140cm2 surf foil stabilizer
I wanted a replacement stabilizer for a specific rig — a KT Atlas 790 front wing on a 70 cm fuselage, currently running a Nomad 140 tail, 80 kg rider, prone. Three goals that pull against each other: unlock roll, maximise pitch stability, minimise drag.
Rather than pick a tail off a spec sheet, I sized it against the speeds I actually ride. Every sweep I ran crossed over inside that range, which means there is no tail that is best across it — and that is the actual finding here.
(Disclosure: foil.tools is my own tool. The Atlas 790 and Nomad 140 are the gear I happen to ride; nothing here is an endorsement or a test of those products.)
Get the speed band from the water, not from a guess
The design point came from a 40-minute Columbia River Gorge downwinder on 20 June 2026 — 2,370 GPS points, 10.4 km, a genuine one-way run (+6.97 km net east, 95% of moving time eastbound).
The run tracks east, which is upstream. USGS gauge 14105700 at The Dalles read 156–159k cfs during the session, worth roughly +0.5 kt on speed through water, applied as a per-sample vector subtraction. If you size off raw GPS speed in a current, you are optimising for the wrong number.
That correction is an estimate: turning discharge into a local velocity needs a channel cross-section I do not have, and The Dalles is an upstream proxy for Hood River. I ran the whole study across 0.18–0.60 m/s of assumed current. The band moves the absolute drag by ±0.15 m/s worth and no ranking changes.
The measured distribution came out at p10 9.7 kt, p50 12.2 kt, p90 15.8 kt through water. Those three are the dashed lines on every chart below. Each sweep is scored by weighting drag across that distribution rather than reading it at a single design point.
Sweep one — aspect ratio
Six aspect ratios at a fixed 140 cm². Five are plotted; the ramp separates five steps cleanly and AR 6 sits between AR 4 and AR 8 without adding to the shape. All six are in the weighted table below. Dashed lines are the 10th, 50th and 90th percentiles of the measured speed-through-water distribution.
data
| speed through water (kt) | AR 4 | AR 8 | AR 10 | AR 12 | AR 14 |
|---|---|---|---|---|---|
| 8.75 | 4.15 | 2.96 | 2.74 | 2.60 | 2.50 |
| 9.72 | 3.61 | 2.72 | 2.57 | 2.49 | 2.44 |
| 10.69 | 3.37 | 2.75 | 2.67 | 2.65 | 2.64 |
| 11.66 | 3.28 | 2.87 | 2.85 | 2.86 | 2.89 |
| 12.63 | 3.28 | 3.04 | 3.06 | 3.11 | 3.16 |
| 13.61 | 3.34 | 3.22 | 3.29 | 3.37 | 3.46 |
| 14.58 | 3.44 | 3.44 | 3.54 | 3.65 | 3.76 |
| 15.55 | 3.58 | 3.68 | 3.81 | 3.93 | 4.06 |
| 16.52 | 3.74 | 3.92 | 4.07 | 4.24 | 4.39 |
| 17.50 | 3.95 | 4.18 | 4.37 | 4.55 | 4.73 |
| 18.47 | 4.17 | 4.48 | 4.68 | 4.88 | 5.07 |
| aspect ratio | span (cm) | weighted drag (N) | vs. best |
|---|---|---|---|
| AR 4 | 24 | 3.477 | +11.4% |
| AR 6 | 29 | 3.196 | +2.4% |
| AR 8 | 33 | 3.123 | +0.0% |
| AR 10 | 37 | 3.122 | +0.0% |
| AR 12 | 41 | 3.153 | +1.0% |
| AR 14 | 44 | 3.196 | +2.4% |
The curves cross at about 12 kt. Below that, high aspect ratio wins on induced drag; above it, the smaller chord's Reynolds penalty flips the order.
Weighted over the real distribution the optimum is AR 8–10, and the basin is flat from AR 8 to AR 12 — inside 1%. That flatness is free design freedom: AR 8 is 4 cm narrower in span than AR 10 at identical drag, and span is what resists roll. If you want the tail to let the board roll, buy that with span you were not paying for anyway.
