Mads Tech
First Officer
Flight distance : 2741657 ft
United Kingdom
Offline
|
Over the last week I have been doing a lot of testing on the DJI Mavic 3 GPS module to try and understand it’s behaviour in regards to how it performs as well as address some miss information that’s been posted around a number of places.
Hardware
The DJI Mavic 3 GPS module has been compared and tested against a number of others including the Here 2, Here 3 as well as some other random ones I have.
Based on my tests of my module my conclusions are there is no hardware/ design issue with the module , the chipset DJI have chosen the UBX-M8030 - KA is performing identically to other modules and there is zero reason the automotive variant is having any impact on the GPS performance. It’s difference between it and the other versions is simply maximum package temperature.
I can confirm there is no external SQI flash connected to the M8030 however is nether unusual or out of place for DJI. They have been using the M8030 series since the Mavic Pro and the fleshless setup has been the same all along with them pushing the config at boot.
Testing this it’s not having any impact with both almanac and ephemerides data being retained on the internal memory on the short term. While external flash help with some specific needs it is not something that is going to be causing issues with the M3. What’s worth noting is DJI have been using this GPS setup on the M1, M2, M Air and Mini series with zero issues. In fact something interesting is the mini series does not have battery backup for retaining anything yet performance remains ok.
From a sensitivity standpoint the module is a little less sensitive than a Here 2 from Cubepilot with it averaging 3-5dbhz down however it’s not something that I believe is causing any impact and it’s performing smaller modules such as the M8Q , Hardware wise it’s as good as any other module and large 25mm antenna and ground plains and sensitively of the front end again appears good overall.
Software
As there is no SQI flash module configuration is not stored and DJI push it to the module on boot, at this time we don’t know what the actual setting is DJI are using but they have a walled garden they have to play in and we can test in that garden as I have already.
GPS Constellations
The Mavic 3 is the first consumer module to make a change in the satellite systems it’s using. Traditionally DJI have used GPS and Glonass however the Mavic 3 is the first to brake form that and use three systems with BeiDou and Gallalio replacing Glonass.
Having spent a lot of time testing the module in both system configurations ie GPS/Glonass and GPS/BeiDou/ Gallalio and learning the quirks of the module and Ublox I have found some interesting things.
The M8030 series is capable of up-to 10hz scanning rates in single constellation mode, 5hz in dual mode with GPS and Glonass. In a three system mode with Galileo Ublox state the maximum rate is 3hz. While you can clock the module faster with these systems I have observed some issues if you do. What’s interesting about this is there is a trade off when choosing 3 constellation’s and this reduced rate can have an impact on both overall HDOP, acquisition time and even est horizontal accuracy. For instance I have observed on a number of occasions the module failing to get to the same levels of low HDOP sub 0.8 and horizontal accuracy below 1m at 3hz compared to using GPS and Glonass at 5hz that will quickly get down to 0.6 HDOP and less than a 0.6m. This is somewhat dependent on satellite count of the Galileo and BeiDou system at the time.
Having tested the power modes on the module too from full power to balanced ect I’m not seeing any major impact on performance and it’s unlikely DJI are cycling it lower as the power saving in this application is minimal.
It’s also an observation that sat count for Galileo and Beidou seems to be more variable and there are many occasions where the two combined add up to the same as what you would get with Glonass. When all the systems are strong with good sat count positional accuracy is very good but it does appear to struggle sat system as times once they are not as strong. This would also tie in with what DJI are saying in high latitude areas.
There have been several occasions I have observed it seem to take a long time to bring in the BeiDou sats into the mix with gps and I do wonder if this is to do with the fact of it’s documented poor performance at times in certain part of the globe.
Overall there is nothing fundamentally different or wrong with either the module I have at least or the choice of system with Galileo and BeiDou but I have seen enough of a difference to come to some opinions on what I think may be going on.
I personally believe the biggest cause of the longer that most would like wait times is simply the switch from GPS/Glonass to the 3 sat system DJI is now using. Coverage of these two systems with the reduced rate of 3hz in my tests is having an impact at times for sure, Beidou and Galileo system alone may be what’s largely at play here. Most of DJI testing would have been done in China where BeiDou has the best coverage and this would have masked the issue.
We don’t know how DJI are configuring the M3 GPS module and that’s something I would love to see however we know what walled garden they are in as they can’t push too far outside the boundaries of the Ublox limits, or at least we would hope they are not.
While there is still the possibility there being a specific hardware issues with some craft personally I think most of the people with longer times is simply as I have said above. Perhaps there is something fundamentally wrong with the config DJI is using but tbh that’s very unlikely. I guess it’s possible there could be an issue at times where at boot the Ublox config is not accepted by the GPS module and as such it would boot to standard base config with GPS/Glonass and 1hz. But that could still out perform what we are seeing still so not really likely tbh.
Also does DJI do checks on this and may be performing retires. It could explain some people seeing no sats at all for a few minutes. It’s far more likely that there would be an issue with loading the config from the FC to the GPS at boot than the actual Ublox config it’s self. I could see this being something that could happen.
At the End of the day DJI have been doing GPS a long time and you can even see that in their PCB designs with it very much being optimised over the years. They ain’t new to this big issue in their own side is made likely.
However DJI statement is interesting when you look at it about high latitude areas and again imo it’s more system choice than anything at play here.
It’s also worth pondering that the use of the Gps, Galileo and BeiDou is not well used and there could be something in the Ublox firmware causing issues. It was last updated in 2017 and a lot has changed since the. It also worth noting that may be Ublox could be able to improve things I’m not sure the Ublox core firmware on the Ublox module in the M3 can be upgraded without SQI flash anyways.
Looking at what I have seen i think the mix of BeiDou/ Gallalio compared to Glonass is just takes longer to lock in at low HDOP levels is affecting the most. We don’t know what DJI are using to set their point of it being good to go but if its HDOP or positional accuracy then I can see why it’s taking longer.
As for why the GPS performance has gotten worse after the firmware update in late 2021, my own conjuncture on this is i think DJI did a tweak to improve GPS stability in flight and that’s resulted in longer acquisition times. Perhaps they broke something but at release a number of people noted that the hover and flight performer of the Mavic 3 was not as good as previous models and there were even sporadic reports of big issues after take off. I even noticed this my self in my review and while we don’t know what measure DJI specifically uses for it’s home lock ie horizontal accuracy or HDOP it does make sense that to address these complaints DJI would tighten down the minimum levels of accuracy for the GPS but the result of that is longer wait times on the current GPS mix.
The question then is why are some people affected and others not. Honesty I think this is simply location and sat count on the day for BeiDou and Gallalio even with good Almanac data.
This would also explain why it’s not been a simple as a firmware fix as they are balancing good lock time with overall GPS accuracy.
As for what they will do next. DJI have said there is firmware coming now, I would not be shocked to see an algorithm where Glonass is brought back into the mix and is chosen based on location or sat signal.
I am continuing to test and I would love a craft to hook up to but this is where I’m am at for now, I will be posting the video tomorrow but this covers it all. |
|