A detector gives a clear tone over an ordinary patch of yard. The display settles on a number that the owner’s chart associates with a coin. A second sweep changes the number. Turn ninety degrees and the good tone breaks into a scratchy pair of sounds. The easy story is that the machine cannot decide what is buried. The more useful story is that the machine has measured different parts of a complicated target scene. It has not looked through the soil and seen a coin. It has driven an electromagnetic field into the ground, received a changing response, and compressed that response into sound and a number.
Before any shovel comes out, this distinction is worth learning in a place where the answer is known and recovery is allowed. Put a modern coin, a clean iron nail, and an aluminum pull tab separately on a surface clear of other metal. Pass the detector over each at a steady height. Then move the nail near the coin without allowing them to touch. The exercise is small, but it exposes nearly every mistake that makes metal detecting frustrating: treating an identification number as a metal assay, believing iron rejection makes iron disappear, or assuming that a quiet pass means the desired object is absent.
This is the first article in a six-part guide. The example is a controlled demonstration, not a report of a real buried find. Its purpose is to make the physics visible. The manufacturer’s explanation of how detectors work and its technical guide describe a search coil that transmits a field, metal targets that respond, and a receive signal processed into audio and visual information. Garrett’s AT Pro manual illustrates how a nail and coin together can return a response unlike either object alone. Those accounts are product-specific in details but share the physical starting point: a signal is evidence about an interaction, not a photograph of an object.
What happens under the coil
The coil is the detector’s active part. Current through its transmitting winding produces a changing magnetic field. A nearby conductive object responds with circulating electric currents, often called eddy currents. Those currents produce their own field, and the receiver senses the change. The control unit interprets timing, strength, phase, and other features of the return according to its design. A basic machine may communicate mostly through tones; a more elaborate one adds a target-identification number, ferrous indication, depth estimate, or trace. None of those displays is a direct observation of the target’s printed date, shape, or legal status.
The soil is part of the measurement. Iron minerals, salts, moisture, and nearby electrical equipment can create responses of their own. The machine also sees a changing target position as the coil moves. A shallow coin flat to the coil will not present exactly the same signal as the same coin on edge, deeper down, beside an iron nail, or under a pocket of strongly mineralized soil. The detector is solving an inference problem with incomplete data. Two objects can have similar electromagnetic behavior; one object can behave differently under different conditions. This is why a target ID is useful as a clue and dangerous as a promise.
Many modern detectors classify signals partly by electrical conductivity and ferrous behavior. The Minelab general FAQ explains that ferrous and conductive properties inform discrimination, while its Gold Monster manual stresses that gold nuggets can occupy different target IDs depending on size and shape. The exact number scale differs by model; a number on one brand is not a universal metal name. A silver-looking number could be a coin, a piece of folded aluminum, or several adjacent items. A low number could be foil, a small gold item, or a tiny fragment of something else. The detector’s language is a set of measured categories, not a catalog of buried things.
The known coin and the unknown sound
Return to the three test objects. The coin usually offers a relatively repeatable response when it lies alone, flat, and close to the coil. That is not because coins possess a special property called “coin tone.” Their material, size, thickness, and shape combine to make an electromagnetic response that many detector programs find convenient to classify. Change the coin to a different denomination or metal and the number may shift. Tilt it on edge and the coupled area changes. Raise the coil and the signal weakens. Turn the sensitivity down and a weak target may vanish. None of these changes alters what the object actually is.
The nail introduces magnetic behavior and an elongated shape. Depending on the machine and its settings, it may give a low iron tone, a broken tone, a surprisingly high edge response, or silence if that range is rejected. A bent nail can complicate the result further. Rejection does not remove the nail from the ground or its effect on neighboring signals. It only tells the software to suppress or reclassify particular measurements. Garrett’s manual gives a concrete example: with one discrimination setting, a nail is rejected, but a nail over a good coin produces a combined response that can be heard. Other combinations can hide a desirable signal. The important observation is not that one fixed setting solves all iron. It is that two objects create a scene different from two isolated air tests.
The pull tab is the most honest nuisance in the demonstration. It is nonferrous and conductive. Many detectors cannot confidently distinguish every aluminum tab from every small piece of gold jewelry simply by a target ID. A program that rejects the tab’s range may also reject some desired objects. A program that accepts the range will ask the user to examine more trash. There is no free filter that keeps all good targets and discards all bad ones, because “good” is a human judgment and the fields overlap. The manufacturer’s detection FAQ states the tradeoff bluntly: no discrimination is the only way to be certain that the machine has not excluded a potentially desirable response, though that entails more unwanted signals.
This does not make discrimination useless. It is a tool for managing time and choosing which ambiguous signals to pursue. In a modern yard with permission to recover recent lost items, a person may choose to prioritize stable coin-like signals. In a controlled survey for a specific metal object, another user may accept a broader range. In nugget prospecting, excluding low conductive signals would be a serious mistake if the expected gold is small. Settings should follow the question, and the losses imposed by each setting should be understood.
