Field density testing: nuclear gauge, sand cone and cores

Three methods, three different jobs. The disputes happen when someone treats them as interchangeable.

Choosing the method

Nuclear density gauge (ASTM D6938 for soil and aggregate, D2950 for asphalt) is for rapid, non-destructive control while compaction is happening. On a high-production job it is what tells the roller operator whether to keep rolling, within a minute, while the mat is still hot.

Sand cone (ASTM D1556) is direct physical verification of dry density in soil, subgrade and aggregate base. Use it where nuclear gauges are restricted or unavailable, as a check on gauge results, and on coarse variable material where the gauge seats badly. It does not apply to asphalt.

Cores (ASTM D2726 / D1188) are the contractual benchmark for finished asphalt. Acceptance, payment and dispute resolution all end up here, because a core is a physical piece of the pavement and nothing else in the argument is.

Frequency and where the tests go

Typical requirements, subject to your own specification:

• Unbound layers: one test per 500 to 1,000 m², or three to five per shift; on linear work, one test every 100 to 200 m per lane.
• Asphalt: one core or test per 500 to 1,000 m² per lift, or five randomly spaced cores per daily lot of roughly 500 to 1,000 tonnes.

Locations for acceptance must be chosen by stratified random sampling (ASTM D3665): divide the lot into equal sub-lots and generate random offsets. Picking your spots is not testing, it is negotiating. Deliberately targeting suspect areas is fine and useful — but for process control, not for acceptance.

What you are measuring against

• Unbound layers: 95% to 100% of maximum dry density from a laboratory Proctor (ASTM D698 standard or D1557 modified). Base courses typically want 98% to 100% of modified Proctor MDD.
• Asphalt against Marshall or gyratory density: 97% to 101%.
• Asphalt against Rice density (Gmm): 92% to 97%, which is the same thing as 3% to 8% in-place air voids, with 4% the usual target.
• Longitudinal joints: 90% to 92% of Gmm, because an unconfined edge will never match the mat.

Always confirm which reference your specification uses. "97%" against Marshall and "97%" against Rice are completely different requirements.

Running a nuclear gauge

1. Prepare the surface. Smooth about 300 × 300 mm with a scraper plate. Fill surface voids with a thin layer of fine sand and sweep off the excess — the sand fills voids, it does not form a pad under the gauge.
2. Standard count. A four-minute count on the reference block each day, at least 10 m from vehicles, walls and other sources. Check the drift against the manufacturer's limits (about ±1% density, ±2% moisture).
3. Drive the rod hole. Guide plate down, pin driven vertically at least 50 mm deeper than the test depth, then pulled straight up.
4. Seat and read. Lower the source rod to the depth notch, pull the gauge body firmly toward the rod so the source presses against the hole wall, and take a one-minute count.

What beginners get wrong

• Rocking the pin to free it. That makes an oval hole, and the air gap around the source reads as low density.
• Too much filler sand, lifting the gauge so it reads loose sand rather than the compacted layer.
• Direct transmission into fresh asphalt instead of backscatter.
• Not pulling the gauge back against the hole wall, leaving a void between source and detector.
• Testing within a metre of a trench wall or a parked pickup, which changes the gamma scatter.

Spotting a false reading

Causes include hydrogen-rich material (organics, binder, clay minerals) and heavy metals distorting the counts, air gaps under the plate, surface water from a dust suppression truck reading as moisture, calibration drift, and backscatter off trench walls.

The quickest check in the field is the rotation test: take a reading, rotate the gauge 90° or 180° about the same hole, and read again. More than about 30 kg/m³ (2 pcf) difference means air gaps, bad seating or local segregation — not a real density change. Also treat any implausible moisture value as a fault, not a result: dry-looking gravel reading 12% moisture is telling you something is wrong with the test.

When gauge and cores disagree

They rarely match straight out of the box. On smooth dense lifts a gauge usually tracks within 1.5% to 2% of core density once a project offset is applied. On open-graded or coarse mixes an uncalibrated gauge underreads badly, because of the surface voids sitting directly under the plate.

The correct procedure (ASTM D2950) is to build your own correlation: take 5 to 10 gauge readings, core at exactly those footprints, test the cores for bulk specific gravity, and program the mean difference into the gauge as an offset. Factory calibration alone is not acceptance-grade for asphalt.

When there is a contractual conflict, cores govern. Three referee cores in the contested sub-lot, tested in the laboratory, is the binding result.

Safety and licensing

None of this is optional, and it is where an inspector's authority is most clear-cut.

• Operators must hold formal nuclear gauge training approved by a Radiation Safety Officer. The organisation must hold a valid radioactive material licence.
• Transport in the locked Type A container, double-locked, strapped down, as far from the cab as possible, with the bill of lading, emergency response guide, yellow labels and licence copy in the cabin.
• Storage in a locked, posted room at least 5 m from continuously occupied desks, with a sign-in and sign-out log.
• Field: dosimeter badge worn at all times, source retracted and locked except during the reading, and ALARA — minimise time, maximise distance, use shielding.

A density dispute, and how it ended

On a 100 mm asphalt base course, contractor gauge readings averaged 98.2% of Marshall density. Client cores at the same locations came back between 94.5% and 95.8%, against a 97.0% requirement. The contractor argued the core bit was damaging the porous bottom of a thick lift.

A joint audit found two separate errors, one on each side.

• The contractor had never established a project correlation offset for a newly approved 28 mm NMAS base mix. The coarse texture created air gaps under the plate and the gauge was overreading.
• The laboratory was testing open, interconnected cores by standard SSD (ASTM D2726). Water was draining out of the void channels before the wet weight was taken, inflating the calculated volume and understating density.

Both were corrected: core testing moved to vacuum sealing (CoreLok, ASTM D1188), and ten fresh correlation cores gave a new gauge offset of +28 kg/m³. Re-evaluated properly, the disputed 8,000 m² sub-lot averaged 96.6% — genuinely below 97.0%, but structurally sound.

The outcome was a 5% pay-factor deduction under the specification rather than removal and replacement, plus two extra passes of a 24-tonne pneumatic roller on every subsequent lift. Both numbers had been wrong, in opposite directions, and neither party could have proved anything without correcting the test methods first.

Calculators for this work

• Road Base & Subbase Calculator — Calculate compacted volume and delivered tonnage of granular base or subbase, including the bulking allowance between loose and compacted material.
• Backfill Calculator — Calculate backfill volume around a structure and the material you need to import, using whichever compaction convention your source figure came from.
• Excavation Calculator — Calculate excavation volume in bank measure and the swelled volume that has to be hauled away.

This guide describes general practice and typical figures. Your project specification, job mix formula and laboratory results govern the actual values on your job. Nothing here is a design or a substitute for the engineer of record.