Rob Marcelis (BE Precision Technology ‐ The Netherlands), “H3D Profiler for Contact Less Probe‐Card Inspection”:
Probe cards require inspection since they are consumables subject to wear. Changes in probe position or shape can damage the semiconductor devices they are testing. As probe cards increase in size and probe count, the probe cards themselves are becoming more expensive to test in terms of test time and complexity. Each new test system typically requires an expensive “motherboard” for the probe card metrology tool to simulate the mechanics of the tester and provide electrical interconnect to the card for electrical testing.
BE Precision Technology took a different approach by focusing on only the probe positions (X & Y position and Z planarity) to inspect for wear. This allows for non-contact inspection using a laser-based system. The H3D system uses a “line scan” laser unit to scan the probe array in the Z and X directions. The unit is mounted on a motorized assembly to move in the Y direction.
For the prototype system they selected a lower cost laser system resulting in an initial measurement time of 16 seconds per pin. By reducing the scan area resolution, less than 4 seconds per pin was then achieved. In order to achieve the desired measurement rate of less than 1 second per pin, they will evaluate a higher cost laser unit with faster processing speed. (Though this may not be fast enough since a 20,000 pin probe card would take approximately 5.6 hours at 1 sec / pin. A target of ten pins per second is likely necessary.)
Using a micro-electro-mechanical system (MEMS) probe card as a test case the X & Y positions measured with the H3D were correlated well with measurement data from a traditional probe card analyzer. Since traditional analyzers use electrical planarity (which varies with contact resistance (Cres)) to measure Z position, these values did not correlate directly. For new probe technology under development to probe multiple stacked die for 3D packaging, it is no longer possible to measure Z position electrically as done on traditional probe card analyzers since each die’s set of probes are located at different heights. These new technologies will require a non-contact measurement.
One application of this type of tool is to scan the probe card before and after each use. Then by comparing these measurements, one can predict wear and lifetime of the card. Additional potential usages include integration with wafer probers to measure probe cards in-situ and measurement of test sockets since there are currently inadequate measurement tools.
Richard Portune (DCG Systems, Inc.), “Case Study: Integrating a 300mm Probing Solution with a Diagnostic Emission Microscopy tool (Meridian WaferScan)”:
Jointly Awarded Best Data Presented
DCG Systems builds photoemission electron microscopes (PEM) for failure analysis and device characterization. These tools operate in the infrared spectrum using indium gallium arsenide (InGaAs) cameras that can detect individual photons as electrons change energy levels. Since silicon is “virtually invisible” to infrared, this allows measurement from the backside of the semiconductor device while it is operating. And unlike an electrical probe which requires an accessible probe location – in the circuit itself not on a bond pad like traditional wafer probing – and changes the circuit itself due to the capacitance of the probe, optical techniques like PEM do not change the operation of the device under test (DUT).
The basic Meridian products image DUTs in modified packages (the package top is typically removed to allow access to the backside of the silicon) connected to an automated test system via a socket on a load board. There is also a need for the ability to test devices while still in wafer form without dicing of the wafer, so DCG Systems extended the product line to create Meridian WaferScan.
Due to the complexities of the PEM camera system, the wafer needs to be positioned and held so that the desired DUT is above the camera optics. This requires the probe card (while interconnected via flexible cables to the test system) to be aligned to the wafer and brought into contact with the wafer. There were significant mechanical and operational challenges being the opposite of typical wafer probing where the wafer is brought in contact to a stationary probe card.
Aligning the probe card to the wafer without the traditional upward and downward vision system found in a typical wafer prober was one of the challenges. Most advanced technology probe cards no longer support downward alignment like older cantilever style cards since space transformers block the optical path. This was solved by using their existing high resolution camera system (which is upward looking) to image the probe tips before a wafer is loaded. Then the wafer is imaged through the backside to identify pad locations. The images of the probe tips, wafer pads, and computer aided design (CAD) data (if available) are compared to provide the proper probe card movement in-plane and rotation in theta to align the probe card to the wafer.
Since the wafer is unsupported under the die being tested (unlike a wafer loaded on a wafer chuck) due to optical requirements, the wafer tends to bow when contacted. The “effective overdrive” of the probes therefore is significantly less than the movement of probe card (programmed overdrive). To compensate users have to program the system to greater than the specified maximum probe card overdrive. Since there is not full compression of the probes due to the bow, there is no damage to the probe cards.
Currently the system supports up to 45 kgF of probe force. As probe counts and the resultant probe force increase, they are developing a next generation system to support higher forces. Currently, direct dock probe cards are problematic due to their extreme size of the printed circuit boards (PCB).
John Strom (Rudolph Technologies), “Approaches for Reducing the Cost of High Pin Count Probe Card Test”:
Probe cards with a large number of probes require a large force to place the wafer in contact with the probes. The force required is a function of the number of probes since each probe acts as an independent spring. With today’s high force probe cards (some requiring upwards of 450 kgF), probe cards, wafer prober, and tester must be structurally stiff to avoid deflection. It is very expensive for a probe card analyzer (PCA) to apply a large force with the required stiffness to fully emulate a wafer sort cell for these high force cards. Over the past several years, there have been many papers on the challenges of high force probe cards at SWTW. At the same time probe card interfaces (“motherboards”) for PCAs are becoming even more expensive due to the increasing complexity and difficult requirements. (The number of different motherboards and the low utilization are significant contributors to the high non-recurring engineering costs faced by probe card manufacturers, as discussed in my SWTW 2011 presentation.)
Rudolph performed a study of the change in probe Z position as the probe card is deflected from the loading of the probes. Of greatest interest is the path traveled by each probe, i.e. the probe starting and ending positions and the scrub length. Since each probe has a slightly different height (planarity), each will contact the wafer at a different time as the wafer is pushed upwards. This variation in probe height also changes the force applied to each probe since force is a function of the compression or overdrive on the individual probe determined by the displacement of the tip from it’s initial position.
A typical measurement of a probe card is first to last touch planarity from where the first probe makes electrical contact to the point at which the last probe makes contact with the wafer. If it is not possible for the PCA to properly emulate the test cell (tester & wafer probe) with the same mechanical stiffness (spring rate) the result will not be accurate due to the difference in programmed versus actual over drive. Not only has emulation been difficult to achieve, there is often variations between identical test cells which creates variation in the actual usage of the probe card difficult to simulate on the PCA.
It was shown that the unloaded planarity (probe height without placing a load on the probes) actually correlates to scrub uniformity and probe force uniformity. While the loaded planarity does not correlate to probe performance, unloaded planarity measurements are more important than loaded planarity. In order to determine the actual probe travel (direction and length of travel as a function of over drive), Rudolph’s latest Vx4 PCA has the ability to over drive a small section of the probe head at low force to measure the movement of individual probes.
Therefore a paradigm shift between fully emulating test cell to determining what really should be measured is required. This will reduce complexity and cost of PCAs. It will also allow scalability for the future including larger wafers such as the 450 mm standard under development.