A number is compressed evidence
Target ID displays can create an illusion of precision. “Twenty-eight” feels much more exact than “a medium nonferrous response.” But it is a processed summary of a signal that may have changed along the sweep. Many machines average or select parts of the return. The display may update more slowly than the tone. The center of the coil, its speed, target orientation, and nearby objects can all influence the value. It is rational to record the number, but irrational to treat it as a chemical analysis.
Try moving the coin farther from the coil in the controlled setup. The audio may weaken before the displayed number becomes unstable, or the number may jump while a faint repeatable tone remains. Different detectors behave differently, but the point is general: signal-to-noise ratio falls as the target response weakens. A weak field makes any inference less secure. The machine may know “something conductive is present” with more confidence than it knows exactly how to classify it. A depth indicator is similarly conditional. It often assumes a coin-sized target; a larger item deeper down can imitate a smaller item nearer the surface.
Try the same coin with the nail nearby. Move the nail to the side, then over the coin, then beyond it. Sweep from two directions. The display and tone may change even though the coin never moves. This is target masking and target combination in a form the eye can verify. The coil can receive more than one object’s response within its field. It cannot always separate them into neat independent signals. Minelab’s explanation of target separation distinguishes the ability to hear separate adjacent targets from the ability to classify or reject a target. A smaller coil or suitable recovery-speed setting can help in some trashy situations, but neither turns the detector into an X-ray camera.
The thought experiment also shows why a “perfect ID” story on a video is weak evidence. The audience sees one successful dig and a number after the fact. It may not see the ground balance, the rejected signals, the exact sweep speed, or the many other holes. A single demonstration proves that a machine identified that scene under those conditions. It does not establish a universal setting for all yards, soils, or depths. The best demonstration includes misses, mixed signals, and known test objects. It makes the machine’s uncertainty visible rather than editing it away.
Listening for repeatability
Audio deserves as much attention as the screen. A signal that repeats at a consistent point as the coil crosses from multiple directions is more informative than a one-way chirp that appears only at a particular edge. This is not a rule that all one-way sounds are junk; a deep, tilted, or iron-masked desired object can produce an imperfect signal. It is a way to gather more evidence before disturbing the ground. Sweep with overlapping passes, hold the coil level, and change direction without wildly changing speed or height. If the response moves as the coil moves around a surface rock, the ground or the rock itself may be part of the signal. If it remains centered on one spot, a discrete target becomes more plausible.
Careful listening can also expose the costs of aggressive settings. Turn iron rejection high in the controlled nail-and-coin setup. The nail may go silent, but the coin’s tone can change or vanish at some positions. Raise sensitivity until the detector chatters from electrical interference or soil, then lower it until the background becomes interpretable. Maximum sensitivity on the control panel does not necessarily produce maximum useful information. The goal is to hear a faint real target against a stable background. A machine that reports every fluctuation may make an apparently busy patch less intelligible.
In an ordinary permission-based modern yard, the result of this listening may be a decision to pass on a target. That is a valid choice. Metal detecting is full of tradeoffs between time, damage, curiosity, and recovery. The article’s central lesson is not “dig everything.” It is “know what your rejection and uncertainty cost.” If the aim is to find a lost modern ring, a pull-tab-like signal may be worth investigating. If the aim is to locate a known lost steel tool, ferrous rejection would defeat the task. The target question should be stated before the settings are chosen.
Air tests and buried ground answer different questions
The coin-nail-tab exercise is best begun aboveground because the objects and spacing remain visible. It isolates how the machine responds to known targets. But air tests do not replicate mineralized soil, moisture, compact layers, or depth. They may exaggerate the clarity of target ID. A shallow test area on property where the owner permits it can add realistic ground, provided the objects are placed and retrieved without harming anything historic, planted, or buried. If digging is inappropriate, surface tests still teach the classification problem. The test is not a license to bury objects on public land or at a historic location.
A proper comparison keeps the variables visible. Change one thing at a time: object orientation, depth, separation from the nail, coil height, or discrimination. If all change at once, the result is dramatic but uninterpretable. Note both the target ID and the audio, including moments when the display says nothing but a faint tone repeats. Note any nearby electrical sources or metal structures. The purpose is to learn the machine’s response under known conditions, not to produce a viral treasure-hunting clip.
The most revealing result is often a mismatch. The pull tab may sound more coin-like than expected. The coin may be masked by the nail. The number may change when the coil is raised an inch. Each surprise replaces a simplistic rule with a better model of the signal. Later, in a real search, that model helps a user describe an unknown target without claiming to know it. “Repeatable medium-conductive sound from two directions” is better evidence than “definitely silver.” It also leaves room to stop when recovery would be unlawful or damaging.
There is a useful negative control as well. Sweep the same clear surface with none of the three objects present, then with only an empty plastic container or a wooden spacer. If the machine sounds on the supposedly empty setup, look for a hidden fastener, nearby fence, phone, electrical cable, or a ground response. Without a negative control, the first beep in a demonstration may be credited to the object being tested even when the environment produced it. A test area need not be a laboratory, but it should answer the simple question of what happens in the absence of the target.
