RTK Facet DGPS Accuracy & Disabling DGPS

I purchased an RTK Facet GNSS receiver awhile back and I have been using it to (try to) survey my property, particularly finding my corners etc. At first I used it without any RTCM corrections at all, and the HPA (which I have since learned is better described as an estimate) was reliably claiming 9.8 cm accuracy with a DGPS type fix, which I regarded as good enough for what I wanted to do (I was planning to move my fence line in about 2-4 ft from the property line which would only mean a sacrifice of one-half or one acre of my land being outside my fence lines). However, after going around and establishing where I thought the corners were (based on county GIS coordinates) I found out the coordinate source I was using was wrong, so I re-investigated what I needed to do to get the right coordinates and in the process also checked the actual accuracy of the DGPS fixes and realized they were entirely wrong - by an order of magnitude. The actual accuracy (based the coordinates it was actually putting out over time for a fixed receiver) was closer to +/- 1-2 meters in both latitude and longitude, which is honestly more what I should have been expecting anyway. I also noticed that while the single fixes were around the same level of accuracy as the DGPS fixes or worse, they weren’t wandering around as much - and were more repeatable within the span of an hour or so. All that said, I am wondering if it is possible to disable the DGPS fix altogether and make it stay in single or RTK float/fix and ignore a DGPS fix?

Using rtk2go, the closest publicly available base is around 155 km away, and using it I have gotten an RTK fix (claimed) occasionally with accuracy of around 3 cm (which I also no longer believe, although the output coordinates move around by around that much). Nevertheless I have used that and gone out and gotten coordinates off the corners of the section my land is within and then interpolated to check the measured coordinates for side monuments against the location they should be at based on the corners and I have gotten within less than a foot pretty reliably (with RTK fix/float coordinates collected the same afternoon). I think that’s pretty good given how those monuments were originally placed over 80 years ago, and it also makes me think that the RTK float/fix based on that faraway base is probably fairly accurate within a short window. In other words, the data is accurate if it is all collected at approximately the same time, but it isn’t necessarily repeatable at a different time.

Therefore I am in the process of establishing a fixed base on my land, and will go down that route going forward to get both accurate and precise data that is repeatable. But I still think it would be useful to be able to disable the DGPS fixes, if that is possible. Thanks.

was reliably claiming 9.8 cm accuracy with a DGPS type fix

I generally see 0.3m HPA (30cm) off the shelf with the RTK Facet. Yes, it’s an estimate. But I have never seen ~0.15m or below. Your reported 0.098m is pretty amazing.

is possible to disable the DGPS fix altogether and make it stay in single or RTK float/fix and ignore a DGPS fix?

3D/DGNSS fix is sort of the most basic type of fix. I don’t think it’s possible to ignore this type of fixe and have the module only output RTK Fixed locations. I know we can turn off constellations for example BeiDou, QZSS, or SBAS, but I don’t think you can suppress lock until RTK Fix is achieved.

In other words, the data is accurate if it is all collected at approximately the same time, but it isn’t necessarily repeatable at a different time.

You are correct and very keen to have arrived at this. RTK is always based on the accuracy of the base. If the base is situated inaccurately 5.441m to the left of reality, all your rover readings will be 5.441 (+/- 0.014m) to the left.

TLDR: Turning off SBAS should disable DGPS. You’re on the right track by setting up a local RTK base. Consider hiring a surveyor before building.

  1. There are a plethora of GPS solution types. They are reported in the 6th data field in the GGA statement

From http://lefebure.com/articles/nmea-gga/

Fix type. This is always a single number. Reportable solutions include:
    0 = Invalid, no position available.
    1 = Autonomous GPS fix, no correction data used.
    2 = DGPS fix, using a local DGPS base station or correction service such as WAAS or EGNOS.
    3 = PPS fix, I’ve never seen this used.
    4 = RTK fix, high accuracy Real Time Kinematic.
    5 = RTK Float, better than DGPS, but not quite RTK.
    6 = Estimated fix (dead reckoning).
    7 = Manual input mode.
    8 = Simulation mode.
    9 = WAAS fix (not NMEA standard, but NovAtel receivers report this instead of a 2).

Autonomous is the most basic solution type.

RTK Fixed is is usually the most accurate and precise.

Note that list isn’t totally in order of accuracy.

DGPS requires a differential correction source (hence the “D”), generally a Space-Based Augmentation Service (SBAS) such as, in the US, WAAS.

