Testing and characterization of optoelectronic devices on the factory floor or in the wafer fab can be done sequentially, multiplexed, in parallel, or concurrently, for a wide range of sophistication and complexity. Let's examine each of these strategies, review how they work, and show where they're best applied.
In all automatic test strategies, communication between the test equipment and the component handler, robot, or wafer prober is critical. A long delay between placing a device under test (DUT) in the test fixture and signaling the test equipment to start wastes time, while starting the test before the DUT is connected generates flawed data. Keeping the chain of command as short as possible reduces the likelihood of miscommunicating the 'start test' and 'end of test' events. Sequential and Multiplex
In sequential test, the component handler puts a component in the test fixture and signals the test equipment to begin the test sequence. Once the sequence is complete, the test equipment sends the results to the component handler, which bins the part and places another in the fixture. Sequential test is appropriate when the component handler costs less than the test equipment.
In the multiplex test method, the component handler puts multiple components in the test fixture at once, and the test equipment incorporates a multiplexer to switch between devices. This approach increases throughput, especially if the component handler can move all components simultaneously, as with tape-and-reel systems or wafer prober systems with multisite probe cards. The method is appropriate for test sequences that are short compared to the time required to index to the next set of test sites on the wafer or position the next set of DUTs in the fixture. Parallel and Concurrent
Adding multiple test sets allows parallel test, in which the system tests all components at the same time. This is good when throughput is vital and the test sequence execution time is significant compared to the component handler or wafer prober index timeand especially when the test equipment costs significantly less than the component handler or wafer prober. To accomplish more than about a two-time improvement in throughput over sequential and multiplex test, parallel test equipment must operate autonomously during execution, preferably using embedded real-time controllers.
In this example of synchronous concurrent testing, a robot component handler tends a four-site test system. The robot loads modules in each test head in turn, triggering test equipment as it does so.
Robotic component handling at the module level requires more sophisticated test strategies. Consider a robot component handler tending a four-site test system (see figure). The robot loads a module in each test head and signals the test equipment to begin testing on that head, then goes on to the next. Each set of test equipment executes concurrently on different modules. As the robot finishes loading the module on the last test head, the first test head is reporting results so that the robot can put the module in the pass or fail bins. This strategy is called synchronous concurrent.
Most module testing requires a number of components to be tested. If a component in the module fails, it's best to stop the test and reject the module rather than continue testing it. This calls for an asynchronous concurrent test strategy.
The choice of test strategy depends upon the relative speeds and costs of the test equipment, component handler, robot, or wafer prober. When deciding on a strategy, make sure that the equipment has all the capabilities required and also provides an upgrade path as test throughput needs change. oe
Paul Meyer is a product marketer in the Small Systems Group at Keithley Instruments Inc., Cleveland, OH.