Electronic – GPS parity decoding, how to get D29 and D30 of previous bits

A full size patch antenna (normally about 25 × 25 × 4 mm) will typically have a gain of around 5dB and for an SMA connector pair insertion loss for even fairly cheap connectors is likely to be under 0.5dB at 1.5GHz. For coax RG-174 is a pretty popular (and cheap) cable often used on consumer grade GPS antennae. Running the numbers through this Attenuation and power handling calculator over that short distance the loss at GPS frequencies is only 0.3db. So you should still have a net gain of over 4dB.

Most modern receivers will work well with that and for example the Fastrax IT520 documentation recommends a loss of less than 1dB when using a passive antenna which the above combination achieves. In a vehicle application sometimes the effect of too much gain can be mixed, you can for example start picking up weak reflected signals that can cause multipathing effects. Other than extended testing with a particular module there's not really any way to predict that reliably.

For ESD protection I don't really see an issue if you're using an encapsulated antenna and the SMA connectors are normally always mated. But a little extra ESD protection isn't a bad thing and while the low capacitance protection diodes made for that purpose are small they do normally meet all the usual ESD standards.

So in general I'd say give the external passive antenna a shot and compare against an active in some real-world tests. Some of the cheaper antennae are actually quite good, so grab a few of each from a variety of manufacturers and unlike reworking a PCB design they can easily be compared. It sounds like you've talked yourself out of an on-board antenna for other valid reasons, but I don't see signal loss to be significant enough over that short distance to worry about that side of things.

Let us break up the problem into four parts: GPS, Barometer, the Naza, and terrain.

1. GPS:

   The altitude error of GPS is known to be around 1.5x horizontal error, due to the simple fact of geometry: In order to obtain higher accuracy, the angular spread in altitude of the GPS satellites from point of observation (the sensor) needs to be high. The constraint on this desired spread is due to signal attenuation and signal path distortion when a satellite is close to the horizon.

   For civilian GPS (such as the Naza), the best case scenario for altitude accuracy turns out to be around +/- 25 meters, for 95% confidence. This holds true only when the GPS antenna has unobstructed view of the sky at all times, and a number of GPS satellites (at least 4, preferably 6 or more) are accessible throughout the flight duration. Also, this height is relative to the center of rotation of the respective satellites - for a reasonable approximation, take it to be relative to the center of the earth. There is no correspondence to height above the actual ground wherever you might be.

   This 50 meter spread is pretty much unusable for low-level flight. Even if this variance were acceptable, for up to 5% of sensor samples the variance would exceed this, ending up with the hexacopter trying to embed itself into the ground at some point.

   On the bright side, this will happen just once, since the copter would need to be replaced or rebuilt after that.

2. Barometer:

   A typical low-cost MEMS barometer promises +/- 0.8 to 3.5 meter accuracy, with the Naza claiming +/- 0.8 meters. This sounds great in theory, but in practical use, air pressure variations due to weather activity, and even more so on mountainous terrain due to rapidly moving pressure vortexes, reduces the actual accuracy to perhaps +/- 10 meters. The barometric height reading is based on air pressure, and thus relative to an arbitrary "mean sea level", varying widely with air pressure conditions. Again, no direct correlation to the actual height of the ground wherever you might be.

   This is still significantly better than what GPS alone can deliver. For programmed flight, this would suffice if the copter were to be consistently flown 20 to 30 meters above the height of any ground-based obstructions, in level flight. By introducing ground contour variations, all bets are off.

3. The Naza:

   This is where things get better. The combination of barometer and GPS in the Naza controller provides checks and measures to reject outlier samples and jitter from one or the other of those sensors. In addition, the integrated accelerometer provides a further set of inputs for the damping algorithm of the controller. In tandem, and with a fairly conservative damping setting in place, the hexacopter should behave tolerably well, with effective vertical accuracy of a couple of meters in stable weather.

   Gusty wind conditions, and especially low-pressure pockets during impending storms, would reduce the accuracy to perhaps +/- 10 meters in the real world. Further, calibration of the sensors would be needed each flying day, perhaps even once every few hours, to keep the altitude readings meaningful.

4. Terrain:

   Unlike flying over flat terrain, flying over ski slopes can make things rather interesting. For one thing, the terrain contours will necessarily be sloping, so the flight path design needs to take that into account, that too with an accuracy of a few meters. Further, the tops of trees and jutting rocks need to be taken into account.

   In addition, signal distorting factors such as water bodies (reflection of GPS signal causing false readings) and metal cables of cable cars or ski lifts, will adversely impact the accuracy of the GPS.

   Finally, the obstruction of line-of-sight from GPS antenna to individual satellites, due to mountain tops, trees, or rocks, will cause wide variation in GPS accuracy over the duration of a flight.

The solution:

Fly by eye. Control the hexacopter manually, do not depend on GPS or barometric settings. Also keep in mind that if some form of sensor dump takes place during a flight, and the hexacopter drops out of the sky, one must ensure that it does not endanger any skiers.

Ideally, one would fit the hexacopter with a camera with a medium telephoto lens - not the ultra-wide of the GoPro cameras, for example - and fly it at well over 20 meters above local terrain contour at all times. Also, one would ensure that the flight path, and especially a fall trajectory in case of failure, would be far removed from the route taken by skiers.