Not sure if the photo that Henning sent me with this reply will be sent to you all but lets try it and see what happens. (Henning, the article on conformal coating reliability is at this link //www.freelists.org/archives/tinwhiskers/03-2008/msg00010.html) -Bob Landman Dear Bob, I have placed some comments between some lines of what you mailed. You may post this, if you wish... Best, Henning > -----Original Message----- > On Behalf Of Kane, Joseph E (US SSA) > Sent: Thursday, March 13, 2008 6:14 PM > Subject: Re: Conformal Coating? When Reliability Goes Astray > > In order to cause a failure, whiskers must grow straight enough and > long enough to contact an adjacent conductor. The whisker must bridge between two conductors at different potentials, and breakdown any insulating films. We have seen this happen when the whisker bent into an arch! And, there would have been NO shorting, had all the whiskers grown straight. As for "long enough" --- yes, indeed. On the other hand, this short was caused by a whisker that was a few mils long --- i.e., some 0.1 mm long. This is detailed at http://nepp.nasa.gov/WHISKER/photos/pom/2003dec.htm > They have to > contact the adjacent surface with enough force to make electrical > contact, which in some cases means penetrating through a second layer > of conformal coating on the other surface. Well said. Many whiskers are doubtless in mechanical --- but NOT in electrical --- contact: the tin oxide coating is thick enough to block current flow in many cases. We, and others, find that the tin oxide coating does not break down (to allow current flow) until the potential exceeds some 0.05 V. Most coatings require 0.3 V to 15 V to break down. A few require more than 15 V, and the champion-to-date is 45 V. The electronics of the Shuttle Endeavor (OV-105; launched a few days ago and now docked to ISS) was found in 2006 to be threatened by between 100 and 300 million tin whiskers more than 0.5 mm long; many were over 5 mm long. Many of these had probably been present for more than 15 years. But NO electrical faults were logged by the Shuttle program as caused by whiskers. And clearly the Endeavor has not suffered any major electrical outages. Clearly, the conformal coating was useful. > And they have > to do this in the time frame that the product is in service. Well, "intended service" anyway. "Service" may be over when the first serious whisker-induced short happens. Jay and I have seen whiskers on two-year old products, and lots of them on five-year old products. > Looking at the available literature, it seems like the following > things are at least partly true, depending on circumstances. Of > course, your mileage may vary. > > 1. In some cases, conformal coat can delay the onset of whiskers > (longer incubation period). We saw a modest difference between both latency and density, between coated and uncoated sides of our specimens. Statistically significant, but not enough to be really helpful to reliability. > 2. In some cases, it can prevent whisker growth altogether. We have not seen that. The whiskers formed under the conformal coating, and then grew at roughly the same density as on the uncoated side. (I do not mean that this growth prevention does not ever happen; I only mean that this is not what our specimens showed.) > 3. In most cases, conformal coat slows down the growth rate. Our whiskers extended their lengths about the same in the coated and uncoated sections. But the whiskers growing on the coated side "crumpled" and so their "heights" stopped after reaching about one mil. So some folks might say, "NOT slowed", while other folks might say, "Slowed", depending on whether they were thinking of "arc length" or "elevation above coating". > 4. For whiskers that grow through coating, most are gnarled, > kinked, or otherwise unlikely to cause a direct short. It's > possible that the coating causes this, e.g. maybe the stress of > penetration affects the morphology. No whiskers at all have penetrated the 2.0 mm thick Uralane 5750 coating on our specimens. Many tens of thousands have tried, and crumpled. > 5. Whiskers may grow through > coating, but may have a harder time penetrating the coating on an > adjacent conductor without buckling. For realistic dimensions, the whisker may dimple into the distant coating by up to a few dozen angstroms; then, it buckles. Once it buckles, then the force it exerts on the distant coating drops to a low value, making further action on the distant coating unlikely. > Any one of these effects may be pretty good mitigation. It kept the Shuttle Endeavor working correctly. > As far as statistical models go, it seems like there are still > too many unknowns to make meaningful predictions. Depends on what you require to be "meaningful". It seems to me that there are enough "knowns" to say that pure tin often grows whiskers that reach lengths of millimeters. And that these have caused a few hundred documented cases of major losses. And it seems to me that we know that the density "D" of tin whiskers is NOT a sharply defined number. Rather, we know that "D" is a probability distribution that extends from "D=0" to "D ~ 100 to 140 whiskers per square millimeter" for bright tin on copper alloy; and, "D=0" to "D ~ 8 to 10 whiskers per square millimeter" for matte tin on copper. This range presents a huge problem! And I agree completely that this range is a major "unknown" in the sense that we do not know how to predict the value of "D" for a given tin plated surface any better than that; on the other hand, we know that this random variable has this range. Yesterday, I had to estimate the number of tin whiskers that might