Today you can’t buy an integrated computer/phone/fax/copier with internet and video conferencing capabilities that fits in a shirt pocket and costs less than $500. Right now, this device is only a virtual product. But consumer electronics companies are leveraging technology aggressively toward the day when gadgets like this become realities.
The most obvious test problem associated with the very compact porta-office is the lack of test access. As physical access is reduced, it’s no longer possible to employ traditional bed-of-nails test fixtures in manufacturing.
“The use of fixtured ATE is limited by the lack of spacing between probes due to the diameter of the probe shafts and the barrel that houses the shaft,” commented Jack Ferguson, general manager at ITA. “Techniques are available to use smaller-diameter probes for hitting targets as small as 10 mils, but the implementation is rather expensive. It can only be justified on very high-volume production runs.”
And the problem of limited physical access is not likely to go away. “Physical access using traditional bed-of-nails fixtures will continue to decrease over time. As a first consequence, we are observing an increasing use of automatic optical inspection (AOI),” added Elie Bouskela, product manager at Teradyne. “However, when traditional in-circuit test (ICT) is removed from the test process, AOI fault coverage usually is insufficient to assure a high yield before functional test. Wrong value components are not detected by AOI and can rapidly overload the functional test repair loop.”
Limited test access is a big factor in the diverse life of a project manager, but not the only one. Assuming his project is on schedule and the window of opportunity has remained open, it’s critical that the product launches on time.
Studies have shown that a six-month slippage in time can result in as much as a 33% lifetime profit reduction.1 Slippage at the beginning of a product’s life-cycle curve subtracts dollars from the peak sales period. This reasoning assumes that the forces acting to end a product’s life occur at the same point in time whether the product is late or not.
So, getting through the prototype stage smoothly is important. Extraneous work that can be minimized includes design and build of pre-production test fixturing.
Flying Probers Provide a Solution
A few companies are producing so-called flying probers that avoid the cost, build time, inflexibility, maintenance, and storage of a bed-of-nails fixture. Jimmy Green, a product manager at Texmac, noted, “When test access declines, the benefits of the flying-probe tester increase. They do not have to use vias or test pads, but can go right to the termination or solder joint of a component. The most efficient use of flying-probe testers would be with PCBs with high-density surface-mount devices and fine-pitch components where manufacturing process faults tend to be the highest.
“A company that produces many different, complex boards with potentially high levels of manufacturing faults would see a fixtureless tester as a lifesaver,” Mr. Green said. “Such a company knows that before the product can go to high-volume production the R&D group must deliver a defect-free prototype board. A flying-probe tester would allow R&D to quickly isolate the board faults. It could shorten the time from prototype to production, which gets the product to market quicker.”
As with conventional ICTs, the test program is prepared from board layout coordinates and the corresponding bill of materials. The time to program a flying prober often is a day or less, compared to a week or more for design, build, and test of a conventional bed-of-nails test fixture. As a result, flying probers are ideal for prototype, short production run, and field return testing.
Mr. Ferguson of ITA takes a very wide view of the functions that a flying prober provides. “The ultimate goal of a prober should be to replace a fixture-based machine. To do this, the prober must have the same electrical test capabilities of the fixture-based machine and sufficient numbers of probes to accomplish the electrical test requirements. Including the motion time for placement of the probes, the tests must be performed in nearly the same time as the fixture-based machine.
“When considering the properties of flying probers,” he added, “capabilities should be separated into two categories: electrical test, and mechanical interface.” See Table 1.
When using a prober made by a company that also offers conventional ICT, an important consideration is compatibility between programs. Of course, if the prober is being used to gain access, it also may be used in production. But for those cases where the main advantage of the prober is quick, fixtureless prototype testing, the transition to production will be faster if the same test program can be used.
Count the Legs
Unlike insects that, by definition, always have six legs, most flying probers have four to eight probes on the component side of the PCB or as many as 20 in the case of ITA’s FlyingScorpion. Some models also provide up to 16 stationary probes on the opposite side of the board.
Positional accuracy is a more complicated parameter than it might first appear. The motions of probes are controlled by separate mechanisms or shuttles. More than one probe may be mounted on a shuttle, but a shuttle will position at least one probe.
