There’s been a wave of sensor-based ore sorting systems installed at copper operations in the past two years. The promise is compelling: reject waste rock before it enters the mill, reduce processing costs, improve head grade, increase throughput.

For operations mining 1.5-3% copper ore, it works exactly as advertised. For operations targeting 0.4-0.8% copper, it’s been a very expensive lesson in physics limitations.

I’ve talked to three different operations that installed ore sorting systems expecting similar performance to the published case studies, only to discover the technology doesn’t perform below certain grade thresholds. Here’s what they learned.

How Ore Sorting Actually Works

Sensor-based ore sorting uses X-ray transmission (XRT), X-ray fluorescence (XRF), laser-induced breakdown spectroscopy (LIBS), or near-infrared sensors to detect element concentrations in individual rocks on a conveyor.

High-grade rocks get sorted into the “accept” stream heading to the mill. Low-grade rocks get rejected to waste. In theory, you increase mill feed grade while reducing total tonnage processed.

The critical variable is grade contrast: the difference between ore and waste grades. If your ore averages 2% Cu and waste averages 0.2% Cu, that’s 10:1 contrast. Easy to discriminate with sensors.

If your ore averages 0.6% Cu and waste averages 0.3% Cu, that’s 2:1 contrast. Much harder to reliably discriminate, especially at the 1-2 second scan time typical of high-throughput ore sorters.

The Grade Threshold

Based on vendor data and operational reports, sensor-based ore sorting starts breaking down below these approximate thresholds:

• XRT sorting (copper): 0.8-1.0% Cu minimum ore grade
• LIBS sorting (copper): 0.7-0.9% Cu minimum ore grade
• XRF sorting (copper): 1.0-1.2% Cu minimum ore grade

Below these thresholds, the sensors can’t reliably distinguish ore from waste at production sorting speeds. Error rates climb, you start rejecting economic ore or accepting barren waste, and the whole system economics collapse.

The vendors don’t advertise these thresholds clearly. Marketing materials show successful installations at high-grade operations (2-4% Cu) and imply the technology works equally well at lower grades. It doesn’t.

Case Study: Failed Sorting Installation

An operation in South Australia installed a $12M LIBS-based ore sorting system in 2024 targeting a 0.5% Cu ore body with 0.25% waste grade.

Expected performance (based on vendor projections):

• 35% mass rejection to waste
• 25% head grade improvement
• 30% throughput increase (from reduced mill tonnage at higher grade)

Actual performance after 18 months:

• 22% mass rejection to waste
• 11% head grade improvement
• 8% throughput increase
• 17% of sorted “ore” was actually waste (sensor error)
• 12% of sorted “waste” was economic ore (sensor error)

The sensor error rates meant they were losing roughly $1.8M per year in copper sent to waste dump. Combined with underperformance on grade improvement and throughput, the system never came close to delivering projected returns.

They’re now bypassing the sorter for 60% of throughput and only running it on coarse, visually distinct ore zones where sensor performance is better. The $12M system is operating at 40% of design capacity.

Why Low-Grade Sorting Fails

The physics problem is signal-to-noise ratio at low concentrations. Copper at 0.5% concentration produces weaker sensor response than copper at 2% concentration. That weaker response is harder to distinguish from background noise and natural mineral variation.

Compounding this, ore sorting happens on heterogeneous rocks with variable mineralogy, not prepared samples in lab conditions. A rock containing 0.7% Cu isn’t uniformly distributed—it might have small chalcopyrite blebs in otherwise barren host rock.

The sensor scans the surface of the rock passing on conveyor. If the chalcopyrite blebs are on the interior and the surface is barren, the sensor sees low grade and rejects an ore rock. If the reverse happens (chalcopyrite on surface, barren interior), waste gets accepted as ore.

At high ore grades, these errors average out and don’t matter much. At low grades approaching waste cutoff, these errors kill the economics.

Where Sorting Works

Ore sorting delivers strong results in specific conditions:

High-grade deposits: 1.5%+ Cu for base metals, 3+ g/t Au for gold. Strong grade contrast makes sensor discrimination reliable.

Coarse-grained mineralization: Visible chalcopyrite or bornite is easier to detect than fine disseminated copper. Porphyry deposits with strong stockwork veining work well.

Simple mineralogy: Deposits with one or two dominant copper minerals are easier to sort than complex polymetallic ores where sensors get confused by multiple element responses.

Consistent ore characteristics: Ore from a single geological unit with predictable texture and mineralogy sorts better than variable ore from multiple geological domains.

The best installations I know of are high-grade skarn or vein deposits with coarse, easily identified mineralization. They’re achieving 40-50% waste rejection with <5% error rates. That’s transformative for mill economics.

The Marketing vs. Reality Gap

Ore sorting vendors naturally showcase their best installations in marketing materials. What you see are case studies from 3% Cu operations achieving spectacular results.

What you don’t see are the installations at low-grade disseminated deposits where sorting underperformed and operators quietly stopped running the systems.

I’m aware of at least five ore sorting installations in Australia that are either bypassed entirely or running well below design capacity because actual performance didn’t match projections. Those don’t appear in vendor case studies.

Due Diligence Questions

If you’re evaluating ore sorting for your operation, ask vendors:

1. What’s the minimum ore grade where this technology performs reliably? Get specific numbers for your commodity, not generic answers.

2. Show me installations with similar ore grades and mineralogy to my deposit. Case studies from high-grade operations aren’t relevant.

3. What are your actual sensor error rates at different grade levels? They track this data—make them share it.

4. Can we do extended pilot testing on representative ore samples? Lab tests on ideal samples don’t predict production performance. Test on actual mine run-of-mine ore.

5. What happens if performance guarantees aren’t met? Get financial protection if the system underperforms.

Alternative Approaches

For low-grade deposits where sensor sorting doesn’t work, other pre-concentration methods might:

Visual sorting: Manual or automated optical sorting based on rock appearance. Lower throughput but works on some oxide copper deposits where malachite/chrysocolla is visibly distinct.

Dense medium separation: Uses specific gravity differences to separate ore from waste. Works where copper minerals significantly increase rock density.

Screening and size separation: Sometimes waste is preferentially concentrated in fines or coarse fractions. Simple mechanical separation can improve grade.

Geological domain segregation: Mine different geological units separately and process optimally for each, rather than blending everything into a single mill feed.

None of these are as elegant as high-tech sensor sorting, but they’re often more practical for low-grade, complex ores.

The Bottom Line

Sensor-based ore sorting is a proven technology for high-grade deposits with distinct mineralogy. It’s not a universal solution for every low-grade deposit looking to improve mill economics.

If your ore grades are below 0.8-1.0% Cu (or equivalent for other metals), be very skeptical of vendor performance projections. Demand extensive pilot testing on representative samples. Build conservative business cases that account for sensor errors and underperformance.

The operations that succeed with ore sorting are the ones with inherently suitable ore characteristics. The ones that struggle are trying to force a technology into applications where the physics don’t support vendor promises.

Figure out which category your operation falls into before you write a $10M+ check for sorting equipment.