be growing on a 120 square millimeters of pure tin on electrical feedtroughs in some space equipment. Now, three years after the plating. And at the planned end of the mission, nine years from now. Allowing for the solid angle filled by other conductive surfaces, and their distance --- but using the upper limits of "D" --- these estimates ranged from 30 to 3,000 whiskers. And Jay and I have in our offices, some whiskering specimens for which even the "3,000 whiskers" would be apt. But we also have some specimens for which more like "zero whiskers" would be apt. The darned tin coatings show this wide range of whiskering densities. > But since the > entire field seems to thrive on example and anecdote, Data gathering is hard work. But there are a number of things we know pretty well --- * whiskering happens. * the density of whiskering has a very broad distribution, but has an upper limit that depends on some plating parameters, and on coating thickness. * the distribution of lengths is close to log-normal with median length growing roughly as 1 mm/yr, and standard deviation of log(L) ~ 0.8 +/- 0.2. (Log to the base "e".) * the thickness ranges between 0.1 um to 10 um. * whiskers are often kinked. * CALCE has useful tables for the distribution of "whiskering angles" --- the angle between the surface and the whisker. * whiskers have oxide coatings that act as insulators: a whisker can be in mechanical contact, but not in electrical contact, with a different surface --- until the potential difference exceeds the dielectric breakdown strength. This has been measured by Hilty, CALCE, Brusse and students and me, and Karim Courey: The breakdown strengths range between (roughly) 0.3 V and 15 V, with the largest seen so far at 45 V. CALCE has two dissertations published with a number of tables of data. Two more excellent works are in progress. Karim Courey is publishing a doctoral dissertation on dielectric breakdown strengths of the coatings on whiskers, and has had a summary accepted for publication in an IEEE journal. > I think it's worth asking this: > Does anyone know of a failure of a fine-pitch electronic > component that has been conformal coated? I do not. But there are some pretty wimpy coatings in use, and the literature reports at least some escape rate for them. One problem is that many folks who have failures, clamp down rigidly on the news, so we do not learn of these failures! Jay and I meet that on a weekly basis. This field has many many skeletons silently littering the closets. One example: The SWATCH group had a 5% failure rate (and inspections showed a 30% whiskering rate) for their electronics. We know this because the petition for relief by SWATCH lawyers became distributed on the web. But have you heard of this from SWATCH? I doubt it. This is a field in which Absence of evidence is NOT evidence of absence. > Furthermore, has anyone seen a picture of a long, straight > whisker that's grown through a coating? > With all of the pure tin terminations that are out there, we > should have some examples by now. I am attaching an image, "TomWhiskerThruCoating.png" from the article by Tom Woodrow. It may be what you are looking for. You can find the two articles by Tom on Jay's web site: look for Tom Woodrow and "conformal coating". > Joe Kane > BAE Systems > Johnson City, NY > > > -----Original Message----- > Ray > Sent: Thursday, March 13, 2008 5:30 PM > Subject: [tinwhiskers] Re: Conformal Coating? When Reliability > Goes Astray > > Bob, > > The harder the conformal coat the more it will slow done the > propagation of the tin whiskers, slow down is NOT the same as > stop. It is the un-contented nature for tin to want to be in a > crystal string, not in a ball. It will penetrate, just like a > trees roots will penetrate through concrete. I suppose a whisker will push its way through a soft enough coating. Perhaps silicons are soft enough? We do not use them, and so we have not studied soft coatings. The Uralane 5750 that we studied is "cohesive" enough so that none of "our" tin whiskers have plowed their way through the coating, bursting out of the other side. Rather, each whisker we have examined, has carried out the following program: * The young whisker is stubby and exceedingly stiff. It pushes upwards on the conformal coating, and detaches the conformal coating from the tin coating. This forms the conformal coating into a shape like a "circus tent" with a single main pole. At least, this happens for whiskers emerging from the tin coating in a nearly vertical direction; for a sufficiently large tilt (ie. for whiskers growing nearly parallel to the surface) the whisker wedges between the coatings and grows like a "gopher". Well, we suppose that would happen --- we have not seen it. * As the whisker grows longer, it detaches a wider disk of conformal coating. The upwards force needed to do this grows about as the circumference of the disk, which is roughly proportional to the length of the whisker. (The conformal coating stretches under creep, and this spoils perfect proportions; also, the peel-angle between the conformal coating and tin coating changes, and so the peeling force is not perfectly proportional with the circumference.) And the Euler buckling force needed to snap the lengthening whisker into a bow, decreases as 1/length**2. * There comes a time then the force needed to peel away even more conformal coating (increasing like "length"), exceeds the Euler buckling force (decreasing like "1/length**2". Then the whisker buckles. --- We have watched the whiskers push up from under the conformal coating, and make "circus tents", and then buckle and form "igloos" with