Fiddling with the process increases its variation. Never do it!

The term “tweaking” has many different definitions: 34 according to one Website I found. The definition that best fits electronics manufacturing in general and printing in particular is “to touch something up, fiddle with the finishing touches or make tiny little changes.” In our industry, tweak is usually used in such a way as, “I just gotta tweak a little bit here and there and it will be perfect.” That in turn usually leads to a response such as, “I just tweaked this and this and this and this and this and ohhh no, now this!”

Customers often ask, “What printing operating parameters can I allow my operators to adjust to optimize printing process performance?” The simple answer: None! That typically provokes this response: “But my operators are well-trained and have worked with the printing process for several years. With their knowledge and experience, shouldn’t they be given the authority to change operating parameters to optimize performance?” The answer again: No. No process operator, no matter how knowledgeable or experienced, should under any circumstances be allowed to adjust any operating parameter.

I am very much a traditionalist in the area of process development, optimization and control. I am a traditionalist because I have seen traditional formal process development, optimization and control methods work. During my 18-year career as a Motorola process engineer and manager, I experienced how effective well-designed and -executed experiments and studies are in identifying, quantifying and optimizing critical operating parameters, for printing or any other process. I transformed from skeptic of formal process development and optimization tools to supporter simply by witnessing their effectiveness firsthand.

I have seen high-volume paste printing processes operate in the single-digit parts per million defect level with absolutely no tweaking of the defect calculations as is so often done by many organizations. What do these extremely high-yield processes have in common?

• A formal design of experiments (DoE) approach to process development and optimization.
  
  
• Careful evaluation and use of statistical studies and laboratory testing to select suppliers of paste, stencil, tools, etc. They understand that printing solder paste, or any material, is a process, and to optimize it, all process inputs and operating parameters must be understood, quantified and optimized.
  
  
• An understanding that a machine is not a process. They use sound engineering and investigation to select printing equipment suppliers, not just the equipment itself.
  
  
• SPC process monitoring. Product is not monitored by excessive inspection.
  
  
• Data from internal studies and investigations to drive supplier selections, operating parameters, etc. They do not rely solely on price, hearsay or vendor-supplied data to make selections.
  
  
• Continuous study; material, equipment and tool evaluations; and process performance data reviews, which further optimize the processes. They understand continuous improvement has no end.
  
  
• No process tweaking. They stop the process for out-of-control conditions, investigate the “special cause,” and implement corrective action or a containment plan. Only then is the process resumed.

If the aforementioned formal process development and optimization work is performed, the process monitored using SPC, and the concepts of common cause and special cause variation understood, then we should be adamant about never tweaking the printing process. Every process has variation, but we must understand the types of process variation to be able to take accurate corrective action.

In review of common cause and special cause variation, look at the differences:  

Common Cause Variation

• Defined by variation within process’s upper and lower control limits on the control chart (see the Shewhart Rules for defining out-of-control conditions).
• Always present.
• Natural, expected variation.
• Difficult to identify.
• Cannot be eliminated.
• Examples: Solder paste characteristics such as viscosity variation over time, temperature and humidity variation.

Special Cause Variation

• Defined by an out-of-control condition on our control charts (see the Shewhart Rules).
• One-time event.
• Unnatural, not expected.
• Easily identified.
• Can be eliminated.
• Examples: Paste dried out on stencil; machine failure or out-of-calibration condition; logged stencil aperture(s).

Common cause variation always exists and can be reduced by formal process development and optimization, but never totally eliminated. Special cause variation is a unique event that drives the process out of control and must be understood, identified and corrected before the process can continue. Making process adjustments (tweaking) to correct special cause variation never should be done. Tweaking will only drive the process more out of control.

We can examine the concept of process tweaking from work done by Dr. W. Edwards Deming. Deming discusses four different methods commonly used to adjust a process. Deming used a funnel experiment to illustrate that variation of a process reacts to adjustments made to improve the process. We can illustrate the four approaches to process tweaking via a simple example of traveling to work.

Your goal: To arrive at work at 7 AM every day.

Rule 1. Choose a reasonable route and consistently follow it, leaving home at the same time each day.

Day    Departure   Arrival

1      6:30 AM     7:05 AM

2      6:30 AM     6:58 AM

3      6:30 AM     6:55 AM

4      6:30 AM     7:00 AM

5      6:30 AM     7:02 AM

Each day we may or may not encounter a variety of situations as we travel, such as traffic lights, normal weather variation, school buses, traffic (vehicular and pedestrian), etc. These situations are always present and are common cause variations to our travel time, affecting the arrival time.

Rule 1 is by far the best choice. It produces a stable distribution of points and minimum variance to your arrival time at work.

Rule 2. Compensate for the amount of time you arrived at work early or late by leaving home that much earlier or later, and continue adjusting your departure each day based on the previous day’s departure.

Day    Departure   Arrival

1      6:30 AM     7:05 AM

2      6:25 AM     6:50 AM

3      6:35 AM     7:15 AM

4      6:20 AM     7:00 AM

5      6:20 AM     6:50 AM

Tweaking! Never do it. It produces a stable process, but higher variance than Rule 1.

Rule 3. Compensate for the amount of time you arrived at work early or late by leaving home that much earlier or later than the original departure time, and continue adjusting the departure each day based on how early or late you were the previous day.

Day    Departure   Arrival

1      6:30 AM     7:05 AM

2      6:25 AM     6:50 AM

3      6:40 AM     7:15 AM

4      6:15 AM     6:40 AM

5      6:50 AM     7:25 AM

Worse than tweaking. Arrival time “explodes.”

Rule 4. Compensate for the amount of time you arrived at work early or late by leaving home at the previous day’s arrival time, and continue adjusting the departure each day based on the time you arrived the previous day.

Day    Departure   Arrival

1      6:30 AM     7:05 AM

2      7:05 AM     7:30 AM

3      7:30 AM     8:15 AM

4      8:15 AM     8:45 AM

5      8:45 AM     9:20 AM

Major “explosion” of arrival time.

If not understood and controlled, variation can and almost always will result in defects, with all the negative impacts defects create. Tweaking a process only increases process variation. Never do it!

Joe Belmonte is project manager, advanced process development, at Speedline Technologies (speedlinetech.com); jbelmonte@speedlinetech.com. His column appears bimonthly.