Home Roasting8 min read

How to Roast Ethiopia Yirgacheffe at Home: Gentle Filter After 14 Roasts

A data-backed case study of dialing in washed Ethiopia Yirgacheffe: batch size, failed dark F-WSH runs, and the Gentle Filter corridor that worked.

BrewedLate Coffee

#Ethiopia Yirgacheffe#washed Ethiopian coffee#home coffee roasting#Kaffelogic#Gentle Filter#roast profile#origin clarity

The short answer: for washed Ethiopia Yirgacheffe Grade 2, the best home-roast result in this log was Gentle Filter at level 2.4 / 100g — about 11% DTR, ~14% roast loss, smooth and lightly fruity without green. Going darker on F-WSH or NordicLight did not fix underdevelopment.

This is a case study from 14 logged roasts (John Burton, heirloom, ~1,700–2,200m, goal: origin clarity), with klog metrics and tasting notes — not a generic Ethiopian guide.

Ethiopia · Yirgacheffe · Washed · 1,700-2,200 MASL · G2 · John Burton

JB Ethiopia Yirgacheffe Grade 2

Goal: origin clarity16 roasts logged

Aroma of dried red fruits, citrus and vanilla biscuit. Flavours of toasted tea, lemon zest, caramel, spiced cookies and floral notes. Roasted nuts and caramel aftertaste. Low acidity and medium body. Score: 82.5+

Best so far: Bright Filter at level 2.2, 100 g — tasted good 1 of 1 time.

The journey

  • Tasted good
  • Had a fault
  • Not tasted yet

What the roasts show

  1. Roast 1: F-WSH 1500-2200 at level 2.2, 100 g. 1 with a fault (grassy).
  2. Then level 2.2 → 3.3, load 100 g → 120 g: roast 2. 1 with a fault (grassy).
  3. Then level 3.3 → 2.8, profile F-WSH 1500-2200 → NordicLight: roast 3. 1 good.
  4. Then level 2.8 → 3.0: roast 4. 1 with a fault (grassy).
  5. Then level 3.0 → 3.1: roast 5. 1 with a fault (grassy).
  6. Then level 3.1 → 3.3: roast 6. 1 with a fault (grassy).
  7. Then level 3.3 → 2.8, profile NordicLight → Gentle Filter, load 120 g → 100 g: roast 7. 1 with a fault (roasty).
  8. Then level 2.8 → 2.3, profile Gentle Filter → Rich Filter: roast 8. 1 with a fault (grassy, roasty).
  9. Then level 2.3 → 2.2, profile Rich Filter → Bright Filter: roast 9. 1 not tasted.
  10. Then level 2.2 → 2.4, profile Bright Filter → Gentle Filter: roast 10. 1 good.
  11. Then level 2.4 → 2.2: roast 11. 1 good.
  12. Then level 2.2 → 2.1: roast 12. 1 with a fault (flat).
  13. Then level 2.1 → 2.3: roast 13. 1 with a fault (grassy).
  14. Then level 2.3 → 2.2, profile Gentle Filter → Bright Filter: roasts 14–15. 1 good, 1 not tasted.
  15. Then load 100 g → 150 g: roast 16. 1 not tasted.
  • Weight lost: good roasts 12.3%–14.2%, roasts with a fault 12.4%–14.0%.
  • Time in the hot zone: good roasts 1:43–2:43, roasts with a fault 1:46–3:01.
  • End temperature: good roasts 214 °C–223 °C, roasts with a fault 213 °C–222 °C.
  • Good cups were tasted after 8–13 days of rest (4 tastings).
  • The first good cup on Gentle Filter at level 2.4, 100 g was roast 10 of 16.

This shows what differed in the roast. It does not say why the cup tasted different.

