Why I preach 3 minute observations in canopy...

Shawn Billings

Shawn Billings
5PLS
Yesterday I surveyed the residential lot outlined in red...

aerial.png



There is a 3/8 inch rebar stake I needed to tie in near the Southeast corner of the Lot next to the 7' tall brick column post seen in the street view picture. There was some sky to the South, but the column blocked a lot to the Northwest and there is a thick holly tree to the East, and of course all of the trees seen to the Northeast. To the South, on the other side of the road running East-West is a thick wall of pine trees. The only open sky was somewhat high to the South.

street view.png


I started the observation at 14:08. After about a minute and a half, I finally started getting a repeatable fix. After about a minute (66 seconds) I acquired my required confidence of 10 with 11 independent fixes (shown in the screen shot in dark blue).

L327_20170425-14.11.15.png


At 14:12 I had almost 2 minutes of fixed RTK. I believe I had three engines at one point, but I cannot recall for certain. It was a minimum of 2 because I have my receiver set to only log epochs with at least two engines fixed with this profile. Notice the time is 114 seconds. Eventually the fix was lost...

3_COLLECT_20170425-14.12.28.png



I waited for a new fix, expecting it to agree with what I already had and collecting the necessary epochs to finish the shot. I got several fixes that did not agree, but for some reason, the receiver did not reject this average as it should have. So I manually stored this point and started again...

L327_20170425-14.21.02.png
 

Shawn Billings

Shawn Billings
5PLS
On my next attempt, I could not even get past the first phase. I did however have a handful of epochs that were spread apart by more than two minutes of time... Notice that the reddish brown group has two epochs that are separated by 142 seconds.

3_COLLECT_20170425-14.28.28.png



After a bit more time, the group grew to five epochs. There was four minutes of separation (243 seconds) between these five epochs. After waiting an not getting anymore fixes. I manually stored this shot as well.

L328_20170425-14.30.27.png



These points, L327 and L328, were on the same point, so they should theoretically inverse to zero. Without knowing what I know today about time separation in canopy, I would have sworn that L327 was superior to L328. I had more epochs and more fixes. But in reality, L328 proved to be the right point. I verified it by moving to the sidewalk in the open and taping the point in from two points as a check. L328 was definitely correct. Unfortunately, neither of these points processed in DPOS. For a tricky canopy shot, I really should have given it more time to insure success with DPOS, but I knew I had what I needed with five epochs spread out by four minutes.

The inverse between L327 and L328 is (which can also be seen in the distance to last white box for the L328 screen capture):
AZ 330°11'23"
HD: 5.540'
dU: -9.992'
 

Shawn Billings

Shawn Billings
5PLS
So what is the moral of the story...
  1. When you are working in canopy make sure that you use validate. Validate gets a check on the fix at the end of the observation. This represents the longest elapsed time from the start of an observation to the end of the observation.
  2. Make sure that you have your minimum observation time set to 180 seconds. This will insure that the validate doesn't validate the fix until at least three minutes have elapsed.
  3. Don't manually override the process unless you know what you are doing. Here I never was able to collect my full observation setting for control, but I knew that once I had fixes repeat for more than 180 seconds that I had a good fix.
Where does this 3 minute number come from?
It's based purely on my own experience with the LS. In canopy, it's not rare to see a bad fix repeat for 30-60 seconds. It's less likely to see it repeat for 120 seconds, but I've seen it. I've never seen a bad fix repeat after 180 seconds. I can only speculate that the satellites move just enough in that period of time for the conditions that create a bogus fix point to unravel. I don't know if this 180 second timeframe applies to other receivers. I've tested a lot of receivers, but never as much as I've pushed the Triumph-LS. If you see a bad fix repeat itself after 180 seconds, I'd like to know about it. But for now, I feel confident in repeat fixes with 180 seconds between them.
 

