Why do I lose GPS lock?

The symptom

Pick your flavour. The satellite count crawls to four and stops. You get a lock on the bench, then it evaporates the instant the motors spin. It holds beautifully on the ground and quietly rots in flight. Or you have a solid fix, engage position hold, and the aircraft draws slow circles in the sky, drifting round a point it never quite arrives at — the toilet bowl.

They look like four different problems. Mostly they are one problem.

Read this before you buy another module

A GPS module is a radio receiver listening for signals from space that arrive at roughly the noise floor. The satellites are 20,000 km away and transmitting on a modest power budget; by the time their signal reaches your aircraft it is extraordinarily weak. Meanwhile, sitting a few centimetres away, you have a video transmitter, four ESCs switching tens of amps at tens of kilohertz, and a metre of unshielded power wiring acting as an antenna.

Everything in your build radiates broadband RF. All of it is louder than the satellites.

So the overwhelming majority of "my GPS is bad" is not a bad GPS. It is your own aircraft drowning out the sky. Once you accept that, the whole diagnosis reorders itself: you stop hunting for firmware settings and start hunting for the noise source.

Check this in 60 seconds

1. Power the aircraft outdoors, under open sky, and just leave it alone for five minutes. Do not touch it. Watch the satellite count climb. This is your clean baseline.
2. Now turn the VTX on at full power (or if it is always on, turn it off and repeat step 1). Does the count collapse, or hold?
3. Spin the motors up to a hover throttle, props off, aircraft restrained. Does the count collapse then?

Whichever step kills the count is your fault. If nothing kills it, you have a placement or a patience problem, not an interference one.

1. VTX interference — start here

This is the single most common cause and it is not close.

The GPS L1 civilian signal sits at 1575.42 MHz. Your video transmitter's fundamental is elsewhere, but transmitters are not clean, and harmonics and spurious emissions from certain bands and channels land unpleasantly close to L1. Add a VTX mounted 30 mm from the GPS module, radiating at several hundred milliwatts, and the receiver is trying to hear a whisper next to a leaf blower.

The classic tell: the satellite count is perfect on the bench with the VTX off or at 25 mW, and collapses the moment the VTX is transmitting at power. If your fix disappears when you switch from pit mode to full power, you have found it in one move.

Fixes, in order of how much they buy you:

• Physical separation. Move the GPS away from the VTX and its antenna. Centimetres matter here.
• Drop VTX power and see whether the count recovers. This is a diagnostic first, a fix second — but a fix that works.
• Try a different channel or band. Some are markedly worse than others for L1.
• Shield the VTX — copper tape, grounded properly, is crude and effective.

2. Placement — the module needs sky

A GPS antenna is designed to look up. Buried in the middle of the stack, tucked between the battery and the ESC, mounted flat and low, it is looking at your own electronics.

Carbon fibre is conductive. It is not a neutral material to a radio; it blocks the signal. A GPS mounted under a carbon top plate, or with carbon between it and the sky, is deliberately handicapped. On a heavy platform with big carbon plates and thick arms, this is very easy to do without noticing.

What good placement looks like:

• Up on a mast, above the electronics, above the carbon, with an unobstructed view upward.
• Ground plane underneath it — a small metal plate under the module both improves the antenna's pattern and shields it from the noise coming up from the stack below.
• As far from the VTX, ESCs and power wiring as the airframe allows.

Yes, a mast looks fragile and yes, it is one more thing to break in a crash. It also works, and nothing else does quite as well.

3. ESC and power-wiring noise

Switching regulators and ESCs are broadband noise generators by construction — which is what the capacitor on your ESC's power pads is there to absorb. Unshielded, unfiltered battery leads and motor wires radiate that noise, and if your GPS cable runs alongside them it will pick it up directly.

• Route the GPS cable away from power wiring. Do not zip-tie it to the battery lead because it was tidy.
• Use shielded cable for the GPS, with the shield grounded at one end.
• Cross power wires at right angles, not in parallel, where you must cross them at all.

4. "It only happens when the motors spin"

This one narrows things down usefully. If the fix survives everything until you spool up, you are looking at ESC noise (electrical, appears with motor current) or a physical problem — a module or connector shaken loose, a marginal solder joint, a cable chafing. It is not the sky. The sky did not change when you pushed the throttle.

Distinguish the two by spinning the motors slowly with props off: if the count collapses with almost no vibration but real current draw, it is electrical.

5. Cold start — wait longer than you think

A module with no valid almanac, in a new location, genuinely takes minutes to produce its first fix. It has to find the satellites from nothing.

A fix in 30 seconds is a module with recent ephemeris still in memory. A truly cold module can take several minutes. New module, first flight, or a location hundreds of kilometres from where it last flew — power it up under open sky and go and do something else. Impatience here gets misdiagnosed as hardware failure constantly.

6. The FC and the module never really talk

Less common, but trivially checkable and it costs you an afternoon if you miss it: wrong UART assigned, wrong baud rate, wrong protocol selected for the module you actually have. The FC reports no satellites because it is not receiving anything intelligible, not because the module cannot see anything.

If the count is a permanent, unmoving zero and the module's own status LED is happily blinking, this is where to look.

7. Compass interference — the toilet bowl

Adjacent problem, worth its own paragraph, because it produces a distinct and alarming symptom.

If your aircraft has a magnetometer, it is even more sensitive to your build than the GPS is. A compass mounted near power leads reads the magnetic field of the current flowing through them, so your heading estimate moves with throttle. The position controller then thinks the aircraft is pointing somewhere it is not, corrects in the wrong direction, and the aircraft circles the target instead of holding it. That is the toilet bowl, and it is a heading fault, not a GPS fault.

On heavy platforms this is a serious problem, because the currents are large. Same fixes: get the compass up on the mast, away from the power path, and calibrate it on the aircraft, in its final position, with everything mounted.

A fix is not the same as a good fix

The most dangerous state is not "no fix." It is a marginal fix that the aircraft is willing to arm on.

Look at satellite count and HDOP, not just at whether the flight controller says it has a lock. Six satellites with a poor HDOP is a position estimate that can be tens of metres wrong and can wander. Arm on that, engage a return-to-home, and the aircraft will fly confidently towards a home point that was never where it thought it was. That is how fly-aways start, and it is why the arming check exists.

How to know you actually fixed it

Not "it locked once."

• The count holds under load. VTX at full power, motors at hover throttle, count steady and high — not merely present.
• HDOP is good, and stays good, rather than sliding as the flight goes on.
• The count does not change when you change VTX power. If it still does, you have improved the interference, not removed it.
• It survives a real flight. Check the log afterwards: satellite count over time, not a snapshot on the bench.

Fix the radio environment and the GPS problem stops being a GPS problem. Almost nobody needs a better module. Almost everybody needs six more centimetres of separation and a mast.