What kind of absolute position accuracy is typical?

So I bought a couple reach units for my drone mapping (my drone doesn’t have RTK) for when I need to correct the absolute position of my maps using ground control points. Since I’m not a Land Surveyor doing work to be used to set property lines etc, I don’t need like <3cm survey accuracy with my application. However I would like to get <11cm accuracy (or about 4-inch accuracy.) I thought these units would be perfect for that. They are a bit more finicky for RTK then I anticipated due to varied job site conditions and I think primarily because they are only single frequency L1.

I had a guy do some PPK work for me on a job that I didn’t have a fixed known point to put my base on. (In that case I’m learning I probably don’t need to use two units anyway and just use one unit as a rover and a CORS station as the base). He has much more experience than me and his comment/recommendation has got wondering if these units are really going to fulfill my need going forward.

I’m wondering how realistic people think achieving 10cm absolute position accuracy without having a known survey point (thereby require PPK with CORS) is considering site conditions are “average” (meaning not optimal but not some of the worst case scenarios). This was his comment that got me thinking:

If you need to know 100% that the coordinates achieved by survey are within 10 cm of actual location, emlid reach is probably not the way to go, long time to fix can lead to setting changes which if you’re not very careful can lead to bad results (I am guilty of this in my early days of using it), need to understand fully what you’re doing when changing the settings. If you need to know with 100% confidence and your volume of jobs is low/infrequent, I think you would be better off renting a Trimble Network Survey Kit for $300/day, they will give you the rover and RTK will be computed using VRS (multiple CORS stations). Survey equipment like this is $50K + and is extremely reliable. I personally use this option when I need to survey a large quantity of locations, if coordinates need to be within a couple of cm, or and the most important case when there is a high risk of multipath signals (survey undertaken near/under trees or near buildings). Reach is mainly used successfully with drones for this reason (no multipath problems). The more expensive survey gear avoids multipath through use of L2/L5 frequencies (emlid only L1) and also through very expensive antennas (choke rings, I’m imagining you’re only using a base plane).

In basic terms either way you can easily achieve better than 10cm accuracy according to whatever coordinates you use for your GCPs. It depends on your definition of Absolute Accuracy. Globally or Locally? Are you using base/rover, or a service/rover? You can use GCPs derived from native GPS coordinates as provided by the Reach Units for global accuracy or you can use the coordinates of your local site as set by a land surveyor/engineer to achieve absolute local accuracy. This includes using a local benchmark as your elevation datum and not the Ellipsoid/GEOID (not real) elevations.

I don’t know your use-case, but in construction this, in my opinion, is the only way to do it. Otherwise you are creating a pretty 2D map and 3D model that’s not relative to the site, CAD or BIM. Nice picture to look at though.

For me, 90% of the time it’s going to be absolute world coordinates or state plan coordinates.

I think you are overthinking it.
In my experience with my RS+ units, is that they are more precise than my ability to level a rover-pole on a bi-pod or taking into account the possible flexing of the GCP plate.

With a recent experiment I compared the baseline of the Reach RS+ combo with a total station edm measurement. They agreed to within under 1 cm at 680 meters: Reach RS+ baseline vs total station EDM

A few days ago I flew 2 drones (Phantom 4 Pro and M600 Pro with a Phase One iXM 100 mp) over a 436x176 meter area, using 15 GCP’s. Combined 3D RMS reported by Photoscan was ~1.6 cm for both projects.

As long as you got decent sky view (doesn’t have to be excellent as such), you are in good shape.

So, relative precision is not an issue.
The absolute precision is again just up to either your baseline to a NTRIP station, or your own ability to mechanically line up your own base over a known point, given the known coordinates are still valid (which is probably much more likely to be a significant error than RS+ precision).

