Here3+ Inconsistent Cold Start

I’m using a here 3+ with a cube orange (not +) and ardupilot 4.1.5. I’ve done 100+ successful flights without any manner of symptom or error. However, the cold start of the GPS/drone rarely takes a long time to acquire a home position. I’d like to know if this is expected and normal behavior.

The procedure is:

  1. start with a drone that hasn’t been turned on in 12hr+ to ensure cold start conditions
  2. turn on controller or laptop to see when it acquires home position
  3. turn on the drone and time the acquisition of home position

Obviously step 3 is a composite score of the autopilot booting, gps booting, gps acquisition rate, and autopilot algorithms. Qualitatively, I have seen no indicators that it is autopilot driven.

Typical results are a 80-100 second cold start. Occasionally it will take more - 3 minutes, 5 minutes, or even 10+ minutes (after which I stop timing).

The frequency is wildly inconsistent. At worst (week of Oct 16) it exceeded 2 minutes 4/15 times at 134, 381, 410, and 600+ seconds. I flew after the 410 second acquisition and observed no performance degradation in the air. Those were with AP 4.1.5. I’ve since updated to 4.5.6 and upgraded 1 drone with a here4, and I am 1/9 over 2 minutes, with a 180s start on the here4. Before August, I can only recall this happening 1 time (the drones were manufactured Q2 2024), but I also hadn’t noticed it was happening at all so that may not be reliable information.

Hot starts (any start within a few hours of a cold start) are not affected and are consistently 45-55 seconds. I can collect parameters if needed, but seeing that it works correctly in flight I’m guessing that isn’t the issue. Even if a cold start is slow, I can’t eyeball any degradation in the GPS position’s stability after a hot start.

Lastly, I have experienced this across 3 here3+ and 1 here4. I have controlled my data for the following and seen no correlations:

  • Atmospheric and solar weather
  • Orientation: both drone and GPS right side up and level
  • Environment: the drone elevated 18" off the ground on a plastic bucket, on asphalt or grassy field, with no obstructions in a 45 degree cone pointing upwards from the drone
  • I have avoided any emitter strong than my telemetry radio (RFD900)
  • Geographic location: tested at 8 locations across two states in the mid west USA
  • Satellite count/HDOP: when home position is slow to acquire, the satellite count is slow to go up, but does go up. In the 600 second case, it acquired 14 satellites within a few minutes and still wouldn’t acquire home. I’ve never seen any anomalous behavior between HDOP with respect to satellite count.

I did see strong correlation when flying at the edge of lake michigan, but that exhibited the additional symptom of an abnormally unstable GPS position. Myself and those I worked with attributed that to GPS multipathing off the water, which have seen before in other products we have that use a different uBlox module. I believe it was a different issue, but it is the only other issue we’ve had with the Here3+.

I have not tried connecting the drone/GPS to u-center via droneCAN. I have no idea what I would look for, or how I would compare that data to my existing data.

Thanks for any insight and a mostly exceptional product.

Might it be to do with the position of the constellations in the sky, back in the day when there were fewer groups of satellites to choose from we used to look up what we might expect at a given time. Many projects, things like Zipline Zips that need to just go at short notice never let their GPS switch off they are powered up all the time. This is not really an answer is it!

I haven’t checked the constellation positions, but at first thought it does make sense. I’ll look into it (I won’t be done in time to continue talking on this post). If that’s the case, I would expect the Here3+ to exhibit the problem more than the Here4 because the uBlox module in the Here4 is more capable in that regard.

It dawns on me that sky access won’t be uniform here. Simply guaranteeing a cone of access 45 degrees off vertical may not be enough control. I’m at 43N, so I wouldn’t be surprised if, statistically, there is better coverage if my sky access is more open to the south than to the north. GPS should be “fine” up to 50-60N, but as you approach it more of the satellites will be in the southern portion of the sky than the north. Thus, a short obstruction to the south is more impactful than the north. For any future readers - it would be reversed in the southern hemisphere.

I’m suspecting I need to start testing the GPS with u-center to remove the autopilot variable if I really want understand this.

I am sure it will be a factor this is the sort of thing I used to use. https://www.gnssplanning.com/

I don’t smoke, but I tell folks when they ask me, once you get a good lock, imagine you are having a cigarette and then fly.