Discrimination and notch settings: silencing junk without losing gold
Discrimination is the setting beginners reach for first and understand last. The promise sounds perfect: tell the detector to ignore junk and it will only beep for treasure. The reality, documented by every serious guide on the subject, is narrower and more interesting. A detector does not identify objects; it analyses a target's electromagnetic response and places that response somewhere on an identification scale. Discrimination is the decision to suppress responses in parts of that scale. Used surgically — as notch discrimination, which rejects a narrow band while accepting targets on both sides — it genuinely cleans up a trashy park. Used bluntly, it silently deletes gold, because gold and the junk being rejected often share the same stretch of scale. This guide explains the mechanism, the overlap problem in numbers, and a field procedure for setting a notch from evidence gathered on the site itself rather than from a chart copied off the internet.
Two different tools sharing one name
Standard (threshold) discrimination rejects everything below a chosen point on the scale: set it high enough and iron and foil go quiet, along with everything else that reads low. Notch discrimination is selective instead — it creates precise “holes” in the detection pattern, silencing a specific conductivity range while targets above and below still sound. Serious Detecting's notch guide gives the scale its usual shape on machines that display target ID from roughly 1 to 99: low numbers are iron and foil (around 1–15), the mid-range covers pull tabs and small gold (around 20–35), and high numbers indicate silver and copper coins (around 70–90). When a segment is notched out, the detector still senses targets in it — the field still detects the object — but the control box suppresses the audio and visual alert entirely.
That last detail matters in the field. A notched target is not absent; it is unreported. On a site where the notch was set from yesterday's junk, today's different junk — or today's gold ring reading like yesterday's pull tab — passes under the coil in silence. LP Metal Detecting's discrimination explainer makes the underlying point bluntly: different objects can produce similar responses, and a target's response is shaped by size, shape, depth, orientation, ground conditions, mineralisation, nearby targets and corrosion, not by material alone. Discrimination can therefore never draw a perfect line between good targets and rubbish; it draws a line through a cloud of overlapping responses.
The overlap problem, in the published numbers
| Scale band (typical 1–99 ID) | Junk that reads there | Good targets that read there too |
|---|---|---|
| 1–15 | Foil, small iron | Smallest gold items and fine chains can read very low |
| 20–25 | Some pull tabs | Nickels (kept deliberately in published park patterns) |
| 25–35 | Pull tabs — the classic notch target | Small gold, including rings, in the same 20–35 mid-range |
| 35–45 | Bottle caps in published park patterns | Larger gold and some tokens can reach into this band |
| 70–90 | Comparatively little common junk | Silver and copper coins |
Read the middle rows twice, because they are the whole argument. Published park patterns notch pull tabs at 25–35 while carefully keeping nickels at 20–25 — and the same guides place small gold at 20–35. The pull-tab band and the small-gold band are not neighbours; they substantially overlap. A worked example from XP Deus II guidance makes the precision point from the other direction: if pull tabs consistently hit one specific ID on a given machine, notching just that number leaves gold rings reading a couple of points either side still audible. The lesson generalises: the narrower and more site-specific the notch, the less collateral damage. The wider and more generic the notch, the more it functions as a gold filter that occasionally also removes tabs.
A field procedure: notch from evidence, not from charts
- Start in all-metal or minimal discrimination for the first twenty minutes. Dig a honest sample of the site's actual junk. Success at this stage is a finds pouch that shows what this ground really contains, not what a chart predicts.
- Record the IDs of the junk you actually dug. Note the exact numbers your machine assigns on this site, in this ground — the same tab can read differently at a different depth or orientation, so use the range your own recoveries show.
- Notch the narrowest band that covers the repeated offender. One or two ID points of proven junk, not a ten-point swath. If the junk's IDs sprawl across the gold range, that is the site telling you a notch will cost more than it saves — hunt by ear instead.
- Seed a test if the site allows it. Where permitted and practical, burying a known low-conductor (a thin gold-coloured test piece of your own) at shallow depth and sweeping it with the notch active verifies the notch has not silenced the class of target you came for.
- Re-check the decision hourly. Sites change character across a single field. If good targets start appearing at the edges of the notch, narrow it or remove it.
What success looks like: the junk rate in the pouch falls while the detector still audibly reports low and mid conductors either side of the notch — silence has been applied to a proven nuisance, not to a category of metal. What failure looks like is quieter and worse: a wonderfully peaceful hunt. Total silence from a trashy park usually means the detector has been told not to report the site, not that the site has been cleaned.
When to leave discrimination off entirely
Three situations argue for ears over filters. On a new site, the first visit is reconnaissance — the junk is information about age and activity, and notching it away destroys the evidence before it has been read. On a site with genuine gold potential — old swimming areas, historic fairgrounds — the overlap table above is the argument: the cost of a missed ring dwarfs the cost of digging tabs. And in iron-heavy ground, iron-masking already hides good targets near ferrous junk; adding aggressive discrimination compounds the loss, which is why features like iron audio, which let iron be heard faintly instead of silenced, exist as an alternative philosophy: hear everything, decide with the brain, and let the spade arbitrate the doubtful ones.
Frequently asked questions
What is the difference between discrimination and notch? Discrimination rejects everything below a threshold; a notch rejects a specific narrow band while accepting targets above and below it. Can a detector tell gold from a pull tab? Not reliably — both sit in the same mid-range of the ID scale (roughly 20–35 on a 1–99 scale), and a response also depends on size, depth, orientation and ground. What do published park patterns notch? Typically foil (1–15), pull tabs (25–35) and bottle caps (35–45), while keeping nickels (20–25) and high coin IDs (70–90) — patterns to adapt from your own dug junk, not copy blindly. Does notching lose depth? The detector still senses notched targets; it suppresses the report. The real cost is not depth but silence over anything sharing the band.
Sources, accessed October 21, 2026: Serious Detecting, “Notch Discrimination: How to Filter Unwanted Targets” and metal detector technologies library; LP Metal Detecting, “Metal Detector Discrimination Explained.” ID ranges are the typical scales those guides publish; your machine's numbering may differ — verify against its manual and your own test targets.