What damage does the assembly process do to a pcb? (part 6)

In this final planned post of the “What damage does the assembly process do to a pcb?” series we shall discuss copper diffusion.

What is copper diffusion?

When soldering, copper diffusion is a process in which copper atoms are removed from the copper surface and redistributed into the solder over a wide area.

Back in 2005 I was running some thermal and solderability tests on multilayer boards using various types of cured laminates, different surface finishes and various solder alloys. I was doing my RoHS and Lead-Free due diligence. I had a whole range of tests that I was running. One test consisted of solder dipping a test board in 288°C solder for 20 – 5 second intervals. I was trying to make a multilayer board built on phenolic laminate delaminate. When I reached the tenth dip I had to stop my test. I didn’t make the board delaminate but I did dissolve away much of the copper circuitry into the solder pot.

This is an extreme case but it explains my point. The soldering process results in a copper thickness reduction. Every time the board passes through the hot air solder leveling process at the printed circuit board fabrication stage a little copper thickness is removed. When the board is wave soldered at assembly a little copper thickness is removed. How much copper is removed depends upon the type of solder and temperature applied. Test data available from various sources indicate that this rate is approximately .2 mils (0.005 mm) for roughly three typical solder cycles.

Understanding that copper is removed from each soldering excursion should be a concern to all involved in the supply chain. Printed circuit board manufactures need to limit the number of hot air solder leveling cycles. The same is true for wave and re-flow cycles at the assembly level.

What’s the worst thing that can happen?

Low copper in the hole comes to my mind.

The IPC requires .8 mils (0.02 mm) minimum with a 1 mil (0.025 mm) average of continuous copper plating to be within a plated through hole. It is common to see this value specified on fabrication drawings to be specified as 1 mil (0.025 mm) minimum of continuous copper plating to be within a plated through hole.

Why is this important?

When plated copper in the hole passes below a critical minimum thickness, under the strain of the assembly process, the copper cracks. In part 5 of this series I briefly explain the forces applied to a plated through hole when exposed to the assembly process. Thick copper may take a lot of strain. The force is distributed across the area of copper. Thin copper in the hole has less area to distribute a load and fails. Take two rubber bands, one very thin and one very thick. Pull them both as far apart as you can. The thin elastic snaps with very little effort. The thicker rubber band stretches further and requires more applied force to snap it. This is more or less what happens in the plated through hole. The thinner the copper, the easier it is to crack it in the assembly process.

A printed circuit board supplied with .8 mils (0.02 mm) minimum copper in the holes may have trouble if exposed to more than three assembly cycles. This assumes that the copper thickness shall go down by .2 mils (0.005 mm) in thickness over three solder cycles and is in contact with solder. Copper below .6 mils (0.015 mm) runs the risk of cracking from the strain applied on the hole through the assembly process. Reworking or solder touch-up with a soldering iron could provide enough localizes stress to make the hole fail. The thinner the copper the greater the risk of failure.

What can I do to minimize the potential problem?

Some things to consider are as follows…

  1. Limiting the number of soldering cycles where solder is in contact with copper is a good place to start.
  2. Use a nickel barrier between the copper and the solder. ENIG is a possible solution here. You can also have the board fabricator solder level the ENIG if you prefer solder as your surface finish of choice.
  3. Use an alternative final finish such as immersion tin, immersion silver or OSP. Not as much copper is removed by these finishes when compared to solder leveling.

Tags: , , ,

2 Responses to “What damage does the assembly process do to a pcb? (part 6)”

  1. ftimm Says:

    How long can I expect the pcb driver in an industrial LED light fixture to last?

    Some LEDs may perform effectively for 100,000 hours (or possibly as long as 200,000 hours). That is 11.4 to 22.8 years of 24×7 operation.

    The fixtures will be installed in a manufacturing plant. The operating temperature will likely not exceed 125F. There is metal dust from shearing, forming, cutting, grinding, and deburring operations. Smoke and fumes are present. Overhead cranes, forklifts, and the presses produce vibrations.

    The only comment I have found is that pcbs can last a long time. The author mentioned a 1957 transistor radio that is still operating, but I’ve not been able to find any typical service life.

    ftimm

  2. David Duross Says:

    How long can a pcb driver in an industrial LED light fixture last?

    I’m sorry to say that there is no easy way to pinpoint an answer for you on this question. This is because the answer depends upon several key variables.

    1) Design and PCB Materials.
    2) PCB fabrication process.
    3) Assembled Components.
    4) Type of assembly technique used.

    Design and PCB Materials are important since you need to calculate your thermal management and select the proper materials. For example, LED applications typically use thermally conductive laminate to allow heat to disapate from the hot running device to a heat sink structure like aluminum. The LED light fixture market is very competitive and preasure is on to shave cost. Hence the temptation exists to use cheaper materials. You may need to use a material with a thermal conductive rating of 3 W/m-k. The right material makes you uncompetitive in price compared to your competition. Switching to a 2 W/m-k material cuts your cost but is not as efficient. Heat builds up and the system exceeds you designed thermal limit. Over stressing leads to a pre-mature failure.

    The PCB fabrication process is important since no two fabricators are alike. We all use different equipment, materials, chemicals and techniques to manufacture a PCB. A PCB made by one fabricator shall be better than a PCB made by another fabricator. The boards are all qualified to a minimum standard (IPC specifications). Fabricator #1 rinses the boards with DI Water prior to shipping. Fabricator #2 rinses with filtered city water prior to shipping. The board from fabricator #1 solders very well while the board from fabricator #2 requires slightly higher temperatures to make the board solder just as well. The higher temperatures can be damaging to the product and lead to premature failure.

    Assembly components are a wild card since there are dozens of components from multiple sources. An assembly shall only be as strong as its weakest link. You may find that the board is stable but its a component soldered to the board that fails in the field.

    The type of assembly technique is a critical factor. Is the device soldered with tin-lead solder or with a Lead-Free solder? With tin-lead solder we know that the solder joints are robust since we have decades worth of history and experience working with the solder. That 1957 radio still working today cited by an author in your question makes sense since tin-lead solder forms a very good solder joint. Lead-Free solder joints are more brittle compared to tin-lead solder joints. Solder joint failure through metal fatigue is a potential concern. Unfortunately we don’t have enough experience with Lead-Free solders to determine the longevity of the solder joints. Prior to RoHS almost no one used Lead-Free solders. After RoHS you can’t buy electronics anymore with out it having Lead-Free solders in it. From my experienc, electronics built today don’t last as long as electronics built before RoHS. Something to think about there.

    When you consider the variables that go into a complex structure it is not feasable to project the end of life of a product. There are a lot of variables. If an estimation for end of life is truely needed there are simulation tests that can be run on devices in a controlled environment. I would go that route since a simulation would stress the entire assembly once its been through the manufacturing process.

Leave a Reply