Catching Up (Literally) with Uyemura’s George Milad at SMTAI 2016
I managed to catch up with George Milad of Uyemura at the recent SMTA International conference, which wasn’t easy to do. But he did take a few minutes to fill me in on his schedule for the week and it certainly was packed, with a tutorial, presentations and IPC committee meetings.
Patty Goldman: Hi, George, it’s great to see you again—about twice a year, right? Let’s talk a little bit about what you’re presenting here at the SMTAI conference.
George Milad: I have had a very busy schedule. Yesterday I gave a presentation, a four-hour tutorial, and I will never do that again. To stand up and talk for four complete hours, it's very challenging staying alert and relevant and keeping the audience’s attention and making sure you're delivering value, because these people paid money to sit in that meeting. That was very challenging but it went very well and I'm happy. I get energized by the feedback from the attendees in the class, but then when I got back to my hotel, I was trying to listen to the presidential debate and I fell asleep after the first five minutes. (Laughs)
Goldman: I heard it was not that exciting of a debate, anyhow. But that does sound exhausting. How many people were in your audience?
Milad: We had approximately 15 people who signed up and paid money. That is very rewarding, because I gave the same training course at PCB West and I think if you signed up it was free, and I had three people show up. I talked forever for three people and it kills your energy, but this one was a very rewarding experience.
Goldman: You imparted a lot of knowledge.
Milad: Yes, they thought it was good. They filled in the forms and so I'll get some feedback from SMTA. Then first thing in the morning tomorrow I'm going to talk about the “Via Fill Copper Plating Process” and what it takes to optimize the via fill process. I-Connect007 published an article along these lines recently and this is basically the same concept but in the form of a presentation rather than an article.
Then I have the IPC 4-14 Plating Processes subcommittee meeting. Our committee meetings are not so critical because we do our work by conference call.
Goldman: Yes, I was very impressed when we talked about that before. I've been very impressed with how you guys conduct those meetings every…how many weeks?
Milad: Every other week.
Goldman: You get a lot accomplished that way.
Milad: Indeed. Then you set things in motion, and you say, "Okay, we need to just make a test vehicle (TV) ." Somebody manufactures it and then every couple of weeks we find out the status, and when the TV is complete, we send it out to people to plate it. Then we send it back out for evaluation We're always following these activities and evaluating results, and then we begin to write the specification and get interaction from people who sometimes like what's being written or some people take a strong objection to what's being written. And for good reason—I'm not saying they will just object for the sake of objecting.
For this, we were rewriting the ENIG spec. We wanted to set the upper limit for ENIG. It was written in 2002.
Goldman: That's a long time ago.
Milad: Yeah, 14 years later we're revising the specification. In 2002, we agreed that gold should be a minimum of two microinches. We did not put an upper limit, but we wrote that it should typically be 3−5 microinches. Typically, 3−5 is not the specification. Minimum of 2 microinches is the specification. Anybody who read the specification said the IPC calls for 3−5 μins, which was a horrible thing. We never called for 3−5 μins and it created a lot of problems.
Now, we’ve come back and we are redefining the gold thickness. We put the lower limit of 1.6 μins instead of two. Then, we wanted to put an upper limit. We tried very hard to pass 3.2 μins as the upper limit. We got a lot of resistance from manufacturers who do ENIG surface finish for the medical industry. They called for a minimum of 3.0 μins. They are the biggest printed circuit shop in the United States, I don't feel I should mention their name, but they said they have medical customers who will require a 3.0 μins minimum. We wanted to put 3.2 as the maximum, leaving no room at all to work between 3.0 and 3.2 μins. They kept arguing and finally we conceded and we put 4.0 μins as the upper limit. That is what the new specs will say, 1.6 to 4.0 μins. This is the first time we put an upper limit and as far as the 3−5 μins, we're hoping that will go by the wayside.
Goldman: I imagine it is a problem to put too much down, isn't it?
Milad: Absolutely. The problem with putting too much down is you're staying in the gold bath for a prolonged amount of time and that gives a lot of opportunity for nickel corrosion, or black pad. It's always been our objective to say, “Don't do that. You need only 2.0 μins. Get them deposited on the nickel and get out of the bath. There is no reason to stay any longer.”
It's not that it's guaranteed that if you stay longer in the bath that you're going to have corrosion, but if the nickel is a little bit compromised, it's not the perfect nickel, and you stay too long you will get corrosion. If you have the perfect nickel, it can withstand prolonged dwell in the gold bath, but nobody makes perfect nickel 100% of the time. We felt it's very important to restrict the gold thickness.
Goldman: This corrosion you're talking about is black pad, right?
Milad: Yes, the nickel corrosion and black pad are synonymous. If it is too extensive, then the pad is not going to solder, and the whole objective of the ENIG finish is that it’s a solderable finish. That's the story there. We don't want you to stay too long in the gold baths.
We also are including a corrosion chart in this new revision of the ENIG specification, which was quite challenging because when you have corrosion and you do a cross-section, it appears as spikes. When you look at a cross-section under 1000x magnification and you see less than 10 spikes in the field of view, that's acceptable. If you have more than 10 spikes, that's not acceptable. If the spikes are joined together, meaning they blend, one spike runs into the other, that's not acceptable. If the spike is too deep, that's not acceptable. We specified that and we ran into a huge tizzy of argumentation.
Is it 10? Is it 8? Is it 9? Is it 11? Is it 12? We finally said that we're just going to define what is acceptable and we're going to define what is rejectable. We're going to admit that there is a gray area between acceptable and rejectable.
Goldman: Yes, like saying 10 is good but 11 is bad? It seems that people might say, "Wait a minute. That's not that bad, is it?”
