There was a big study done well before I became part of the industry some 30 years ago that determined for the typical two-layer, 0.062-inch board, the amount of copper required for a reliable connection in a plated through hole (PTH) was 0.0007 inches or 0.7 mils. Anything less could question PTH reliability and anything more would not make much of a difference.
Simply put, the .001 inch or 1 mil as stated on most PCB fabrication drawings is a safety factor.
Some corporate fab specs will state an average of 1 mil, with 0.8 mils as the minimum—again higher than the originally determined 0.7 mils.
Now that you know a PTH hole can be reliable with as little as 0.7 mils of copper in the hole, that doesn’t mean there can’t be other plating defects.
Barrel cracking was a not an uncommon defect when I joined the industry in the early 1990s—a defect that might not manifest itself as much if there was more copper in the hole.
One of the main reasons for the defect was the difference in the Z expansion rate (height) of the copper as compared to the fiberglass/epoxy material. The epoxy material would expand faster than the copper in the hole during the thermal cycle(s) of the assembly process—that difference in would cause copper separation from the hole wall.
The thicker the board the more copper required and likewise the thinner the board the less epoxy meaning less copper required for reliability.
However, with improvements in raw materials, plating chemistry, PCBs designs and manufacturing technology, barrel cracking is seldom seen today.
DirectPCB specializes in micro-electronics by building plenty of two-layer, 3 mil thick finished PTH boards and many other projects requiring blind and buried vias where 0.6-0.8 mils of copper plating is all that is required.
NOTE: Many in our industry confuse mils with ounces! A mil of copper is not an ounce. An ounce of copper is 1.4 mils.
Need help with PCB purchasing? Reach out to me at greg@directpcb.com.
Click here to get a free copy of my book, PCB Basics for Buyers.
Was involved in this process while I was with IBM Microcircuits in the 90’s
Thanks for commenting–we have come a long way since the 90s for sure! Greg