Drop Tests vs. Shock Table Transportation Tests
Should relatively inexpensive drop tests replace shock table transportation tests?

by Matt Daum and Wayne Tustin

Can all transportation shock tests (on "packages" handled by common carriers such as Airborne, DHL, FedEx, Greyhound, the Postal Service, UPS, etc.) be performed with free-fall package drop testers exemplified by the unit of Figure 1?

Figure 1 - Freefall Drop Tester from Lansmont.

A. Shock Test Standards
Some shock tests are very simple, informal. Some computer manufacturers, for example, evaluate hard drives by resting one side on a wood block. Pulling out the wood block allows one edge of the hard drive to drop onto the work bench. Normally, nothing except drop height is measured. Nothing is known about the impact energy profile.

Such procedures are often shared at technical meetings. Other companies in that business adopt the procedures. Before long it becomes an "industry standard". It is documented and henceforth extremely difficult to change. It might even be adopted by and published as a national and possibly even as an international standard. Here are a few organizations that offer shock test standards:

• American National Standards Institute (ANSI)
• American Society for Testing and Materials (ASTM)
• Deutsche Institute fur Normung (DIN)
• Institute of Environmental Sciences and Technology, which now offers several Recommended Practices, including RP-030: Classical Shock and RP-031: Non-Classical Shock.

Besides the preceding organizations, we have the USA Department of Defense, source for the current F revision of Military Standard MIL-STD-810 (now available on CD-ROM). Method 516.5 deals with shock testing, including such test topics as Functional shock, Material to be packaged, Fragility and Transit drop.

Section 2.2.2 of 810F identifies the purpose(s) of each procedure, so that military and naval personnel can choose which procedures should be applied to the particular hardware being considered. That permits Section 2.3 "Determine test levels and conditions" to be quite general. The goal here is to provide test personnel with information so they can "tailor" tests to the "real world" to which hardware will be subjected.

Early versions of MIL-STD-810, by contrast, were more like a "cookbook". They were much shorter and simpler because they offered far fewer test procedures. Many called for the same tests no matter how the affected hardware would be shipped nor where the hardware would be used.

Unfortunately, most of the world's test standards today rather closely resemble the early 810 "cookbooks". These test standards, like the early 810 versions, focus on classical pulse shapes.

Figure 2 - Classical Uniaxial Shock Pulses

Figure 2 shows three classical shock pulses, with the first, the half sine (usually the beginning and end are rounded into the haversine pulse) required most often. These pulses share several features:

• none of them has ever been found in actual shocks in the "real world"
• they lack the "hashy" oscillatory behavior of "real world" shocks
• they can be accomplished on relatively simple mechanical devices
• they contain excessive (compared to "real world" shocks) energy at low frequencies, overexciting the lower natural frequencies within the DUT (device under test).

What is the meaning of that last point? Let the simple resonators of Figure 3 represent how your automobile responds to vibratory forces. The lower frequency resonators at the left represent body bending and body torsion. The higher frequency resonators at the right represent localized deflections such as doors and instrument panel. All real structures act like a variety of resonators. We want our shock test to appropriately (similar to real-world shocks) excite the various resonances in our test article.

Even with all the "negatives" listed above, engineers have adopted these highly arbitrary, nonrealistic classical pulses, and have even applied tolerances such as ±3 dB.

Figure 3 - Collection of simple resonators representing the various natural frequencies of a complex structure.

How did our shock test standard pulses get so far from the "real world"? Mainly because of shortcomings in early ('forties) instrumentation and data processing. Instrumentation specialists of that era lacked today's small, light weight, broad-frequency-capable accelerometers. They were forced to use heavy, limited-frequency velocity sensors, which could only be mounted on ship, aircraft and land vehicle structures, rarely on equipment items.

For example, recording pen-on-paper oscillographs were only useful 0-50 Hz. Higher frequencies, although present, were not recorded. The advent of magnetic tape recording and the use of galvanometers writing onto photographic film extended the frequency range to perhaps 2,000 Hz. The higher-frequency components of "real world" shock pulses (such as Figure 4) were not recorded.

Figure 4 - "Real world" shock event; note the "hash" content.