I am also not a land surveyor, and I haven’t tried any L2/L5 gear, but cost vs performance, is, as you can see from the numbers above, unbeatable by a 30k usd setup. In decent sky view, their performance will be neck-to-neck I am sure. Only in limited sky view or noisy environments the multi-band solution will have an advantage.
It too will have limitations in limited sky view, where a total station simply is a better option (hence they are still made for and used outside

That’s exactly what I am talking about. Global vs State Plane is not the same thing. You will be very lucky if you get a perfect translation from WGS84 to a local EPSG that actually matches what is on the ground.

We’re not talking about the theoretical accuracy of the units. Yes they are that accurate, but when someone asks a question about “What accuracy am I going to achieve?”, they need to understand that there is a lot more to it than just going out and shooting points so I don’t quite agree on you’re “overthinking it” statement.
This is just another instance of the fact that there are two completely different ways to do drone mapping. One is using theoretical data from satellites, ellipsoids and geoids and another that uses practical ground information that is actually going to be used for RPLS surveys and construction projects.
Lastly for a comparison of a survey-grade GPS receiver versus an L1 only receiver all depends on how much work you’re going to be doing. I choose to continue with the Reach units because I have other initiatives in mind, but if it were for my personal use in setting ground control points for drones there is no comparison between a dual-channel Topcon/Trimble system and the Reach. The labor time that I would save in 18 months in my scenario pays for the survey grade units. Being that I would plan on keeping my receivers for 3 to 5 years then it’s a no-brainier. I have never post-processed a survey grade receiver in 10 years and can match within hundredths of a foot of a robotic total station every time.
So do some thinking, LOL. About you or use case and your labor.

The labor you are talking about is in terms of transformations between WGS84 and locals EPSG’s?

Can you give an example of the difference in accuracies between your Emlid system and your Topcon system?

Labor as in it takes at least 2-3 minutes (or more) to be confident in a point, IF I have an RTK fix. Since I have yet to gain that absolute confidence I stakeout the point after it has been shot. Mostly good, but not always. Even then the Emlid system still just averages the data. The Topcon shots take about 5 seconds and in that time it collects almost 30 separate points and does crude SnR processing on-the-fly. Because of the actually quality of the fix I am almost certain that I probably still need to post-process the Emlid (still learning to do so) to get to the same accuracies.

The kicker is what accuracy do you NEED? This is obviously a GIS and existing or want-to-be drone mapper community from my experience so <5cm is pretty good. If you are a registered surveyor and are able to use this gear suffieciently for your needs then kudos. My construction tolerance is 3cm and my in-house QA/QC tolerance is 1.5mm so the Reach are not yet (until I figure them out a little better) good enough for day-to-day, but are fine for my drone.
So, a couple of minute deltas (at best) for shooting with an average drone flight needing 12 GCP’s X 250 flights/year is allot of money to me and this is just GCPs. IF I can maintain a fix.

“Labor as in it takes at least 2-3 minutes (or more) to be confident in a point”

Is this with fix or float because the default collection time for float is more like this, the default on the Reach for fix is 40 seconds. When collecting points with a good fix and lots of satellites, I see the RMSE on the rover read out to be around n the 0.001m order of accuracy or 1cm. I would think then if you were able to establish a base coord position <5cm accuracy then your collected rover points should be <10cm? With a good fix, I’m wondering if even 10 or 20 seconds of collection would get you 10cm absolute position accuracy? No conversion of coordinates, talking only WGS84.

So are you using two Topcon units (base and rover) or are you using a Topcon as a rover and a Reach as a base, or are you just using one Topcon that is using other means of correction (NTRIP, multiple CORS) or some other sort of correction network subscription? CORS can only be used for PPK right, not RTK and you say you are only doing RTK, no post processing?

Correct. Trying to get anywhere close to L1/L2 accuracies is what I am testing and giving you the real numbers on. The RMSE can say whatever it wants because that is what it has at that point in time. Try shooting the point, walking off about 10m and then staking that point out again. I think 60 seconds with a good fix will give you good confidence for less than 10cm. Because of the Topcon software I run I can be repeatable withing 7-10cm with a good float on the verge of a fix.

