not necessary mean that it works as a model for the other ovens. Creating a profile at one point in time does not necessarily mean that the profile will apply at a later date, even for the same oven. As a result, two types of experiments were needed to better understand oven behavior. One experiment is aimed at analyzing the variations within a single oven, while the other explores variations across multiple ovens. In relation to circuit board paste specifications, there parameters of interest were maximum temperature, time above temperature, and maximum slope.
It was felt that a meaningful number of samples for a statistical analysis of oven performance for reflow soldering is about 30 samples. For performing an analysis at a facility with four ovens, about 120 samples would be needed. For such an analysis, developing a method to collect and accurately manage the data is as important as the data itself. In simulating profiling for PCBs, special attention must be given to the tooling material, since it will exhibit multiple heat fluxes throughout its lifetime. The tooling must include a data logger, thermocouples, and a holder, while providing data as its primary output.
Temperature data loggers are used throughout the electronics industry for profiling PCBs. They are typically sold with dedicated software that accurately processes and stores the data after profiling. The use of such data loggers might be the most cost-effective means of validating oven programs for a PCB assembly.
The medium between the operating environment and the data logger is a K-type thermocouple. The use of multiple thermocouplers is important for measuring temperatures of a defined mass and the surrounding air. To best capture the thermal properties of a PCB, several thermocouples can be embedded into different masses. The embedded material should be of the type that can handle extreme heat and significant temperature cycles. PCBs consist of copper layers and dielectric material, such as FR-4 which is a composite fiberglass material. Such material is not well suited as a holder because of modifications that occur over a small number of thermal cycles. The tooling holder brings these several test components together into an assembly that can handle repetitive thermal cycles without significant degradation in material properties. The holder design should consider the importance of thermocouple position, air temperature position, and data logger position.
This provides an idea of the model that will be used to generate data for the oven models. Repetitive runs on a designated tooling device will simulate a PCB. Individual runs will be tabulated for variations across time and multiple ovens, and the variations will be compared to the paste mix tolerance. DURAPOLE® was used to fabricate the holder. This is the same material used to make solder pallets. The body for the tray was 9 x 18-in., with enough room for a data logger. The packet includes four K-type thermocouple hookups: two aluminum pucks with thermocouples epoxied into them and an air thermocouple wedged between ultrahigh-temperature Garolite to hold it in place.
Repeatability is measured by consecutively running the pallet through the oven and tabulating the data over a short period of time. This measurement would also be conducted once each month to record long-term variations. Each measurement run requires about 20 minutes: 5 minutes to set up and run through the oven and 15 minutes to cool down in preparation for the next run. Acquiring 30 samples of data equates to about 10 hours of line time to complete a set of control data for one oven. In a production environment, this can be accomplished by scheduling sufficient downtime or running the set when there is an adequate break in normal product. The latter approach was chosen in this instance. The second set of measurements, for analyzing variations across different ovens, requires a great deal of data management.
|These plots of long-term data help identify performance trends for the four ovens.
For these data gathering exercises, the initial process window is defined by the reflow solder paste. For a CM, many reflow solder pastes can be run in a facility, depending on the assembly in production. It is important to know all of the solder paste requirements and evaluate specification limits for the smallest range among those requirements. This will provide a realistic process window based on the mix of pastes.
When comparing ramp-to-spike (RTS) profiles for reflow solder paste, three key parameters are temperature, time, and slope. Following the initial control set, these variables can be compared to determine each range as a whole. In addition, the control set can be evaluated over an allowable statistical window. When developing a Six Sigma process, given the data collected, the ranges can be broadened to support the 99.99 percent expectancy of a Six Sigma process. By applying these equations for upper control limit (UCL) and lower control limit (LCL), new bounds can be found (see Fig. 4)..
Since small differences can exist between ovens from different manufacturers, it can be useful to record the manufacturing information for each oven. By using control limits set by a baseline run, it is possible to develop a sense of month-to-month consistency and how the ovens perform over time. Data snapshots of different months can provide a closer look for each oven in relation to the other ovens.
An analysis of initial data results revealed a 10°C reflow solder paste window with a 5.3°C temperature range on the control set for maximum peak temperature. This effectively makes a new paste window of 4.7°C. A range of this magnitude can be more difficult to hit than the original paste specification (over 50 percent). For peak temperature improvements, improvements can be made in the oven or ovens or improvements can be made in the paste specifications.