Shock Tables
How, in early test labs, were mechanical shocks generated to meet the classical shock pulse shapes in Figure 2? We will examine a few "moving carriage" shock test machines. The DUT was attached to the carriage, the carriage was hoisted, then released. Gravity increased DUT + carriage velocity. An arresting mechanism provided the shock pulse. Many test procedures (such as ASTM D3332) have been written around "falling carriage" shock test machines.

The carriage of one pioneer machine fell into a bed of sand. With fewer wooden blocks on the carriage bottom, greater penetration into the sand gave lesser acceleration and longer shock pulse duration.

Figure 5 suggests a less-ancient "drop carriage" type of shock tester. The DUT is attached to the carriage, which is pneumatically elevated, then pneumatically accelerated (increases impact velocity) downward. Again, stopping the carriage produces the test shock pulse. The time history (waveform on an oscilloscope) of that shock pulse depends upon the material placed upon the target surface, e.g. rubber for half-sine pulses, pointed lead cylinders for terminal-peak sawtooth pulses, etc.

Figure 5 - Stopping the Carriage.

Alternately, machines may be equipped with "programmers" as in Figure 6. These allow more adjustments to the severity and the duration of the arrest. These programmers provide two different arresting surfaces for the falling carriage. Filling the cylinders with pressurized gas gives a trapezoidal shock pulse shape. These are known as "gas programmers." When the gas in the cylinders is bled out, the carriage will contact the stiff plastic surfaces above the cylinders, giving a half-sine shaped shock pulse. These are called the "plastic programmers".

Figure 6 - Drop-Carriage Shock Test Machine. Note "programmers" for arresting carriage motion.

Where we are today: shock tables widely used for package tests
The field of packaging has borrowed from the work of early pioneers and their reliance on the shock table for creating shock events. The shock table is now commonly used for several applications, including

In all three cases listed above, fundamental limitations interfere with idealized use of the shock table. Instead, a much simpler and less expensive way to create shock events is to use a free-fall package drop tester similar to Figure 1. The free-fall drop tester works simply. It pulls support away from the package, allowing the package to fall freely. For the three uses mentioned above, the free-fall drop tester (and data acquisition software) gives us the same, and sometimes better, test results at a fraction of the cost, since shock tables are expensive and occupy large amounts of precious lab space compared to a simpler free-fall drop test machine.

Case 1: Eliminate Shock Tables for Generating Damage Boundary Curves (DBCs)
Fragility assessment for years has been based on modeling fragile components within a product as linear, undamped spring/mass systems inside a rigid frame. The component is said to have failed when

According to accepted theory [1], an "input" shock pulse to a product must have a critical velocity change DVcr and a critical deceleration Gcr in order for a component within that product to fail. Modeling components as simple, linear spring/mass systems led to the development of the Damage Boundary Curve (DBC) of Figure 7. The DBC shows pictorially the combination of velocity change and deceleration of an input shock pulse which is needed to damage the component. ASTM D3332 [2] describes the use of a shock table to obtain the DBC. The procedure works by placing the DUT onto the falling carriage, and dropping the carriage at increased heights onto the plastic programmers until damage occurs to the critical element. This is depicted in Figure 7 with the x's finally crossing the critical velocity line. A new product in the same orientation is then dropped at increased gas pressure on the gas programmers until damage occurs to the same critical element. This is depicted in Figure 7 with the x's finally crossing the critical acceleration line. Now the damage region bounds velocity change and G level for input shock pulses that will cause failure of the critical component.

Figure 7 - Traditional DBC.

Note that this traditional DBC is limited as to what pulses may be used to break the critical element:

Unrealistic inputs
We will agree that the trapezoidal and half sine pulse shapes generated by a shock table are fairly repeatable and "mathematically tractable." But we question the representative nature of these waveforms. These "pure" events are rare (if they exist at all in real world packaging events). This question is not new - even ASTM D3332 [2] points this out. The packaging engineer seeks to analyze shocks resulting from "real world" free-fall drops and the effects they have on his products. Furthermore, the trapezoidal wave is generally considered the most damaging of waveforms. To visually understand this, Figure 8 predicts the response of our range of spring mass systems (Figure 3) to several input shock pulse shapes, including a square (comparable to trapezoidal) wave input shock in the second example. Note the large frequency range over which the response is twice the input level. This demonstrates why the trapezoidal shock is considered so damaging; it conservatively estimates product fragility.