Sweep two — area
Six planform areas at AR 8. Five plotted, 120 cm² omitted from the chart and kept in the table. Note the spacing: the curves bunch on the small side and spread on the large side.
data
| speed through water (kt) | 100 cm2 | 140 cm2 | 160 cm2 | 180 cm2 | 200 cm2 |
|---|---|---|---|---|---|
| 8.75 | 3.91 | 2.96 | 2.76 | 2.71 | 2.73 |
| 9.72 | 3.29 | 2.72 | 2.73 | 2.79 | 2.87 |
| 10.69 | 2.83 | 2.75 | 2.84 | 2.94 | 3.06 |
| 11.66 | 2.72 | 2.87 | 3.01 | 3.15 | 3.29 |
| 12.63 | 2.75 | 3.04 | 3.20 | 3.38 | 3.57 |
| 13.61 | 2.85 | 3.22 | 3.44 | 3.65 | 3.86 |
| 14.58 | 2.98 | 3.44 | 3.69 | 3.93 | 4.16 |
| 15.55 | 3.13 | 3.68 | 3.95 | 4.22 | 4.51 |
| 16.52 | 3.30 | 3.92 | 4.23 | 4.56 | 4.88 |
| 17.50 | 3.49 | 4.18 | 4.55 | 4.91 | 5.26 |
| 18.47 | 3.69 | 4.48 | 4.88 | 5.27 | 5.66 |
| area (cm²) | span (cm) | weighted drag (N) | vs. 140 cm² |
|---|---|---|---|
| 100 | 28 | 2.998 | -4.0% |
| 120 | 31 | 3.018 | -3.4% |
| 140 | 33 | 3.123 | +0.0% |
| 160 | 36 | 3.269 | +4.7% |
| 180 | 38 | 3.434 | +10.0% |
| 200 | 40 | 3.610 | +15.6% |
This is the sweep that changed my answer. Going from 140 down to 100 cm² saves about 4% drag and costs 29% of pitch authority. Going up is punished much harder — the curves spread on the large side and bunch on the small side.
So area should be chosen for stability, not drag. It is nearly free in the direction that matters, and expensive in the direction that does not help you.
Sweep three — section thickness
Four symmetric NACA sections at AR 8, 140 cm². The order at the top of the speed range is not the order at the bottom.
data
| speed through water (kt) | NACA 0006 | NACA 0008 | NACA 0010 | NACA 0012 |
|---|---|---|---|---|
| 8.75 | 3.55 | 2.96 | 2.88 | 2.99 |
| 9.72 | 2.98 | 2.72 | 2.84 | 2.98 |
| 10.69 | 2.70 | 2.75 | 2.90 | 3.06 |
| 11.66 | 2.73 | 2.87 | 3.03 | 3.19 |
| 12.63 | 2.86 | 3.04 | 3.20 | 3.36 |
| 13.61 | 3.03 | 3.22 | 3.39 | 3.57 |
| 14.58 | 3.24 | 3.44 | 3.61 | 3.81 |
| 15.55 | 3.47 | 3.68 | 3.86 | 4.06 |
| 16.52 | 3.70 | 3.92 | 4.13 | 4.36 |
| 17.50 | 3.96 | 4.18 | 4.41 | 4.68 |
| 18.47 | 4.24 | 4.48 | 4.70 | 5.00 |
| section | t/c | weighted drag (N) | vs. best |
|---|---|---|---|
| NACA 0006 | 6.0% | 3.050 | +0.0% |
| NACA 0008 | 8.0% | 3.123 | +2.4% |
| NACA 0010 | 10.0% | 3.271 | +7.2% |
| NACA 0012 | 12.0% | 3.443 | +12.9% |
Thinner wins on average — and inverts at the low end. NACA 0006 is the cheapest section from 12 kt up, and the worst of the four at 8.7 kt, where trim alpha climbs to about 5°. An average would have hidden that completely. You need the distribution.
Where it lands
AR 8, around 140 cm², a 8% section — chosen for a flat drag basin, the narrowest span inside it, and no low-speed inversion. Not because any one of those won a sweep outright, but because none of them lose badly anywhere I actually ride.
What would change this
Stated plainly, because these are the assumptions the numbers rest on:
- Trim load is assumed, not measured. 60 N of tail downforce, about 7% of all-up weight. It scales the drag magnitudes; it barely moves the rankings.
- Current velocity is an estimate, per the cross-section caveat above.
- This is AeroBuildup — strip theory with NeuralFoil sections, not CFD and not a tank test. It is appropriate for a trade study between shapes, and it is not a substitute for measuring the real thing.
- The speed distribution is one session on one river. A different spot, or a different rider, moves the weighting and can move the answer.
None of this required a wind tunnel or a CFD licence. It required a GPS track, a public river gauge, and about twenty analysis runs.