TECHNICAL / FLOORING

The Six Windows That Decide a Dry-Shake Floor

From specification sheet to failure mechanism: reading a timing-driven process

 

A dry-shake hardener floor is usually written down as a checklist - thirteen operations, each with its numbers. What the checklist never says is why the numbers sit where they do. Read the thirteen as one timeline and the whole process collapses onto six windows, all of them governed by the same variable: the few millimetres of water at the surface.

 

TECHNICAL DESK | 2 November 2026 | TECHNICAL ANALYSIS

---

The specification for a dry-shake hardened floor arrives as a list: thirteen operations, each with its thickness, its timing, its grit. Follow it and the floor should come out right.

 

What happens on site is different. Same mix, same crew, same pallet of hardener - one bay dries to an even sheen, the next delaminates inside six weeks. The difference is rarely whether a step was performed. It is whether it was performed half an hour early, or forty minutes late.

 

This article puts the numbers back where they came from: the physical processes they are guarding against. An operative who knows what a number is defending can still make the right call in the cases a specification never covers.

 

6 critical windows	40 min bond coat to pour	2/3 + 1/3 the aggregate split
24–36 h the saw-cut window	≥1/3 depth of slab thickness	3000 grit at final burnish

Values throughout reflect common practice for heavy-duty industrial floors. Temperature, humidity, wind, slab thickness and hardener system all shift them.

 

1. Thirteen steps, one water curve

The flow chart reads as a chain: scarify and clean, set out, erect formwork, apply the bond coat, place concrete, install mesh, work up the cream, broadcast and grind the first pass, broadcast and finish the second, saw-cut, cure, densify and polish, protect.

 

Read it a second way. All thirteen steps serve a single objective - controlling water content and paste maturity within the top five millimetres of concrete. Scarifying lets water in. Eight hours of pre-wetting lets the substrate drink its fill. The bond coat glues one generation of concrete to the next. Working up the cream lifts paste to the surface so the hardener has something to sit in. Broadcasting locks aggregate into that paste. Saw-cutting releases the stress that drying builds. Curing stops the water leaving too soon.

 

Three physical processes run along that curve at very different speeds.

 

THE FULL CHAIN
(1) Scarify and clean	(2) Set out	(3) Formwork
(4) Bond coat	(5) Place concrete	(6) Install mesh
(7) Work up cream	(8) First broadcast	(9) Second broadcast
(10) Saw-cut	(11) Cure	(12) Densify and polish
(13) Protect

 

Steps 4, 7, 8, 9, 10 and 11 carry hard timing windows. They are what this article is about.

 

Table 1 Three concurrent processes at the surface

Process	Driven by	Timescale	Cost of losing control
Bleed and evaporation	Gravity pulling water down, evaporation lifting it up; they meet at the surface	Minutes to hours	Surface water/cement ratio drifts: too wet dusts, too dry yields no cream
Hydration	Temperature, cement composition, admixtures	Hours; irreversible after initial set	Miss the working window and nothing brings it back
Aggregate settlement	Density difference between aggregate and paste; vibration energy	Minutes, mostly within 30 min of placing	Aggregate sinks, paste rises, the wear layer is no longer at the surface

Where the three overlap is what this article calls a critical window. Every operation has a period during which it can be done. Before it, or after it, the result is not a slightly worse version of the same floor - it is a different material.

 

Worse, the width of that window is set by the weather rather than by the document. The same method statement that can be followed calmly at 15 °C under cloud will compress to a third of its allowance at 32 °C, force 4 wind, on a roof in direct sun. A standard gives you a median. A site needs a correction.

 

2. The forty-minute clock

The bond coat is the step most often treated as optional. It looks like brushing on a bit of cement slurry; plenty of crews skip it and rely on the concrete's own paste. Yet on a retrofit, or at a construction joint between two pours, the interface strength is almost entirely decided here.

 

The numbers are hard. A styrene-butadiene rubber (SBR) latex-modified cement slurry, about 3 mm thick, applied within 20 minutes of mixing, covered by concrete within 40 minutes, in areas no larger than 50 m². Each of those three numbers answers a different question.

