Current Issue An Independent Journal Dedicated to the Advancement of Chip - Scale Electronics July 2001 Email the editor

Paul M. Sakamoto Contributing Editor

In past columns, I have promoted the advantages of non-singulated device handling and testing.

There are various names used for the movement of devices still mechanically attached to the leadframe, but electrically isolated in such a manner that they can be tested as individual units.

Better Throughput

A few of the names used are strip handling, Matrix and panel test. I am sure there are others. They all share a common advantage, however: better throughput than singulated handling equipment.

This improvement over singulated device handlers occurs because singulated handlers typically move devices in serial, test them in parallel, and then bin them out serially. Strip handlers, however, always move and test the devices in parallel.

There are, however, some issues that non-singulated handling processes should address that are unique to the format.

The first is that some device types beg to be retested or re-screened if they fail testing the first time through. One example is the high-density flash device.

Some manufacturers' yield curves and processes make it possible to recover some amount of the devices that fail packaged test the first time out. If this quantity is even one or two percent of the total, many companies find it profitable to retest the fallout.

Retesting is easy to do if the test process consists of singulated packages. If strip handling is used, retesting fallout devices in the strip means that the entire strip would need to be run through the line again, just to recover some amount of the "lost" percentage.

To put it another way, the good devices are going to have to go through processing that they didn't need, which causes loss of throughput and possible loss of quality while being handled again. Since the advantages of strip handling are worth pursuing, a solution to this issue is worth pursuing too.

One answer to the re-screening of strip-processed devices is to go ahead and test, mark and trim and form only the good devices. The marginally failing devices that are to be re-screened are left unmarked on the top, but are still trimmed and formed. The re-screening is then performed on a separate low-volume singulated handler.

This works if the tester employed for the strip testing is available in a low parallel configuration, perhaps as low as a single site. This saves a separate program development effort and reduces the extra correlation efforts that may otherwise be required if a totally separate platform is used.

An alternative is to use a high-volume singulated handler for re-screening, but this only works if the quantities are high and if more programming effort is not required. Remember, a typical single site handler and interface costs less than a thirty-two site handler interface by itself.

Another issue in strip handling is that the tight packing density of devices on the strip tends to increase the average distance between the tester channels and device pins, which promises to be a big issue for high performance and for low-power devices with relatively low drive outputs.

Wireless Applications

This issue encompasses most of the devices we care about for wireless and other battery-powered applications.

There have been a few possible solutions posed for this set of problems, but I am not happy with them yet, since they usually involve really exotic interfaces or really expensive tester pin electronic designs (or both in combination).

I think the name of the game is economy, and if an economical solution is not available, we still have a lot more work to do.

In the meantime, working around the performance AC test at strip test issue in the test process flow is what I suggest. Correlated test limits, a rescreen flow (if first pass yields are high), and performance test at probe are a few possibilities. An inexpensive single-site handler and tester that is program compatible with the strip test cell is another.

Thanks to Bob and Earl, you both know who you are, for bringing up these core issues. The problems are often more interesting than the solutions, because they give us more to think about.