Home » KnowledgeCenter » How to: Wire Pull » How to: Wire Pull (WP)
This extensive how-to consists of 13 paragraphs and 4 appendices. Please use the navigation on the right.
This manual is a guide advising what to consider and how to best perform a Wire Pull test. The main focus of this manual is on wire- and ribbon pull using wire hooks. Loop height measurement is also covered. The manual briefly touches on surrounding topics like tweezer pull and peel testing of wires, which is the subject of a separate how-to by xyztec.
Wire Pull Testing applies an upward force under the wire, effectively pulling it away from the substrate or die. The purpose of the test is as MIL-STD-883 2011.9 describes it: “To measure bond strengths, evaluate bond strength distributions, or determine compliance with specified bond strength requirements”. A wire can be pulled to destruction, but there are also non-destructive variants whereby one tests whether the wire can withstand a certain force. Non-destructive test methods are typically used for 100% testing of safety critical, high quality and high cost products, avoiding damage to the acceptable wired bonds tested.
The term Wire Pull usually refers to the act of pulling a wire with a hook mounted on a pull sensor on a bond tester. However, to promote certain failure modes, wires can be cut and then pulled by tweezers, also mounted on a pull sensor on a bond tester. Wire Pull is sometimes abbreviated to WP. Our pull test type page contains videos of both thin wire pull testing and thick wire pull testing.
Usually wires up to 75µm diameter (3 mil) are classified as thin wire. Beyond that size, we speak about thick wire testing.
The cross sectional shape of wires are either round or rectangular then being referred to as ribbons. Slightly more exotic than pulling regular wires and ribbons is pull testing on SMT gull wing leads. In this how-to we also share our thoughts on testing those.
Wires can be made of various materials, most common in semiconductor are gold, aluminium, copper and silver. In principle every wire can be tested, although some wires may be difficult to reach because of surrounding wires and/or components. In some cases, sample preparation is required.
Also, especially ribbons sometimes have very low loop heights and require customized tooling to effectively get underneath and pull successfully.
As is always the case xyztec’s Golden Rules of Bond Testing apply;
For more information on the Golden Rules refer to Science of bond testing module BTM1.1.
Wire pull is done using a hook. Hooks come in many shapes and sizes. Standard hooks are made from bent round wire and defined by their foot diameter, foot length and shaft diameter. A nice long hook length like 30mm is useful especially when deep access is a requirement, but also for vision lines and because length has very few downsides.
Hook diameter is an important consideration. As a rule of thumb, one should choose a hook with a diameter at least 3 times the diameter of the wire to be tested. This reduces the stress in the wire bent over the hook, so that its strength is similar to the maximum tensile load of the wire ensuring the highest load on the bond possible. The hook should not cut into the wire.
The foot length is normally about 3-3.5 times the hook diameter. However, for fine pitch applications, a shorter length is required to avoid pulling multiple wires at once. The angle is typically 85% to pull the wire into the hook so it does not slip off.
A suitable hook material must be:
Tungsten often meets these requirements and is widely used. The material pulled by a hook is typically softer than the hook and so hook wear is rarely a problem. Poor hook quality is normally due to damage.
Most hooks are delicate and easily damaged if you touch them. The most common causes of damage are being hit by the sample or work holder when changing the sample to be tested.
A separate hook from the hook body is preferable because there is more chance of successful repair after a collision. Also, the concentricity of the hook can be readjusted (see below). A damaged hook should not be used, as it will affect your measurement results.
Examples of damaged wire hooks that are not suitable for use. The one on the left is bent, probably because of a heavy landing or by a collision with the sample. The one on the right was probably too small for the test force.
A concentric hook is one where the main shaft of the hook, in the vicinity of the foot, does not move when the hook is rotated. It is important to have a concentric hook, especially when using some form of automation, like auto hook or full automation. If the hook is not concentric and you rotate it when you are below the wire, you may bend the wire and test two wires at once, or miss the wire altogether. Good hook concentricity also makes manual alignment much easier.
Given the very small gap between fine pitch wire bonds hook eccentricity may need to be no greater than 5µm. Because it is not possible to manufacture a sensor spindle and hook to this accuracy, xyztec developed the Concentric Tool Holder. Using this tool holder, an experienced user can obtain concentricity easily. Of course, all hooks are shipped concentric.
