AIAA 98-3626, Proceedings of 34th Joint Propulsion Conference, Cleveland, OH, July 1998.

Lessons Learned from Safe and Arm Detonation Transfer Test Program


John Sudick
Special Devices, Inc.
Newhall, CA 91321

Special Thanks to DJ

Barry T. Neyer, Member AIAA
EG&G Optoelectronics, Inc.
Miamisburg, OH 45343-0529

Current Address
Barry T. Neyer
Excelitas Technologies
1100 Vanguard Blvd
Miamisburg, OH 45342
(937) 865-5586
(937) 865-5170 (Fax)


This paper summarizes the lessons learned from explosive interface testing between a detonator and an FCDC end tip. The most important lesson is that the test setup must be IDENTICAL to the device under test in all physically important parameters or the results of the test may be meaningless. For example, the test setup can inhibit detonation transfer at large standoffs by creating a geometry that does not exist in the real explosive interface. Secondly, radiographs of fired hardware can be a good indicator of "prompt" detonation. Many test fixtures can be designed with some type of "witness" material near the FCDC end tip. High order detonation causes deformation of this material indicating the point at which high order detonation was reached. Prompt detonation is indicated by evidence that high order detonation was realized from the initiation point. This evidence can provide additional evaluation criteria besides the traditional pass or fail criteria used today. Finally, statistical programs may incorporate assumptions regarding data distribution and selection of test standoffs; assumptions that can yield artificial indications of failures at small gaps in the actual design gap range. Proper attention to the statistical methods used to test designs, and to the analysis methods used to analyze the data from these tests is critically important.

Technical Papers of Dr. Barry T. Neyer


The Safe Arm Device, hereafter referred to as SAD, manufactured by Special Devices, Incorporated (SDI), is used in the Atlas launch vehicle for flight termination and solid motor ignition, as well as in the Atlas and Titan Centaur for flight termination. As shown in Figures 1 and 2, the SAD contains a pair of detonators which are recessed in the end of a cylindrical body. The body contains a permanent magnet which serves as the rotor in a brushless, bi-directional torque motor enclosed in a housing. A reversible positioning current causes the motor to rotate the body between Armed and Safe positions, and thereby establish or interrupt axial explosive transfer paths from the detonators to a corresponding pair of Flexible Confined Detonating Cords (FCDC) ported through the housing cover. An internal switching circuit disables the detonators when the body is in the Safe position and connects the detonators to the firing circuits when the body is in the Armed position.

Figure 1: Safe and Arm Device P/N 106679-6

A detent ring secured to the rotatable body, and a pair of spring loaded detents in the housing index the rotatable body in the Armed and Safe positions and prevent its accidental rotation. Mechanical safing means are provided for manually overriding an electrical arming signal. The device may be remotely Safed and Armed by electrical signals and may be manually Safed by means of the safing pin. Arm/Safe status of the SAD may be determined remotely by electrical indications or visually by the visual status indicator.

Figure 2: Safe and Arm Device P/N 106679-6

As can be seen in the figures, the SAD incorporates three electrical connectors, one for Safe/Arm commands and status and two for firing circuits, each leading to an independent detonator. The SAD also incorporates two ports for attachment of FCDC end fittings. These ports capture the FCDC ends in a manner which maintains proper standoff gap between the ends and the SAD internal detonators.

Between the detonators and FCDC end tips is a .005 inch aluminum barrier which is integral to the SAD housing. This barrier provides a hermetic seal of the SAD, at the port interface, when the FCDCs are not installed.

If the device is in the Arm position and a firing signal is applied to one or both firing connectors, the current is routed to the corresponding detonator. The detonator fires and produces a shock and flyer output. This output ruptures the aluminum barrier and the flyer from the detonator and barrier impact on the end tip of the FCDC. The impact produces a detonation in the HNS contained within the FCDC. The interface utilizes the end to end detonation transfer mode.

Summary of Testing

The SAD explosive interface consists of a donor (an Electro Explosive Device (EED) internal to the SAD) and a receptor (an FCDC connected to the SAD) . A thin aluminum membrane (barrier) is placed between the EED and the FCDC to provide an environmental seal for the device. Previous interface testing (conducted for Atlas) was done on the EED side of the barrier. The testing reported here was done on the FCDC side of the barrier; hence, all test gaps referenced herein refer to the relationship between the FCDC and the barrier.