 

Why latex and not neat cement. SBR latex is a copolymer emulsion of styrene and butadiene. It works on the interface through two routes. The first is physical: as water is drawn into the substrate or evaporates, the emulsion particles flocculate and coalesce into a continuous polymer film - a film with some flexibility, able to bridge the small movement differences between old and new concrete. The second is chemical: hydration products react with the calcium hydroxide and ettringite on the old surface, forming secondary bonds. A neat cement slurry offers only the second route, shrinks as it hardens, and typically ends up as a brittle weak layer rather than a connection.

 

Why twenty and forty minutes. Neither is a folk number. Both mark the initial set of the slurry. By penetration resistance, initial set sits at about 3.5 MPa; with a polymer emulsion in the mix, that point typically arrives between 40 minutes and an hour after gauging. Once the slurry crosses it and loses plasticity, fresh concrete vibrated against it cannot interlock with it. The interface degrades from one slab into two slabs laid on top of each other.

 

Why fifty square metres. This is arithmetic, not regulation. Area per batch equals placing rate multiplied by the time from application to cover. If a crew lays 1.2 m² a minute, forty minutes covers 48 m². Set the batch larger and the far edge of every batch is over-age; set it smaller and the crew spends its rhythm shuttling back for materials. The figure of 50 is a chemical deadline converted into a working area.

 

THREE WAYS THIS STEP GOES WRONG ! Slurrying onto dry concrete. An unwetted substrate pulls the water straight out of the slurry; the polymer has no time to film and dries to powder. Bond strength falls to near zero. Pre-wetting should run for at least eight hours.   ! Turning "wetted" into "flooded". No standing water may remain at placing. Free water dilutes the local water/cement ratio of the bond coat and produces the same weak layer by a different route. The two requirements look contradictory; they share one objective - controlling absorption, not soaking the slab.   ! Substituting neat cement or a general-purpose PVA adhesive. The first shrinks and is brittle; the second is not water-resistant. Neither survives a winter-summer cycle.

 

3. Working up the cream: a state, not a time point

The instruction is the vaguest in the whole document: begin once the concrete has entered initial set, and the surface no longer sinks or collapses under people and machines. In practice this is the step most often misjudged.

 

Start with what the operation is for. The job is to lift paste to the surface so that the hardener about to be broadcast has a continuous bed of adequate depth. Aggregate carries no strength on its own; it develops wear resistance only by being locked into paste. Too thin a bed and the aggregate sits on aggregate, gets torn out under the trowel, and the floor shows exposed stone. Too thick a bed and the surface never reaches hardness - it dusts. The requirement for at least 3 mm of neat paste after levelling is simply a quantity for that bed.

 

What happens if you start early. Within thirty minutes of placing, aggregate is still settling and bleed channels are still open. A trowel pan at that moment accelerates the sinking and drives water upwards: the machine manufactures a high-water/cement laitance layer at the surface. That layer hardens weak and porous, and it is the direct origin of later dusting.

 

What happens if you start late. Once the surface has lost water, capillary tension locks the paste into the voids between aggregate. The pan is no longer stirring paste; it is rubbing dry. What comes up is usually powder that has already lost its reactivity - physically piled on the aggregate, chemically unconnected to it. This is why some floors look as though a layer of dust has been laid on top rather than finished in.

 

So "no longer sinking" needs translating into a measurable state. Three criteria are used in practice, listed in order of reliability.

 

The window is acutely sensitive to conditions. Above a wind speed of 3 m/s, in direct sun, or above 30 °C, surface evaporation can run three to five times the normal rate and the window can close from forty minutes to under fifteen. In those conditions waiting will not deliver the right moment. Crews have to be pre-positioned, bays shortened, and windbreaks or shade erected before the pour starts.