Adam

Well-Known Member
5PLS
So what is the moral of the story...
  1. When you are working in canopy make sure that you use validate. Validate gets a check on the fix at the end of the observation. This represents the longest elapsed time from the start of an observation to the end of the observation.
  2. Make sure that you have your minimum observation time set to 180 seconds. This will insure that the validate doesn't validate the fix until at least three minutes have elapsed.
  3. Don't manually override the process unless you know what you are doing. Here I never was able to collect my full observation setting for control, but I knew that once I had fixes repeat for more than 180 seconds that I had a good fix.
Where does this 3 minute number come from?
It's based purely on my own experience with the LS. In canopy, it's not rare to see a bad fix repeat for 30-60 seconds. It's less likely to see it repeat for 120 seconds, but I've seen it. I've never seen a bad fix repeat after 180 seconds. I can only speculate that the satellites move just enough in that period of time for the conditions that create a bogus fix point to unravel. I don't know if this 180 second timeframe applies to other receivers. I've tested a lot of receivers, but never as much as I've pushed the Triumph-LS. If you see a bad fix repeat itself after 180 seconds, I'd like to know about it. But for now, I feel confident in repeat fixes with 180 seconds between them.
I bumped up to 4 minutes for harsh enviroments for a factor of safety. Just in case. I have seen one go over 2 but not 3.
 

Shawn Billings

Shawn Billings
5PLS
I feel like I should probably say this. This post is specifically about pushing the LS in bad locations. In the open, three minutes is not really necessary to prove the fix. I will often occupy a control point for three minutes in the open anyway because I can refine my position just a little bit better. This three minute rule is for working in canopy.
 

Nate The Surveyor

Well-Known Member
I can also comment that if you stop an observation, short of the full three minutes go ahead and press "reset rtk". Thats U1 on my unit. Then, let it gather a few more epochs. This is ok, where sky is pretty good. This artificially gives a validate at whatever time you have given it. (if it continues with the same group).
I sometimes do this at 90 seconds, or 125 seconds. (fairly open areas). This is a "use at your own risk and judgement", depending on how obstructed it is.
 

John Evers

Well-Known Member
5PLS
Shawn,

In your second to last screen shot, why am I seeing (2,112,1) for the blue group. It looks to me that the values for the blue group should be (2,2,1) based upon the vertical time plot.

Am I missing something?

Regarding the "reading the x-ray" comment, at quick glance I can now see the brown group as truly the superior group. What an awesome machine.
 

Shawn Billings

Shawn Billings
5PLS
I'm glad you mentioned that blue group, John. I had a thought on it also. It's 112 seconds between two fixes that agreed with each other and that were not correct. 112 seconds is dangerously close to 120 seconds (or two minutes), which is why I have been using 180 seconds (minimum).

As far as 112 seconds or 2 seconds, the time is not to scale on the vertical chart. It only shows the chronological order, but does not represent the duration.
 

Darren Clemons

Well-Known Member
Very good detailed information Shawn, thanks for sharing. Several things stand out to me in you photos and comments.

First off, this is why the change in the way the LS "holds" on to the data in the groups implemented last year was so crucial. You probably wouldn't have had a chance to store a point like this before with the LS throwing everything out and re setting after 10 groups...most of which would be one click epochs. It was a critical change those of us who used it in heavy canopy areas saw. The engineers obliged and made the changes happen. No other GPS design team does that!

I have had/stored many, many shots as you describe here with very similar experiences. I have also noticed a few times where, in a situation similar to your second screen shot (233,114), the LS will begin "clicking" away with new epochs but the screen will NOT be changing/updating or adding to the total number of epochs at all. It basically goes into a "freeze" mode. Every time it has done this, the point it has "held" currently has been incorrect. If I see this happen now, I just stop, reject and start over. I will also confirm, I have seen a few bad shots of around 2 minutes store, but never one with 3 minutes or more. No matter how many groups I have or how long I've been on a point, as soon as a group shows up with even 2 epochs separated by over 200 seconds, I'm totally confident that's a good shot.