The first set of tests consisted of 4x maximum gap tests. Five test firings were conducted at 4x maximum standoff (0.254 inch minimum gap) with worst case tolerances applied to all secondary test parameters (i.e. EED to FCDC misalignment and offset were both set to maximum drawing tolerances and the firing temperature was -65º F). These tests were conducted in fixtures that increased the stand-off (gap) dimension between the FCDC and the SAD seal barrier by simply increasing the port depth. The typical SAD FCDC interface is shown in Figure 3, while the first gap test set-up is shown in Figure 4. It is important to note two things about this test set-up that later became significant; first, the "tunnel" created between the FCDC and the barrier has a relatively large L/D ratio, and second, there is no provision for venting trapped gas within the tunnel.

Figure 3: SAD FCDC Interface

Four (4) of the five (5) tests were successful; one (1) failed to propagate high order. A failure analysis was conducted; the conclusion of which was that the FCDC simply did not receive sufficient stimulus to detonate.

Figure 4: Gap Test Set-up / FCDC Interface

The 4x gap tests were repeated with the test set-up shown in Figure 5. The following secondary test parameters were reduced from worst case to nominal:

Figure 5: Gap Test Set-up / FCDC Interface

The test results were the same as observed earlier: four (4) tests were successful as defined by high order detonation of the FCDC. One (1) failed to propagate high order.

The gap test was repeated again; only this time, the gap was reduced to 3x max design gap (0.189 inch minimum gap). The secondary parameters were again nominal including ambient firing temperature. The test results were the same as observed earlier: four devices transferred high order detonation and one failed.

At this point, the failure of one sample in each of the three series of max gap tests seemed to defy explanation. By all accounts, these three failures should not have occurred!

A concern was raised regarding the design of the gap test fixtures. In these fixtures the test gap was non-adjustable, created by simply increasing the FCDC bore depth in the test block. The problem was that this design also created a long bore between the barrier and the FCDC endtip that was not representative of the SAD design. It was speculated that the tunnel effect of the long bore could artificially attenuate the transfer pulse.

To gain a further understanding of the problem, all fired test blocks were subjected to X-Ray examination. The FCDC endtip is surrounded by a cylindrical bore. Since high order detonation causes the bore to swell its inside diameter, the point on the radiograph at which swell is observed is the point at which high order detonation was reached. The radiographs revealed that many of the successful firings had not experienced "prompt" detonation.

Prompt detonation would be indicated by a swelled ID that essentially matched the original FCDC end tip length; i.e. evidence that high order detonation was realized from the initiation point. The fact that many successfully fired units had experienced non-instantaneous (non-prompt) detonation transfer was further indication that something was causing attenuation of the transfer pulse.

The next logical step was to conduct a sensitivity test. Two sensitivity test methods were considered, the Neyer D-Optimal test method and the Langlie test method. The Neyer method held the promise of better convergence, i.e. better results from the same number of firings. The Langlie method was more familiar, having been used in the Atlas test some three years earlier.

Several computer simulations of the test program were conducted using the Langlie test method. Results obtained during previous gap tests were used as a basis of predicting the most likely mean. Successes and failures were then distributed about the predicted mean with a normal distribution. In all simulations, the predicted All-Fire value was larger than the minimum gap dimension (standoff) for sample sizes of 50 or greater. Based largely upon the confidence in the Langlie simulations, this test method was selected for the subsequent sensitivity testing.

The Langlie tests could not be conducted with the same type of fixtures as had been used in the earlier series of gap tests. In those tests, the standoff is a machined, non-adjustable dimension built into the test hardware. The Langlie test required an easily adjustable method for setting the test gap. In addition, concerns raised earlier relative to the "tunnel effect" created in the gap test hardware prompted a different type of test fixture for the Langlie testing. This fixture is shown in Figure 6.