If I’m out running my Facet rover in RTK and I lose the RTCM link, it will drop down from RTK Fix to DGPS in 60 seconds. If I’m down in a valley and my Facet doesn’t have good sky vis to the south (I’m in the US and the WAAS satellites are geostationary–that is, over the equator) I won’t even get a DGPS fix and the Facet will report an Autonomous solution.

  1. You can turn off DGPS by disabling SBAS (which includes WAAS) satellites. Some surveyors running RTK explicitly turn off SBAS. Here’s why:

  • If you get a DGPS solution using the WAAS differential corrections, the receiver’s reported solution (coordinates) will be in ITRF, or so I’ve come to believe. I haven’t found a authoritative source for this, but that’s what I’ve gathered from surveyors posting on one of their most active web forums. You can web search (this is Google syntax) “waas datum site:surveyorconnect.com” to read what I’ve read.
  • - US CORS stations provide RTCM messages based on NAD83. If you're running network RTK utilizing a US CORS station, the rover wil output NAD83 coordinates.
  • - Likewise, if you setup your own base and configure the base with fixed coordinates in NAD83 or ITRF, your rover's output coordinates will be in either, respectively, NAD83 or IRTF.
  • - Some surveyors running RTK in NAD83 will TURN OFF SBAS so their receiver's solution (output coordinates) doesn't jump to ITRS and then jump back to NAD83 when they briefly lose and then reestablish the RTCM feed. If your're working at the limits of your RTCM comm link, this gets very old very fast. NAD83 vs. ITRF are about 5.7 feet different in my part of the US, and that's significant.
    1. Standard errors provided in marketing literature for just about every GNSS receiver sold are well-known to be very optimistic. Ditto for standard errors reported by the devices while in operation. It’s just the market; like horsepower ratings on vacuum cleaners and cold-weather temperature ratings on sleeping bags.

    Some of the reason for the optimism is that the receiver calculates a standard error based on what it knows, but there are error sources that it might not know about or that it doesn’t model accurately. It’s doing what it can, just take it with a grain of salt.

    On top of the optimism, the standard errors are, unsurprisingly, generally reported like standard deviations, 1-sigma, expressing a 68% confidence interval. So 32% of the readings will fall outside of the +/- 1-sigma range. And 1% of the readings could likely be terribly off. So to get a 95% confidence interval, you need to approximately double the standard error. And then add in the optimism factor.

    So if the Facet is reporting 3cm, that can be interpreted as 68% of the readings will fall +/- 3cm of the actual coordinates. So 68% of the readings will fall in a 6cm range (a 3cm diameter circle). ~95% of the readings will fall withing 12cm. And those are optimistic estimates, so add in (or multiply in) a factor to account for the optimism.

    I do a bit of post-processing and least squares processing of GNSS data. Due to the well-known optimism factor, a few widely-used least squares adjustment softwares include a configuration setting to account for the optimism. The s/w I use takes the setting I input, raises 2 to that setting, and then multiplies the reported error from the GNSS receiver by that number. The take-away is that the reported errors are so optimistic, widely-used post-processing software packages provide an exponential scale to convert the optimistically-reported errors from the GNSS receivers to more realistic errors.

    1. Municipal GISs are not known for their accuracy. They generally exist as land tax maps, 911 systems, code enforcement, etc. (Yes there are municipalities that fly drones periodically and look for unpermitted swimming pools!) The vast majority of municipal GISs are not accurate enough for surveying property boundaries. I doubt any surveyor would use GIS coordinates for anything more than a rough starting point to search for evidence in the field.

    2. Setting up a local RTK base is the way to go. The short baseline from base to rover will help a lot. By far the limiting factor on the a Facet RTK Base-Rover setup is the RTCM radio link. If you have cell/internet availability over your property you have an alternate to the radio link. Run a couple static sessions for the base on different days, get multiple post-processing solutions, compare the results.

    3. If you think you’ve found the corner monuments, you can always run static on each corner and post-process that data with OPUS or CSRS-PPP, etc.

