My Part 2 Emlid journey vs the DJI RTK2 Base Station

Recoveryone · Friday at 7:38 PM

adm_geomatics · 2024-11-04T09:15:00-0500

Great video Ray, thanks for the in-depth report! Just to add on a few notes.

I can't stress enough the importance of your point about running a drone on a cloud NTRIP service. It's just a a bad idea. There may be places where cell data stability and constellation visibility makes it more plausible. I am guessing that your NTRIP issues are due to the cell stability or your baseline length.
It's a little unclear what you are talking about with the Emlid vs the DJI obtaining and maintaining a fixed position. I'm assuming you were using NTRIP. I have never had an Emlid receiver take more than 10 minutes to fix on an NTRIP that was within the recommended 20km baseline and stable enough to trust. The DJI system requires less satellites correcting which in turn means a less globally accurate position but is the reason it fixes so quickly. This is not the total number of satellites in view but the number that meet the SNR, PDOP and time of corrections. You might have 30 satellites in view but only 7 actually being used for corrections. The contrast is that the DJI only requires 10 satellites and only uses it's position relative to the entered/known point to determine that it is good enough to call a fix. The Emlid system uses many more factors including the number of satellites actually accepted for corrections. I have zero issues when the conditions are acceptable running the Emlid from my phone's hotspot.
That said the Emlid system does not need to maintain a fix during flight. You should be acquiring a point whether that be by some sort of averaging or entering a known point. If you have a known point it would still be best to shoot it in and occupy it with Emlid's software and use local NTRIP. This is basically the same protocol as the DJI. Many people think the DJI is sending corrections directly to the drone but it is not. The corrections are being transmitted to the RC and the RC in turn passes those corrections to the aircraft. We know this because the D-RTK2 does not transmit nor does the drone receive corrections via a direct radio signal. If you lose aircraft to RC connection the RTK fix will also be lost.
Your comment about the drone use of the DJI vs Survey use of the Emlid this is where the real separation happens. The localization feature of the Emlid is a huge factor when you need to rectify your data to that of another party. Also, the ability to choose which constellations the base is using for corrections can be the difference between 5cm and 2cm relative accuracy. If you have dirty satellites you are getting bad data mixed in. You can see the stats if you use PPK and analyze the streams.
The statement about the number of GCP's is a little misleading. Per ASPRS spec you should have 30 independent control points at all times. That can be GCP and checkpoint inclusive which means that you could have 10 GCP's but you need 20 checkpoints. I agree that this is a bit ludicrous on sites smaller than 100 acres but if you are working with Surveyors and Engineers they should be requiring this. If they are not then they are not doing their due diligence. If you are doing this outside of that industry then I agree that you don't need as many but the larger the subject tract the more control points you need. As a matter of fact if you are on less than 100 acres you really don't need to use GCP's if your photogrammetry software has the ability to use a GEOID. I would recommend GCP's and checkpoints for Survey work but otherwise you only really need checkpoints to rectify the data when using RTK/PPK. Don't destroy the relative accuracy of your native data if you don't have to.

Recoveryone · 2024-11-04T12:43:11-0500

adm_geomatics said:
Great video Ray, thanks for the in-depth report! Just to add on a few notes.

I can't stress enough the importance of your point about running a drone on a cloud NTRIP service. It's just a a bad idea. There may be places where cell data stability and constellation visibility makes it more plausible. I am guessing that your NTRIP issues are due to the cell stability or your baseline length.

It's a little unclear what you are talking about with the Emlid vs the DJI obtaining and maintaining a fixed position. I'm assuming you were using NTRIP. I have never had an Emlid receiver take more than 10 minutes to fix on an NTRIP that was within the recommended 20km baseline and stable enough to trust. The DJI system requires less satellites correcting which in turn means a less globally accurate position but is the reason it fixes so quickly. This is not the total number of satellites in view but the number that meet the SNR, PDOP and time of corrections. You might have 30 satellites in view but only 7 actually being used for corrections. The contrast is that the DJI only requires 10 satellites and only uses it's position relative to the entered/known point to determine that it is good enough to call a fix. The Emlid system uses many more factors including the number of satellites actually accepted for corrections. I have zero issues when the conditions are acceptable running the Emlid from my phone's hotspot.