To test a wire the hook is ideally positioned underneath it without touching any other wire or part of the sample. This enables an accurate sensor null (force zeroing) before the measurement is taken. Manually aligning the hook into this position can be difficult and time-consuming. If you want to test all the wires in a fan-out it is simple to start at one end and let the tested wires build up on the hook as you proceed. For fine wires, their build-up on the hook has a negligible effect on the test result. This is not true for heavy wires. You can start anywhere and test in any direction. It is a fast way to test both manually and automatically.
The next step up in terms of automation is using a functionality called auto hook. This makes it possible to rapidly test each wire without disturbing the wires next to it. It is very useful, one could say essential, for non-destruct testing. Auto-hook is one test done automatically. A tester with auto hook enabled does the following when the operator presses the test button:
Auto hook can be started from a position where the hook is oriented along the wire, or across the wire. In the case of the latter, extra rotations are performed before and after the test.
Auto hook along the wire is faster than across the wire because the hook rotates 2 times rather than 4 times. The advantage of starting across the wire is that you can see how the hook will engage with the wire when it will be below it. For both it is important to test away from the previous test so the hook can be aligned low and close to the wire to be tested for easy alignment and to minimize Z movements.
Especially when doing non-destructive wire pull, it is essential to avoid hitting wires other than the one being tested. Hook concentricity is a requirement for auto hook (see chapter 6).
We also have to align the hook along the length of the wire; where do we pull? Common practice is to pull at the center of the wire. The MIL standard 2011 specifies pulling between midspan and loop apex in order to avoid adverse wire deformation (previous versions of the same standard specified pulling mid span).
Alternatively if your interest is only in a particular one of the two bonds you can increase the load on it and increase its chance of a bond failure by testing close to it. If your failure mode of interest is the first bond, try aligning your wire hook close to the first bond.
It is important to pull in the same place. Not pulling in the same place will:
Automation ensures tool alignment is always the same. Only highly advanced bond testers with sophisticated automation functionality can achieve the most reliable results to optimize your process.
The alignment along the wire and the wire length affect the angle of the wire pull. The angle in turn changes the load on the bonds. Standard DVS 2811 advises that the pull position should be such that the angle of the wire at the first and second bonds are equal. It then normalizes the results around a “standard” angle of 30° using a correction factor K. The load on the bonds is calculated as the measured pull force multiplied by K. Find K in the table below.
DVS 2811 enables the strengths of different wire bonds with a variety of pull angles to be compared more meaningfully.
While DVS 2811 sets the angles equal and therefore the forces too, it is possible to pull in many positions and creating different forces on the first and second bonds. This requires some fairly straightforwards mathematics.
In Appendix 1 of this how-to, the various formulas required to calculate the forces by triangulation are given. It is possible for an advanced bond tester to calculate all of these values for you.
So far this how-to has focused around manual wire pull testing, whether or not assisted by the auto hook functionality. Experience tells us that the influence of the operator on the measurement results is relatively high. This can be eliminated by automation.
A sufficiently advanced bond tester, like the Sigma, can perform automatic tests by running automation programs that may include fiducial mark recognition. Using automation, all wires are pulled in a very consistent position, so the distribution of results goes down and the quality of measurement data goes up.
The next step up, and one in which the Sigma is truly unique is when the bond tester optically detects the wire and corrects for “wire sweep” before going on to pull the wire automatically. This ensures more repeatable positioning even when the placement of wires is not consistent.
The illustrations show how a traditional bond tester without wire recognition may miss the wire when executing a simple programming macro. Only by measuring changes in the wire position, the tester can correct the test position and ensure consistent results. The third illustration shows how the hook is driven to a corrected position after the wire was detected at the position of the green crosshairs.
Now, when the bond tester performs the pull test, it does not miss the wire but pull it in exactly the same position as every other wire. This enables a higher degree of consistency than possible when doing manual tests.
On the previous capture, triangulation was discussed. In order to make the desired force calculations, a loop height measurement is required. Bond testers are traditionally used for force measurement only, but one can also use the positional accuracy to measure distances with the same wire pull sensors. Advanced bond testers can even combine the measurement of loop heights with wire pull tests. Click here for more information on loop height measurements.
