Glass Stop & Resin Starvation – The Main Culprits Behind CAF Failure

by | May 20, 2021 | PCB Failure Modes

As we increase our focus on spanning the gap between PCB design and fabrication, on the design side, we start to come up against fabrication terms and technical issues with which we may not be that familiar. This is the first in a series of blogs that will address how and where these terms and technical issues come into play, the problems associated with them and how best to address them as part of an overall, successfully implemented PCB design process.

The first term we will define is glass stop. This blog describes what glass stop is, what causes it and how to make sure it doesn’t negatively impact your design.

An Overview of The Basics

First, it’s helpful to do a quick overview of the basics. In simplest terms, the three main components in a PCB are:

  • The copper foil that is used in the signal and power layers.
  • The glass reinforcement.
  • The resin systems.

The combination of the resin and glass is termed a laminate or core when the copper foil is bonded to one or both sides. When the resin is only partially cured and no copper foil has been bonded to it, it is designated as prepreg.

Prepreg is available with a variety of resin contents. Today’s complex, multilayer boards have a lot of features that all need to be filled in with resin. In addition, the high current parts that are now crucial to these designs need to have one- or two-ounce copper thickness (1.2 mils and 2.4 mils respectively) and the prepreg that needs to flow into these thicker copper layers has to have enough resin in it to completely fill the voids.

Bottom line: A crucial part of the PCB stackup design process is to make sure that the prepreg(s) you have selected for your design will be of such resin content that glass stop won’t be a problem.

Figure 1. (a) Left. This is a case of sufficient resin to provide a “butter coat” between the copper and glass.
(b) Right. Here, there is insufficient resin to fill the glass, copper roughness, and etched away parts of the copper, leading to direct contact between the glass and copper (“glass stop”) and voiding.

The Origin of Glass Stop and The Problems Manifested by It

During lamination, the resin will flow out of the cloth to fill the voids in the adjacent layers. When this flow occurs, if there isn’t enough resin in the prepreg, the end result is that the glass fibers will come into contact with the copper layers. This is the definition of glass stop, also often called resin starvation.

Glass stop, in and of itself, isn’t necessarily a cause for concern, unless high voltage is applied directly to the board. The risk here is dielectric breakdown or hi-pot failure, because the glass has a much lower dielectric breakdown threshold than the resin. One larger issue is that if there isn’t enough resin to fill the voids in the adjacent copper layers the result will be air-filled bubbles (aka voids). Further, if the drill goes through one of these bubbles during manufacture, chemicals can become trapped within them and this, in turn, can lead to degradation of the plated through holes. Additionally, if enough resin is not available to bear the stresses during drilling, the glass resin interface may exhibit microfractures, which can lead to Cathodic Anodic Filamentation or CAF. This is referred to as crazing and is another significant issue that can be caused by glass stop. Crazing can also be caused if there is a poor bond between the resin and glass due to a poor selection of the finish but that should be addressed by the material manufacturer before release of the product.

Perhaps the biggest issue that glass stop can cause, however, is Conductive Anodic Filament (CAF) failure. The majority of actual CAF growth caused under accelerated humidity and temperature testing conditions can be traced back to a lack of resin covering the glass – a much smaller number of defects can be traced back to hollow fibers, which are a defect in the glass caused during the yarn manufacturing process. In recent times, with glass manufacturers using optimized formulations and better yarn drawing technology, coupled with improved inspection processes, the incidence of hollow fibers has gone down significantly.

Glass stop means there is insufficient resin to cover the etched away copper, roughness, and glass. This creates a pre-existing pathway for moisture ingress in a humid environment, which, under applied voltage bias, can lead to CAF growth and failure.

I have personally seen multiple instances of products being certified as poor CAF performers, when, in reality, it was all a function of the availability and mobility of resin. I remember a case where a supplier was trying for 2 years to get a dielectric product qualified in a Test Vehicle, but it kept failing on CAF. At my recommendation, the construction was changed and the product no longer failed. As a result, the stackup review became an integral part of the process at that manufacturer to build TVs resulting in almost zero CAF failures during TV testing.

Of course, CAF can be caused by issues with the resin electrochemisty, but, in such dielectric materials, CAF problems would be ubiquitous and nearly unavoidable and such a product would be unlikely to ever make it to production. It is, in a majority of cases, not a resin chemistry issue, as is commonly assumed, but simply a question of availability of resin and its mobility. CAF avoidance boils down to making sure enough resin is available to fill the etched copper and roughness. The resin availability and mobility is related to the filler content and viscosity, and, if the product is not properly designed, a question of compatibility of resin to glass (though, such a product should never make its way out of the lab!). Most Resin systems are designed to reduce the coefficient of thermal expansion and, as such, are laden with inorganic fillers. High levels of filler reduce the availability and the mobility of the resin, and, therefore, lead to higher levels of voiding and glass stop and increase the potential for CAF.

Avoiding Glass Stop

From the mechanical side, the way to address glass stop is to be sure that you select a prepreg that has a high enough resin content for your design. As noted above, today’s complex multilayer boards now require copper thicknesses of one or two ounces to support the high current components which are mounted on the PCB. This means that the prepreg you select for your design needs to have lots of resin. The good news is that there all kinds of choices of prepregs available and some are really resin rich just to support these thicker copper layers. (As an example, 1078 glass prepreg is available upto 78% resin.) There are, however, tradeoffs with this approach, since a higher resin content also leads to a higher expansion in the Z-direction, which is highly detrimental to plated through hole reliability and can also lead to delamination of the laminate material in tight pitch areas.

When Should Glass Stop Be Addressed?

As with all of the aspects associated with designing a multi-layer board, the more aspects that you can address during the design cycle, the more reliable and manufacturable your design will be. This applies to proactively addressing glass stop problems.

Avishtech’s Gauss Stack stackup tool accurately calculates how much resin is needed to avoid resin starvation and it enables you to calculate exactly how much resin you need in your prepreg. It also estimates the amount of copper that is missing in the opposing layers and it warns you when a stackup change is needed. In addition, Gauss Stack will catch cases of glass stop with high accuracy by not only assessing available resin against etched copper, but also by accounting for the effects of dielectric filler (to catch cases of filler damming) and conductor roughness. The design can further be checked for thermal reliability to make sure that one has not overcorrected for glass stop to only end up with an issue on the reliability side. This process can be iterated so as to achieve the optimal balance between CAF reliability and thermal cycling reliability.