 

Table 2 Correcting the window for weather (baseline: mild, still air)

Condition	Effect on the window	What to do
Above 30 °C	40–60% shorter	Smaller bays, avoid midday, chilled mixing water
Wind above 3 m/s, including through-draught indoors	30–50% shorter	Windbreaks, seal openings, misting (never direct wetting)
Roof slab in direct sun	About 50% shorter, edges drying first	Shade the work; treat edges and column bases ahead of the field
Below 10 °C	50–100% longer, but strength gain is slow	Extend curing, avoid accelerators by default, no night work

Percentages are field estimates, given to show direction and order of magnitude rather than regulated values. The actual window should be established from trial work on the project, measured evaporation rate and a penetration-resistance curve.

 

4. Two thirds and one third: structural zoning, not a saving

The hardener goes down in two passes, split two thirds to one third. Most crews know there are two passes but read the reason as "one pass does not spread evenly", and broadcast accordingly. The two passes carry entirely different structural jobs.

 

The first pass is structure. Its task is to let aggregate sink into the paste and key into the substrate. At that moment the paste is still generous and the surface still damp, so aggregate can be wrapped and can part-settle. This layer sets the floor's compressive and impact resistance, and decides whether there is a reliable load path between aggregate and base. It goes down when a footprint sinks 3–5 mm, and is trowelled in once the aggregate has drawn in moisture and visibly darkened.

 

The second pass is function. Its task is fine levelling, making good the unevenness of the first pass, and forming the final surface. The base is now stable - no further sinking underfoot - so the aggregate stays at the surface and builds a concentrated wear layer. This layer sets surface hardness, abrasion resistance and colour uniformity.

 

Put both down at once and here is what happens. The upper aggregate has no paste left to wet it and forms a dry, loosely piled layer; the lower aggregate, loaded all at once, sinks further and squeezes paste upwards. The surface then looks densely covered in grit while actually being loose grains floating on a paste-rich layer, lifted away by the trowel to expose the paste underneath. That is the classic mechanism by which exposed aggregate and colour variation appear together.

 

A few details deserve separating out.

 

 

Table 3 Division of labour between the two broadcasts

	First broadcast	Second broadcast
Share of total	Two thirds (structural)	One third (surface)
Trigger	Footprint sinks 3–5 mm	No further sinking underfoot
What the aggregate does	Partly sinks into the paste, keys into the base	Stays at the surface, does not settle
What it delivers	Compressive and impact resistance, load path	Surface hardness, abrasion resistance, colour uniformity, fine levelling
When to trowel	Once the aggregate has wetted out and darkened	Immediately after broadcasting, at least two passes
Direction	Along the long axis of the bay	Perpendicular to the first pass

Cement Grout Application on Concrete Floor Substrate

 

Installation of Steel Mesh Sheets

 

5. Twenty-four to thirty-six hours: a race

Saw-cutting gets five lines in the manual, and is the only operation that cannot be put right afterwards. It deals with a contradiction built into concrete: hydration liberates heat and then the slab cools and contracts, while friction against the base prevents it contracting freely, so tensile stress accumulates. When that stress exceeds the tensile strength of the moment, the slab cracks.

 

Saw-cutting is therefore not the act of cutting a groove. It is appointing a crack before the crack appoints itself - trading one controlled, straight, sealable joint for a whole bay of uncontrolled ones. It is a scheduled release of stress.

 

The cost of cutting early. While flexural strength is still low, the blade drags aggregate out with it and leaves spalled, chipped edges. The defect is exaggerated on a hardener floor because the surface aggregate is hard and its elastic modulus differs sharply from the paste, so the interface is the first thing cutting forces tear apart. The 24-hour floor is really the threshold at which flexural strength can resist the disturbance of cutting.

 

The cost of cutting late. The hydration peak usually arrives 12–24 hours after placing, after which the slab cools and thermal contraction begins to accumulate. Past 36 hours the tensile stress has often already crossed the tensile strength, random cracks appear first, and the saw is then only adding straight lines to an already broken floor - the stress has long since gone.

 

Where the parameters come from.

 

TWO TIMING TRAPS AT THE SAW ! Delaying the cut to make it look neat. Neatness comes from the chalk line and the blade, not from concrete age. Two hours of postponement for appearance can buy a crack across the full bay.   ! Not washing the slurry out. Cutting slurry is loaded with fines that set to the same colour and hardness as the floor, are very difficult to remove afterwards, and clog the joint faces so the sealant will not bond. Flush while the slurry is still wet.

