GNSS RTK Accuracy
Jhframe. I have not yet been able to use a javad. But have talked with people who work for javad and or distribute it and use it personally. Javad i have no doubt probably has some of the most stringent algorithms and methods for testing before accepting. When the man created the ashtech z models the proved to be some of the most sophisticated and durable receivers on the market and i will have to be completely honest i love Trimble but those old receivers for geodetic absolute work derived some of the best data. I think my geek side would love to use the Javad system but i know I would probably never get anything done work wise as i would probably push it to its limits and get so caught up in tweaking things beyond needed for every day work. I will at some point probably get one and test it for a week or so. Trimble is great for reliability and accuracy and production in many different ways for many different tasks. I know it and know the leica and novatel. I know which one is truly more accurate and precise but thats not all that it takes. So i use trimble because thats what we have and can make the field to office flow well with my limited abilities. Novatell recievers are probably better than trimble in many situations but neither of them meet all the requirements listed for use so understanding that is why I develop field and office procedures to catch that 5% failure under different circumstances. The javad i have only seen results back years ago in a true testing environment along with many others. It soared above all the rest. In almost all cases. The man was a genius. A true scientist. Not a marketing wiz. His passion was seen through his creations for sure. He saught the most accurate and reliable in all circumstances. Thats why I think those solo folks once seen it work grabbed it and held it close . Its a unique system outside the box no pun intended. If all i was doing was setting rtk control and rural boundary work thats all i would own honestly. No other equipment needed maybe a steel tape. I honestly might buy a cheap total station is all but even then I might just rent one when needed. Now the r12 is catching up in those hostel environments and i am getting amazing results in situations i shouldn’t be getting results. I set 3 points in nasty area all able to be seen between and two more in wide open i could see into two of the 3. 3 observation 2 first day 4 hrs apart 1 2nd day at the 2hr mark between 1st day all different base locations. I had .03 hz vt error after comparing robot results measured rounds 4 with traverse kit running through them all and cross tied where I could. I get better than that a lot of times but this was flat out bad gps area . At a power plant under main power lines and tall metal buildings. Next like literally to power grids like 50 ft. Now my base stations were set outside in good areas just the control for mapping set in the bad areas. I did some shots to the building corners with r12i and robot reflector-less as well. Those creeped up closer to a tenth when compared independently. Rtk has become the tool of choice. By my crews and they seem more productive in most cases even when re observing control and bldg corners twice. For the mapping we used the robot as some items were located inside the buildings themselves . I imagine those in the midwest and such hardly even need a robot or total station with open skies. Here on east coast lots of trees from pine to holly magnolia are the worst. Oaks hickory poplar leaves don’t seem to be as bad except after a rain when leaves are wet Or from a heavy dew once the hardwoods leaves dry it tightens down pretty quickly. I do wish i had more info on to look at like javad gives. But i can barely get crews to monitor the sn and azimuth and elevation of sats rms during the minimum 180 epochs. . So getting them to watch all the other javad gives might be difficult. Lol. If i am observing personally i watch and might remeasure or stop and wait and the. Start agy. Seeing a satellite go out of horizon or one coming in in middle of an observation I usually will not hold that as gospel until i have more information. Topo is a little different as i get into production mode.
I was basing my initial thoughts on a video about RTK observations where the surveyor ran a similar test but got very different results
I didn't see a histogram in the youtube video, but looking at the time series plots I would bet it would have been more or less Gaussian if they plotted it out that way.
They may have only been observing one or two constellations, which could account for that data wandering a lot more than yours. Their rover was also 4 miles from the base station - how far was yours? They also only gathered about 15 minutes of observations, and who knows what kind of receiver and what RTK engine they had running.
It's tough to compare these sorts of experiments unless you have the exact same setup and under the exact same conditions. Like Jim mentioned, RTK results with newer receivers in open sky are darned reliable these days, but the proof is in how they react under adverse conditions.
Another thing to consider is that for most RTK surveys, the rover is seeing continually different conditions as the operator moves around the site, regularly dropping & adding SVs and frequencies, losing initialization and regaining it. When we set up a receiver in ideal conditions and proceed to log tons of data at one single location, we're not really replicating what happens during a typical RTK survey.
Take a look at this excellent paper from ODOT concerning the incorporation of NRTK observations into control networks:
https://www.oregon.gov/odot/Programs/ResearchDocuments/SPR304-821_UpdatedSurveyStds.pdf
It's focused on RTN observations, but is probably the best and most recent analysis of how long to observe and how long to wait before repeat observations. I think they are a little conservative, but then again a lot of my experience is with base-rover and full-constellation RTN, and they are trying to replicate static-level results. Regardless, there is a lot of good info there...
"Another thing to consider is that for most RTK surveys, the rover is seeing continually different conditions as the operator moves around the site, regularly dropping & adding SVs and frequencies, losing initialization and regaining it. When we set up a receiver in ideal conditions and proceed to log tons of data at one single location, we’re not really replicating what happens during a typical RTK survey."
I completely agree, I cant think of a straight forward way to incorporate this factor into reliable testing and observe enough data to make the test as sounds as possible, so I'm all ears if you have any suggestions?
Thanks for the link, 108 pages long... I know what I'm doing for the rest of my Sunday.