Figure 6: Threshold Test Set-up / FCDC Interface

The results of the Langlie test were enlightening:

As can be seen from the earlier gap test data, failures had occurred at gaps of 0.254 inch or smaller in 20 percent of the tests. With the redesigned test fixture, no failures occurred at gaps of less than 0.457 inch. If the transfer probability was independent of the fixture, one would have expected some failures at or near 0.254 inch based upon the earlier data. The test data clearly showed that the design of the test fixtures had significantly impacted the detonation transfer margin.

There was one problem with the Langlie test results. There were several successful transfers at large gaps. These large gap successes, typically considered a design asset, caused the Langlie program to diverge and select large step sizes for successive tests. These large gap successes also resulted in a large standard deviation. The large standard deviation caused an unrealistic All-Fire level prediction. For the All-Fire prediction to be valid, the transfer function used to normalize the distribution of the test data must be normal relative to the mean. The problem with the Langlie test was that this distribution was not normal with a standard logarithmic distribution.

Based on the Langlie test data, a decision was made to perform 5 additional (constant gap) 4x tests using the Langlie test fixture. This test fixture eliminates the narrow bore, producing a condition more representative of the S&A design. The actual test gap was set between 0.272 and 0.274 inch, greater than 4x the maximum design gap. The test conditions were as follows:

This final set of maximum gap interface tests consisted of 5 shots conducted at 4x maximum standoff with nominal tolerances applied to all other test parameters, i.e. misalignment and offset set to nominal drawing tolerances and ambient firing temperature.

The results were that all five (5) units successfully transferred with high order detonation.

Post Fire Radiographs

During the failure investigation, radiographs were taken of the post fire test setups for all tests performed during this margin test program as well as the SAD to FCDC margin test program conducted for the Titan IV Solid Rocket Motor. Both programs utilized an aluminum FCDC holder with a cylindrical bore surrounding the FCDC end tip similar to that shown in Figure 4. Inspection of the radiographs revealed that some had evidence of prompt detonation initiation in that the cylindrical bore inner diameter had swollen adjacent to the full length of the detonating end tip. Steel debris from detonator flyer and FCDC end tip, visible in the bore, was comprised of very small pieces. Radiographs of setups in which detonation transfer had not occurred clearly showed no swelling of the bore and large pieces of shrapnel. Some radiographs of setups in which detonation transfer had successfully occurred revealed a lack of prompt detonation initiation in the FCDC end tips. This was characterized by a ramping up of the bore’s inner diameter over a length of 0.050 to 0.100 inch to the final steady state swell and by larger pieces of shrapnel than that evidenced in the prompt initiation units. Although the output from these setups indicated a successful detonation transfer, the radiographs revealed that the transfer was actually marginal at 4x. Prompt detonation was observed in all units at gaps of nominal or smaller standoffs; but, for the tests performed at three or four times design gap, with the original test fixture, more than half exhibited marginal initiation of the FCDC.

Whereas the test success was defined as go / no-go criteria, i.e.; detonation output from the flexible cord of the FCDC, the radiographs revealed that although a test might have produced a successful output the initiation might have been marginal. These results indicated that post test radiographs are a good

investigative tool to provide additional information concerning the results of margin tests.

Test Setup

The original gap test fixture as shown in Figures 4 and 5 was a replication of the design interface in the SAD except the bore containing the FCDC end tip was increased in length in order for the FCDC end tip to stand off from the barrier by the required test gap distance. After three series of tests in which one out of five test firings failed to propagate, the design of the gap test fixture was reevaluated. The bore containing the FCDC end tip had a diameter of 0.231 inch nominal and a length to the end tip of 0.252 inch minimum, resulting in a length to diameter ratio of 1.09 whereas the nominal L/D ratio in the SAD is 0.28. Concern was raised about the longer bore, down which the flyer had to travel prior to impacting the FCDC end tip. It was thought that the long bore attenuated flyer velocity by friction between bore and flyer or by resisting air pressure. In addition, radiographs of post firing test setup, as described above, always showed prompt detonation initiation for setups with a short bore, similar to that of the flight SAD. Modifications to the test fixture were considered to eliminate the potential effects of the long bore.

As seen in Figure 6, the modified fixture required the flyer to travel down a bore length equal to that in the SAD followed by open travel the remaining length. All other test conditions remained the same as those of the earlier gap tests. Results of testing were significantly different in the two test setups. The original setup showed a 20% failure rate at standoffs of 0.252 and 0.189 inch. The modified setup produced the lowest failure at a standoff of .457 inch and successes at standoffs as high as 1.086 inch.