All field tests have been with a Topcon Hiper V base/rover kit vs the Emlid Reach RS+ base/rover kit. Our scenario is probably different than allot of you all because I (1) Do not have allot of trust in our CORS system after testing unless I post-process everything, (2) I need site acuracy and (3) The CORS system is not fast enough for our machine control. I think I will make this a disclaimer on my account as a little different of a user-base trying to push the Reach system or at least take it where it may not be intended.

Interesting scenario, I’ll try that soon!

What is your take on mechanical inaccuracies? I.e the levelling of rover-pole, threshold of a levelling bubble and so on?
I see a lot of images of people just hand-holding the rover-pole. Having tried that, and thus knowing how hard it is to keep that thing perfectly level for just 60 seconds, aren’t we then past 2-3 cm error quite quickly?

Yeah, that’s a big no-no and I highly encourage everyone to get get a rover pole with a tripod attachment for their base stations and a bipod for their rover. Don’t get me wrong, level tripods and a tribrach plus a mounting adapter are great, but for less cost you get an unmistakable, on-point, fixed height level in 30 seconds every time. Of course like anything else they need to be calibrated once every 6-months or so. I like the Seco yellow aluminum snap-lock poles. You can hunt around though and find better pricing.

https://www.tigersupplies.com/Products/Seco-Construction-Series-Thumb-Release-Tripod__SEC5218-40-FLY-.aspx

https://www.tigersupplies.com/Products/Seco-2m-Snap-Lock-Rover-Rod__SEC5125-20-YEL.aspx

https://www.tigersupplies.com/Products/Seco-Construction-Series-Thumb-Release-Bipod__SEC5217-40-FLY-.aspx

Further on the testing. I setup each base on known points and have a quick-lock setup for my rovers so I can push-button move them from my vehicle mounts to my rover pole. I drive to the point, take the Reach off the car, lock it to my rover pole, shoot it and while leaving the pole there do the same with the Hiper. The first time I tried this I was using magnet for the Hiper and Reachview for the RS+ and I thought I had just wasted my money because my elevation deltas were huge, but what I quickly relaized is that the Reach is a theoretical ellipsoid where as my Topcon was on a localized datum. I had 6 feet of error from the two! I used the NOAA site the try to transform from the ellipsoid to my GEOID and that was nonsense. It was much closer, but still not the same as the level-loop.

That’s also what I use (tripod for base and pole+bipod for rover).

So for the sake of making a proper comparison, how much time do you spend with the pole at center of the GCP?
2 components as I see it, levelling and measuring, placing and leveling taking 15-30 seconds and then the time spent collecting data in a now static environment. Is it roughly comparable with your time spent ?

Sorry to hijack the thread…

Car Mount - https://www.tigersupplies.com/Products/Seco-Single-Mag-Mount-with-Quick-Release-Tip__SEC5114-051.aspx (They also make a tri-magnet for rough conditions.)

Rover pole - https://www.tigersupplies.com/Products/GPS-Quick-Release-Adapter-(Leica)__ADI765-11.aspx

Rover - https://www.tigersupplies.com/Products/AdirPro-GPS-Quick-Disconnect-Adapter__ADI765-02.aspx

Leveling and placing not included, just the collection. I was trying for 2 minutes Emlid and my typical (don’t laugh) 3-second Topcon. I think I am not quite the normal user on this forum simply for the fact that I want to be relative and right on the ground inside the boundaries of my subject. Not the planet.

1 Like

All that being said, I totally support Emlid. I think it is an amazing product for the cost-point and if I can get it to do what i want it to do with our Magnet Field software with minor user training I will order 20 sets tomorrow.

2 Likes

So in relative terms, you would expect 1.5 mm XYZ combined RMS from the topcon from a topcon multiband system with 3 seconds, if I understand you correctly ?