Report 226 · Defense Tech
Military GPS antennas don't out-shout a jammer. They go deaf in the jammer's direction. The rule of thumb is that an array of N elements can null N-1 jammers, so seven elements handle six. That rule is real, and it comes straight from the engineering papers. It is also a ceiling, counted in directions, with conditions attached. Here's what it leaves out.
GPS arrives weak. Park and Seo, in a 2020 IEEE paper on anti-jam methods, describe the received signal power as "extremely weak," and note that even very low-power jamming can overwhelm it. You can't fix that by turning the satellites up from the ground. So the military answer has long been to change how the receiver listens.
A normal GPS antenna hears the whole sky about equally. A controlled reception pattern antenna, or CRPA, is a small array of antenna elements whose signals are weighted and combined so the pattern has a deep notch, a null, pointed at the jammer. The satellites elsewhere in the sky stay audible. The jammer, in its direction, mostly doesn't.
NAVSYS, a company that built these arrays for the U.S. military under research contracts, stated the rule in a 2000 conference paper on its miniature arrays:
The maximum number of jammers, which can be nulled by a GPS array, is equal to one less than the number of antenna elements (N-1).
Each extra element buys one more independent direction the array can turn its back on. Four elements, three jammers. Seven elements, six. In the same paper, NAVSYS reported that its four-element, six-inch array "was able to null up to three jammers while still providing gain across much of the field of view on the GPS satellites tracked."
Read that sentence slowly, because the qualifier is the honest part: much of the field of view. Not all of it.
Park and Seo put the same rule in signal-processing terms. An N-element array has N-1 spatial degrees of freedom, so it "can steer up to N−1 spatial nulls" toward jammers. Note the words up to. Every null spends a degree of freedom, and an array that has spent them all has nothing left for the next emitter.
Here's the part that matters to anyone on the other side of the problem, and this is my reading, not the papers': the array counts jammers by direction. It doesn't care how much each one cost. If a defender's antenna has seven elements, the cheapest answer is not a bigger jammer. It's a seventh small one, from a new bearing. In counter-IED and EW work, that's the same arithmetic as any capacity limit. Once you know the number, you know what to bring.
Park and Seo are specific that the N-1 count applies to narrowband jamming signals, which they define as a signal whose bandwidth is under 2% of its center frequency. A spatial null is set with one weight per element, and that weight is tuned to a frequency. Jamming spread across the band is usually handled by adding time taps to each element, an approach the papers call space-time adaptive processing. That works, but it is more hardware and more computation, and Park and Seo point out that with poorly chosen constraints, mitigating a jammer can cause "undesired attenuation of GPS signals" as well.
So "six jammers" assumes a particular kind of jammer. That isn't a flaw in the engineering. It's a condition the marketing tends to drop.
A null is a direction. Anything arriving from that direction gets suppressed, wanted or not. Park and Seo state the reason plainly: for an ordinary single-polarized array, the gain depends only on the signal's direction and frequency. If a jammer sits in the same direction as a satellite, there's nothing for the array to separate them by.
Their paper is about a way around this: an antenna that also uses polarization, which can suppress a jammer that isn't polarized like GPS while keeping a satellite in the same direction. In their tests, a jammer placed exactly along a satellite's line of sight was the special case worth an experiment of its own. For a conventional array, that geometry just costs you the satellite.
In practice, losing one or two satellites near the horizon is often survivable, because a receiver needs only four to fix a position. But "the antenna nulled the jammer" and "the receiver kept its full constellation" are different claims.
Classic array design spaces elements about half a wavelength apart. At the GPS L1 frequency, that's roughly 9.5 centimeters. NAVSYS's 2001 paper describes the seven-element GPS Antenna System then in Defense Department use as 14 inches across. The company's contribution was a dielectric lens that let it shrink a seven-element array to 7 inches, small enough to consider for aircraft and munitions.
Even then, look at what that 2001 paper actually measured. It showed the small array tracked satellites as well as a full-size reference array, with roughly 7 dB of gain over a single element on average (its conclusion rounds that to +8 dB). Its jamming tests were still ahead: they were "planned to be conducted later this year." I point this out not to knock NAVSYS but because a lot of anti-jam claims ride on papers that measured something adjacent to anti-jam.
Park and Seo sum up the trade from the 2020 side: arrays outperform single antennas against jammers, but their use "remains limited because of their size, cost, and computational complexity." That's why the phone in your pocket, and most civilian receivers, still hear the whole sky.
Counter-IED and electronic warfare are where I learned to ask what a countermeasure assumes about the threat. I wrote about the mirror image of this in the report on why every transmission is a bearing: a direction is information, for both sides. And "jamming-resistant" doesn't mean one thing made the same point about drone links. A CRPA is a very good tool, and it's a bounded one. The bound has a number, and the number has conditions.
If you want the basics of what jamming does to a receiver in the first place, start with jamming vs. spoofing. Note that nulling is a jamming defense. A spoofer that arrives from a believable direction looking like a real satellite is a different problem.
Fielded performance. I found no open, current test data on how specific fielded military CRPAs perform against multiple real jammers. The sources here are a company's 2000 and 2001 conference papers and a 2020 academic paper. Current systems are certainly better, and their numbers aren't public.
Wideband detail. I'm stating that wideband jamming is handled with space-time processing at a general level, as the Park and Seo paper describes the field. I didn't open the STAP papers it cites, and I'm not giving a count of wideband jammers an array of a given size can handle, because I didn't find one I could verify.
Inferences. The "add one more emitter from a new bearing" point and the civilian-receiver observation are my reasoning from the sources, not findings in them. The 9.5 cm figure is my arithmetic from the L1 frequency, 1575.42 MHz. NAVSYS's 2000 adaptive-pattern results used a jammer power of 50 dBW and a 47.6 dB jammer-to-noise ratio. I didn't check how they'd change at other levels.
Not read in full. A 2020 University of Massachusetts dissertation on man-portable CRPAs for dismounted soldiers looked relevant, but I couldn't open its text this run, so it isn't cited.
An anti-jam GPS antenna works by aiming deafness, not by being louder. The N-1 rule tells you how many directions it can go deaf in, at most, against a particular kind of jammer. It doesn't promise that every satellite survives, and anything that shares a jammer's direction goes deaf with it. When someone says a receiver "handles six jammers," the useful follow-ups are: what kind, from how many directions, and what did it lose doing it?
Scope note: this report explains published, unclassified antenna engineering at a general level. It describes no specific fielded system's vulnerabilities and gives no instructions for jamming. Jamming GPS is illegal for civilians in the United States and most other countries.
U.S. Army combat veteran (Counter-IED / Electronic Warfare), peer-reviewed researcher in microwave spectroscopy, and founder & CEO of Shroombiosis. Consults on laboratory operations, AI, and supplement formulation.