Evaluate fixture strategies for ICT

In-circuit test (ICT) is a time-tested and proven method of testing PCBs. During ICT, a bed-of-nails fixture provides test-instrument access to PCB nodes. Each nail is positioned such that, when the board is placed on a fixture and pulled down (or pushed down) by using vacuum, pneumatic, or mechanical force, the nail contacts its target test point without shorting to neighboring circuits. ICT electrically isolates each component from other components and then tests it as if it were the only component on the board.

Today's large boards can have more than 6000 electrical nodes. You can perform ICT on such boards in one of two ways:

• You can employ a two-fixture approach, in which you split a board into two parts logically, using a low-node-count tester and two test programs to sequentially test the two parts of the same board on each respective fixture.
• You can use one high-pin-count fixture and a single test program running on one high-node-count tester to test a board in one pass.

Your primary concern in choosing between these approaches is to provide adequate test while controlling costs in accordance with the National Electronics Manufacturing Initiative (NEMI) 2000 roadmap (Ref. 1).

The use of two low-pin-count fixtures can seem attractive because of the problems that can plague a single high-pin-count fixture:

• Its increased probe density and resulting probe forces can increase localized stresses, resulting in component damage and contact problems because of board flex.
• A fixture having 4000 to 5000 wires can easily weigh 200 to 250 pounds, and fixture height increases to accommodate the mass of wires.
• Long-wire technology in a high pin-count fixture reduces signal integrity despite the use of twisted pairs.

The two-fixture strategy is widely used when a tester can't handle a high-node-count board. In one study (Ref. 2), a 5700-node board was split between two fixtures having 3300 and 3100 nodes. Probe density in each fixture was less than in the case of one large fixture, and the two fixtures together required fewer delicate 50-mil probes.

Figure 1. In-circuit test aims to test components individually, but to do so, a tester must be able to control the components that surround the target device under test (DUT). If you place probes for U3 and U8 on different fixtures, you cannot reliably test U8. In this case, you could backdrive the output of U3, but you wold have no control of either N1 or N2, making it impossible for you to force U3 into an appropriate state in which it could tolerate backdriving.

To test U₈, the tester applies a series of digital test vectors to its input pins and measures the output pins of the component to make sure the appropriate response is obtained. To ensure that the tester can drive the input pins of U₈ to the required logic states, the output of U₃ (a 2-state-output component) must be inhibited in such a way that it can tolerate a backdrive signal from node N₈, and the output of U₁ (a 3-state component) must be disabled. In addition, notice that node N₁₁ is a bused node—it connects to two component output pins (U₈ and U₅). To measure the output of U₈, node N₁₁ needs to be electrically isolated from the effects of U₅ 's output. Hence, U₅ must be disabled.

To employ the two-fixture approach, you must successfully identify circuit clusters that function together before you specify your fixtures, and this process is entirely manual (Ref. 4). In high-node-count boards, it's difficult or impossible to isolate components correctly. Also, ICT cannot detect shorts between nets whose probes lie on different fixtures. You might require fixture rework to solve testability problems created by multiple fixture testing. Experience has shown that overall up-front time can more than double.

After you have solved the test-coverage problem on two-fixture test, you have to contend with throughput considerations. If you employ only one tester, the time required to test one board (inverse of beat rate) using this approach is calculated as follows:

T_BSF= {[2T_H + 2(T_T/2)(1 + R)(1 + O)]B + 2T_FC}/ B

where
T_H= handling time for one fixture,
T_T= test time per fixture,
R = fixture-induced retest rate (%),
O = test overlap rate involved in the multiple fixture approach (%),
B = batch size, and
T_FC= fixture change time.

To maintain the throughput rates needed in today's mass production environment, you can team two low-node-count testers with your two low-node-count fixtures, but that entails additional capital investment and continuing expenses for labor.

Maintenance of multiple fixtures is expensive. For example, for a 6500-node board, you can expect to deploy 7500 to 8000 nails due to overlap (that is, nodes requiring nails on both fixtures). Because probes need to be replaced every several thousand cycles, 1000 to 1500 additional probes would have to be replaced. The total cost of a multiple fixture can be calculated as follows:

C_F= 2C_FK + 2C_A + C_W

where
C_F= total cost of fixture,
C_FK= cost of fixture kit,
C_A= cost of actuation mechanism, and
C_W= cost of per-point drilling and wiring.

Figure 2. Cost of a fixture kit and actuator are mostly independent of node count, while wiring costs per node tend to decrease with node count.

The one-fixture approach employs a single tester capable of testing more than 6000 nodes in conjunction with one high-pin-count fixture. This approach offers several advantages:

• It improves PCB production beat rate. The test time obtained using this approach is

T_BHPC= [(T_H + T_T + RT_T)B + T_FC]/B

Here, handling time and fixture change time is cut in half, test overlap time is eliminated, and larger batch (lot) sizes result in better beat rates.

• It requires only one tester.
• It cuts test development time in half, because only one program needs to be developed for one board.
• It makes overlapping probes unnecessary.
• It lowers fixture costs—you need only one fixture kit and one actuation mechanism.
• It facilitates automatic test generation and automatic probe placement.
• It can find all possible shorts, including non-adjacent shorts.
• It simplifies debug. For example, in the case above, if the relationship between U₂ and U₃ is realized after a fixture has been built, it is not difficult to control the output of U₂ or U₃ because they co-exist on the same fixture.
• Cost = C_FK + C_A + C_PCB where C_PCB= cost of the PCB that replaces wires.

You must address are several issues when using this approach. A high pin-count fixture cannot render itself useful and reliable independent of several factors.