    4. Consider hiring a surveyor. Besides having the equipment and experience to run a straight line between two corners, they also have the knowledge and expertise to check that the monuments in the ground are actually your property corners, and not something else. They know how to interpret deeds (that’s a complex topic all by itself, and the numbers in the deed descriptions often appear “wrong” if you think like an engineer; land surveyors know how to sort things out. ) There are three different legally-recognized definitions of a “foot” in the areas I work in; they are pretty close but I gather your fence is going to be thousands of feet long, so minor differences matter. The good ones have insurance in the event things get built in the wrong places. They will generally also survey enough of the parcels adjoining your parcel to check that things are in the correct place for you and for your neighbors. Property lines and corners aren’t yours alone, they are shared with the neighbors, and it’s important that no one gets shorted. Finally, if your neighbor is unhappy with the outcome, the first thing you might say is “I hired a surveyor, why don’t you hire your own surveyor and we’ll let them sort it out.” It’s way easier then hiring attorneys and sitting in court rooms.

    sparky:

    was reliably claiming 9.8 cm accuracy with a DGPS type fix

    I generally see 0.3m HPA (30cm) off the shelf with the RTK Facet. Yes, it’s an estimate. But I have never seen ~0.15m or below. Your reported 0.098m is pretty amazing.

    is possible to disable the DGPS fix altogether and make it stay in single or RTK float/fix and ignore a DGPS fix?

    3D/DGNSS fix is sort of the most basic type of fix. I don’t think it’s possible to ignore this type of fixe and have the module only output RTK Fixed locations. I know we can turn off constellations for example BeiDou, QZSS, or SBAS, but I don’t think you can suppress lock until RTK Fix is achieved.

    In other words, the data is accurate if it is all collected at approximately the same time, but it isn’t necessarily repeatable at a different time.

    You are correct and very keen to have arrived at this. RTK is always based on the accuracy of the base. If the base is situated inaccurately 5.441m to the left of reality, all your rover readings will be 5.441 (+/- 0.014m) to the left.

    DGPS Fix

    As you can see (hopefully - it’s been awhile since I’ve used hosted images before) I do get DGPS fixes in the 10 cm HPA realm, but the problem with it is that it’s totally wrong - the coordinates move around a lot (around +/- 1 meter or so). Averaging the points over time does yield a better and more repeatable result, but unless I capture all of that and then post process it all for every point there isn’t much benefit to that. It certainly isn’t very useful for finding specific points (because it would be a slow iterative process).

    However I have been making a lot of progress with my PPP surveyed GNSS Base. Thanks so much for the DIY tutorial on that, it has been immensely helpful - there have been some variations but in general it has been dead on. I’ve been going the minicomputer route, and tonight I just completed an initial seven hour survey (the data is now being processed by CSRS too). The whole setup (except for the fixed antenna mount which is de-galvanized all-thread welded to a steel pole and concreted into the ground) is mobile and I have to run the mini computer and everything else off a battery-inverter setup, and eventually when it is in use I will have to power it with a small solar system and give it internet access via an LTE router. It has all been quite the experience so far.

    I appreciate all the replies I have gotten to this. I am still reading through them.

    toeknee:
    TLDR: Turning off SBAS should disable DGPS. You’re on the right track by setting up a local RTK base. Consider hiring a surveyor before building.

    1. There are a plethora of GPS solution types. They are reported in the 6th data field in the GGA statement

    From http://lefebure.com/articles/nmea-gga/

    Fix type. This is always a single number. Reportable solutions include:
    
    0 = Invalid, no position available.
    1 = Autonomous GPS fix, no correction data used.
    2 = DGPS fix, using a local DGPS base station or correction service such as WAAS or EGNOS.
    3 = PPS fix, I’ve never seen this used.
    4 = RTK fix, high accuracy Real Time Kinematic.
    5 = RTK Float, better than DGPS, but not quite RTK.
    6 = Estimated fix (dead reckoning).
    7 = Manual input mode.
    8 = Simulation mode.
    9 = WAAS fix (not NMEA standard, but NovAtel receivers report this instead of a 2).
    
    
    
    
    Autonomous is the most basic solution type. 
    
    RTK Fixed is is usually the most accurate and precise.
    
    Note that list isn't totally in order of accuracy.
    
    DGPS requires a differential correction source (hence the "D"), generally a Space-Based Augmentation Service (SBAS) such as, in the US, WAAS.
    
    
    
    If I'm out running my Facet rover in RTK and I lose the RTCM link, it will drop down from RTK Fix to DGPS in 60 seconds. If I'm down in a valley and my Facet doesn't have good sky vis to the south (I'm in the US and the WAAS satellites are geostationary--that is, over the equator) I won't even get a DGPS fix and the Facet will report an Autonomous solution.
    