Recipes tried

  • Bright Filter

    Settled · 4 roasts · 2:42–2:46

    • 2.2 · 100 g · 3 roasts, Tasted good
    • 2.2 · 150 g · 1 roast, Not tasted yet

  • Gentle Filter

    Moved on · 5 roasts · 2:22–2:34

    • 2.1 · 100 g · 1 roast, Had a fault
    • 2.2 · 100 g · 1 roast, Tasted good
    • 2.3 · 100 g · 1 roast, Had a fault
    • 2.4 · 100 g · 1 roast, Tasted good
    • 2.8 · 100 g · 1 roast, Had a fault

  • Rich Filter

    Moved on · 1 roast · 3:01

    • 2.3 · 100 g · 1 roast, Had a fault

  • NordicLight

    Moved on · 4 roasts · 1:43–1:54

    • 2.8 · 120 g · 1 roast, Tasted good
    • 3.0 · 120 g · 1 roast, Had a fault
    • 3.1 · 120 g · 1 roast, Had a fault
    • 3.3 · 120 g · 1 roast, Had a fault

  • F-WSH 1500-2200

    Moved on · 2 roasts · 1:55–2:26

    • 2.2 · 100 g · 1 roast, Had a fault
    • 3.3 · 120 g · 1 roast, Had a fault

The recipe that stuck

Profile
Gentle Filter
Level
2.4
Batch
100g
13.6%
2:28
End temperature
222°C
Winning batch
R-090 · 2026-07-09
This is not bad, it's smooth, it's nice, it's a little bit fruity… better than the 91 roast.

Compare two recipes

Comparewith
  • Hot zone about the same (2:46 vs 2:44)
  • 4.1 points less weight lost (10.0% vs 14.1%)
  • End temperature about the same (223 °C vs 223 °C)

This shows what differed in the roast. It does not say why the cup tasted different.

Numbers are averages over each recipe's roasts.

What the roasts taught

  • 60g F-WSH loads stayed green — thermal mass mattered as much as level.
  • Cranking F-WSH / NordicLight darker (3.0–3.3) did not fix underdevelopment; it often stayed grassy or went bland.
  • Gentle Filter at ~2.4 / 100g first delivered smooth fruit without green (R-090) — better than the Bright and Rich Filter profiles at similar levels.
  • Still a work in progress: fruit is there but bloom still a touch dark; R-096 at 2.2 is the lighter fine-tune.

Where the good cups live

Roast level against batch size. Colour is the profile; a larger dot is a roast that tasted good.

Roast mechanics across the roasts

Roast loss, hot zone and end temperature for each logged roast. Each line has its own scale, so compare shapes, not heights.

Roast goal

Roast light to medium-light (ending around 400-410°F / 204-210°C) with a gentle approach. Drop just after first crack completes to preserve the delicate tea, floral, and citrus notes. Avoid pushing into second crack or developing heavy caramelization. Aim for a bright, clean cup that showcases the Yirg character.

Start your own roast log

The counts in this guide are from when it was written. Roasts added since then are in the Ethiopia Yirgacheffe roast log, which lists every attempt and how it tasted.

What is this Yirgacheffe lot?

Yirgacheffe is one of Ethiopia’s signature washed origins — tea-like, citrus, florals when roasted for clarity. This Grade 2 lot aimed at origin clarity: preserve delicate notes, avoid heavy caramelization or second-crack territory.

Supplier cup notes pointed at dried red fruit, citrus, vanilla biscuit, toasted tea, lemon zest, and light florals (~82.5+).

The journey

1. Small loads and green cups (early F-WSH)

Early F-WSH 1500-2200 at 60g tasted very green. Notes blamed both profile and underload. Bumping level on small batches did not magically produce fruit.

2. Darker is not always more developed

Pushing F-WSH and NordicLight into 3.0–3.3 / 120g was a natural instinct (“more development”). Cups often stayed grassy after rest, or went subtle/sour without clear sweetness. One NordicLight run showed fruit but still felt underdeveloped; later higher levels were “still terrible” weeks later.

Lesson: for this bean, underdevelopment was not a simple “roast darker” problem.

3. Gentle Filter unlocks clean fruit

Gentle Filter @ 2.4 / 100g (R-090) was the first note that sounded like a keep recipe: smooth, nice, a little fruity, no green — better than a parallel Bright Filter roast that still wanted more fruit. A later 2.2 batch is the lighter fine-tune direction.

