Home › Experiments › Medium vs coarse grind

Medium vs coarse grind on a V60: a simulated comparison

One variable changed · simulated on the CoffeeScope V3 engine · updated 25 July 2026

Grind size is the single most powerful dial on a pour-over, and the debate is old: grind finer to pull more out of the bed, or coarser to keep the cup clean and avoid bitterness? We ran a medium and a coarse grind through the CoffeeScope simulator on one identical V60 recipe — changing only the grind — and read off what the model pulls out of the bed.

The question

A finer grind exposes more surface area, so each ground gives up its solubles faster; it also produces more fines that pack the bed and slow the water down, extending contact time. Both effects push extraction up. Coarser grind does the opposite: less surface, a looser bed, faster drawdown, less extracted. The barista's worry is that fine over-extracts into bitterness and astringency, while coarse under-extracts into a thin, sour cup. This pairing probes the coarse half of that trade: a medium grind that should land in the sweet spot against a coarse grind that gives extraction away.

The setup

Both lanes are the simulator's stock V60 recipe (16 g, 1:16, 94 °C, spiral pour). The only difference is grind median — 600 µm (medium) vs 850 µm (coarse) — with the simulator's fine/medium/coarse size-class model enabled so fines can migrate and clog the paper.

Exact recipe (reproducible in the simulator)

{
  "brewer": "v60", "doseG": 16, "ratio": 16,
  "waterTempC": 94, "roastLevel": 0.4, "roastAgeDays": 7,
  "grindSpread": 0.45, "bloomWaterX": 2, "bloomTimeS": 45,
  "pourRate": 4, "pourHeight": 8, "pourPattern": "spiral",
  "bedLevel": 0, "pourHeadroom": { "stopMm": 10, "resumeMm": 13 },
  "sizeClasses": true, "resolution": "medium",

  "grindUm": 600   // Lane A — medium
  "grindUm": 850   // Lane B — coarse
}

Every number on this page is read from these two runs, taken to brew completion (the drawdown stall detector) on the deterministic V3 engine.

Simulated results

Physics-based simulation, calibrated to move in the right direction — not a measurement of your grinder and beans. How to read these numbers →

Measure	Medium · 600 µm	Coarse · 850 µm	Takeaway
Extraction yield (EY)	18.2 %	15.5 %	Medium lands at the low edge of the ~18–22 % Gold Cup window; coarse sits just below it.
Strength (TDS)	1.27 %	1.08 %	Medium brews a noticeably stronger cup; coarse lands at the thin edge of filter strength.
Drawdown time	176 s	147 s	Fines slow the paper: the medium bed drains about 1.2× slower, and that extra contact time itself drives more extraction.
Cup volume	229.9 ml	229.9 ml	Same water in, the same beverage out — grind changes concentration, not volume.
Cup temperature at finish	70.9 °C	73.4 °C	The longer medium drawdown loses a little more heat before the brew reaches the cup.
Caffeine	171 mg	162 mg	Caffeine is the fastest-extracting solute, so it strips to near-completion given time — the longer medium drawdown pulls a little more. It is insensitive to grind itself; the contact-time gap drives this.

Predicted taste axes

The taste axes are directional calibrations, not sensory-panel scores. A difference is flagged tasteable when it clears the model's own threshold of 0.045 (0–1 scale).

Axis	Medium	Coarse	Direction
Sweetness	Sweet, developed	Underdeveloped	Large shift
Clarity	Slightly denser	Cleaner	Moderate shift
Aroma	More aromatic	Muted	Moderate shift
Bitterness	More bitter	Cleaner	Small shift
Body	Fuller	Lighter	Below threshold
Astringency	Smooth	Smooth	Below threshold
Acidity	Balanced	Balanced	Below threshold

Exact taste axis values (0–1 scale)

Axis	Medium	Coarse	Δ
Sweetness	0.44	0.27	0.17
Clarity	0.73	0.80	0.07
Aroma	0.79	0.72	0.07
Bitterness	0.36	0.31	0.05
Body	0.56	0.52	0.04
Astringency	0.09	0.07	0.02
Acidity	0.53	0.53	0.00

Four of the seven axes clear the tasteable threshold. The headline is sweetness: the medium cup reaches the extraction where the sugars develop, the coarse cup stops before them. Clarity and aroma move together in the other direction — the medium cup is a touch denser and more aromatic, the coarse cup cleaner but quieter on the nose. Bitterness ticks up with the extra extraction, though only a small shift. Acidity, by contrast, doesn't move at all here: both cups carry the same amount of dissolved acid, so the coarse cup's cleanness comes without any extra brightness to show for it. Body and astringency stay under the tasteable threshold too.

How it builds over time

Cumulative extraction yield and cup volume, sampled during the brew:

Time	Medium — EY	Medium — in cup	Coarse — EY	Coarse — in cup
60 s	7.7 %	35.0 ml	6.5 %	42.8 ml
120 s	14.4 %	143.6 ml	13.2 %	180.2 ml
Final	18.2 %	229.9 ml	15.5 %	229.9 ml

At two minutes the two brews sit close on extraction (14.4 vs 13.2 %) but the coarse bed has already passed noticeably more liquid to the cup (180.2 vs 143.6 ml). The gap that decides the final cups opens in the tail: the coarse brew simply runs out of contact time at 147 s, while the medium bed keeps extracting for another half minute.

Why it happens

Two mechanisms compound. First, surface area: the extraction kinetics scale with grind surface, so finer grounds hand over their fast and slow solubles more quickly per second of contact. Second, and larger here, permeability: bed permeability follows a Kozeny–Carman shape that scales with the square of grind size (k ∝ d²), and the finest particles migrate down and cake against the paper. Going from 850 µm to 600 µm roughly halves the bed's permeability, and the drawdown stretches from 147 s to 176 s. That extra contact time is what carries the medium cup from close-to-level at two minutes to 18.2 % EY at the finish — enough into the late-arriving sugars to develop real sweetness, with only a modest bitterness cost. The free-draining coarse bed stops at 15.5 %: clean, but before the sweetness fully arrives. See the methodology for the Darcy/Kozeny–Carman and two-timescale extraction models this rests on.

Caveats

This is a simulator, not a measurement of your grinder and your beans. Read the numbers as a well-reasoned direction, not a lab result.

The taste axes are calibrated to move the right way by a plausible amount, not validated against a trained sensory panel (see the calibration note).
Real grinders produce a whole particle distribution with a fines tail specific to the burr; the model uses a median plus a fixed spread and a three-class fine/medium/coarse split, not your grinder's true curve.
The clog model is a tuned fines-caking throttle on the paper, not a resolved particle-by-particle filter — it captures the direction and rough magnitude of the slow-down, not an exact seconds figure for a specific paper.
Channeling or paper-wall bypass — a common real-world cause of uneven fine extraction — is not modeled.

Run it yourself

Load this exact A/B in the live simulator and slice the rig open to watch the fine bed clog and the drawdown stall:

Run medium vs coarse live →Get Early Access →

Related experiments: Light vs dark roast 1:14 vs 1:17 ratio 85 °C vs 96 °C

Methodology, validation & limitations →