    
    
    
    
    2. <B>**You can turn off DGPS by disabling SBAS**</B> (which includes WAAS) satellites. Some surveyors running RTK explicitly turn off SBAS. Here's why:
    
    
    <LIST type="lower-alpha"><LI>If you get a DGPS solution using the WAAS differential corrections, the receiver's reported solution (coordinates) will be in ITRF, or so I've come to believe. I haven't found a authoritative source for this, but that's what I've gathered from surveyors posting on one of their most active web forums. You can web search (this is Google syntax) "waas datum site:surveyorconnect.com" to read what I've read.</LI>
    <LI>- US CORS stations provide RTCM messages based on NAD83. If you're running network RTK utilizing a US CORS station, the rover wil output NAD83 coordinates.</LI> 
    <LI>- Likewise, if you setup your own base and configure the base with fixed coordinates in NAD83 or ITRF, your rover's output coordinates will be in either, respectively, NAD83 or IRTF.</LI>
    <LI>- Some surveyors running RTK in NAD83 will TURN OFF SBAS so their receiver's solution (output coordinates) doesn't jump to ITRS and then jump back to NAD83 when they briefly lose and then reestablish the RTCM feed. If your're working at the limits of your RTCM comm link, this gets very old very fast. NAD83 vs. ITRF are about 5.7 feet different in my part of the US, and that's significant.</LI>
    </LIST>
    
    
    
    3. <B>**Standard errors**</B> provided in marketing literature for just about every GNSS receiver sold <B>**are well-known to be very optimistic**</B>. Ditto for standard errors reported by the devices while in operation. It's just the market; like horsepower ratings on vacuum cleaners and cold-weather temperature ratings on sleeping bags.
    
    
    
    Some of the reason for the optimism is that the receiver calculates a standard error based on what it knows, but there are error sources that it might not know about or that it doesn't model accurately. It's doing what it can, just take it with a grain of salt.
    
    
    
    On top of the optimism, the standard errors are, unsurprisingly, generally reported like standard deviations, 1-sigma, expressing a 68% confidence interval. So 32% of the readings will fall outside of the +/- 1-sigma range. And 1% of the readings could likely be terribly off. So to get a 95% confidence interval, you need to approximately double the standard error. And then add in the optimism factor.
    
    
    
    So if the Facet is reporting 3cm, that can be interpreted as 68% of the readings will fall +/- 3cm of the actual coordinates. So 68% of the readings will fall in a 6cm range (a 3cm diameter circle). ~95% of the readings will fall withing 12cm. And those are optimistic estimates, so add in (or multiply in) a factor to account for the optimism. 
    
    
    
    I do a bit of post-processing and least squares processing of GNSS data. Due to the well-known optimism factor, a few widely-used least squares adjustment softwares include a configuration setting to account for the optimism. The s/w I use takes the setting I input, raises 2 to that setting, and then multiplies the reported error from the GNSS receiver by that number. The take-away is that the reported errors are so optimistic, widely-used post-processing software packages provide an <B>**exponential scale**</B> to convert the optimistically-reported errors from the GNSS receivers to more realistic errors.
    
    
    
    
    
    4. <B> **Municipal GISs are not known for their accuracy.** </B>They generally exist as land tax maps, 911 systems, code enforcement, etc. (Yes there are municipalities that fly drones periodically and look for unpermitted swimming pools!) The vast majority of municipal GISs are not accurate enough for surveying property boundaries. I doubt any surveyor would use GIS coordinates for anything more than a rough starting point to search for evidence in the field. 
    
    
    
    
    
    5. Setting up a local RTK base is the way to go. The short baseline from base to rover will help a lot. By far the limiting factor on the a Facet RTK Base-Rover setup is the RTCM radio link. If you have cell/internet availability over your property you have an alternate to the radio link. Run a couple static sessions for the base on different days, get multiple post-processing solutions, compare the results.
    
    
    
    6. If you think you've found the corner monuments, you can always run static on each corner and post-process that data with OPUS or CSRS-PPP, etc. 
    
