On the HZ rates of GNSS systems

Omfg not again?!

?ÿ

Nate!!! Did you have a copy/paste incident?

No. But I think that discussion of this subject may yield some core information.

I'd think that all surveyors, both dumb and smart, would want to stay current with equipment.?ÿ

The reason I brought this up, is I've got a limited life span. Somebody somewhere will retrace my work. They will have the very best equipment that 2041 has to offer.?ÿ

Every survey we do, will eventually be retraced. Every error found. Wouldn't it be cool, if we could argue over 2mm someday, like we do 0.045 feet today? I think so.

Knowledge of equipment will be important.?ÿ

Some make comments disparaging Things that they don't understand. I've done it too. Dad used to talk about the vicious cycle, "get in debt, to get more work done, to pay off the debt". Don't skip personal time.

That's why I talk. It's disengenious to try to block people from equipment that:

Really does real time static processing.?ÿ Javad does.

That takes advantage of .2 second rates, deriving real answers, in between the even seconds. Javad does.

It works.

Time marches on.....

Someday, all gnss will be better than it is today.?ÿ

N

Posted by: @nate-the-surveyor

The reason I brought this up, is I've got a limited life span. Somebody somewhere will retrace my work. They will have the very best equipment that 2041 has to offer.?ÿ

0.19' error circle is not a result I would call that came from 'the very best equipment that 2021 has to offer. '

?ÿ

This is a VERY practical post. I am not from academia. I am approaching this from the very practical side. ie, experience with use. These are my own practical applications. Somebody from the Theoretical side of academia, could possibly re write this, and say this much better than I can.

There is a guiding principal that I apply. I call it the Casino theory. That is, the HOUSE wins all ties.

This means that, GIVEN enough time, (Large enough sample time) the GPS will EVENTUALLY "get it right".

Understanding what a GPS observation is, is important.

It is a small slice of TIME, with a small slice of GPS data, GPS receivers use this small collection of material, to GENERATE a coordinate. When you use RTK GPS, you get these small "Windows" of data. If the sample is small enough, it is not going to yield enough data, to gather an honest error estimate. If it is LARGE enough, then it gives you an honest answer to It's own error estimate. So, the size of the sample is critical. And, time is critical.

When you are out in the open, it is pretty easy to GENERATE this window of data. And, it's error estimate is usually quite accurate.?ÿ And, the resulting coordinate is usually within 0.03' to 0.05', in a short time. As the actual error increases, so does the physical size of the pile of epoch data, used to generate the actual final coord.

The trouble starts when you go into the woods. The amount of TIME needed to GET the shot, goes UP. And, the pile of data grows in it's "Dilution" as in, it has data in it, that yields a LARGER error estimate.

What all this says is: "To gather an honest shot, in the woods, you MUST enlarge the SAMPLE, of data taken, and properly review it, to determine it's location".?ÿ In the old days, this required copious amounts of time, and careful analysis of the data. I have taken observations that were 36 hrs long, and picked through the data, Not really trusting the data, but just messing with it. I later set control points in the wide open, and used a total station to observe the above 36 hr shot. Less than a hundredth difference. The software (Ashtech Solutions) only yielded a FLOAT. But, this became a philosophical key to understanding it.

Now that we have a mechanism that CAN work, we want to apply it.

I know this is too wordy, but this is the PATH I took.

What I have discovered, through trial and error methods, is that:

IF you let a GPS receiver sit long enough, and collect enough data, then it can get it.

What I am excited about is that:

A single observation, with the Javad unit, can give you the needed pile of information in about 1-1/2 to 4 minutes. This is frequently true.The biggest deal is the INIT coord value. This is the BIG ONE. Errors of 3-7 feet are typical. Javad has almost completely eliminated this error. This is done by various mechanisms to statistically determine the INIT coord. (Remember the Casino illustration above?)

