N776R — G3X log review22 Aug 2026

79 flights · 7 Feb – 21 Aug 2026 · engine 344.5 → 407 hrs · airframe 1182.8 hrs
Prepared for the condition inspection starting 31 Aug 2026

Flights reviewed

+3 since 17 Aug

CHT #4 fault rate

79%

26 of 33 flights since 29 Jun

Oil-temp excursions

in 28 flights since 3 Jul

Peaks over 450 °F

in 30 flights since 30 Jun

Squawks

left sender cleared this pass

Squawk list for 31 Aug

Two changes since 17 Aug, both from re-measuring rather than from anything the aeroplane did: the left fuel sender comes off the list, and the autopilot watch item is dropped. Nothing was changed on the aircraft between assessments — oil added and fuel as needed.

CHT #4 probe — replace, don't just re-seatPresent on 79% of flights since 29 Jun, and reading about 7 °F low on average even between dropouts. Probe + retaining spring, adapter torque, lead chafe, GEA pins.
Right fuel senderTypical-flight noise up from 0.14% to 2.30% since spring and holding there, with excursions to 10.8% and 5 gal. The left sender was on this list as a suspected twin; it is now cleared on unmasked data and does not need pulling.
Cylinder 3 coolingStill the hottest cylinder on 72 of 79 flights. Inter-cylinder baffles, #3 baffle seals, air ramp.
Oil cooler circuitVernatherm seat, cooler core, ducting. Quiet for 28 flights since 3 Jul — do it because the cowl is off, not because it is pressing.

1CHT #4 probe — persistent, and not resolving

The 13 Aug detector gated on RPM > 1500, which threw away taxi and idle — and that is where 60% of the fault time lives. Re-run at RPM > 900 with a −40 °F threshold, the same logs tell a much bigger story.

	13 Aug detector	Corrected detector
Flights affected since 29 Jun	13 of 33	26 of 33
Episodes	17	63
Flagged time	~44 min	175 min

Flagged seconds per flight

Each column is one flight, ordered by engine hour. Nothing before 29 Jun; near-continuous after.

Before 29 Jun (46 flights, 1 flagged) 29 Jun onward (33 flights, 26 flagged)

The onset date survives the looser threshold. The 46 flights from 7 Feb to 29 May show zero probe-signature episodes even at −40 °F. The deepest deviation anywhere pre-onset is −43 °F on 7 Feb — and there EGT #4 was also 165 °F below the pack, i.e. a genuinely cool cylinder in cruise, not the probe. The fault starts on the first flight after the month-long layup.

The three new legs are more of the same. 20 Aug outbound was completely clean (deepest deviation −37 °F) — the seventh clean flight since onset. 20 Aug inbound ran two episodes totalling 449 s with a worst deviation of −146 °F, the deepest since 11 July. The 21 Aug night circuits produced five separate episodes, 360 s in total. EGT #4 stayed normal on all of them (+0 to +19 °F). Prevalence has held near 80% of flights for eight weeks with no sign of settling.

16 Aug, holding short at KMLB — the clearest single episode

35 s at 1380 RPM. CHT #4 sheds 106 °F while cylinders 1–3 climb; EGT #4 never moves. Takeoff power at t=25 s and the channel snaps back within 4 s.

Cylinder 4 Cylinders 1–3

A cylinder head cannot shed 106 °F in 35 seconds at idle and recover it in 4. The EGT panel is the control: combustion in that cylinder is unchanged throughout. This is the instrument.

A standing low bias, separate from the dropouts

Mean CHT #4 minus mean of cylinders 1–3, level cruise only. Above the line = #4 reads warm, below = #4 reads cold. It wanders flight to flight, but the centre of the distribution has moved.

#4 reads warmer than 1–3 #4 reads colder than 1–3

Correcting the 17 Aug wording here. That pass said the offset "now sits at −7 to −15 °F," which was three flights read as a trend. With 79 flights the honest statement is a mean shift of −7.7 °F (pre-onset +0.5 °F, n = 43; post-onset −7.2 °F, n = 32; t = 3.7) around which individual flights scatter widely — the last five read −14.9, −8.0, −6.3, −3.6 and −1.3. The shift is real; a fixed offset it is not.

The useful control is the seven post-onset flights with zero detected episodes. If the offset were only dropouts bleeding into the cruise window, those flights should sit near zero. They average −6.5 °F, against −7.4 °F for flights that did have episodes. So the channel reads low even when it is behaving — which still argues for a degrading junction rather than a seating problem, and so for replacement rather than a re-seat.

