Richard, please send the photos you describe below. We have also considered replacing electroless with direct metallization, but I have already been down the road where fancy expensive equipment had to be yanked in favor of simple dip lines, and I have been fighting the change. I greatly appreciate your sharing this story with Technet. Connie Huffman Process Chemist Avionics Production Division Warner Robins Air Logistics Center > -----Original Message----- > From: Richard Mark Mazzoli [SMTP:[log in to unmask]] > Sent: Friday, October 01, 1999 06:28 > To: [log in to unmask] > Subject: [TN] Direct Plate > > DIRECT PLATE PITFALLS > > IS IT REALLY AS ROBUST AS ELECTROLESS COPPER? > > RICHARD MAZZOLI > > > BACKGROUND STORY > > Recently there was a shop that was interested in replacing their > electroless > copper with one of the new alternative methods making all the headlines. > They interviewed prospective vendors, listening to the good and bad points > of > each system on the market; some had long processing lines, some were > relatively unproven, and some were just poorly supported or represented in > their part of the country. > > Decisions were made to test one supplier's product based on a number of > factors such as 1) it was a "tried and proven" process boasting "millions > of > square feet" of high quality boards, 2) a name-brand supplier to the > industry > made the process, 3) product support was considered to be the best > available > and 4) the equipment and pricing was in line. Testing commenced when the > equipment was installed. > > Test data, as is often the case, was conflicting. Basically, in order for > this process to be approved it had to stand up to 500 thermal cycles in a > liquid-to-liquid chamber. Tests showed that if 1 mil of copper was plated > on > test panels there would be a 1% to 10% failure rate. A separation would > develop - the boards would develop circumferintial voids at the oxide > interface of the inner layers. So, the process supplier and the board > shop > mutually decided to test the panels with 2 mils of copper. Subsequent > results were good (no separation occurred) and the board shop's largest > customer approved the process based on the latest tests. > > The shop soon switched over to the new direct plate process and threw out > their old electroless copper line. Cycle times were reduced, some > unhealthy > chemistry was eliminated and everyone was happy. Until field failures > began > turning up. > > The shop's largest customer began to find a significant number of failures > due to open circuits. Analysis showed them to be wedge voids. Things > went > from bad to horrible as the shop scurried to find a solution. They began > a > series of tests designed to find the cause of the voiding and, more > importantly, how to stop it. They began extensive quality checkpoints > using > many times the normal amount of destructive testing out of copper plating > in > an effort to filter out these escapes. Tests were conducted to find out > how > they could escape from electrical test. > > Of course, to the customer, all of these efforts were exercises in > futility. > They understood the PCB shop's problem but had problems of their own. The > customer was quickly moving to the decision of canceling all orders and > putting the shop on a "no order" status. They were out of patience and > had > their own business to protect. The PCB shop took 60 days to discover the > cause of the defects and construct a more robust process. Too long in > this > day and age - the customer replaced them. All orders ceased. The shop > replaced the direct plate process for electroless. Too little, too late. > Several months later the shop filed for protection under bankruptcy laws. > One year later they closed their doors. > > If this story sounds like fiction to you, then you'd be wrong. It > happened. > Of course, this is only the abridged version. The lessons learned by this > shop have left an indelible mark on its personnel. The tests that > resulted > from the 60 days of failure analysis and trials have been reconstructed > here > so that other PCB shops might not make the same mistakes. > > THE FAILURES > > Figure 1 shows an example of the failures this study evaluated. This PCB > has > been thermally cycled in a liquid-to-liquid chamber for 500 cycles. > Figure 2 > shows the same sort of defect observed at only 100 cycles. Failures were > found at 50 cycles as well. It's important to note that these same parts > passed original electrical test then failed after cycling. There were not > many failures per board. In fact, out of more than 2,000 holes only 2 to > 4 > holes would fail. Finding this level of defect is akin to stumbling > across > the proverbial needle in a haystack. > > There were many questions. What is the exact cause? How to stop them? > How > to detect them? Oddly enough it was only on certain board types although > all > board types were being processed through direct metalization. Why were > these > parts different? > > A study was conducted to determine what was different about these parts. > Taken into account was: > > § Hole size > § Aspect ratio > § Drill type > § Drill supplier > § Material and pre-preg style > § Lamination techniques > § Drill parameters > § Material supplier > § Layer count > § Process differences between these and other parts > § Sequential lamination techniques > § Unique board features > > Analysis showed that there was no difference in these boards than any > other > product being built except that the failed parts had non-functional pads. > But > other than that, there were no common features that could have accounted > for > the failures being only on these particular PCB's. > > THE NATURE OF THE BEAST > > This direct plate process uses a micro-etch as its final chemical bath, > followed by a rinse then dry. Panels are then ready for dry film coating. > This micro-etch bath is to remove the metalized coating from the copper > surfaces - outer layers surface and interconnections. During failure > analysis several indications pointed to this portion of the process as > being > a major cause of concern. A good many boards were found to have slight to > moderate negative etchback. A "wedge" or crevice was formed at the top > and > bottom side (the oxide and base side) of the interconnects. It was at > these > places that failures