How to Find the Perfect CSP Test Handler (and Other Impossible Tasks)

It often seems that the rule in test is that equipment to handle parts for testing must lag demand by six months, a year or more.

When I'm asked, "What's the best way to handle and test CSPs?" my initial reaction is to fire back with, "What's the best way to handle and test any package?" But I stifle the urge (most of the time) and try to put myself in the questioner's shoes.

After 26 years in the industry, working in packaging, assembly and test, perhaps I should be more understanding.

But the problem is that, after almost a quarter of a century, the industry, at least with regard to test, is pretty much where it was when I started at National Semiconductor in 1974. The rule in test more often seems to be that equipment to handle parts for testing must lag the need by 6 to 12 months or more.

The CSP handler solution is obviously no exception. (Or perhaps the CSP handler is an exception, since it has been more than three years since the industry really started looking for solutions to the handling problem of BGAs and near-CSPs like the TSSOP and MSOP 8-lead SMT packages.)

Unfortunately for the user, existing trays, tubes, handlers and contactors/sockets will not work with new CSPs without new designs, kits and tooling. This is due in part, to the fact that the package is now essentially leadless and package outlines are not truly standardized.

The "standard" CSP design is a PC board with a molded cap up to the edge of the board. The contacts that take the place of leads are now at the extreme edges of the package (for SON-type packages) or under the edge of the PC board, as in a BGA. No longer is there a body and set of leads to locate and guide the package.

Whose fault is it that handlers are lagging CSPs? Can we scream at the handler vendors?

With all fairness to the handler industry, the typical vendor is an unwitting victim of circumstances who can supply handlers for packages only after receiving production samples and tubes or trays.

"UNFORTUNATELY FOR THE USER, EXISTING TRAYS, TUBES, HANDLERS AND CONTACTORS/SOCKETS WILL NOT WORK WITH NEW CSPs WITHOUT NEW DESIGNS, KITS AND TOOLING."

(Aha! a possible hint to management to help resolve future handling problems with new packages: work with the handler vendor and packaging supplier to get the parts and carriers to the vendor in sufficient quantity and in first-class condition before you start shipping to your customer. And maybe management should not assume that the customer's test staff has a solution for handling and testing that new package.)

All right then, given the difficult situation, what is the best handler for CSP? Why is it the best? And when and where can you get one?

The answer, unfortunately, is that there is no "best" handler for any package, old or new. We all have different requirements, based on our company's installed base of handlers and test systems as well as on existing flows. In addition, the handler design depends on the device type you are running and the ability and quantity your test system can test in parallel, if it can test in parallel at all.

We should remember too that most of us are using subcontract assembly vendors and that the packages can and do vary from one vendor to another and, in many cases among the factories of a single vendor.

With no time or funding to plan and implement a better way, we are forced to fall back on the tried and true pick-and-place handler as the solution. Of course, we all realize this is just a "temporary fix" until we can get a production handler situation.

Yet this industry workhorse and temporary fix has enabled us to handle and test any package developed, as long as the parts are supplied in a tray, or on a flat plate, or the handler is supplied with a tube kit and a socket/contactor that has been custom designed.

The trade-off for the flexibility of the pick-and-place system is its inherent slowness, which makes it prohibitively costly when handling inexpensive parts. This additional cost is due to handling speeds in the neighborhood of one second per device, as well as the added cost of designing and manufacturing custom kits, trays, sockets or contactors.

All of which result in added costs and low throughput per system, as well as delays of months, and even quarters in getting the needed hardware.

For what it's worth, I suggest that you and your management address the following issue now, not later: With all the new package variations able to be tooled and in production in a couple of weeks or less, it is an absolute must that test start working farther back in the organization-whether internally or with a subcontractor.

The ability to review new package designs, obtain advance samples and work with the assembly staff, as well as with handler vendors, is key to meeting the next new package challenge.

The giants-Motorola, IBM, and the like-are no longer funding new handler designs like Motorola did with Delta's initial pick-and-place and the new FICO in-line systems. Worse, the other big guys and the subcontractors merely promise to order at least one, but only if and when it works at someone else's factory.

So where can you turn for help in selecting a CSP test handler now?

Short term, I am sorry to say, you are stuck with the pick-and-place solution for most CSP packages. And with some ingenuity you can utilize gravity feed for the near-CSP designs and for some of the SON replacements for leaded packages.

Long term, I suggest you look at the new "strip handlers" being developed by Aetrium, Aseco, Delta Design, FICO, MCT (which has units in the field), and the rest of the handler and trim-and-form tooling vendors. Socket vendors ELH, OzTek and others are supplying contactors to test multiple units in parallel in strip format. This "strip test" strategy promises to be the basis of a new era of testing, especially for low-ASP, long-test-time, high-volume, multisite testable packages.

Of interest here is that ASA, ASM, FICO and other traditional assembly tooling companies have realized that for leaded packages and the strip format, they have a definite advantage with their experience in the assembly of these packages.* They have spent the last decade or more handling and processing semiconductors in a leadframe or strip format. To test they say they "just" need a contactor, and they can supply you with a "strip handler." If only it were this simple.

Can you buy a handler today for the CSP package you just got? Absolutely. Are the available pick-and-place handlers the best solution? Probably not, but they beat hand testing, by a little bit anyway.

Can you test in gravity fed handlers using tubes? In many cases, again the answer is yes, but this will require new tube and handler kits. Aetrium, MCT, and Uptech are working on or already provide gravity-fed handlers for some of the new CSP packages.

"THE TRADE-OFF FOR THE FLEXIBILITY OF THE PICK-AND-PLACE SYSTEM IS ITS INHERENT SLOWNESS, WHICH MAKES IT PROHIBITIVELY COSTLY WHEN HANDLING INEXPENSIVE PARTS."

By replacing the softer, cheaper polyvinyl tubes normally used with the new polycarbonate tubes from Senyo Corp., Tokyo, the jam rates improve immediately, from the low one in hundreds to one in the thousands. And even better, the tube can be supplied in 20 inch lengths, further increasing productivity. The initial tube cost, over the polyvinyl tube, is infinitesimal when calculated against the uptime and UPH increases seen to date.

Are the new strip handler designs, like those used by FICO and MCT, the ultimate solution to testing? Not at the present for many of us, and maybe never for your particular application and package.

Some of the delay in acceptance and supply is due to lack of a support structure, hardware and tooling needed to allow strip handling; and some is due to the existing culture and the "we've always done it this way" syndrome.

As with anything new, we all want a guarantee it will work before we buy in. Thankfully there is always at least one brave soul who will convince his company that something new needs to be done to meet the production and cost goals being given to test.

In the interim, most of us will just have to work with what we have, pick-and-place and gravity-fed handlers.

Keep your sense of humor, because the future of test handling for the smaller, cheaper, finer-pitch packages and true chip-scale packages, such as flip-chip, will continue to challenge all of us, and for longer than we'd like.

*See the May/June 1998 issue of Chip Scale Review for my article that details some of the other handling options that have been used in low-volume production testing of CSPs.