 

Grinding and Polishing of Hardened Flooring

 

6. Trowelling and curing: locking in surface density

Trowelling is usually understood as smoothing the surface. That is half of it. The steel blade does two things: it compacts, eliminating pinholes, air voids and locally loose material; and it seals, bending and shrinking the surface capillaries to narrow the paths by which water and contaminants later enter. On a hardener floor those two actions set surface density, and surface density sets abrasion resistance and permeability.

 

One line in the manual is easy to skim past: hand trowelling should be at least as many passes as machine trowelling. That is not a request for more labour. It points at a specific defect source - colour variation. A power trowel has a working radius, and there are permanent blind spots along walls, at column bases and in doorways. Worked only by machine, those areas compact less than mid-bay, and after hardening their porosity and therefore their reflectance differ. The same slab then reads as a patchwork of light and dark. Hand trowelling exists to bring the blind spots back up to the compaction of the field.

 

Curing is not adding water; it is controlling a moisture gradient. Surface concrete must not lose water so fast that plastic shrinkage cracks it, nor stay saturated so long that strength gain stalls and the surface dusts. Curing holds the surface inside an acceptable band for long enough.

 

Which explains why the manual splits the method by location. In direct sun, apply a curing compound, because evaporation is the dominant loss path there and a film suppresses the rate. Underground, with no sun, cover with geotextile or felt and keep it wet, because evaporation is not the problem and what is needed is a continuing supply. The two methods address two different physical processes and are not interchangeable.

 

Where the seven days come from. Ordinary Portland cement concrete reaches roughly 60–75% of its 28-day compressive strength at seven days, and the first week is exactly when plastic shrinkage - hours after placing - overlaps most violently with drying shrinkage - days to weeks as the surface loses water. Banning traffic during curing is not only about damage from loads. It is about early vibration disturbing a skeleton that is still forming. That damage is silent, invisible at the time, and shows up months later as a network of fine cracks.

 

The manual's field indicator is a good one: water should stand on the cover when you walk on it. No instrument needed - standing water means the cover is in a saturated evaporating state and the internal gradient is flat. A cover that has gone white and dry means supply is short and watering is overdue.

 

7. Grit sequence: each step does one job

Densifying and polishing is compressed into six lettered lines in the manual, and it is where most of the technical content of the whole process actually sits. There is one governing rule: each grit exists to remove the scratch left by the previous one, not to make the floor shiny.

 

Hold that rule and the reason you cannot skip grits becomes obvious. Diamonds on a disc have a fixed size and cut grooves of the same order. Jump from 200 to 800 and the 800 grit, smaller than the depth of the 200 scratch, can only work the ridges and never reaches the trough. What results is a surface that has been smoothed over rather than cut back: the deep scratch is still there, the later polishing passes round off its edges, the floor looks glossy, but the gloss is scattered rather than transparent and a haze of texture shows in raking light.

 

400 grit is an inspection gate, not a formality. By that stage the surface is flat enough that any trowel mark, depression or pitting is fully exposed. It is also the last point at which repair is still economic: defects can be dressed out locally with a diamond cup wheel and the earlier steps repeated. Past 800 those defects are wrapped inside successive layers of polish, and rework costs multiply. The manual places "inspect the surface" at this step, and the placement is correct.

 

What 3000 grit actually buys. From 2000 to 3000 the removal is minimal. The job is to press the depth of the remaining scratch down to a level invisible to the eye and to raking light, then to lay a gloss enhancer over a surface of even reflectance. "Transparent gloss" is really uniformity of surface roughness, not absolute smoothness. Any local step in roughness scatters light, and the floor reads as patchy.

 

One cross-material comparison is worth making here. The same graded diamond sequence, the same grit numbers, run on a cementitious hardener floor will go all the way to 3000 and take a mirror; run on a resin-bound engineered stone it starts to smear and lose gloss past 800. The discs are not the variable - the substrate is. Polyester resin softens at around 150 °C, and the frictional heat of fine-grit grinding is enough to soften, smear and blind the surface resin, so the tool is no longer cutting stone but pushing softened polymer. Grit number is never the only variable in gloss; the temperature a substrate can survive is the hard ceiling.