In some machines, a camera is provided for each shuttle, which means that calibration can directly remove positional offsets for all probes. In other machines, there is only one camera, and it is associated with a particular shuttle and its probes.
These probes can have very high resolution and repeatability, but probes attached to other shuttles can only be indirectly calibrated. As a result, a relatively large minimum contact dimension is required to ensure that any probe, not just the directly calibrated ones, will hit the target.
Positional accuracy is strictly the error between the absolute position of where the probe should be and where it actually is. Cameras improve accuracy because they can recalibrate the prober to local fiducials, eliminating tolerance buildup. Having positioned a probe to a particular point, repeatability is a measure of how closely that point can be hit again. Resolution is a fundamental feature of the encoders or stepping motors used in the mechanism.
Generally, repeatability is less than 0.002″, and two probes can be positioned as close to each other as 0.004″ to 0.008″, depending upon the model and manufacturer. The probing speed can be >100 tests/s, but typically is 10 to 60. Speed depends upon the distance moved, so part of the programming software job is to order the probe movements to minimize travel time.
Control software ensures that the multiple probes avoid running into each other, but each probe also must avoid topside obstacles. Program optimization becomes a 3-D problem when the height of an object may take longer to get over than to get around.
The ITA FlyingScorpion Prober can reposition its probes within a 2″ × 2″ target area because of its joystick-style probe modules. Individual probes can change their inclination from the vertical and their angular bearing with respect to the X-Y axes of the board being tested. By combining this degree of flexibility with almost 2″ of vertical motion, probes can be programmed to work around or over large component obstructions.
Probes from other manufacturers have fixed angular positions and rely on X and Y repositioning of the probe shuttles for accurate placement. Teradyne uses the 6° inclination of its probes to prevent slippage. An interesting variation is the mixed angles used in the Texmac Models APT 83/8400: -16° X and -10° Y for probe 1, -5° X and +5° Y for probe 2, +5° X and -5° Y for probe 3, and +16° X and +10° Y for probe 4.
Complementary angles provide additional clearance so all four probes can be brought close together. Table 2 compares several major specifications of six flying probers.
Flying Probers Have Other Benefits
In his introduction to the panel discussion on flying probers at ITC ’98, Bob Russell of Bull Electronics, the panel chairman, said: “In at least two areas, flying probers hold more promise for future expansion than bed-of-nails units. The first is boundary scan, where a shrinking geometry may necessitate probe repositioning to achieve contact. Another area is optical inspection, a feature available on some models.”
Opinions differ regarding the value of providing boundary scan capabilities in a flying prober. Some models do provide it, allowing scan nets to be probed even if they don’t actually terminate in a conventional test contact or pad. Because many components are programmed via boundary scan, having this capability means that a board can be tested and programmed at the same time.
Fiducial cameras are standard on most flying probers because manufacturers claim it is the best way to ensure high local probe-placement accuracy. However, for small boards not subject to relatively large dimensional tolerance buildup, a camera may not be necessary. For example, the Texmac Model APT83/8400 Fixtureless Tester uses mechanical alignment as standard with an optional fiducial camera. No manufacturer supplies cameras for AOI as standard.
Proteus makes special-purpose flying probers for automated fault insertion and verification of signal integrity. These two activities are very time-consuming if done manually, but they also are key to design validation and verification.
The software self-test routines in a product are developed to detect faults and respond appropriately. An example is the fault protection required in high-reliability computer and disk storage systems. By physically inserting temporary faults at pre-determined circuit nodes, the response of the self-test routine can be noted and the routine’s effectiveness evaluated. If untested faults are discovered, tests can be added accordingly.
Similarly, in boundary scan testing, vectors can be revised to increase coverage, if required. Through this method of fault injection, a minimum set of test vectors can be developed.
Designs that are operating at high clock rates—above 50 MHz—must have the integrity of each signal verified. The twin probe head developed by Proteus offers 1.5-GHz bandwidth and can automatically measure dynamic signal characteristics for ASIC acceptance testing, for example. In addition, a comparison of actual and simulated performance can be made. Design simulation models can be altered to agree more closely with real, measured values.