2Oil temperature — three excursions, quiet since

Peaks of 230 °F (29 Jun), 220 °F (30 Jun) and 228 °F (3 Jul) — the three highest in the dataset, against an ordinary range of 176–218 °F. Nothing in the 28 flights since. The shape of those three is what tells you where to look.

Cruise oil temperature, normalized for OAT and CHT

Residual after regressing cruise oil temp on outside air temp and average CHT — i.e. how hot the oil ran relative to how hot it should have run that day. Zero is the fleet baseline.

Hotter than expected Cooler than expected

These are isolated spikes on an otherwise flat baseline. Exactly one flight — 3 Jul morning — sits far off the line at +28 °F. Every other flight in the window is inside the ordinary ±6 °F scatter, and the baseline itself has not moved: mean residual is −0.4 °F for February to May (n = 43) and −0.9 °F for 10 July to 21 August (n = 28). Nothing is drifting. Whatever caused those three flights switched on and then off again.

3 July flew two legs, and only one of them ran hot

Morning and afternoon of the same day, about four hours apart. The morning leg climbs until it pins at 228 °F and holds for 16 minutes. The afternoon leg settles at 204 °F and stays there.

3 Jul morning — KCLW 3 Jul afternoon — X04

A 24 °F difference in settled oil temperature between two legs of the same day points at something that changes flight to flight, not at a condition that builds up. The morning leg also never stabilizes at a lower plateau — it climbs monotonically until it reaches 228 °F and parks there, which is what a bypassed cooler looks like. Oil pressure on that flight was normal (+0.9 psi against a 1.9 psi fleet spread), so supply and flow were fine; the problem is downstream, in the cooler circuit. That reads as a thermostatic valve that stuck on one leg and seated on the next.

It has been quiet since: 28 flights since 3 July, peaks 189–209 °F. The 21 Aug night circuits peaked at 206 °F, the warmest of the three new legs and entirely normal for the conditions (29 °C at the surface). An intermittent valve that has not stuck in 21 engine hours is exactly what you would expect to see, which is why this stays on the list but at the bottom of it — worth doing while the cowl is off, not worth chasing on its own.

3Cylinder 3 — still hottest, but quiet

Cylinder 3 holds the flight's peak CHT on 72 of 79 flights. Nothing has crossed 450 °F since 30 Jun — 30 flights, max peak 445 °F. The 21 Aug night circuits peaked at 435 °F, the highest of the three new legs and unremarkable for pattern work.

Peak CHT per flight

Highest CHT on any cylinder, in flight. The line marks the 450 °F caution threshold.

Peak CHT ⚠ Caution exceeded (labelled)

A correction to the 13 Aug pass. Peak CHT has crossed 450 °F on four flights, not three: 15 Feb (455 °F, cylinder 2, 47 s), 16 May (454 °F, cyl 3, 30 s), 29 Jun (470 °F, cyl 3, 60 s) and 30 Jun (466 °F, cyl 3). The 13 Aug pass missed the February one because it sat on cylinder 2 rather than 3 — which is mildly interesting on its own: the hot-cylinder story is not exclusively #3.

Do not read the two 16 Aug legs (peaks 405 and 401 °F) as improvement — both cruised at 4700–4800 ft rather than the usual 2200 ft, so the cooling air was colder. EGT 3 remains mid-pack, with no sick-cylinder indicators. Keep the baffle inspection.

4Fuel senders — right confirmed, left cleared

Percentage of level-cruise samples sitting ≥2 gal away from the channel's own 61-sample rolling median. This metric had a bug, and fixing it changes the conclusion.

Both channels rail at 18 gal, and a tank sitting on that rail bounces against it — which the old metric counted as sender noise. On the 10 July flight that produced the left tank's headline 18.7% figure, 62% of the flagged samples had the tank at ≥17.5 gal. That was the rail, not the sender. Excluding every sample with the tank above 17 gal — rather than merely reporting how often it was full — gives a clean read on both channels.

Sender noise with rail samples excluded

Upper panel is the typical flight in each era (median); lower panel is that era's worst single flight. 73 flights have enough non-rail cruise data to measure. Note the two panels use different scales.

