Electronic Design

Letters

Why Not Use One Cup Of Water?
After reading your article "Energy Independence—Without Pollution—Lies At Our Fingertips" \[Nov. 6, 2000, p. 165\], I, probably like 10,000 others, wondered, why not use one cup of water for the airliner? If the gases are generated and burned in a closed system, capturing the exhaust water to recycle, no additional fuel would be required. But the question is, can more energy be developed in burning the gas than is used in generating it? I did really enjoy your article.
George Barber

Trekking Through The Minefield...
The assertion made in "Quadrupole Resonance Detects Landmines With 100% Accuracy" \[Oct. 16, 2000, p. 28\], about a QR detection system that demonstrates 100% detection accuracy is technically incorrect and misleading, as any competent communications or radar engineer will attest.

First, the term detection accuracy is a misnomer and meaningless when discussing signal detection. The correct term is probability of detection (Pd), which properly reflects the statistical nature of the real-world detection problem (see any good text on the subject, e.g., A.D. Whalen, Detection Of Signals In Noise, Academic Press, 1971).

This is true because the detection process is subject to the vagaries of propagation and the medium, the effects of random and nonrandom noise and interference, receiver sensitivity, signal-to-noise ratio (SNR), and so on. Thus, the process is probabilistic and statistical, so even under the most favorable conditions, like high SNR, Pd may only approach 100% (arbitrarily closely).

Second, the data Quantum Magnetics Inc. reported is anecdotal, so it's not a proper surrogate for statistically valid field test data. Reference to so-called field trial data is grossly misleading. To extrapolate these results to future real-world (battlefield) performance and claim 100% success is a most tenuous stretch of an overly optimistic and questionable forward-looking statement.

Third, your article improperly suggests to readers that reduction of the prototype to a useful product, compliant with all specified form, fit, and function requirements, will be a straightforward process. But, as one who spent the better part of a forty-year professional career in radar and communications engineering, I can say without fear of contradiction that the road from a prototypical curiosity to a fully qualified militarily useful and acceptable product can be long and arduous. It will most certainly be a dangerous trek through (no pun intended) a minefield of its own.

Also, given the obvious and necessary requirement for a very high Pd (six "9s" at least, i.e., 99.9999% min) and the concomitant tragic consequences associated with a missed detection, it remains to be seen whether the prototype, let alone the final product, will approach the required level of performance under all real-world conditions.

At least feasibility of QR in this application has been demonstrated.
Randolph A. Kallas
Consulting Electrical Engineer

...Blind Tests Prove 100% Accuracy
Mr. Kallas is correct when he points out that an accurate description of signal detection performance must reflect the statistical nature of the problem. Quantum Magnetics Inc. is careful to characterize the performance of its mine detection system in the form of receiver-operating-characteristic (ROC) curves, which plot Pd versus probability of false alarm (Pfa). ROC curves are widely recognized and accepted in the signal detection community as a proper way to present signal detection performance.

However, an ROC curve characterization of system performance is very different from a report of actual field test results. In 1999, Quantum Magnetics' mine detection equipment was subjected to four blind field tests. These tests were organized and supervised by U.S. Dept. of Defense personnel. To be even more objective, the data from these tests were stored in computer memory and analyzed by personnel in the Institute for Defense Analysis (IDA), not by Quantum Magnetics. IDA subsequently provided the results to the company.

In the first three blind tests, the system parameters of the mine detection system were set to provide a very high Pd and a very low Pfa. In each of these tests, 100% of the mines were detected, with no false alarms.

In the fourth test, the system parameters of the mine detection system were set to provide a high Pd and a Pfa of approximately 5%. When an alarm response was returned, the same area was rescanned once for alarm confirmation. This protocol dramatically reduces Pfa at the cost of a small reduction in Pd and a small increase in scan time. Again, in this blind test, the Quantum Magnetics system detected 100% of the mines, with no false alarms.

With regard to the difficulty of moving from a prototype to a fully qualified military product, I believe we are all in agreement. It's a big job, and the U.S. Army and Navy have responded by providing Quantum Magnetics with the necessary resources to do it.

In conclusion, Mr. Kallas asserts that mine detection systems require very high detection probability ("99.9999% min"). This confuses the performance requirements for humanitarian de-mining with that of military mine detection. Humanitarian de-mining requires an exceptionally high Pd, because all of the mines must be detected and cleared, and time isn't critical. On the other hand, the military often needs only a path through a minefield, and time is critical. So the competing requirements of scan time, false alarm resolution, detection probability, and total area cleared must be balanced with the requirements of the mission.
Lowell J. Burnett
President & CEO
Quantum Magnetics Inc.

Hide comments

Comments

  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.
Publish