1 cm level vertical accuracy at 90% confidence

[Tyler]

Is it possible to achieve 1 cm level vertical accuracy with a 90% confidence interval with RTK methods? We have a Triumph 1-M base and the Triumph LS + receiver. Or is this only possible by performing a static session?

I have no issue getting a relative accuracy to meet ALTA/NSPS standards at 2 cm + 50ppm using RTK at our current settings.

Current Settings: Stop after 50 epochs, Auto Accept [H/V = 0.131ft/0.23ft], verify with engine reset, conf level = 5, min RTK engines = at least 1, validate result = at least 1 engine.

Utilizing GPS, GLONASS, BEIDOU, and GALILEO.

 

[Tyler]

Is it possible to achieve 1 cm level vertical accuracy with a 90% confidence interval with RTK methods? We have a Triumph 1-M base and the Triumph LS + receiver. Or is this only possible by performing a static session?

I have no issue getting a relative accuracy to meet ALTA/NSPS standards at 2 cm + 50ppm using RTK at our current settings.

Current Settings: Stop after 50 epochs, Auto Accept [H/V = 0.131ft/0.23ft], verify with engine reset, conf level = 5, min RTK engines = at least 1, validate result = at least 1 engine.

Utilizing GPS, GLONASS, BEIDOU, and GALILEO.

*** ALTA standard at 2cm was for horizontal (my mistake). With these settings it appears that our relative positional accuracy is between 0.03 & 0.09 feet using the Relative accuracy tool on 8 repeated observations.

 

[Matthew D. Sibole]

Yes Alta standards are 0.06561667 + 50ppm. With that said your distances of 0.03 & 0.09 may meet the standard depending on how far the points are apart from each other. When you run the relative accuracy with the ALTA standards selected on just your average points does it say that any points failed?

 

[Tyler]

Yes Alta standards are 0.06561667 + 50ppm. With that said your distances of 0.03 & 0.09 may meet the standard depending on how far the points are apart from each other. When you run the relative accuracy with the ALTA standards selected on just your average points does it say that any points failed?

I think I may have distracted from my original question when I posted my correction. I was wondering if it's possible to get 1 cm (0.03 feet) vertical accuracy using RTK methods. With my current settings I'm only able to get repeat measurements within 0.09' vertical but I have not experimented with other settings.

 

[Shawn Billings]

The best way to improve vertical precision, in my opinion, is longer RTK observations (300 seconds) and repeat them (three observations or more) with at least an hour of separation between them. 1cm is a tough requirement. I have not really needed to get a high precision vertical since multi-constellation became so good, so it may be easier. Try for yourself and see what procedures get you there, but I think your best results will be averaging several solutions together.

 

[Jim Frame]

I was wondering if it's possible to get 1 cm (0.03 feet) vertical accuracy using RTK methods.

Is it possible? Sure! Can you do so reliably? That requires more information. Do you need datum accuracy, or just relative accuracy? Whose specifications do you have to meet? How far between base and rover?

I'm currently negotiating a contract to do a mini height modernization project to bluebook a new driven-rod mark to replace one that got knocked out by a contractor. I have to get 2 cm accuracy relative to 3 existing marks within about 7 km. Cost is going to be just under $30k (includes the new monument, not an insignificant item). I tend to roll my eyes when people talk about mm vertical accuracy and RTK.

 

[Nate The Surveyor]

. I tend to roll my eyes when people talk about mm vertical accuracy and RTK.

Unless it's 30 or 40 mm!

 

[Tyler]

Is it possible? Sure! Can you do so reliably? That requires more information. Do you need datum accuracy, or just relative accuracy? Whose specifications do you have to meet? How far between base and rover?

I'm currently negotiating a contract to do a mini height modernization project to bluebook a new driven-rod mark to replace one that got knocked out by a contractor. I have to get 2 cm accuracy relative to 3 existing marks within about 7 km. Cost is going to be just under $30k (includes the new monument, not an insignificant item). I tend to roll my eyes when people talk about mm vertical accuracy and RTK.

From the inputs in our LS unit, we are looking for elevations based on NAVD88. I've attached our system settings below. The base and rover are typically within 1 mile.

When I look at the results from our raw data file below... is the 95% confidence (sigma h) the confidence in the absolute vertical position on this vertical datum? I had always assumed that these were relative accuracies but now I'm not sure.

 

[Jim Frame]

I'd take that 0.016 height error with a very large grain of salt. And I recommend becoming familiar with NOAA Technical Memoranda NOS NGS-58 and NGS-59 before certifying any elevations derived from GNSS measurements. At the very least, understand what standard you're purporting to comply with.

 

[Tyler]

I'd take that 0.016 height error with a very large grain of salt. And I recommend becoming familiar with NOAA Technical Memoranda NOS NGS-58 and NGS-59 before certifying any elevations derived from GNSS measurements. At the very least, understand what standard you're purporting to comply with.

That is what I figured but I'm still curious where that number is coming from. Several of our other raw files were in the +/- 0.1 foot range so the example I posted may have been extreme but I will take these statistics with a grain of salt.

Thank you for the reference materials.

 

[Shawn Billings]

That 0.016' value will be a relative accuracy from the base, Tyler.

Jim is correct regarding heights from GNSS. As I recall, NGS states the current geoid models are no more accurate than 2-5cm, which I believe would relate to datum accuracy. For most normal projects I'd think that the relative accuracy of the geoid models is much better than that.

The 2022 datums (expected in 2027) will bring an entirely new perspective on heights derived by GNSS, but alas we are not there yet.

 

[Shawn Billings]

Tyler, from my experience, short observations like this (18 seconds) can provide very optimistic statistics because the sample size is so small. Letting the receiver average for a while will allow it to experience some variation. Imagine that the errors of a vertical position generally follow a sine wave like pattern. If you collect data from peak to peak of the cycle, then you have a pretty good estimation of the precision of the observation. However, if you only observe briefly while the position is at the top or the bottom of the cycle, then the graph looks more like a horizontal line with very low variation. The statistics will appear to be much better than the actual position is. As for how long that time needs to be, I cannot comment. Several years ago, I landed on about 4 minutes, but that was with different processors and only two constellations.