    
    
    7. <B> **Consider hiring a surveyor.** </B> Besides having the equipment and experience to run a straight line between two corners, they also have the knowledge and expertise to check that the monuments in the ground are actually your property corners, and not something else. They know how to interpret deeds (that's a complex topic all by itself, and the numbers in the deed descriptions often appear "wrong" if you think like an engineer; land surveyors know how to sort things out. ) There are three different legally-recognized definitions of a "foot" in the areas I work in; they are pretty close but I gather your fence is going to be thousands of feet long, so minor differences matter. The good ones have insurance in the event things get built in the wrong places. They will generally also survey enough of the parcels adjoining your parcel to check that things are in the correct place for you and for your neighbors. Property lines and corners aren't yours alone, they are shared with the neighbors, and it's important that no one gets shorted. Finally, if your neighbor is unhappy with the outcome, the first thing you might say is "I hired a surveyor, why don't you hire your own surveyor and we'll let them sort it out." It's way easier then hiring attorneys and sitting in court rooms.
    

    That was a lot. I really appreciate that too - I generally write a lot myself (irreducible complexity cannot be overcome), and I appreciate it when others do the same. At present I don’t think I can reply to everything - I was out all night running a PPP survey for my base, but I would like to reply to the last item (and maybe one other), specifically “consider hiring a surveyor.” The short answer to that is that I have considered it and investigated it and what they’re asking is a lot more than I want to pay for what they would give me under my present circumstances:

    My land does have private “neighbors” (meaning private land bordering it), but there are absolutely no fences or building sites or activity of any kind in the area at all in the entire section. In 1975 half the section my land is in was subdivided into a tract - and no one bought any of the pieces and now they’re all being held by a holding company that doesn’t have any contact information and is registered to a non-descript office that is closed 24/7 and only has a PO Box that doesn’t accept mail (they auto-return mail to senders). Perhaps someday they will try to develop it, but given where it is, and the market of the area, that isn’t likely to happen in my lifetime. The closest standing structure to my property is over 2 miles away, and the closest paved road is over three miles away. There are legally recognized (by the county) dirt roads providing access to my property, but there are no legally recognized easements through my property (or any adjacent piece) to the isolated pieces behind my land, so anyone that ever does try to become my neighbor (I’m trying to buy as much of it as possible because I don’t want neighbors complaining about me taking off and landing my planes) is gonna have a heck of time even getting legal access to their property along 3/6 of my sides. Additionally 1/6 of my sides borders BLM land, and they aren’t known to squabble (in my area) over a fraction of an acre. That leaves two sides that are of moderate future concern (though again I am trying to buy those pieces if I can find a way to contact the owners). The isolation of my land from real neighbors is a significant part of why I bought it.

    Now that said, the legal description of my land was as you say a bit of mess at first to my mechanical engineer mind, but I do know exactly how to interpret it now. My land is legally exactly 3/4 of the SW corner of the SW corner of the section I am in, so 30 acres, with a perimeter of a mile (give or take a few feet). Additionally, the monuments I am referencing to are the USGLO survey monuments - the same ones that were referenced to for the entirety of the 1975 tract I mentioned above. None of them are directly on my property (because the nominally ten acre SW corner of my SW corner of the SW corner of the section isn’t yet mine), but I could absolutely dig out some stakes and mason line and a 100 ft tape and chain my way to my corners just like old fashioned surveyors used to do. Using the GPS is just so much more convenient and will let me do stuff like positioning things within my property more easily, differential elevation surveys, and keeping track of my roads, runway(s), piping, and wire runs etc.

    In light of all of that, I don’t really see what advantages a proper insured survey affords me. Neighbor concerns are pretty much moot, and if I ever do have neighbors at my fence lines (I won’t fence all the pieces, just the core piece) that complain about my fence placement and they can prove I am wrong, I won’t fight it, I will just move the fence.

    This whole thing is, and is expected to continue to be, a learning experience. I am not afraid of that, and I am not afraid to get my hands dirty or do hard work to avoid paying someone else to do it for me. Thank you for the input you provided and all the time that you took to do that. I really appreciate it.

    Good luck with it! Sounds like a nice bit of property.

    FYI, for important points (like property corners) surveyors will often take multiple RTK measurements separated in time (by hours or days) to check their GNSS results. I do this and also run static and get OPUS solutions for some points as a extra check on my work.

    Sometimes those monuments “move”, due to natural causes and/or people doing things they shouldn’t.

    Sometimes the monuments weren’t originally set accurately by the original surveyor. Determining whether the “incorrectly” set monument now holds as the actual property corner (it’s been there for decades and everyone has acquiesced to it, so it is what it is) or whether the corner is actually at a more “mathematically accurate” location is a decision I leave to the licensed surveyor I work under. That’s a complex topic with a lot of case law.