Dialed-in recipe (current best)

SettingValue
ProfileGentle Filter
Level2.4 (try 2.2 if bloom looks dark)
Batch size100g
DTR~10.8%
Roast loss~13.6%
ROR pre-FC / dev~31 / ~8 °C/min
Rest7–10 days

Interactive charts, full attempt list, and tasting trail:

Open the Yirgacheffe case study →

What not to do

  1. Don’t underload standard F-Series profiles at 60g and blame the green coffee.
  2. Don’t keep raising level into the 3.x band hoping grass will turn into fruit.
  3. Don’t cup washed Ethiopia at day 3–5 and call the profile dead.
  4. Don’t assume the profile that won on Kenya Gatugi (F-WSH 1500-2200) auto-wins here.

Related reading

Track every roast, improve every batch

BrewedLate helps home roasters log green coffee inventory, roast batches, Kaffelogic curves, and get AI recommendations — all in one web-based tool.

Try the Free Roast TrackerBrowse Kaffelogic Profiles

How to roast Ethiopia Yirgacheffe on a Kaffelogic Nano

Home-roast recipe from a 14-attempt log for washed Yirgacheffe Grade 2

1

Load 100g green

Use a full 100g load. 60g F-WSH batches stayed green in this log.

2

Select Gentle Filter at level 2.4

Gentle Filter is a light, delicate filter profile; 2.4 preserved fruit without the green notes that plagued darker F-WSH runs.

3

Check metrics

Aim near ~11% DTR and ~13–14% roast loss if ambient and bean moisture allow.

4

Rest 7–10 days

Judge after full rest before dropping level further (e.g. 2.2 fine-tune).

5

Taste for smooth fruit, not green

Target toasted tea, citrus, light florals — not grassy underdevelopment or flat dark roast.

Frequently Asked Questions

What is the best roast profile for Ethiopia Yirgacheffe on a Kaffelogic?
In this 14-roast log of washed Yirgacheffe Grade 2, Gentle Filter at level 2.4 with a 100g batch delivered the best cup — smooth, lightly fruity, and free of green notes. Bright Filter and Rich Filter at similar levels were less fruity or a touch dark on the bloom.
Why did my Yirgacheffe taste green or grassy after roasting?
This log saw persistent green notes on small 60g F-WSH loads and on darker F-WSH/NordicLight runs (levels 3.0–3.3). Underdevelopment was not fixed by simply going darker; thermal mass and profile family mattered more.
What DTR and roast loss work for washed Yirgacheffe?
The best batch landed around 10.8% DTR and ~13.6% roast loss, with pre-FC ROR near 31 °C/min and development ROR near 8 °C/min. Use these as a corridor, not a rigid target.
Should I use NordicLight for Yirgacheffe?
NordicLight at 2.8–3.3 / 120g sometimes showed fruit but often stayed underdeveloped or bland after rest. Gentle Filter was cleaner for this lot’s origin-clarity goal.
How long should washed Yirgacheffe rest?
Plan on 7–10 days. Several earlier batches still tasted grassy at five days and only clarified with longer rest — or never fully recovered if the roast was underdeveloped.

Related Articles

How to Roast Ethiopia Sidamo at Home: Gentle Filter Dialed-In Recipe

Data-backed home roast case study for Ethiopia Sidamo ETH04: profiles tried, rest lessons, and the Gentle Filter corridor.

Home RoastingRead more →

How to Roast Kenya Gatugi Peaberry: 17 Roasts to a Dialed-In Cup

A data-backed case study of dialing in washed Kenya Gatugi peaberry: batch size as the hidden variable, profile experiments, and the final roast metrics that stuck.

Home RoastingRead more →

How to Roast Colombia ASOSPAC: Bright Filter for Fruit + Chocolate

Data-backed case study for Colombia ASOSPAC COL03: profile experiments toward a fruity-bright goal with real klog metrics.

Home RoastingRead more →

How to Roast Swiss Water Decaf Brazil: F-NAT 0-1200 at 3.4

Data-backed case study for natural-process Swiss Water Decaf Brazil: medium chocolate-nutty goal and the F-NAT corridor that stuck.

Home RoastingRead more →