Repeat the sequence with the coil at several measured heights. A person might expect every signal to fade smoothly. In practice, a weak target near the detection threshold may alternate between a recognized tone and silence. The threshold is not a line painted in the soil; it is a decision by electronics working with noise and a moving field. This matters when someone claims that a detector can reliably reach a fixed depth for “a coin.” The depth depends on target size and orientation, soil, coil, settings, electrical noise, and the criterion used to call a response reliable. A dramatic one-time chirp is not the same as a repeatable signal useful for recovery.
The exercise should end with written observations, not only a favorite setting. A short table can record each object’s orientation, coil height, direction of sweep, audible response, and displayed range. No exact number from that table should be copied to another detector model as a universal coin scale. Its value is local calibration: the owner can compare future sounds with known examples and notice when the ground has changed the machine’s behavior. This is the modest foundation of skill, and it is more durable than memorizing a chart.
The site is part of the instrument
The detector cannot tell whether a target belongs to the user, a landowner, a mining claimant, or the public trust. It cannot distinguish a modern lost coin from an archaeological coin by date. Those questions are answered by land status, permission, and context. A Forest Service prospecting guide illustrates low-impact recreational use and separates mineral prospecting, recent lost items, and historical resources. It specifically warns that archaeological and historic material is protected and that local rules and closures matter. BLM’s public collection guidance likewise distinguishes reasonable low-impact mineral collection from cultural artifacts, which should be left in place.
National Park Service land is different. The NPS archaeology FAQ says metal detecting and possession of detectors in parks are prohibited under its rules absent authorized work. A good target tone at a park boundary does not alter the rule. Nor does a faded mining site become free for relic hunting because its buildings are gone. Historic context can survive as fragments, nails, coins, and soil relationships. The later article in this guide on the land beneath the coil will follow those boundaries in detail. For the present exercise, the safe and analytically clean location is a known, permission-based modern space with no reason to expect protected artifacts.
The point is more than compliance. Context determines what a signal means. A modern coin dropped at a picnic site tells a different story from a coin found in a documented nineteenth-century camp layer. Removing the latter without recording the stratigraphy and associated material can erase the evidence that made it historically meaningful. A detector gives no warning that the object carries that context. The user has to know the site and be willing to leave a find alone when the setting calls for it.
There is also a common inversion in treasure talk: the machine’s uncertainty is used as an excuse to dig first and ask later. A scratchy number becomes “maybe a rare coin,” and that possibility is treated as sufficient reason to disturb a place. A better sequence begins with permission and purpose. At a lawful site, the sound can be investigated with the least disturbance allowed. At a protected or uncertain site, the same sound is a reason to stop and seek guidance, not to escalate. The detector’s limitations make that ethical judgment more important, not less.
The sound after the hole is opened
Recovery can test an interpretation, but it can also change the signal. Imagine the permitted test patch again. After a shallow plug is lifted, a formerly broad tone becomes sharp and high. The target has not necessarily transformed; the coil is now closer to it, part of the overlying ground is gone, and the nearby nail may be separated from the coin in a new geometry. Conversely, a signal can seem to vanish when the target is in the removed plug rather than the hole. Before digging farther, scan the plug and hole separately with the main coil or a pinpointer. This simple step saves needless disturbance and prevents a mistaken story about a “false” signal.
If the recovered object is a pull tab, the original target number was not a broken promise. The response was an estimate based on electromagnetic behavior, not a visual label. Record the original sound and the actual object while both are fresh in memory. Over time, those paired observations teach more than memorizing a chart of supposedly fixed coin numbers. The same habit helps at a site where recovery is not permitted: the operator can describe the response accurately without claiming to know what remains below the surface.
What the first signal teaches
The controlled coin, nail, and pull-tab scene is small enough to repeat and rich enough to change a beginner’s expectations. A detector can reveal that something conductive or magnetic affected its field. It can estimate how the response compares with patterns stored in its electronics. It can help locate the center of a target and sometimes distinguish adjacent responses. It cannot certify a coin, measure precious-metal purity, infer historical ownership, or authorize recovery. The number on the display is the beginning of an interpretation, not the end.
Once this is clear, the next problem is the ground itself. New Mexico has places where iron-rich rocks, variable soil, and electrical noise can produce responses that compete with a target. The next chapter follows that noisy ground and the meaning of ground balance. Later chapters place the same signal problem in a modern coin hunt, a documented gold district, a land-status check, and an apparent historic find. The coil remains the instrument in each story; the question around it changes.
Source notes
- Minelab, “How Detectors Work” and its technical guide, explain the transmitted field, induced target response, and the roles of coil and control unit.
- Garrett AT Pro manual, illustrates a nail and coin responding together under a particular discrimination setting.
- Minelab FAQ and Gold Monster manual, explain target ID limits and the effect of size, shape, ferrous behavior, and discrimination. Model-specific numbers are not generalized here.
- US Forest Service metal-detecting guidance, BLM public collection guidance, and NPS archaeology FAQ, provide the different land-management contexts mentioned in the article.