Applications Are Expanding
All these machines have a motion system capable of positioning one or more probes to within about 0.002″ over a large area. None of them require conventional test pads, instead contacting component leads or vias directly.
Flying probers can be categorized according to their intended use. GenRad, ITA, Texmac, and Teradyne have built ICT equipment that provide guarded measurements on analog components, vectorless IC open lead test, and automatic test program generation. The Polar Model FT100 Prober uses a single moving probe to characterize board nodes by their analog signatures. It does not directly measure the value of a 100-nF capacitor, for example.
The Proteus DVT-100 is intended for fault injection and design verification work, not for complete prototype testing. These two new capabilities are examples that support Bob Russell’s comment, “The range of usefulness of flying probers has yet to be fully explored.”
Reference
1. Bralla, J.G., Design For Excellence, McGraw-Hill, 1996, p. 255.
Desired Electrical Test Capabilities
Desired Mechanical Interface Capabilities
Wide-range current/voltage sources for analog component testing
Highest possible number of probes (for higher test speed and test coverage)
Active and passive analog tests
Top and bottom side simultaneous probing
Network analyzer (for high-frequency L, R, C networks)
360° probe angle placement (for placing probes around objects)
Reverse polarized capacitor tests
Highest possible probe movement speed
Connector tests
Fully automatic probe placement for top and bottom probes
Vectorless IC tests, passive and active
Full 3-D X, Y, Z positioning (independent between probes)
Crystal and oscillator tests
Conveyor system for in-line or manual use
Function test capability (VXI, IEEE, or any bus instruments, power supplies)
Real-time program and test-point editing capability (automatic probe placement)
Boundary scan capability
Smallest possible test-point target size
Shorts test
Highest possible probe retraction height (to clear objects on the board)
Automatic discharge before/after test (to ensure high test stability)
Automatic object position recognition (for fast programming and unrestricted testing on the component side of a board)
Unrestricted allocation of electrical resources to any probe
Highest possible number of cameras (for fiducial find and optical inspection)
Automatic program generation from CAD (test-program and test-point data)
Automated program optimizing tools (such as guard generation, merging parallel device tests, best test point selection for multipoint nets, program resequencing for highest execution speed)
Table 1.
Texmac
APT-83/8400
Teradyne
Flying Prober
ITA
FlyingScorpion
GenRad
GR Pilot
Polar
FT100
Proteus
DVT-100
Maximum Test Area *
19.7″ × 15.7″
23.6″ × 19.7″
25.6″ × 25.6″
15.75″ × 23.6″
20.9″ × 11.8″
18″ × 18″
Number of Probes on Topside
3 or 4
4
4
16 optional
4
1
1 or a signal/ground pair
Number of Probes on Bottom Side
4 optional TEM-10
0
2 standard
2 optional
16 optional
stationary
0
1 or 2 optional
Maximum Height of Topside Components
0.79″
1.18″
1.97″
0.94″
3.94″
3.0″ or **
5.0″ max
AOI Cameras
optional
optional
future option
no
no
no
Fiducial Cameras
optional
1
3 (4 optional)
1
1
1 (1 optional)
Tests per Second
25 max
20 max
100 max
10
5
n/a
Scan Option
yes
no
yes
yes
no
no
Minimum Contact Dimension
0.010″
0.004″
0.004″
0.007″
0.003″
0.002″
Placement Repeatability
±0.002″
±0.002″
±0.0002″
±0.0002″
±0.0003″ typical
±0.001″
Placement Resolution
0.0008″
0.0004″
0.0002″
0.0002″
0.0006″
0.0002″
Self-Learning
yes
option
no
yes
yes
yes
Automatic Program Generation
yes
yes
yes
yes
yes
X-Y only
semiautomatic
ICT Component Testing
yes
yes
yes
yes
signature
analysis
no
Vectorless Opens test
optional
optional
yes
optional
no
no
Price
$180k to $300k
from $185k
from $340k
call
company
from $52,195 + computer
from $150k
Circle
209
210
211
212
213
214
* Maximum area that can be tested, although actual board can be larger for some models.
** Two different Z-axis assemblies are available.
Table 2.
Copyright 1999 Nelson Publishing Inc.
February 1999