 

There is a second assumption buried in any grit sequence, and it is less often discussed: that every block in the sequence behaves like its label. Sequences fail at the tool far more often than at the operator. A block that cuts harder than its neighbours - because its bond is harder, its diamond concentration higher, its segments taller - opens a scratch that the next grit cannot reach, and the floor reads as haze long after handover, with no visible cause. Turning six nominal steps into six real ones depends on bond hardness, diamond distribution and segment height holding steady from block to block, and from pallet to pallet. It is the least glamorous line in the specification, and the one that decides whether the other six windows were worth defending - which is why huicetools builds its diamond floor grinding blocks around bond consistency, controlled diamond concentration and an aggressive, free-cutting matrix.

 

Table 4 Grit sequence and what each step is actually for

Grit	Passes	Job at this step	Chemistry	Inspection point
50 / 100	2–3, cross-hatched	Strip laitance, open capillaries, rough level	Penetrating densifier after 100	Even aggregate exposure, no laitance left
150 / 200	2–3, cross-hatched	Refine flatness, prepare the surface for densifier	Sealing densifier	No visible scratch, even absorption
400	1–2	The last gate before defects become permanent	-	Trowel marks, depressions and pitting surface here
800	2–3, cross-hatched	Remove the 400 scratch, first gloss	-	No directional texture in raking light
1000	1–2	Transition into polishing	-	Gloss begins to even out
2000	1–2	Polish, raise reflectance	Gloss enhancer afterwards	No hazy patches
3000	1	Burnish, flatten the last of the scratch	-	Continuous reflection in raking light, no colour banding

 

8. From a schedule to a set of criteria

Back to the opening question: why does the same manual produce such different floors. The answer sits in the word timing. A manual states moments; a site needs states. The standard says 24–36 hours because it cannot know what time you poured, whether there was wind on the roof, or whether the slab is 120 mm or 200 mm. Those judgements have to be made on site.

 

Turning six windows into something a crew can act on comes down to three moves.

 

One closing observation that is not strictly technical. The real difficulty with a process like this is seldom knowing when to act. It is whether enough people and plant are standing there at the moment the window opens. A forty-minute window for working up the cream means crews on standby from arrival of the first truck, every machine serviceable, and concrete supply uninterrupted. One gap in the truck cycle and the timing for the entire bay is lost.

 

Which makes a timing-driven process, at bottom, an organisational problem. It asks for tighter rhythm rather than higher craft. A manual can teach one person to read a state; only organisation ensures that everything which should be present when that state arrives is actually present.

 

Appendix Six windows at a glance

Reference values, normal heavy-duty conditions

Operation	How to read the window	If too early	If too late
Bond coat	Mixed and applied within 20 min; covered within 40 min; ≤50 m² per batch	No such failure mode	Paste past initial set, no interlock, delamination
Working up the cream	Penetration resistance 0.3–1.0 MPa; footprint 3–5 mm; even film behind the pan	Disturbs settling aggregate, builds a laitance layer, dusting	Surface locks up, lifts powder, aggregate left without a bed
First broadcast (2/3)	Footprint 3–5 mm; trowel in once the aggregate has darkened	Aggregate sinks, wear layer too thin	Paste hardened, aggregate cannot key, exposed stone
Second broadcast (1/3)	No further sinking underfoot; check level with a straightedge first	Lands on an unsettled base, uneven density	No paste left to bond to, dry shake floats
Saw-cutting	24–36 h; cooling under way; flexural strength adequate	Spalling and chipped edges	Random cracking arrives first, the cut becomes decoration
Trowelling and curing	No deformation under hand pressure; cure from final set; ≥7 d; water standing on the cover	Ripples and bleed marks	Surface dries too fast, plastic shrinkage cracks

 

Values are reference figures for normal conditions; the project method statement and the hardener supplier's technical data govern.