# SMT reflow oven-to-oven repeatability: how to adjust an oven so a single recipe will work across multiple ovens for an individual product.

The ability of equipment to produce good product is fundamental to every manufacturing process. The ability to quickly setup and run a product on different process lines can make the difference in being able to satisfy a customer's requirement, missing a delivery date, or making a profit. This process line flexibility requires good equipment with repeatability between lines to be a success. To do this, the process engineer needs to maintain their equipment to high standards and understand its capability.In the 1950s, Wendell Abbott of GE wrote a detailed paper about the distribution of a measurement within a lot of material (or machine) and among lots of material that described the interactions. Although his paper is complicated, the concepts of variation within one oven and among different ovens are important when discussing flexibility across multiple SMT process lines.

Up until recently SMT process engineers with multiple process lines established a different recipe for each reflow oven because of variation between equipment. With modern technology and control systems, it is possible to adjust some reflow ovens so a single recipe will work "among" ovens for an individual product. Once this is done, repeatability among ovens becomes a process tool that saves time and makes production easier. To be comfortable with control among ovens, we first need to discuss the variation within an oven.

Repeatability within an Oven

Engineers are often asked how well their process meets a manufacturing specification. Phrases such as "nearly all the time" or "the majority of the time" are not acceptable in a technical world; thus establishing quantifiable data with measurable attributes is important. Placing a measurement or number on the ability of a process to meet a specification has evolved to what is known as process capability. This process capability measurement establishes a ratio of the process variation to a specification and is expressed as Cp or CpK.

It would be simple if we could make a single measurement of the process and compare it to the specification, but real life is not simple. Processes have variation that needs to be quantified before we can make a statement about how good they are. It takes multiple samples and math to quantify process variation. If we want to be 100% confident of the variation of a process, we would need to measure every piece. But that is impractical, so we use sampling and statistics to estimate the variation in a process. As you can guess, choosing the sample size becomes an issue of time and cost verses the accuracy of the data. We do not intend to turn this into a statistics course, but will give a simple overview of the concepts and how they can be applied to an SMT reflow oven.

Since process capability is a comparison of the process output to a specification, both need to be identified. In the case of a reflow oven, the specification is usually the reflow profile supplied by the paste manufacturer. It consists of ramp rates, soak time to traverse specific temperature ranges, time above the liquidus of the solder, and peak temperature.

The process output includes two pieces of process data for each specified attribute. The first is the average and the other is the standard deviation. The average is simple math (addition and division), while calculating the standard deviation is more complicated. Spreadsheets, statistical programs and some calculators can do the analysis without the need for manual calculations. (Consult statistical handbooks and texts for a detailed explanation of standard deviation.) What process engineers need to know is the standard deviation of sample data can predict the process variability.

The process to specification ratio can be expressed in two ways: One is called Cp and the other is CpK. The difference is that Cp focuses on machine capability, while CpK focuses on the process for an individual product. Cp uses the tolerance range and standard deviation without regard to the actual target to determine if the machine has the ability to make good product. The tolerance range is usually set to an industry standard. On the other hand, CpK uses Cp calculations and the actual target to ensure the process meets the specification for an individual product.

The calculations are:

Tolerance range = upper specification limit minus the lower specification limit

Cp =Tolerance range / 6 times the standard deviation

CpK = Cp times (1 minus K)

where K = the absolute value of the target minus the mean / by 1/2 the total tolerance range

Let's say that we have the following peak temperature data:

The specification given by the paste manufacturer is 240 [+ or -] 10[degrees]C.

Therefore, the target is 240[degrees]C with a tolerance range of 20[degrees]C.

After running 10 profiles, we found that the average peak temperature is 238.5[degrees]C.

And the spreadsheet calculations tell us that the standard deviation is 0.34.

Therefore, the Cp would be 20 / (6 X 0. 34) = 9.8

K would be (240.0 - 238.5) / 10 = 0.15

Then the CpK is 9.8 X (1-0.15) or 8.33.

Statisticians note that when the machine capability (Cp) is 2.0 or greater, we have a capable process, regardless of the process center. They also say that for a process to be in control, the CpK has to be 1.33 or greater, accounting for the process center. In this case the Cp and CpK are excellent.

To help make it clear, we can look at the following figures. FIGURE 1 depicts a perfectly centered process with the target and process average equal to each other. The statistical process variation is well within the specification limits and has a 1.33 CpK. This allows for some movement of the average, while the product remains within the specification.

[FIGURE 1 OMITTED]

FIGURE 2 shows a centered process, but the variation is larger than the specification limits. When this happens, the process is producing conditions that are out of the specification window. In this case, the CpK is about 0.8.

[FIGURE 2 OMITTED]

FIGURE 3 shows two processes that have variation less than the specification limits, but they are not centered. While the Cp (not considering the target) would be good, the processes are producing conditions that are out of the specification window. If they were centered, they would be OK. The CpK in these cases is about 0.85.

[FIGURE 3 OMITTED]

Now let's look at the results from a study of an actual reflow process conducted over a five-day period. Profiles were taken each morning and afternoon on a 10-zone BTU Dynamo reflow oven running a Pb-free solder recipe. The product was a 350g board with six thermocouples attached to critical components.

The ramp, time above liquidus (TAL) and peak temperature data from each run were put in a spreadsheet to calculate the averages and standard deviations. The results are in TABLE 1.

TABLE 1. Average and Standard Deviation of Sampled Product, Oven 1 RISING FALLING TAL PEAK Average of all TCs 1.42 -2.73 53.3 243.76 Std dev 0.07 0.19 2.13 1.82

Comparing results to the Pb-free solder specification, we see the averages are very close to the target (TABLE 2). This indicates the recipe is correct for this product, but does not indicate the process variation. Not knowing the variation reminds us of the guy who had one hand in boiling water and the other in an ice bath; he was uncomfortable, but on the average, the temperature was OK. Thus we need to know the variation of the process. Since we are taking a sample and not measuring the profile for every part, we can use the standard deviation of our sample and the formulas previously discussed to calculate the CpK.