After we get the right init, then:

This yields a pile of epochs, and their coordinates, that vary in size. This is the error ellipse that I am talking about. After much playing with it, I learned that close to 100% of the time, in the woods, The ACTUAL coordinate is within this ellipse. And, often the average of this coordinates is quite close. IF the epoch data shows it is a LARGE pile, it is less accurate. This is large in size.

So, If I get a Javad coord, thats got a little pile of epochs, It's error estimates are nearly always correct, and that the true coordinate is in that pile. If I get 2 observations that are 0.05' apart, with error estimates, that are built from 0.10' size piles, then we have a very close shot. Then, I get a 3rd shot that HAS larger error estimate, (0.26') and it's average is 0.14' from the other 2 points, then it is in agreement, since the 3 shots ALL include the actual coordinate. The error ellipses overlap. They all 3 include the CORRECT coordinate, and the error estimates are accurate. Just one of the observations is poorer.

I am sure that some body else could word this better.

But, that's the best I can do to explain it.

Here is a screen shot from yesterday:

Interpreting this screen shot goes as follows.

I stayed on this observation for 142 seconds. I got 622 Epochs There were 2 little clumps of data generated, from the 4 engines. The SIZE of the left hand square is 0.132'. This is the Horizontal shot spread. The right hand square shows that the vertical shot spread is 0.358'

The Star, in the left hand pile shows that the RTPK (In the field static shot). The RTPK Coord is 0.046 North ,0.022West, and 0.177' Lower than the RTK shot. This is real data, from working yesterday.

This is in a pine thicket.

The REAL coordinate is within that left hand square, almost 100% certainty. This statement comes from HUNDREDS of comparisons with the many old jobs I have tied into, over the years. It's tying into years of closed traverses.

I have not studied the VERTICAL components, to the degree I have the horiz.

Maybe this description can help those who are honestly looking for information. I have a feeling that some of you folks were not looking for the truth, but to pop a top.

That's fine. but, this is real data.

I shot the above point 2x, and averaged them The 2 shots were 0.06' apart. The weighted average was 0.02' from the above point 1500. The second shot had a larger error estimate. The left hand box was 0.18' square. It created a coordinate that was a weighted average, and held closer to the 1st shot.

Maybe this can answer any questions others may have. I can say that I have been playing with these concepts for around 15 yrs. It's been alot of fun.

I like what Javad has given us. It is a fine tool for surveying in hard environments. I recommend Getting 3 observations, (each one containing hundreds of epochs) if a point is CRITICAL. If it gives large error estimates, do it again. It's usually done with one observation is less than 4 minutes. Sometimes in as little as 80 seconds, in hard places.

Nate

?ÿ

?ÿ

?ÿ

?ÿ

At the risk of proving an adage I learned shortly after beginning my employment with Hayes Instrument Company almost 23 years ago, the adage being that the difference between a GPS salesman and a used car salesman is that the used car salesman usually knows when he's lying to a customer, I have a question about Nate's proposition.

Speaking of the base/rover systems I have dealt with the base broadcasts a new position correction once each second.?ÿ For the faster interval you propose Nate, the base would need to be broadcasting at a higher frequency than 1 second intervals.?ÿ I have never seen that option at the base.?ÿ The faster interval at the rover is there, not for greater accuracy but for other purposes, such as speedier navigation to stakeout point, shorter distance intervals between autotopo point, for example.?ÿ And for the higher refresh rate to be of real value, it would need to be based on calculating the instantaneous position more frequently than once a second.?ÿ I frankly after over twenty years of exposure do not know if that is even possible.?ÿ If it were I would think it would be exploited by the manufacturers.

Therefore that leaves me with a question as to whether the shorter interval reference position correction at the base is even a possibility.?ÿ I am thinking not, looking back to what I know from post-processed data (which is not all that much) the minimum interval I ever saw possible was 1 second.?ÿ If my understanding is correct, the faster update rates at the rover are then interpolated, therefore less precise than matched epoch positioning.

@jerrys?ÿ

?ÿ

I'll send you a PM, Jerry.