Well we can test. Through redundancy and building in independent checks but for topo that goes out the window no one is going to shoot a bunch of ground shots twice. I do however every so often put a dot on the ground or x with paint years ago when initialization were not as good as they are today would carry old golf Tees stick it the ground flush with surface and call it a ground shot. Before map screens on data collectors. I kept a pocket field book and whenever i lost initialization would go back to that point for a check. Back then if you lost initialization all points in that time frame were suspect. Now i just navigate back to a point or after swapping batteries or when job is done zig zag my way back spot checking. The reality is it has become so good that 95% of the observations are probably within tolerance and therefore meet any state standards in that regard. I just build checks. I use a vector spreadsheet in office to identify any suspicious points. I catch a few and usually can see the rms or hz vt precision ramping up to the worst point. I had to send one crew chief back to a job he was storing a lot of bad data just was not paying attention at all and overwriting the tolerances. He was having a bad day. I looked at him and said. I don’t care how or when but you have 8 hrs left in the budget. Get it done. He did. And he spent more than 8 finishing the job and correcting his blunders. He admitted he didn’t pay attention had some issues going on the week before. I can remember a sokkia total station and when the battery was getting low it made a distinct sound and rhythm. You could get some weird crazy distance to a point like thousands of feet off. Had one one time literally just for several shots no matter the angle or distance just had the same exact distance every time. So we learned and was intune with the equipment so we usually caught that. No difference for gps. Set up redundancy. Spot ck every so often and just roll with it . It’s probably fine in ideal conditions. In canopy i am more Leary but its not supposed to be good there. I am not afraid to use it for control In canopy because those redundant observations at a gap in time might save me hours of traversing and cutting line if it works. If not well off to traversing. I have had 3 observations per point in canopy for many points and i bet less than 5% were bad that i needed to disable those vectors. What i see is i need 22 sats minimum good geometry pdop vdop etc and just let her burn. I kept a log early on when testing the R12i myself. 98% of the time if it got 180 epochs like clock work in the 3 minutes it was good enough to pass alta specs and good enough for property corners. If it struggled then it was suspect same if I dropped below 22 sats. This all was me on several projects i did on my own time mostly and comparing to robotic traverse . Now i do build a network and perform a least squares on all projects. Which can expose some problems as well.
Looking at Oregon DOT’s summary, I don’t like their 5 minute observation time. It’s not static data, so once precision numbers stop decreasing a longer observation isn’t needed. In fact, too long a RTK observation can degrade if satellites get into worse geometry or you lose them from obstructions.
Given that NGS NOS 92 calls for 5 minute observations when using NRTK for secondary and local control, I'd say that Oregon's DOT standards are apt. I generally instruct crews to collect 3-5' redundant observations for projects I run.
If anything, just run a RTK-LOGGING survey style. If the crew ends up on a point longer than you think is necessary for RTK, just postprocess baselines and disable the RTK vectors.
In fact, too long a RTK observation can degrade if satellites get into worse geometry or you lose them from obstructions.
If that were true, static observations would be "worse" than RTK observations because they are "observed for too long". All GNSS observations bounce around as geometry changes. That's the point of setting those time windows long enough to capture that movement, as well as returning under a different constellation. We have no idea when the end users of this control will be observing, after the values are published, and so we need good network accuracy.
I don’t like their 5 minute observation time. It’s not static data,
Indeed, 5 minutes is not a traditional static observation. The entire point of the study was to investigate what it takes for NRTK observations to be comparable with traditional static observations so as to improve efficiency when running control networks. Several other studies (check out the Allahyari and Weaver papers referenced in the appendix) have found that the ideal window is somewhere between 3 and 5 minutes, especially for vertical. Not to mention that not all classes require 5 minutes, just the higher-order ones.
Yep you nailed it again. Bare minimum 180 3 minutes for my control and boundary corners. Period anything less go back and so it again. I have tested NGS guidelines for base n rover and vrs or nrtk. It freaking works. Falls withen what they state most of the time.
I did that static job you helped me with not long ago. I followed to the T ngs guidelines for gps derived heights. Even when pressure from the top disagreed. Well it worked a minimum of 2 4 hours observations. Held one BM 1st order at first checked to all dini run levels and to 2nd 1st order BM. Yep we hit the requirements like a champ. Found 1 cors station with a huge vertical issue and reported that and as I guessed the antenna had been changed not updated. What’s great about it the client said we want yall on the contract now. No one else had met the requirements after the combed through all the reports of everyone.
The more darts or epochs you throw at the dart board or point. The more data we have to eliminate outliers and derive at a more confident solution.
I like the way you think. In the big picture 5 minutes is nothing. It takes most people longer than that to set a total station up over the point not to mention you might have to set up several times to even get to that point from where you started. In that 5 minutes one can do a lot if they don’t want to look at the dc. I will flag up check for other evidence. Make a sketch. At one time light a cigarette. Don’t do that anymore. Now go find a tree and water it. Lol.
<font style="vertical-align: inherit;"><font style="vertical-align: inherit;">Y QUE CUAL ES LA DIFERENCIA ENTRE EXACTITUD Y PRESICION , CREO QUE ES LO MISMO</font></font>