First is wiring method. Table 1 relates the type of wiring to different performance parameters in light of the high-pin-count scenario. Traditional long-wire technology is not well suited to handle the rigors of high-pin-count scenarios for several reasons:

• You can't ensure that duplicate fixtures will always perform identically.
• Moving wires inside a fixture can cause crosstalk and other problems.
• It's not always possible to have probes in a fixture in close proximity to their corresponding resources on the tester interface. This results in wires pulling or pushing the sockets, increasing lateral loads on the sockets and reducing contact reliability.
• The wire-wrapped endings of probe receptacles may cause wires to get nicked and may result in shorts.

In addition, if wiring changes are made and not documented, it becomes difficult to trace such modifications.

In light of these difficulties, wireless fixtures become the logical solution. Following are some compelling reasons to consider wireless fixtures:

• Wireless fixtures eliminate manual wiring errors.
• Electrical parameters are constant because the fixture has no moving wires, so duplicate fixtures perform identically.
• Probing accuracy is enhanced because there are no lateral loads on sockets.
• You can easily implement complicated fixture electronics and accommodate small-scale ECOs.
• Most important, the fixture size and weight are independent of probe count.

Table 2 relates different actuation methods to several performance criteria. In theory, vacuum can be successfully used to establish contact between test nodes on a board and test probes on a fixture if the downward force caused by the difference in the atmospheric pressure (on the topside) and the vacuum on the bottom side is greater than the upward force exerted by the probes. Mathematically, net force acting downward must be greater than the upward force due to probes:

(P_A– P_V)L_B W_B >NF_P

where

P_A= atmospheric pressure,
P_V= vacuum applied to a board,
L_B= length of board,
W_B= width of board,
N = number of probes, and
F_P= upward force exerted by each probe.

In practice, leakage from gasket seals or from holes on the board needs to be considered and a factor of safety needs to be applied. As the number of probes increases, increased vacuum needs to be applied. Also, as the area of a board increases, additional vacuum must be generated. Care should be taken to maintain clean vacuum lines because contaminants could be drawn in, resulting in contact issues. The same principle (net downward force being greater than the upward force due to probes) holds even when pneumatic cylinders or mechanical presses are used as methods of actuation.

Board-flex due to uneven stresses can be an issue when using vacuum or pneumatic cylinders for actuation. The solution to this problem lies in analyzing stresses on a board (Ref. 5), and in some cases, using software that uses efficient and advanced probe-placement algorithms.

Modern probe-placement software makes it possible to run algorithms that can space probes apart evenly instead of clustering them together. This reduces the possibility of localized stresses. It is also accepted practice to counterbalance upward-acting probe forces by using a suitable number of topside probes in a clamshell fixture. Clamshell fixtures are also advantageous in that they help reduce the probe density. Topside probes in a vacuum actuated clamshell fixture reduce the level of vacuum needed to establish reliable contact between the PCB under test and the probes. If there are N₁ topside probes, then

(P_A– P_V)L_B W_B > (N – N₁)F_P

This means that the vacuum needed would decrease by N₁ F_P, the total downward acting force exerted by topside probes.

Boards as well as the components on them are subject to stresses. Leaded devices have significant compliance through their leads and are typically unaffected by board flexing. But surface-mounted components are subjected to higher, disproportionate stresses since they adhere to boards. Some available tools help you visualize and analyze potentially damaging forces that could affect a board. It is possible to determine exactly what areas on a given board and what components are subjected to disproportionate stresses and therefore to determine where exactly push fingers or pneumatic cylinders need to be placed to minimize these stresses.

Mechanical actuation is also beginning to find use, and they seem to combine the advantages of vacuum and pneumatic fixtures. Usually, a reusable actuation mechanism (a screw press) can be repeatedly used for different fixtures (and different boards). This eliminates the complexities involved with vacuum actuated fixturing, such as board and component damage due to board flexing, contact problems due to flexing, ESD issues, and the need to maintain enough suction in the line. Mechanical fixtures are lighter because individual fixtures do not contain actuation mechanisms (as in the case of pneumatic fixtures). There is, however, a substantial one-time investment in a mechanical press.

Probe reliability becomes an important factor contributing to the overall reliability of the fixture. The logic is simple. A high-pin-count fixture will naturally have many more probes than a low-pin-count fixture. Hence, the probability of having defective probes is higher.

If traditional three-sigma (3ó) quality-control (QC) standards are followed for a tester capable of supporting 7680 test points (and hence 7680 probes), then on an average you would find 21 defective probes (0.0027 x 7680). Test-fixture manufacturers are looking at new and improved methods to test probe quality (plating, impurities, contact surfaces, etc.). Statistically, if 6ó standards were implemented, fewer than 1 probe in 7680 would be defective (0.0000034 x 7680).

Though probe-placement software makes it possible to spread probes to reduce localized stresses, sometimes board design makes it infeasible to do so. In cases where sections of a board are highly dense and alternate access points are not available, manufacturers have to use 40-, 30-, 25-, or even 20-mil probes.

Certain manufacturers have successfully developed innovative techniques to improve the mechanical stability of these delicate probes by providing recesses in the probe plate. Because probes are delicate, traditional drilling techniques cannot be employed. High accuracy is needed. Therefore, several 60-mil thick plates are drilled individually and sandwiched together to guide these probes in areas of high density. Doing so solves the board flex problem and prevents probe damage. Wireless or short-wire technology is used for such high-density situations, preempting the problems caused by traditional long-wire technology. In many instances, this has been found to be the best approach.

Some ICT software has the capability to test and warn about the possibility of faulty fixture probes. This prevents false failures from being reported and helps quickly identify faulty probes.