That said the Emlid system does not need to maintain a fix during flight. You should be acquiring a point whether that be by some sort of averaging or entering a known point. If you have a known point it would still be best to shoot it in and occupy it with Emlid's software and use local NTRIP. This is basically the same protocol as the DJI. Many people think the DJI is sending corrections directly to the drone but it is not. The corrections are being transmitted to the RC and the RC in turn passes those corrections to the aircraft. We know this because the D-RTK2 does not transmit nor does the drone receive corrections via a direct radio signal. If you lose aircraft to RC connection the RTK fix will also be lost.

Your comment about the drone use of the DJI vs Survey use of the Emlid this is where the real separation happens. The localization feature of the Emlid is a huge factor when you need to rectify your data to that of another party. Also, the ability to choose which constellations the base is using for corrections can be the difference between 5cm and 2cm relative accuracy. If you have dirty satellites you are getting bad data mixed in. You can see the stats if you use PPK and analyze the streams.

The statement about the number of GCP's is a little misleading. Per ASPRS spec you should have 30 independent control points at all times. That can be GCP and checkpoint inclusive which means that you could have 10 GCP's but you need 20 checkpoints. I agree that this is a bit ludicrous on sites smaller than 100 acres but if you are working with Surveyors and Engineers they should be requiring this. If they are not then they are not doing their due diligence. If you are doing this outside of that industry then I agree that you don't need as many but the larger the subject tract the more control points you need. As a matter of fact if you are on less than 100 acres you really don't need to use GCP's if your photogrammetry software has the ability to use a GEOID. I would recommend GCP's and checkpoints for Survey work but otherwise you only really need checkpoints to rectify the data when using RTK/PPK. Don't destroy the relative accuracy of your native data if you don't have to.

Very good deep in the weeds info, I didn't want to go that deep, as the video was 14 min long and I cut out about another 5 min talking about training and what to look for when your adding/replacing GCP's to an existing site.

adm_geomatics · 2024-11-04T14:32:08-0500

Recoveryone said:
Very good deep in the weeds info, I didn't want to go that deep, as the video was 14 min long and I cut out about another 5 min talking about training and what to look for when your adding/replacing GCP's to an existing site.

Absolutely. It can get deep pretty quick when analyzing GNSS equipment, especially when you are doing comparisons of two solutions that technically cross over but are designed for completely different use cases. DJI needs to update their device as I am sure the Emlid receivers have already taken a very large chunk of that business. The D-RTK2 is already more expensive so I can't imagine how much it would cost if it had some decent data collection software. I fell off quick when I had to carry a remote with an iPad attached and the rover. It's my understanding that you can no longer collect data with the newer RC's and Pilot 2. Have you figured that out? Or are you still using an old device with GS Pro on it?

Recoveryone · 2024-11-04T15:40:34-0500

adm_geomatics said:
Absolutely. It can get deep pretty quick when analyzing GNSS equipment, especially when you are doing comparisons of two solutions that technically cross over but are designed for completely different use cases. DJI needs to update their device as I am sure the Emlid receivers have already taken a very large chunk of that business. The D-RTK2 is already more expensive so I can't imagine how much it would cost if it had some decent data collection software. I fell off quick when I had to carry a remote with an iPad attached and the rover. It's my understanding that you can no longer collect data with the newer RC's and Pilot 2. Have you figured that out? Or are you still using an old device with GS Pro on it?

The DJI system is practically useless for setting GCP's or verifying data, as there is no software to speak of. In addition to needing to carry around the controller with a second RTK base to mark points (mode 2 or 3). I understand your statement about the RC communicating with the drone for corrections, but the base is what establishes that info. That is why I am not a fan of NTRIP as the info (signal strength) from a station is not consistent in my experience.

adm_geomatics · 2024-11-04T21:30:21-0500

Recoveryone said:
The DJI system is practically useless for setting GCP's or verifying data, as there is no software to speak of. In addition to needing to carry around the controller with a second RTK base to mark points (mode 2 or 3). I understand your statement about the RC communicating with the drone for corrections, but the base is what establishes that info. That is why I am not a fan of NTRIP as the info (signal strength) from a station is not consistent in my experience.

Correct, that is the difference between cloud and local NTRIP. The D-RTK2 operates via local NTRIP and had this functionality before Emlid. Most high-end receivers still don't have it because it is unnecessary and ineffective in any other situation unless you have some long-range Wi-Fi. It transmits the corrections through the RC, not directly to the drone like some people think. I use this same method with the RS3 and between it and my older RS2+ I have never lost a fix in over 2 years. Now that GS Pro isn't compatible with the new controllers the D-RTK 2 is just an overpriced base.