In heavy canopy areas and shots like this, I have gotten used to hitting "reset tracking" and/or "reset engines" several times during the shot attempt. I'm not sure weather or not this many resets actually helps but most times you will get a "click" or two immediately or shortly after resetting. Also, I will sometimes turn the LS at 90 degree intervals (not very relevant in your scenario here with the wall) to hopefully find the "sweet spot" in the tree canopy and can sometimes see the interference levels drop a bit by doing so and that sometimes also leads to some additional clicks.

Points such as this is where I sometimes, depending on the current survey and accuracy, will almost always be collecting in stake mode instead of collect mode. By the looks of the homes, I'd guess there was a fairly recent plat of lots you were following. We'll have design points calculated on those before we go to the field and after shooting two, will have moved and rotated those points to our survey. Then we have a design point for each pin, most times within a few tenths, again depending on the age and accuracy of the old survey. With a very difficult point such as this, I will hopefully have found enough of the subject lot pins or maybe even an adjoiner or two to get a "feel" for how tight the survey is and what to expect, accuracy wise, on each point.

Then I'll begin collecting on this tough point, in stake mode. Let's say, for example, after adjusting to my first two, I found and collected three more pins on or around the existing lot survey and all were within 0.15' or better. I now have a pretty good idea that, if this pin is original and undisturbed, I should hit it with a similar result. My groups start filling up, with the first one being held showing me a DTT of something like 2.3'. After a few more clicks, a group pops in three quick epochs and takes over as the "lead" group and my DTT switches to 0.10'. I know with a pretty high certainty that that's my point I'm looking for. While a lot still has to be done to acquire the amount of data and accuracy I want/need before I would have a coordinate I can store, this definitely gives me extra feedback, very early in phase one of the RTK.

Basically, with all that's being discussed in this thread, what it means to me is, yes, validation (phase 3) is crucial and very critical to the RTK process of the LS. However, in those extremely tough spots where it just won't complete, there are still many, many ways to accurately check, double check and attain the necessary data for a good confident point. Throw in one or two 15 minute PPK sessions on top of any of that for another check and you'll have a coordinate that you'll have absolutely 100% confidence in.

I've only said this about a million times - there's nothing else out there that's even remotely close to giving you that kind of confidence on these kinds of points - nothing!
 

Shawn Billings

Shawn Billings
5PLS
What's happening Darren, when the LS gets a fix and no epochs are being added...

If the LS has a fix and is averaging a position then loses the fix and gets a new fix and the second fix doesn't agree with the first it collects the epochs from the second fix in the background. Once enough epochs are collected from the second fix, it will show fail on jumps and start over. For some reason this is broken in current release, only working intermittently. You can start over, but I would recommend waiting for the LS.
 

Darren Clemons

Well-Known Member
What's happening Darren, when the LS gets a fix and no epochs are being added...

If the LS has a fix and is averaging a position then loses the fix and gets a new fix and the second fix doesn't agree with the first it collects the epochs from the second fix in the background. Once enough epochs are collected from the second fix, it will show fail on jumps and start over. For some reason this is broken in current release, only working intermittently. You can start over, but I would recommend waiting for the LS.
I figured it was something like that Shawn. I've waited for it a few times and it does finally fail jump. When it does this, I'm sure it's getting conflicting info so I most times just choose to stop, reject, reset engines and start over...
 

John Evers

Well-Known Member
5PLS
I also hate to lose any raw data, I rarely reject anything. There is always a chance a "bailed out" RTK shot, will provide a Fix PPK solution, that just might match 3 other RTK shots.

No matter the strategy, repeated shots is what lets me sleep well at night, even without a drink.
 

Darren Clemons

Well-Known Member
I hate to lose the raw data
Very good points everyone on not losing the raw. I was unsure that, when the LS was exhibiting this behavior, if it had "locked" up, so to speak, therefore I chose to start over. Now that I understand more of what's going on, I'll try and be more patient....not one of my strong suits unfortunately :(
 
Top