TABLE 2.Target Solder Specification SPECIFICATION RISING FALLING TAL PEAK TEMP Upper limit 2 -1 65 255 Lower limit 1 -4 35 235 Target 1.5 -2.5 50 245 Target to Average Delta 0.08 0.23 3.30 1.24 TABLE 3. Process to Specification Ratio RISING FALLING TAL PEAK TEMP CpK DY #3 2.14 2.19 1.83 1.6 TABLE 4. Results of Offset between Ovens 3 and 4 RISING FALLING TAL PEAK 10 run average of all 6 TCs in Oven 3 1.42 -2.73 53.3 243.76 3 run average of all 6 TCs in Oven 4 1.65 -2.83 51.29 240.66 Difference between Oven 3 and Oven 4 0.23 0.1 -2.01 -3.1

This indicates that the process run on the third over 3 is in statistical control, because the CpKs are all above 1.33. If the process specification is correct, the reflow process is making good product. (Note: This study was conducted over a five-day period, but the same can be done with 10 consecutive runs or 10 runs in one day.)

Repeatability among Ovens

Now that we know one oven is in control, how do we show that the process is in control among multiple ovens with the same recipe?

We could obtain a single profile on a second oven to show that it is within the process widow identified by the paste specification. But that would not verify that the second oven had the same process variability as the first, and we could be deceiving ourselves into thinking the process was in control. The answer is again in statistics and CpK.

We could complicate our decision making with statistical evaluations such as the student t-test (1) and Mr. Abbott's calculations, but it is possible for the process engineer to keep it simple by comparing the results of CpK calculations on multiple ovens to show there is repeatability between them.

Before we start the 10 runs on the second oven, we need to fine-tune or adjust the profile by using thermocouple offset in the oven control program.

To obtain the offset, we run a few profiles on the second oven and compare it to the averages we produced on the first oven. We are tuning the oven, not calculating the CpK. Although we could get more accurate data with multiple runs on Oven 4, we chose three runs in an effort to save time (FIGURE 4).

The peak temperature and TAL are slightly lower on Oven 4, so if we use an offset in the oven control of 3[degrees]C to increase the peak temperature, the TAL should be slightly longer.

Then obtaining profile data with the offsets from 10 runs and calculating the averages, we get the results in TABLE 5.

The small differences in the averages between Oven 3 and Oven 4 tell us that we chose the correct offsets.

TABLE 5. Average Profile Data with Offsets of 10 Runs, Oven 4 RISING FALLING TAL PEAK Oven 4 Average of all 6 TCs (10 runs) 1.53 -2.92 53.6 243.09 Original Oven 3 results 1.42 -2.73 53.3 243.76 Difference between 3 and 4 0.11 0.19 -0.3 -0.67

Then, using the same process specification as Oven 3 and the standard deviation from Oven 4 data to calculate the CpK on Oven 4, we get the results in TABLE 6. Both the comparison of the averages and CpK confirms that Oven 3 and Oven 4 are in control with the same recipe.

TABLE 6. Process to Specification Ratio, Oven 4 RISING FALLING TAL PEAK TEMP CpK Oven 4 2.05 2.11 1.79 1.69

TABLE 7 shows the same for yet another oven (identical product, set points, offset procedures, and specification limits).

TABLE 7. Process to Specification Ratio, Oven 5 RISING FALLING TAL PEAK TEMP CpK Oven 5 2.27 2.96 1.56 1.58

Therefore, we have three ovens using the same recipe that are all in control. This allows us to say we have repeatability among them.

Let's do a quick review by comparing the data from the three ovens (TABLES 8 and 9). The first thing we see is that the rising, falling, TAL and peak temperature averages are very close to each other; thus we have the correct tuning offsets. The second and most important is that every specified attribute is within control because the CpKs are more than 1.33.

TABLE 8. Averages of Ovens in Control RISE FALL TAL PEAK AVERAGES C/SEC C/SEC SEC C Oven 3 1.42 -2.73 53.3 243.76 Oven 4 1.53 -2.92 53.6 243.09 Oven 5 1.31 -2.32 53.27 243.62 Range 0.22 0.6 0.33 0.67 TABLE 9. Process Repeatability of Ovens in Control CPK Oven 3 2.14 2.19 1.83 1.6 Oven 4 2.05 2.11 1.79 1.69 Oven 5 2.27 2.96 1.56 1.58

We have proven that we have repeatability within each oven and among the three ovens, and the process engineer can be confident that they can produce good product in any of the three reflow ovens.

END NOTES

(1.) A t-test is any statistical hypothesis test in which the test statistic follows a Student's I distribution if the null hypothesis is supported. It can be used to determine if two sets of data are significantly different from each other, and is most commonly applied when the test statistic would follow a normal distribution if the value of a scaling term in the test statistic were known.

FRED DIMOCK is manager, process technology at BTU International (btu.com); fdimock@btu.com.

Printer friendly Cite/link Email Feedback | |

Title Annotation: | CONVECTION SOLDERING |
---|---|

Author: | Dimock, Fred |

Publication: | Printed Circuit Design & Fab |

Date: | Apr 1, 2013 |

Words: | 2273 |

Previous Article: | Advanced block organization in schematic design and layout: reusing circuits using advanced block hierarchy. |

Next Article: | Bend, don't break: EMS companies showed their flexibility in 2012, adjusting to slowing consumer and defense markets. |