Figure 8 - Shock Response for Various Input Shocks.

Shortcomings of traditional DBC Procedure
Traditional DBC, to determine if cushioning will be required, utilizes an accelerometer mounted to the product, usually somewhere on its base structure [2]. The idea is to "capture the product's response to the shock input." We disagree with that tradition. Note that the component of a product that is the most fragile during shipping and handling is generally not what the accelerometer is mounted to. But it is this component that must be protected from input shocks in order to prevent damage. Traditional DBC does not predict or even monitor the actual response of the element, it only describes the velocity change and deceleration (G) level of the input shock that caused the critical element to fail. This causes problems when a real world shock is captured since now the questions of filtering and fairing of the shock pulse come into play. Furthermore the velocity change and deceleration levels are determined by somewhat unrealistic shock pulse shapes which lead to conservative descriptors of damage as explained above.

It would be better to generate damage boundary curves using shocks from naturally occurring events - free-fall drops, and to do this without worries about fairing, filtering or even what kind (shape) of input shock the product is given. Recent research [3] shows it is now possible to create DBCs using free-fall drop test machines, along with Shock Response Spectrum (SRS) software. SRS is a calculation of a component's response to any arbitrary input shock. A typical SRS plot is shown in Figure 9. The Y axis plots the acceleration responses of an imaginary series of single-degree-of-freedom spring-mass systems (such as those in Figure 3) verses a range of natural frequencies on the X axis.

Figure 9 - Typical SRS Output (Maxi-Max, Composite Output), From Lansmont's Test Partner.

Using free-fall drops to obtain SRS gives a much truer picture of critical element fragility, since it is not dependent on or derived from a half sine or trapezoid pulse. SRS allows any shape input shock pulse to be used so long as it damages the critical element - this is what we really want to accomplish. This would allow nondestructive testing (using a dummy product) since component behavior can be predicted if it's resonant frequency is known

Procedure for generating DBCs with free-fall and SRS
Drop the packaged product (as in Figure 1) from some height. After each drop, open the package. If the component is ok, close the package. Increase the height by some increment (say, four to six inches), repeatedly dropping and increasing the drop height of the package until damage occurs to the component. For each drop, record the input shock measured by the accelerometer you placed near the critical element. It is important not to place the accelerometer on the element as we are letting the SRS predict the response of the element. Plus, placing the accelerometer on the critical element may be impossible (element is small), or may change the element's resonant frequency. Obviously accelerometer placement is now critical for success since we are trying to measure the shock to the component. Now use commercially-available software to calculate your SRS from your drop that caused damage. Several companies offer this kind of software, such as Lansmont, GHI, IST and others. All the information needed to construct your DBC is now available. Do this by first noting the peak G on the vertical axis of your SRS plot (see Figure 9) at the component's first (lowest) natural frequency, at the proper point along the horizontal axis of your SRS plot. (You will previously have determined the natural frequencies of the critical element by a sine sweep test, strobe, etc.). The critical deceleration, Gcr, for your DBC will be one half of the peak G value from the SRS plot (Gcr = G/2) [4]. Calculate your critical velocity change, DVcr, by , [5] where G is the value you plotted a moment earlier, g is acceleration due to gravity (units are not important - just be consistent) , and is the critical component natural frequency in hertz. Details and theory of this method are fully described in Daum's work [3]. This method shares one limitation with traditional shock table DBCs: it gives no information about the effects of fatigue (multiple drops). Fatigue can and should be handled, and new efforts are focusing on this [6, 7]. This procedure gives you a less conservative DBC than would a trapezoid-shaped shock pulse. I have now used a real-world pulse for my input instead of an idealized shape. A benefit is these steps may require less packaging material, since the margin of fragility overestimation is reduced.

Case 2: Eliminating shock tables for simulating free fall drops
Shock tables have long been used to simulate common carrier transport free-fall drops. This application is based on an assumption of how velocity change relates to a free-fall drop. Visualize the shock table falling, with the product attached; they fall at the same rate. Now the table hits the programmers and begins to rise, even though the product itself has not yet stopped. The shock pulse duration t must be sufficiently short (compared to the critical component's period T) to make this assumption. In practical terms, a shock duration t, using plastic programmers, is about 2 ms (lowest forcing frequency 250 Hz). Products with component natural frequencies up to 125 Hz will have sufficient time to behave as described - absorbing the full impact and rebound energy of the shock table before reacting.

