Combination
Grid–Prober Test

written by:
Duane Delfosse

Combination grid-prober (CGP) testing is being employed with increasing frequency as the density of SMD lands on boards continues to increase. A number of factors are at work. Grid testers, as a stand-alone test methodology, provide a fast and comprehensive test, but they require costly fixtures and suffer from intermittent “false opens” when testing very dense or fine pitch boards. Flying probe testers, on the other hand, are exceptional in their ability to handle density and fine pitch, but they unfortunately suffer from long test times relative to grids. The test time penalty is so severe, that in some quick turn environments, fixture test is the preferred methodology only because lead-time does not exist to probe the boards. Since the total cost of test is lower for flying probe than grids below quantities of about 500 boards, some board manufacturers are adding flying probe capacity in much the same way that multiple drill machines together provide the capacity to meet production volume. While this solution may be satisfactory for some, it is cause to reassess combination grid and prober testing as a better use of capital.

CGP can serve two primary roles:

• The cost and complexity of grid test can be reduced if dense and/or fine pitch portions of a board can be transferred from the grid to the prober in the form of a split net test.
• Grid test results, including false opens, can be verified quickly and more accurately by a prober than by a human

Until very recently, these two different functions existed as isolated processes. Separate data preparation processes were needed, and in some cases, the entire test mode on the prober was different. This has recently changed, as a new capability exists that allows both functions, split net testing AND grid fault verification, to be performed in the same prober test cycle. This is an attempt to explain CGP in a practical sense. Since CGP is and should be a “push button” function in test data processing, a deeper understanding will maximize it's potential for most users.

The Driving Force for CGP

Several factors might cause one to consider a CGP approach. Higher than usual test point density and/or fine pitch (or small feature size) are the two most common factors. In North America, high density is the more common of the two. Another factor is the need to prevent test escapes. An automated CGP test environment reduces the chance for escapes by minimizing analyze and handling errors while assuring that repairs and retests occur in a data driven environment.

Density, in the absence of fine pitch, is usually driven by the presence of large, high pin count BGAs. BGAs can be positioned side-by-side or “tiled” to create regions of extreme density. Depending on the grid density and the deflection capability of the fixture, dense boards will either solve within an acceptable pin displacement, or will result in “grid starvation”. Some telecommunication and server boards are beginning to exceed the grid density offered by a double density (200 test points/in2) tester with 0.8” (20mm) of pin tilt.

Figure 1 - Regional density results in grid starvation

Typically, if a dense area of about 30 in 2 is populated at greater than 280 test point per square inch, a double density grid solution alone will not work. For single density testers, the threshold is closer to 150 tp/in 2 . If a dense area does not cover so large an area, or if the regional density is lower, leaning pins in the fixture will usually avoid grid starvation.

Fine pitch in the absence of density poses a different challenge for grid testing. Although the pointing accuracy and fine pitch capability of grid fixtures has been improved substantially in recent years, not all PWB manufacturers have a fixture process capable of producing such fixtures. For example, grid testers are regularly testing production quantities of boards with 12 mil (300 m ) pin-to-pin pitch. Automatic board to fixture alignment is usually necessary for these applications.

Figure 2 illustrates an otherwise conventional board with a direct chip attach (DCA) site that is a good candidate for CGP. As the number of DCA sites per circuit increases, the effectiveness of CGP decreases because more and more of the test is moved from the grid to the prober.

Figure 2 - Direct Chip Attach Site

What is the CGP Process?

A2

• Run Split Task
• Run Verification Task

Repairnet

• Read A2 Fault file

Magic Suite

• Fixture Cam
• CGP Split

Grid Tester

• 9090
• Mania
• LM

DPS

• Process for A2

Standard Test Program
IPC-D-356A
A2 Input

Format
Grid Fault File
A2 + Grid

Combined Fault File
Pass Boards
Fail

Boards
Repaired Boards

CGP can be approached from several perspectives. ECT's implementation involves specific capabilities on the part of the test data processing software, the flying probe and the repair software.

Figure 3 below illustrates the ECT process flow.

In ECT's implementation, Magic Suite performs the CGP split on dense and/or fine pitch boards and outputs standard fixture data and grid test programs. It also outputs an IPC-D-356A netlist that is fed to the A2's DPS station for the fixed portion (CGP split net) of the prober test. The role of the grid tester is to test the grid portion of the CGP split and produce a fault file and bar code tag in it's native barcode format.

The A2 prober accepts processed data from the DPS pre-processing station that contains the full netlist and adjacency data so that verification and analysis can be performed on random grid faults. If desired, a full board test can also be accomplished with the same input data. The A2 reads the grid tester bar code, retrieves the appropriate fault file and performs a combined test consisting of the CGP split portion, and any grid verification that might be required. After completion of testing, a new bar code and fault file are produced for use at repair and subsequent re-test.

The repair software must read the bar code from the prober (and/or grid testers as well), retrieves the appropriate fault file, and steps the user through the faults. After repair, the prober must be capable of retesting boards with sufficient test coverage that boards “ship from the prober” and never go back for a retest on the grid.

A very important requirement for CGP is the use of bar code input at the flying probe and repair station process steps. A robust and automatic process cannot rely on people to key in a complicated alphanumeric string to achieve proper test. In the ECT implementation, all data retrieval is driven by bar code to minimize the chance for error and test escapes.