In the illustration, two tests are simulated. The one on top shows the best measurement, when the sample is clamped to the base. In the one on the bottom, the sample is not clamped. Hence,
If the sample is not firmly and rigidly clamped, test time increases, displacement data is wrong and test data becomes less tight and reliable.
Vector pull is a variety on wire pull, where the hook does not move up at a 90° angle relative to the sample, but at a programmable angle. On a bond tester, this is achieved by simultaneously moving the X- and/or Y-axes with the Z-axis. This enables different loading conditions between the first and second bonds that can help to promote the failure mode of interest.
Alternatively, one can clamp the sample in a work holder which is inclined. See an example of this under SMD gull wing leads testing.
The best way to test the bond of an SMD gull wing lead is to cut the body from the leads and then test each lead with a tweezer pull sensor. This way, the bond is tested with no effects from the body and the other leads.
The downside of this approach is the preparation time involved. If you pull test without cutting the body off and use a hook there are 3 problems, namely:
Because the hook is closer to the package than the pad, a lot of the test load goes into the package rather than the bond of interest. However, a weak lead to pad bond will still give a lower result and failure mode analysis of the tested bond will show poor soldering.
To minimize the effect the body has supporting the bond it may be better to use a vector pull, using a special hook and a shear sensor. This also allows the hook to be made stronger but it might still be weaker than the lead, depending on the clearance under the lead for the hook. With a vector pull, a higher test load is applied to the bond pad than with the regular wire pull.
A strong wire hook is required. Xyztec can support you with extra strong hooks that will withstand relatively high forces.
Another method is to incline the sample. Using this approach, one can still use a regular pull sensor in combination with a strong wire hook. This has the following limitations, though:
The test method allows you to program the test variables. The basic settings contain the test distance, the test speed and whether the test is destructive or non-destructive. There are many more settings though and advanced engineers can sometimes combine them in unexpected ways to solve their challenges in a smart way.
The hold time is most relevant for non-destruct tests. Before almost every test type, you may want to null the sensor. The next settings all expand with more options when enabled.
A touchdown before the test, for example, which is mostly used for tweezer tests in case of pull, is defined by a landing force, maximum landing distance and a landing speed.
If you define a fallback to determine the end of test, the force difference is the most relevant value to set. The bond tester will also give you the option of “overtravel”, which is the distance moved after the fall back has been achieved. This is a useful way to clear the pulled wire away from the test area for failure mode analysis.
Pull tests can be performed with or without auto hook and be combined with loop height measurements and configured to output bond stress (pressure) data instead of force data. For all these options, additional settings are displayed only when in use.
The Sigma uniquely has all the test settings in the same and right place and makes it easy to track what the user has been doing. Previous test method settings are remembered behind a ‘show history’ checkbox.
Like all other test methods such as push, peel, shear and probe, the engineer can set SPC warnings for the pull method. Depending on the measurement result the measurement can include a warning or a fail and the operator can be forced to enter a remark. The software can even block the tester until an engineer checks the results and releases the tester to continue. This advanced functionality is only available on the Sigma bond tester.
The test method settings also include a spectrum of statistical values. Of course everything from simple mean or median calculation to Cpk values, lower and upper spec limits are crucial for most quality assurance processes.
In order to make the right decisions based on the test results, sometimes more analysis is required. It is a great help when the bond tester assists in this process by, for example, storing pictures and/or videos of before, during and after the test with the measurement data and grading info.
The Sigma software comes with a sophisticated query screen, which allows the engineer to sift through massive quantities of data. Also very helpful is the possibility to overlay the force or hysteresis graphs for multiple measurements. This makes comparing the graphs from multiple measurements possible. Forms also offer an auto filter feature that will only show the measurements and samples for a specific field value or set of field values.
Some company procedures require strict data collection. An auto print feature automatically prints the selected report for this dataset as a pdf file when a certain number of samples has been reached.
It may be important to export your data to external systems. Make sure to take your measurements with a bond tester which can export standard reports in the file format (like XLS, DOC, PPT, PDF, XPS, CSV, XML, DBF, etc.) that you need. Also, it is very useful to be able to create your own export format, without having to rely on third parties.
MIL-STD-883 2011.9 lists 26 failure modes (which they call failure categories), depending on the type of sample. We list the most common failure modes for gold wire and aluminum wire below. We also refer to our grading library, where you can download result codes.
This how-to has four appendices, namely:
Tom Haley