As an example, to simulate a 30" free-fall drop, set up your shock table to produce a velocity change equal to the impact velocity of a free-fall drop: , where "g" is the acceleration due to gravity and "h" is free-fall drop height. This calculates to 152 in/sec, or 3.9 m/s. Stated in words, this relationship is:

Shock table velocity change = Free-fall drop impact velocity

The primary shock table limitation is found in the theory itself - products with all natural frequencies greater than 125 Hz cannot be properly equated to free-fall drops using a shock table. This is a concern for modern products, especially for high frequency electronic components. Newer machines can overcome this limitation, such as the GHI linear and rotational shock testers, but these come at considerable expense.

Case 3 for eliminating shock tables - replicating field damage.
Apart from generating DBCs or from explicitly simulating freefall drops, shock tables are sometimes used for input shocks to replicate field damage. But how well can the shock table do this, given the limitations mentioned in the previous sections? Consider the following example, which suggests why free-fall testing can often provide a better solution. Paper sacks intended for storing and handling sand (as at construction sites) were being evaluated in a laboratory for strength when wet. A number of bags filled with sand had been placed out in the rain. One at a time, they were placed on a shock table. No combination of velocity change and G level ruptured the bags. Yet a simple drop test (two men holding, then releasing the four corners) ruptured most of the bags. The center of the bag first impacting the ground, and ruptured. In this case, the real free-fall had much more value than using the shock table.

Inside a bag, what is the natural frequency of the sand? We would need that information to satisfy assumptions for free-fall simulation on a shock table. Obviously it is a stretch to fit this kind of product to the spring/mass system required by the DBC theory. Other non-ideal spring/mass systems are similarly suspect, such as food products. Free-fall drop testing should instead be used for non-ideal products to eliminate model-specific limitations.

Other Reasons to use free fall drop instead of shock tables:

Rotation while falling
Another practical concern is the argument for eliminating rotational package movement during the drop. But this is an advantage of the free-fall drop tester: some products such as disk drives react more to torsional forces than they do to orthogonal forces. Since the goal of lab testing is to re-create environmental conditions, we want to account for all forces, rotational included.

Repeatability
Let's now consider repeatability. Free-fall drops seem prone to "repeatability" issues and may seem to lack repeatability. However, shock tables suffer from a similar limitation. From drop to drop, shock tables can vary ±5% in velocity change. In this regard, shock tables have no advantage over free-fall testing - both will have some velocity change variability.

Corner and edge drops
Let's now consider another supposed advantage of shock tables: corner and edge drops. Fixturing for corner and edge drops (on a shock table) is surprisingly difficult and time-consuming. Since the intent of our lab testing is to re-create real world events, the simpler free-fall drop test provides a cleaner, more realistic test. Repeatability for free-fall edge and corner drops can certainly be questioned, but we maintain that the variability in these drops can be considered an advantage, similar to the argument for rotation.

Summary
Shock tables are large and expensive, and produce shock pulse shapes not found in real world events. Instead, free-fall drop machines and the appropriate data acquisition software can yield results that are as good and sometimes more appropriate than those obtained with a shock table. For the cost-conscious and space-limited packaging engineer, a move from shock tables to simple free-fall drop machines can be a great cost savings opportunity.

1 Newton, Robert E. "Fragility Assessment Theory and Test Procedure." US Naval Post Graduate School.

2 Selected ASTM Standards on Packaging. Fourth Edition. Philadelphia, PA: ASTM, 1994.

3 Daum, Matthew P. "Application of the Shock Response Spectrum to Product Fragility Testing." Masters Thesis. Michigan State University, E. Lansing, MI, 1994.

4 Ibid, p. 23.

5 Ibid, p. 21.

6 Daum, Matthew P. "Shock Response Spectrum and Fatigue Damage: A New Approach to Product Fragility Testing." Ph.D. Dissertation. Michigan State University, E. Lansing, MI, 1999.

7 Burgess, Gary J. "Extension and Evaluation of Fatigue Model for Product Shock Fragility Used in Package Design." Journal of Testing and Evaluation, JTEVA, Vol. 28, No. 2, March 2000, pp. 116-120.