would occur. > > Controlling the micro-etch bath in a direct plate process is easier said > than > done. It's not as simple as most other micro-etch baths since it relies > on > impingement to work. The spray actually has to get underneath the > metalized > coating in order to "lift" it off by removing the copper underneath it. > In > that manner it remains on the non-copper areas - the areas it is intended > to > metalize. Understanding this, it's not hard to comprehend that there is > going to be some amount of copper missing from corners (knees) of > interconnects and surface copper where it meets the resin/glass. This > "gap" > is normally miniscule. But nevertheless, it exists even under the best of > circumstances. It is the nature of the beast. Fundamentally it becomes a > non-conductive portion of the hole that must be plated. The only way this > can happen is by current "jumping" the gap. Copper must literally bridge > across this fissure and make connection to the interconnecting foil. > > Figure 3 shows a drawing of this copper removal. Further, it shows how > this > condition can become amplified when coupled with any amount of nail > heading > from drill. It's not hard to imagine the metalized coating being removed > by > the undercutting action of the sprayed-on micro-etch. It stands to reason > then that when the copper interconnects have been, in effect, "widened" by > nail heading that the resulting removal of the copper will now leave an > even > longer gap. The normally miniscule fissure now becomes even harder to > plate. > The fissure now becomes a full-blown gap - a break in continuity. > > Direct metalization works by way of a "moving front". Copper travels from > each surface, through the holes until it "meets" (Figure 4). This is > commonly referred to as initiation. When the moving front meets, > initiation > occurs and you have successfully encapsulated the hole wall and begin to > build thickness evenly. Until that point of initiation, copper is > building > in thickness on areas that are conductive but obviously not on other areas > - > like the fissure. A wedge-like formation begins to develop which can be > miniscule to very wide, depending on how large the non-conductive fissure > is. > > > Conversely, electroless copper has no fissure issues since it completely > encapsulates the interconnect, resin, glass and surfaces. It begins to > plate > thickness immediately when current is applied. No moving-front is > involved - > no initiation. It's already plating on all surfaces. > > > NON-FUNCTIONAL PADS > > Rather than debate the pros and cons of non-functional pads (NFP's) the > testing in this study included both types of designs - NFP's and > Functional > Pads. Like the moniker implies, NFP's are not functional in the sense > that > they are not connected to any other pad. They are blind pads left on by > designers for a variety of reasons but that really serve no function > insofar > as electrical conductivity is concerned. > > Analysis on defective parts showed that it was at the NFP that the wedge > was > at its worst. All defects were found to be at the interconnect of a NFP. > Further analysis verified that even after more than one hundred micro > sections the wedge-like defect (and subsequent separation) occurred only > at > the interconnect of a NFP. A test vehicle was devised to prove or > disprove > this finding. > > A test coupon was developed that included functional and non-functional > pads > side by side. The coupon was built as an 8 layer board, one hole > completely > functional (linked by circuits to another pad/hole) and the next hole > completely non-functional (all interconnects "dead"). To test the > possible > role of aspect ratio in the creation of the defects, hole sizes were > gradually increased across the coupon. Tested were 5:1, 3:1, 2.5:1 and > 1.6:1 > aspect ratios. Figures 5 and 6 show the coupon configuration. > > Using this design the test hoped to find plating variances between > functional > and non-functional interconnects and perhaps some variance due to aspect > ratio. Since the direct plate process plated by means of the moving-front > it > was decided that the initiation time delay (the time the panel entered the > plating tank and the time it actually had thru-hole contact > surface-to-surface) was important. Going even further, it was felt that > since the fissure/wedge created extra resistance thereby making it harder > to > plate, the test should include current density variables. Therefore, the > test was outlined as described below: > > 1. Panels metalized with direct metalization. > 2. Panels metalized with electroless copper. > 3. Panels pulled from copper tank every 5 minutes and micro-sectioned > continuing for 60 minutes. > 4. Panels to be plated at 10 ASF. > 5. Panels to be plated at 20 ASF. > 6. Panels to be plated at 30 ASF. > > Panels were 18x24 and had numerous coupons merged into actual customer > parts > - the same type parts that were failing in the field. This allowed the > test > vehicle to be a relatively accurate representation of the exact parts that > were failing. > > The direct plate supplier was invited in and asked to perform process > audits > on the metalization line. In fact, they assisted in generating a good > portion of the data collected in the overall testing. No significant > modifications were made to the process but it was optimized to their > recommendations. Further, the copper baths were optimized for brightener, > copper and acid content. Rectifiers were checked for throw and all > connections on the tanks measured for connection. > > RESULTS > > The results were definite in that they clearly showed the difference in > plating thickness between functional and non-functional pads. The > explanation for this is that functional pads are "networked" together in > various configurations. These networks actually assist copper plating by > becoming conductive then "charging", thereby helping the moving front > reach > its initiation point. One thing was clear; wherever there was an existing > wedge there was extra resistance and the moving front of copper was forced > to > somehow get across the fissure. NFP's could have actually become anodic > (since they were isolated on both sides by the fissure), therefore > offering > even more resistance to the moving front. > > The tests were