The physics associated with the different results from the two setups is complex and not yet understood; but, clearly the test setup significantly impacted the test results. The lesson to be learned from these tests is that modification of flight configuration and standoffs in test setups to perform margin testing must be carefully considered in order to attain applicable data.

Although the community initially believed that merely lengthening the bore and maintaining all remaining features identical to those in the SAD was a valid test setup, test results proved that the lengthening the bore artificially impacted the test results.

Discussion of Sensitivity Testing

The sensitivity test data was analyzed with various transformations, from linear, square root, log, 1/square root, 1/x, and inverse square. The analysis shown in Table 1 clearly shows that the best fit to the data is obtained when using an Xp transformation where the power, p, is between -1.5 and -2.0. Because the test was not optimized for determining the transformation function, the confidence level associated with determination of the best transformation function is not large. Using the likelihood ratio test yields a confidence that the physics is better represented by an inverse square transformation than by a linear function by a confidence of approximately 43%. That is, approximately 57% of the time that you make such comparison of two samples drawn from the same population, the test would yield such a difference. While the data is suggestive of an inverse square response, it is not possible to establish the correct transformation with this data set.

Table 1: Table of Likelihood Ratios


Log of Likelihood Function

Relative Ratio




Square Root (X0.5)



















The inverse square response suggested by the test data has a basis in the physics of the device. The interface consists of a detonator, an air gap, a metal barrier, another air gap and a receptor.

In this test, the distance between the barrier and the second air gap was varied. Work using VISAR and flash X-ray on other devices shows that explosively driven flying plates travel for large distances without appreciable slowing. Thus, it is rather improbable that increasing the distance results in a slower flying plate, at least not at the distances of this test.

However, the greater distance could have an effect if the flying plate created from the impact of the detonator flying plate with the metal barrier produced a flying plate that had a number of pieces or one that left at an angle. In such a case, the flying plate or major pieces could "miss" the receptor; thus there would be no detonation. A reasonable assumption is that the flight angle of the piece(s) does not change with distance. Thus, the area that contains fragments with enough mass and velocity to cause detonation grows with the square of the distance.

Although both the statistics and physics suggest that the above discussion is reasonable, additional tests could be performed to verify or challenge these assumptions. Flash X-ray work would show if the flying plates broke into many pieces or left at an angle. Additional statistical testing, at levels chosen to optimize determination of the transformation function would shed additional light on the interface.

Assuming that the inverse square variation of the threshold is proper, the data can be analyzed to compute the desired confidence levels. Figure 7 shows the results of such analysis. The data clearly shows that the interface is reliable. The statistics say that the reliability is 99.9999% at 95% confidence for all second gap distances less than 0.2 inches. The graph also shows that there is a relatively large potential for initiation of the receptor even with a large air gap distance. Of course, as in any statistical analysis, the assumption is critically dependent on the probability distribution function. Because the same dependence on probability distribution function is present in the testing for all detonators, the interface studied in this test has the same statistically demonstrated reliability as a detonator tested by varying the ignition current.


Figure 7: Probability Levels for Barrier + Air Gap Data

One further note is that there are most probably two different detonation transfer mechanisms. The radiographic measurements discussed earlier showed that all detonations as short air gap lengths were prompt, while many at large distances were not prompt. If the difference between prompt and delayed detonation transfer is an indication of different transfer mechanisms the simple probability distribution assumptions used in the statistical analysis must be revised. The data could be reanalyzed to define as failures those sets of data in which detonation was not prompt. The revised analysis could give greater insight into the interface. However, such work is beyond the scope of this paper.

Conclusions / Lessons Learned

The test program and associated failure investigations for the SAD revealed valuable lessons learned which should be kept in mind

for design of future test programs. These lessons include the following:


H. J. Langlie (1965), "A Reliability Test Method For "One-Shot'" Items," Technical Report U-1792, Third Edition, Aeronutronic Division of Ford Motor Company, Newport Beach, CA.

Barry T. Neyer (1994) "A D-Optimality-Based Sensitivity Test," Technometrics, 36, 61-70.