Left tank Right tank

Right — real, and holding. The typical flight went from 0.14% in Feb–May to 1.60% in Jun–Jul and 2.30% in August, a sixteenfold rise in the median that has now persisted across three months. Worst flights reach 10.8% with 5 gal deviations. This is a genuine degrading sender.

Left — cleared. The typical flight went the other way: 0.28% → 0.09% → 0.00%, with an August maximum of 0.36% and a worst deviation of 2 gal across ten unmasked flights. Its apparent spring "spikes" were rail artifacts. The 20 Aug outbound leg is the cleanest test in the dataset — neither tank near the rail at any point in cruise — and both channels read 0.36% and 0.12%.

The left sender comes off the squawk list. It was merged with the right on 17 Aug on the reasoning that its low numbers were unproven rather than clean; with the rail excluded they are now proven clean, over ten flights and three months. Squawk the right sender alone and save the access work on the left wing.

5Clean over 60 engine hours

Re-verified across all 79 flights.

Parameter	Result	Read
Oil pressure, normalized	+0.09 psi / 25 h · r = 0.03 · mean 69.4	Flat. Last three 69.0 / 65.1 / 70.9
Oil temp baseline	raw +6.2 °F / 25 h · r = 0.51	Seasonal OAT, not the engine
Electrical	14.18 V mean · +0.019 V / 25 h · r = 0.29	No engine-running low-voltage events
Mag / runup	160–220 °F EGT rise, all four	No dead or lazy plug signature
Combustion stability	FF jitter 0.13–0.19 gph · EGT jitter 1.2–2.0 °F	No cylinder standing out
Cruise efficiency	+0.11 gph / 25 h · r = 0.09	Scatter. No power loss
Fuel pressure	mean 5.14 psi · −0.38 psi / 25 h · r = −0.35	Last three 4.25 / 4.09 / 5.50 — no continued fall
Carbon monoxide	0 ppm, sensor reporting ~100%	Clean
Autopilot pitch torque	+0.09% / 25 h · r = 0.12 · 35 flights	Watch item dropped — see below

The autopilot watch item was a measurement artifact, and it is dropped. The "2.2 → 3.5% creep" reported on 13 Aug averaged pitch torque across every phase of flight, so it moved with how much climbing and altitude-capturing each flight happened to contain. Measured consistently — autopilot engaged, level cruise only — it is +0.09% per 25 hours at r = 0.12 across the 35 flights with enough AP-engaged cruise to measure. Individual flights run 2.6% to 5.3%, and the highest of them was 10 April, near the start of the dataset. There is no trend. Nothing to look at on the pitch servo or trim.

Method notes for re-running

Parse with skiprows=1, drop the short-name row, then address columns by the G3X short name (E1 CHT4, FQty1) — the long headers change between log versions, the short names do not.
CHT #4 detector (corrected): RPM > 900 and CHT1 > 250 °F; flag where CHT4 − mean(CHT1..3) < −40 °F; merge runs separated by <15 s; keep merged runs ≥5 s. The old RPM > 1500 / −80 °F gate under-reports by roughly 5×.
Bias metric: mean of CHT4 − mean(CHT1..3) over level cruise (RPM>2000, IAS>60, |VS|<200, |roll|<4°).
Sender noise (corrected 22 Aug): level cruise as above; drop every sample with the tank ≥17 gal before computing anything, then take the % of remaining samples ≥2 gal from a 61-sample centered rolling median. Require ≥120 usable samples or report nothing for that flight. Reporting the full-tank % alongside an uncorrected figure — the old approach — is not enough: a tank resting on the 18 gal rail bounces against it and reads as sender noise, which is what produced the left tank's phantom spikes.
Oil pressure normalization: regress on oil temp (−0.184 psi/°F) and RPM (+0.0105 psi/rpm), ref 190 °F / 2380 rpm. Oil temp normalization: regress on OAT and average CHT, residual against the fleet fit.
Autopilot torque: mean |AP Pitch Torque| restricted to autopilot-engaged level cruise, ≥60 samples. Averaging across all phases mixes in climbs and captures and manufactures a trend that is not there.
Flight filter: airborne (IAS > 60 kt) for more than 300 s. The 13 Aug pass counted 67 using a ~15-minute floor.

Source: g3x_data_log/log_*.csv, 118 files in range, 79 qualifying flights. Assessed 22 Aug 2026. No maintenance performed between the 17 Aug and 22 Aug assessments — oil added and fuel as needed.