also definite in that they clearly showed the difference in > plating thickness between direct plate and electroless copper. Repeated > tests showed that electroless copper yielded 25% more thickness in the > same > amount of time. That percentage climbed to as high as 75% more when > looking > at holes with only non-functional interconnects. > > What was not expected is that the results also clearly showed the > formation > of the wedge voids. Pulling the parts every 5 minutes for micro-sections > showed where the wedge started and how it eventually "bridged". What was > even more interesting was to see the difference between the 10, 20 and 30 > amp > current densities. As would seem clear, the lower the current density, > the > lower the plating. Similarly, the wedge was even more pronounced at 10 > ASF > than it was at 30 ASF. However, with today's designs plating at 30 ASF is > pretty much out of the question. > > Figure 7 shows the difference between direct plate and electroless copper > at > 20 ASF after 5 minutes in the plating solution. Figure 8 shows a close up > of > one of the interconnects from the electroless hole and direct plate hole. > The wedge is clear on the direct plate photo. It's important to note here > that 5 minutes in a plating solution should be building copper thickness. > That's almost 10% of a normal 60 minute cycle. It's clear that this hole > is > not plating as expected. > > Even at 30 ASF similar test panels showed the wedge developing. But they > also showed something interesting. On Figure 8 you'll see the wedge but > then > you can also see that the copper has apparently bridged the gap somewhere > around the circumference of the hole, since copper is starting to plate on > the other side. Or did it somehow "jump" across the gap? > > TEST CONCLUSIONS > > After repeating these tests under close scrutiny and varying the test > conditions certain conclusions were developed: > > 1. Non-functional pads increased resistance when using this direct plate > process. > 2. Electroless copper plated better (faster and more evenly) than this > direct > plate process on all coupons. > 3. Even under the best of circumstances there will be some magnitude of > fissure (broadening into a gap or all-out wedge) using this particular > direct > plate system. > 4. This direct plate system is less robust than electroless copper when > used > on designs with high amounts of non-functional pads. > 5. At low current it was impossible to reliably plate NFP holes using this > direct plate process. > > Does this indicate that direct plating is a poor process. Not > necessarily. > With hundreds of installations worldwide, there's plenty of data and > "millions of square feet produced" to indicate it's a viable alternative > to > electroless copper. Does the testing described in this article indicate > that > direct plate has its limits? Sure, and they prove that this direct plate > process had even more limits than what was already known. Certainly no > one > selling this process knew about its limitations on boards loaded with > NFP's. > Neither the supplier nor the engineers who installed it realized it could > present such a problem. In this shop, direct plate was certainly no > drop-in > substitute for electroless copper. > > BACKGROUND STORY CONCLUSION > > The shop where this activity took place is no longer in business. The > building is closed, equipment gone and the people are working in other > places. The mistakes leading to this shop's demise are many. They > include > poor engineering practices when testing new technology, poor decision > making > for allowing inaccurate test data to factor into its decision to add the > technology, poor quality control systems that allowed the failures to > escape > and poor overall management for allowing all of these things to happen. > The > author is not implying that a technology change is solely responsible for > this shop's fall. It's simplistic to think that this one problem caused > the > shop to ultimately go out of business, when in fact there were many other > reasons along side it. > > But the addition of this new technology should have been more thoroughly > investigated. It had limits that even the supplier didn't know about. > Yet > it was sold as a drop-in replacement. That's nothing new to PCB shops. > We've all installed products that don't work the way they are sold. That > don't quite do what they are purported to do. Experiences such as the one > outlined in this article should remind us to move cautiously when > installing > new technologies. It also reminds us that there are very few, if any, > "plug > and play" replacement technologies. And so it goes in this business, as > in > life, caveat emptor (let the buyer beware). > > Richard Mazzoli has been in the PCB industry for 18 years working as > Process > Engineer, Manufacturing Manager and Engineering Manager. He can be > reached > at [log in to unmask] or [log in to unmask] > > ############################################################## > TechNet Mail List provided as a free service by IPC using LISTSERV 1.8c > ############################################################## > To subscribe/unsubscribe, send a message to [log in to unmask] with > following text in > the body: > To subscribe: SUBSCRIBE TECHNET <your full name> > To unsubscribe: SIGNOFF TECHNET > ############################################################## > Please visit IPC web site (http://www.ipc.org/html/forum.htm) for > additional > information. > If you need assistance - contact Gayatri Sardeshpande at [log in to unmask] or > 847-509-9700 ext.5365 > ############################################################## ############################################################## TechNet Mail List provided as a free service by IPC using LISTSERV 1.8c ############################################################## To subscribe/unsubscribe, send a message to [log in to unmask] with following text in the body: To subscribe: SUBSCRIBE TECHNET <your full name> To unsubscribe: SIGNOFF TECHNET ############################################################## Please visit IPC web site (http://www.ipc.org/html/forum.htm) for additional information. If you need assistance - contact Gayatri Sardeshpande at [log in to unmask] or 847-509-9700 ext.5365 ##############################################################