Nate just needs to move where there aren't so many dang trees.?ÿ

My Chevy is better than your (fill in the blank)

My 300 Super Ultra whizzbang Magnum is better than your (fill in the blank)

My Little Jiffy Ore Finder is better than your (fill in the blank)

My whatit is bigger than your (fill in the blank)

If I had a couple of million dollars (probably more), and a YEAR to travel the country (with my team of operators), then I could probably put together a "meaningful" (side by side) test of the half dozen or so of the most popular RTK Base/Rover sets (which I would have to buy outright obviously).?ÿ

Winter/Spring/Summer/Fall in:

The Temperate Rain Forests of the Northwest Coast,?ÿThe Mojave Desert,?ÿVarious High Mountains in the Mountain West (Aspen & Conifer), various places in the Mid-West, the jungles of Arkansas, the swamps and forests of Florida, the hardwood and conifer forests of the Northeast (I love Maine).

As well as various suburban conditions and Urban Canyons along the way.

When all is said and done, the results would [probably] be interesting, but wouldn't change much. Especially the opinion of those who have already make up their mind!?ÿ

When I first started where I work now, close to 10 years ago, we always took 2 separate 3 minute shots. We were using Trimble r6 base/rover back then. Eventually they decided 1 minute per shot was fine. But we still got bad shots every now and then even if the inverse between shots was good (some guys don't believe in losing and regaining lock as an extra insurance).?ÿ

Nowadays, I rarely use GPS since I do construction layout in subdivisions most of the time..I have the oldest GPS, a Sokkia grx2. Instead of shooting 2 separate points for a minute each I now prefer to shoot the same point number over and over again for 20 seconds each, getting a new lock each time, and look at the deviation between shots and based on canopy or the job I'm doing I could shoot it as many times as 6 times and let it use the weighted average between all the shots. I think this gives me a better coordinate than 2 separate shots. If I'm really wanting good accuracy I'll try to shoot a couple times, move on to other stuff and go back and reshoot control or other important points after some time has passed. It's almost never going to be several hours like some people on here have mentioned, but at least some time passing has to help.?ÿ

I donƒ??t care how many flashy lights and fancy algorithms it has, whether Javad or Trimble, when it comes to heavy canopy youƒ??ll have to pry my gun out of my cold dead hands.

@williwaw?ÿ

Or construction layout. Even when they only have to be within 0.010m, I'm not using GPS. Especially when I have to layout 50+ points for one townhouse.?ÿ

since you asked for pictures.

from a local Tersus vendor during one of his demo sessions. was thinking of getting one of those Chinese brands but have not decided yet. I was looking for the picture where the receiver was inside of a warehouse under a small clear acrylic roof panel and they claimed they got a fix but could not find it.

?ÿ

@jonathan50 Do you still get wrong init on those shots, or do you come re shoot them? How do you know your shot is good?

What's your typical bad shot/good shot ratio, in those places? How do you know that your data is good?

We are in the information business. How do you QC those shots? (Quality Control)

One bad shot can give you a serious problem.?ÿ

In advertising, there is something I call "Distracting Information". It's there to "help you swallow stuff that you ordinarily would not".?ÿ

It's the bikini girl, in the beer commercial. It's the gleam of the paint on a new car. It's the energy rating number on a Vacuume cleaner. It's the distracting information.?ÿ

Even in this, it's not "I got a fixed solution". It's how do we know if it's right, within certain tolerances. Or, what's it's error ratio? Is is 1 bad out of 50? 1 bad out of 20? 1/100?

-Fixed-

-Fix-

That's all good. But what tolerances??ÿ

What's the good / bad ratio?

We are "end users". We don't make this stuff.?ÿ

A poor man is better than a liar, says an old proverb.

Advertisements often avoid the real issues. I want to cut through the DI, and know.

Thanks for posting pics. Now, how many bad inits? And, how do you know??ÿ

Thanks

Nate

?ÿ