crumpled whiskers inside. After nine years, no whiskers have broken thru our Uralane 5750 conformal coating, where it is 2.0 mils thick. Tom's coatings are "leaky". But they captured most of the whiskers. Tom is re-visiting this work, and we look forward to seeing more of his good work. > In eutectic solder > there is enough lead to keep the tin contented and keep it from > forming tin whiskers. Therefore, a hard conformal coat > (urethane), only slows down the growth of the tin-whiskers. The > better the adhesion, the more it will mitigate the growth. We did not see any slowing in the accumulation of arc length. But each of the ones under our conformal crumpled, and never reached above 2 mil, so they "slowed" dramatically in this sense. > In regards to adhesion of the conformal coat, any trace of > silicon on the surface, which may not be detectable with a 30x > power microscope, is the worst contaminate for either urethane or > acrylic conformal coat materials. This is in addition to other > residues. Our experience is that a monolayer of silicone can induce debonding. We work hard to keep silicones far away. We do not have debonding, typically. Every now and then, we do, and we stop work and hunt down the source of the silicone and remove it. > Respectfully, > Raymond Bennett > > President > RNB Enterprises, Inc. > 602-889-3461 Direct > 602-978-0248 FAX > > > > > -----Original Message----- > Bob Landman > Sent: Thursday, March 13, 2008 1:38 PM > Subject: [tinwhiskers] Conformal Coating ? When Reliability Goes > Astray > > We are told that conformal coatings are a successful tin whiskers > mitigation strategy. Correct me if I'm wrong, but I don't > believe it's been demonstrated to any significant degree that any > conformal coating on the market today will "prevent" tin whiskers > from punching through the coating. I don't know what you would say is "significant". But Jay and I have been watching for nine years, and none of the hundreds of thousands of whiskers growing under our 2.0 mil thick Uralane 5750 coatings have made it thru. This is 100.000% successful confinement. And our Shuttle Endeavor was flying for 19 years with tin plated card guides. When we looked in 2006, we found 100 million to 300 million whiskers with lengths longer than 0.5 mm (up to 25 mm) surrounding the 100 conformally coated boards. There is no drizzle of shorting failures. (Jay and I think there was one.) This is "not bad, not bad at all". Naturally, the Shuttle program removed all this pure tin when it was found, and cleaned all the boards, and inspected them, and tested them. and tested the re-assembled boxes. Please understand that I hate the use of pure tin. I work hard to see that it is not used in our equipment. Sometimes, we find that it escapes into our gear, and we are about to launch. Do we kill the launch? "Yes" under some circumstances and "NO" under others --- being able to estimate the risk helps us here. > I note that it's popular to use the word "mitigate" and that's a > word that is not as strong a word as "avoid" or "prevent". I > await someone who can do the math on how statistically > significantly conformal coatings "mitigate" tin whiskers. "Mitigate" has two meanings: 1. Complete fix (or completely cure) a problem 2. Make the problem less serious Thus, it is ambiguous, and so I try really hard to not use this word. As for the math, if one is using 2.0 mil (or thicker) Uralane 5750 coatings like the ones we used, then the risk of a whisker shorting something out is decreased by a factor of 10,000 or more. > A dictionary states that the word means "To moderate (a quality > or condition) in force or intensity; alleviate." > > By how much? The word itself gives us no clue. > > If a whisker can grow from one pin on an IC package, then > certainly, it can also grow from adjacent pins as well and then > don't we have the perfect opportunity for shorts? We have a number of examples of tin whiskers shorting between pins on DIPs. Given the number in service, the probability of this cannot be 100%. It must by lower. I estimate it must be lower than 0.1%, but your milage may vary. We have not seen shorting events for conformally coated pins, but we can certainly imagine various shorting possibilities... * whisker grows out of one coating, thru air, and then collides with the other coating and pushes thru it to contact the other pin --- this is essentially impossible since the whisker could not push thru the second coating. * push open a gap between the conformal coating and the substrate, and "mole-along" to the other pin --- this is possible. We have seen a tin whisker burrowing under a tin oxide coating for 100 mils. Apparently this has only been observed once, of thousands of images acquired over the last few years. So I will not dismiss it, but I suspect it may happen only once per hundreds of thousands of pins, or less frequently. Let us continue to look for examples, and to refine our estimates of how risky this path is. It will get more risky as gaps tighten up. * push out of one coating, thru air, and contact a second whisker moving the other way. This could happen. Bell Labs computed it as 1:100,000 or less, when there is NO conformal coating. If one uses 2.0 mils or more of Uralane 5750, then it should not happen at all since neither whisker could escape. > I just read the article below on reliability of conformal > coatings that I thought worth sharing if we are going to have to > count on such a coating to save our lives. This article did not appear in my copy. Where do I find it? > -Bob Landman/H&L Instruments,LLC
Attachment:
TomWhiskerThruCoating.png
Description: PNG image