Milad: Yeah, so we decided to leave everything in between for the buyer and the manufacturer to sort out. If it's less than 10, we don't want any argument. We want this to be acceptable.
Goldman: If it's over 15, you won't want any arguments.
Milad: We do not have any argument there; also, we want that to be rejectable. The buyer can say, "I'm rejecting your product because...” Or the seller can say, "My product is okay because…” Everything in between can go to a lab for failure analysis to prove that the product is solderable. You can do so many things. It's not like an open and shut case. We produced this corrosion chart and we had three levels. We had acceptable, discussable and then we said, “No, we can have acceptable, rejectable, and then everything in between we're going to leave.”
We felt that it's a significant improvement. Because people used to look at one spike and declare a product rejectable. That was wrong. It would go to failure lab analysis and they’d look at the cross-section, see a spike, and end of story, it's black pad. That's why it didn't solder. There could be 100 other reasons why it didn't solder, but then you stop the investigation because you found that spike and you say, "Oh, it's black pad. I don't have to look any further.” This was a problem, and I'm hoping this corrosion chart will be a significant achievement in helping people understand the product I'm talking about.
There's one other thing that we also included and it’s something that we noticed with the XRF equipment that is out there in the field at the board manufacturing sites. Everybody has an XRF machine and they measure the nickel and gold thickness. We found out that some of these machines are not generating the correct numbers.
Goldman: Oh, that’s not good.
Milad: Especially for very thin gold. For the nickel, at 200 microinches, if you're off by 5% and the range says 120 to 240, that's okay. But when you're looking at gold and you want to differentiate between 1.5, 2.0 and 2.5 microinches, that becomes very challenging. So we have written a section on how to verify that your equipment is working properly. It's very simple. You take a pad on the part and you sit there and read it 30 times—the same pad. Just read it and read it and read it, 30 times. Then, do the stats on the numbers that are generated. If there's too much variability, then you have a calculation now. Because your machine has this much inherent variability, reading the same spot over and over and not running around reading from panel to panel, which means your machine can only be good at this level.
For example, if you have a very good machine then 1.6 to 4.0 is acceptable. Any reading you get between 1.6 and 4.0 is the right reading and that's what we'll use. If your machine has got too much fluctuation in it, then you must narrow the range for your machine. You go to your buyer and you say, "My machine isn’t capable of 1.6 to 4.0, so instead I have to give you numbers between 2.5 and 3.2, where I know the variability of my machine keeps me in the right range.”
I tell you, we had so much flak when we were doing this. People were arguing because you're putting the board shop guys with the compromised pieces of equipment on the spot and they must go out and buy new machines. The new machines that give good readings are in the $100,000 range. It's created a lot of flak but I think we gave them an out.
Goldman: You think you've answered everybody's concerns?
Milad: Yes. I think the way we phrased it is, "If your machine has inherent variability then you have to narrow the range of acceptability.” We inferred that would be acceptable.
Goldman: When is this revision going to be published?
Milad: The revision is in the final draft stage. Meaning it has been written and it has been reviewed. We got a lot of comments, we resolved the comments and now it is in the final draft. We are looking for a few better demonstration pictures. Gerard O’Brien is working on that, very extensively, to try to put some illustrations in there that are meaningful to the people using the specification.
We also had to reconcile with IPC-6012, which spelled out acceptability criteria and methods of sampling for process control. The way we did things was a little bit different. We had to resolve it, and it was resolved. I wasn’t at that meeting but Gerard OBrien worked with the 6012 committee and they've come to an acceptable way of writing it so that 6012 is satisfied and the 4-14 subcommittee is satisfied. It's ridiculous to have conflicting recommendations from committees operating under the same organization.
Goldman: Now you must be about ready to go to final approval. Is this where it goes out to the membership?
Milad: We're getting ready for what we refer to as going to ballot. We had the draft and we had the peer review. Now we have corrected everything and then it's going to go to ballot and hopefully it will be out before the end of the year.
Milad: It's a big achievement.
Goldman: It certainly is. Now, you still have a third presentation tomorrow afternoon, correct? A third paper?
Milad: Yes, our technical director, Don Gudeczauskas, has done some research and put together a paper that was accepted in SMTA but he had other travel plans. So I am presenting the paper for him, and that's challenging, because I must get into his mind and see how he laid it out and what message he wants to deliver. I've been talking to him every day now to make sure we're on the same page.
His paper is interesting because he was looking at three different types of gold over nickel and palladium. He referred to them as standard immersion gold, high efficiency immersion gold and “reduction-assisted immersion gold,” so three categories. Two of them are immersion golds, and immersion means it is an exchange reaction. Nickel goes into the solution and gold comes out of the solution. Then he had a third gold he referred to as reduction-assisted immersion gold. It is a combination immersion and electroless deposition process. This immersion gold is more active at the initiation of deposition; the electroless reaction is continuous throughout the deposition process. It does not corrode the underlying nickel.
Goldman: Very interesting.
Milad: He was presenting data on the solderability of these three types of gold on nickel, and the solderability of these three types of gold on nickel palladium. He did ENIG and he did ENEPIG in an effort to differentiate between them.
He pushed some parameters to see a difference. One of them being, staying in the gold bath too long, because then he can create some corrosion so he sees if there is a propensity to prevent corrosion. It's an interesting paper. The differentiation was in the degree of nickel corrosion in both ENIG and ENEPIG. I think it makes a case for the reduction-assisted immersion gold bath as a better alternative.
Goldman: You might have a little more leeway with it.
Milad: Yes, it doesn't go after nickel as readily as an immersion gold bath would.
Goldman: Exactly. Well, George, it sounds like you have yourself a few very busy days here. Thanks for taking time to stop by.
Milad: Thank you, Patty.