When analyzing peak-temperature data from the different ovens, it became apparent that one oven didn't track the rest of the group, exhibiting a much higher temperature range and median temperature than the other ovens. This can mean something wrong with the control data, or something wrong with the oven, and it provided the means for examining the oven.
When analyzing the paste specifications, all of the temperature ranges were about 20°C or above except for a leaded no clean (Pb NC) oven at 10°C. By increasing this unit closer to the other three ovens, it would yield a much larger window. For the maximum time above temperature data, the results revealed a 40-s paste window with a 13.5-s range for the maximum time above temperature, effectively creating a new time above temperature window of 26.5 s. This is manageable in relation to the paste specification.
The final group of data in the control set involved the temperature slope, with results revealing a 0.90°C/s paste window with a 0.47°C/s range on the control set for slope, effectively making a new slope window of 0.43°C/s. While this may seem like an unmanageable window, it is a bit bloated. The paste specification is an average rise over a selected sample. If the sample is increased, with less resolution, the results are more favorable. If the sample is decreased, higher resolution is provided, but the results are less favorable. This control was performed at higher resolution.
Paste Spec Limits
It should be noted that the limits for the paste specification were set by the lead-free no-clean (LF NC) oven, which has a higher lower limit and low higher limit compared to the other ovens, resulting in a narrow specification window. The LF NC ramp could be improved upon to bring that oven more in line with the other ovens.
By reviewing the monthly control data, it is possible to see that the peak temperature has trended downward over a five-month span, dropping on average by about 1°C. This indicates a cause for action and, at this current rate, shows two ovens dipping below the specified control limits. Some of the data, as with the control data set, indicates something amiss with one of the four ovens (Pb NC) compared to the data from the other three ovens. Because of this collected data, the manufacturer of the Pb NC oven was contacted regarding its peak temperature parameters. The manufacturer agreed that a 20°C window is more representative of the oven's reflow solder paste capability. They explained that the initial specifications were set up a long time again and since that time, the firm has performed numerous tests to validate the claim.
Successfully raising the window to 20°C sets the process in a much more workable window. Taking the original Six Sigma oven variance of 5.3°C and subtracting that from the new allowable paste window of 20°C equates to a more reachable window of 15.7°C, which is essentially the new paste window. For each peak temperature specification, the variance of 5.3°C will be subtracted to yield a smaller window. This will ensure that a profile initially tested for an oven will transfer successfully to another oven and still be within reflow solder paste parameters.
Workable Time Above Temperature
The time above temperature window was found to be workable from the data. On the lowest process window, it was shown to be 26.5 s. The oven manufacturer also shed some light on the reflow solder paste slope. They explained that as long as the slope remained under about 2°C, there should be no problems, even with oven variances.
Having the monthly data for analysis was quite revealing, showing an overall shift for all of the ovens colder on average about 1°C for peak temperature. This was not expected. Troubleshooting theory points to causes that affect all ovens: environment and exhaust systems. For this set of ovens, these areas need to be evaluated going forward.
If the ovens were all to trend 1°C colder during the summer months and then go 1°C hotter during the winter months, it would not pose a problem. If the temperature variations are tied into the control parameters on the profiling setup for the ovens and monitored as monthly deviations, the oven systems can be still operated effectively. But if the average temperature deviates more than 1°C, it can lead to other problems with the ovens and with the solder reflow process for the facility.
In terms of determining the capabilities for the ovens, when trying to achieve peak temperature for the control data set, a delta of ±2°C was found among all four ovens. Calculating their Six Sigma limits put the process expectancy at 99.99 percent that the profiles will match within 5.3°C across all four ovens.
For profiling initial assemblies, peak temperature was found to exhibit too tight a tolerance for conformity of the multiple ovens. But these tolerance issues were corrected following a discussion concerning a questionable paste parameter with the oven manufacturer. The manufacturing paste parameters windows were changed to more reasonable values: the peak temperature of 20°C was changed to a peak temperature of 15.7°C; the time above temperature was changed from 40 s to 26.5 s; and the slope was changed from 0.90°C/s to 1.5°C/s. Changing these windows will allow newly designed profiles to fit standard production paste specifications while accepting variations from all of the ovens. Monthly data will be monitored continually so that any trends of concern can be used to trigger a response for corrective action.
Contact: Inovar, Inc., 1073 West 1700 North, Logan, UT 84321 866-898-4949 or 435-792-4949 fax: 435-792-4950 Web: http://inovar-inc.com