Algorithms for CGP Splits

Whether the driving force for CGP is density or fine pitch, the general strategy for performing a CGP split is the same. Both seek to maximize the isolation test coverage on the grid, while pushing the grid fixture to the edge of its capability. If too many test points are pushed to the prober, a simpler grid fixture results, but the prober pass is lengthened unnecessarily, a cost that is borne by every board tested. Conversely, if the grid fixture exceeds its design capabilities, the result will be poor throughput due to false opens or excessive analyze times. The CGP approach differs significantly from a traditional “split net test” that seeks to produce two relatively balanced fixtures.

CGP for density is typically handled by analyzing the dense areas of the board and putting to the prober pass points from large nets until the density drops to a range that can be handled by the fixture technology employed. The algorithm must take into account the grid density of the tester and the allowed pin displacement of the fixture style. Such an algorithm isn't quite as simple as it would seem because it is impractical to iteratively …remove points, solve, remove points, solve… until an optimum balance is achieved. The processing time would be prohibitive. Typically these algorithms move points from large nets such as power and ground to substantially depopulate test points in BGAs. As an objective, at least two points from each net are included in the grid pass to achieve isolation test coverage, and to assure that if a single pin from any split net fails to touch it's pad (or there is an open associated with that pad) a continuity failure would be found. It is usually possible to achieve 100% isolation coverage if density is the only driving factor for CGP. See Figure 4 for an illustration of how density can be relieved by depopulating a fixture of just a few nets.

Figure 4 BGA Land Pattern with Ground Net (only) Highlighted

CGP for fine pitch usually uses a slightly different algorithm. Two possibilities exist for establishing a fine pitch device as being “too fine” for grid test. In some cases the target pad sizes are too small to be reliably contacted. In others the pad sizes are sufficiently large for contact, but are spaced too close together to permit use of available pins with their associated fixture drilling tolerances. In the case of pads too small for reliable contact, the solution is simple; test all pads smaller than a user specified size on the prober. In the case of large enough pads, but too fine spacing, then selective pads may be left in the grid pass to achieve the maximum isolation (first priority) and continuity coverage (secondary priority). As with CGP for density, every effort should be made to leave two points per net on the grid. When 100% isolation test is not accomplished on the grid , isolation testing for these nets must occur on the prober using adjacency to reduce the test time. In the ECT implementation of CGP, Magic Suite calculates the optimum test coverage between the grid and prober.

Role of the Prober in CGP

The role of the prober in a CGP test environment is to perform the split net task and verify grid faults, in real time, all in one test cycle. Ideally the prober should also be able to perform a full board opens/shorts test using the same data input. When a board has been found defective and repaired, it is desirable to retest on the prober rather than set up the grid fixture again for retest. To avoid going back to the grid after a repair, the ECT CGP implementation on the A2 employs a number of more extensive retest algorithms to not only verify that defects have been repaired properly, but also to assure that no new defects were introduced during the repair process.

Advantages of CGP

There are a number of advantages to using the CGP test methodology. Here are a few:

CGP uses the strengths of each tester:
Flying probe testers are excellent for continuity testing while they are less efficient at testing for isolation between nets. Grid testers excel at fast isolation and leakage tests, and perform “easy” continuity testing cost effectively.

CGP extends the life of existing grid testers.
With CGP, older and single density grids can contribute effectively to the test of high-density boards.

CGP reduces the difficulty in dealing with leading edge grid fixtures.
In the event of board registration or fixture problems, the verification capability substantially reduces the time spend testing boards on the grid. It is no longer essential to get a “PASS” on the grid because the prober can efficiently re-test grid faults and perform re-test and analysis after a repair process.

CGP will limit the workload sent to the prober.
A typical scenario that occurs after the installation of a new flying probe system is for the available capacity to be immediately consumed by long-running full board test. CGP testing is a much shorter test, so flying probe capacity is effectively stretched as the total system throughput increases.

CGP reduces the chance for test escapes.
CGP also reduces the chance of test escapes relative to the alternative, split net fixtures and hand verification of grid faults. Bar code driven prober testing reduces the possibility that a board will be analyzed incorrectly.

CGP maximizes your capability with a minimum capital investment.
Depending on your current equipment, the CGP capability can be added to an existing flying probe and grid tester environment with minimal cost. If you don't currently use a prober, CGP will probably add the efficiency edge that will justify one.

CGP Case Studies
Four boards were selected to illustrate the effect of CGP. Table 1 contains the test point count for the grid test and the prober test, as well as the test time on the prober. Cases 1 and 2 are telecom/server class boards, case 3 is a hand held instrument subpanel, and case 4 is a 2-up “Camcorder” type board. Case 1 had a regional density near the threshold for a double density grid tester, but it was pushed over the edge because that regional density occurred over a 55 in 2 area. The fixture resulting from CGP was very reasonable, and the prober test time very short. Case 2 had a very high regional density, but the resulting splits are also quite reasonable. Case 3 had one DCA site on each of three images in the subpanel, so the extent of the isolation testing pushed to the prober was single point nets in the DCA area. Case 4 is an extremely dense and fine pitch board that shows the capability of CGP. Camcorders are often produced in high enough quantities that another approach may be viable. A quad density grid approach would also work for this part.

Table 1 CGP Case Studies

Practical Considerations for CGP
Most board manufacturers who already perform multi-pass testing will have little difficulty integrating CGP into their process. For those who don't normally perform splits, some care must be given to the way that it is introduced to the test floor. Both tests should be itemized on the shop floor control system routing, and care must be taken in board handling. There must never be a question as to the status of a board in the test department. A bar code tag must be kept with each board in a CGP split test sequence. For jobs that are not split (but receive only verification of grid faults), it is reasonable to keep bar code tags on only the failed boards. Clearly identified staging areas before each tester and between tests and in the repair area are helpful.