How accurate is your total station?

Well, think of angular precision this way.
5" X 100m X 0.000 004 85 = +/- 2mm perpendicular to line of sight.
But you don't get 5 second precision until you've turned two direct and reverse sets.
For one bs and one fs (I.e. not double centering) a 5 second station is only precise to 10" under ideal conditions (and neglecting centering and leveling errors etc.) which equals +/- 5mm perpendicular to line of sight.

5" at 141 m (diagonal of your site) is 3.4 mm. A better instrument wouldn't give much improvement unless you can center it and the target better than that and need that precision.

Be aware, though, that the spec if for a 2-direct, 2-inverse averaged angle.

In order to get 1" results from a 1" gun you are going to need top, top quality tripods, tribrachs, and glass maintained in top, top condition. Even then you will see results only when the environmental conditions are just right. For 99.9% of work most of us do 1" is overkill. I really doubt that you want or need a 1" gun.

And there usually isn't a lot of cost difference between a 3" gun and a 5". So the 3" seems to be the sweet spot. That's what you see a lot of. But you will still have difficulty seeing 3" results unless you really work at it.

fobos8, post: 453974, member: 13166 wrote: Hi guys

I'm going to be buying a robotic total station soon (second hand) and I'm not sure if to get 1", 3" or 5". I mainly do smallish surveys - sites not much bigger than 100m x 100m and setting out jobs on construction sites.

How accurate is your gun and what level of accuracy do you think I need for this type of work?

Kind regards, Andrew

First you need to identify what you mean by 1, 3 or 5 second gun. Is that least count or DIN Spec? Read careful if the math says that's a concern, which brings us to consideration 2.
What sort of layout are you doing? More particularly, what is your total error budget for the type of layout? The direction specs of the instrument are only part of the story. There's instrument and target centering, prism quality, etc. Add to that our own physical abilities. Point being that very tiny number grows quickly.
My experience says the 5 second will do most anything you describe, as long as 'setting out' doesnt include column line control. The business guy in me says think several years out and buy what works for that type of work.
Best of luck, Tom

I should mention that my numbers above apply to the 68% or 1 standard deviation level.
At 95.5% or 99.9% the numbers would be 2 and 3 times worse, respectively.

Regardless of what the specs are, get the most accurate you can afford. If you are working alone with a robot, its harder to check things, i.e check elevations with a level. Makes that fuzzy area around .04' less likely to be a measurement error.

okay guys - many thanks.

how accurate is your gun

I don't do construction layout...ever. My current instrument is a 5" gun. I cover long distances with GPS and just traverse to the corners in deep canopy. I usually traverse less than 1/4 mile. I turn 1 direct and 1 reverse angles to each traverse point; but, I'm not trying to improve precision. I'm just checking for blunders.

I started with a 20" Nikon. I turned 3 direct and 3 reverse angle to each traverse point. I typically had closures better than 1:25,000 using that procedure and instrument.

It depends on your work and procedures. If you're willing to turn more angles, you can use a less precise gun. I don't know if it makes financial sense, though. When I started, I just used what the equipment I had. I was a pack mule when traversing. As a solo operator, I was carrying 2 tripods, the instrument and the range pole on every traverse. It wasn't efficient, and I upgraded as soon as I could.

Mark's advise seems good to me.

fobos8, post: 453984, member: 13166 wrote: okay guys - many thanks.
how accurate is your gun

Mine is 3".
But Mark's comments earlier are 100%.
My TS comes from a line of 3 2 & 1 second instruments all same external looks.
Been told production would be say 3", so pull required number of instruments out and calibrated them to 3".
Next the 1". Pull enough off to fill an order and ensure they met 1" specs.
My 3" could be good for 1",but only confirmed it to 3"

But next is user input.
A 1" instrument might be reading such but due to other factors its anywhere.

Not sure which part of that answers your last question.

There was an article I read, concerning daily procedures for qualifying under an ASTM or ISO certification, which called for three direct and reverse observations to be made within the din specs by the manufacturer. This may have been cribbed or paraphrased from a certified publication.
Is something like that true ?

fobos8, post: 453974, member: 13166 wrote: Hi guys

I'm going to be buying a robotic total station soon (second hand) and I'm not sure if to get 1", 3" or 5". I mainly do smallish surveys - sites not much bigger than 100m x 100m and setting out jobs on construction sites.

How accurate is your gun and what level of accuracy do you think I need for this type of work?

Kind regards, Andrew

Hello Andrew.

buy a leica 1ƒ? instrument if you can afford it, and never have to be worried about the direction readings it is spitting out

We own 3ƒ? and 5ƒ? instruments and I would have liked one of these to be a 2ƒ? or 1ƒ? model for construction layout.

Scott Zelenak, post: 453977, member: 327 wrote:
But you don't get 5 second precision until you've turned two direct and reverse sets.

For the modern leica unitƒ??s Iƒ??ve tested this isnƒ??t true. You get the manufacturers stated accuracy straight up by turning angles in one face. Turning 2 faces makes little difference. You may as well just average 2 pointings for the good it will do with these instruments.

For one bs and one fs (I.e. not double centering) a 5 second station is only precise to 10" under ideal conditions...

For the units Iƒ??ve tested this doesnƒ??t improve by using both faces. So this is not a useful rule.

Scott Zelenak, post: 453981, member: 327 wrote: I should mention that my numbers above apply to the 68% or 1 standard deviation level.
At 95.5% or 99.9% the numbers would be 2 and 3 times worse, respectively.

The units Iƒ??ve mentioned do not produce normal distributions, so these numbers do not apply. The distribution of angular residuals from my adjustments are U-shaped and have no tail. so if you calculate 3-sigma it means nothing.

Iƒ??d be wary of making assumptions about surveying instruments specifications based on perfectly distributed randomness. I donƒ??t see it from modern leica instruments at least.

In metric terms I understand that at around 206m, the order of 1", 2", 3", 5" equates to 1mm, 2mm, 3mm, 5mm. We just do engineering surveys and have 1" instruments, and the main original reason for that was so that the accuracy of the instrument couldn't be challenged (unless someone wanted to dig deep and go for the 0.5" or better). Having said that, to date nobody has challenged our equipment, but it is comforting to do a bit of control with the TS and then check it (or vice versa) with optical level and get 1 to 2mm of vertical difference and then be able to muse that that is what the TS should be able to achieve.

Conrad, post: 454044, member: 6642 wrote: For the modern leica unitƒ??s Iƒ??ve tested this isnƒ??t true. You get the manufacturers stated accuracy straight up by turning angles in one face. Turning 2 faces makes little difference. You may as well just average 2 pointings for the good it will do with these instruments.

I use my Nikon similarly. I was bought up through Theodolites and when EDM arrived believed F1 F2 was the ONLY way to achieve accuracy.
I decided sometime to trial F1 only taking just as much sighting care as I did with 2 faces.
Closures never suffered and its very rare to read 2 faces now.

You need to know your instrument and watch your procedures. Good results follow.

Richard Imrie, post: 454047, member: 11256 wrote: have 1" instruments, and the main original reason for that was so that the accuracy of the instrument couldn't be challenged (unless someone wanted to dig deep and go for the 0.5" or better). Having said that, to date nobody has challenged our equipment, but it is comforting to do a bit of control with the TS and then check it (or vice versa) with optical level and get 1 to 2mm of vertical difference and then be able to muse that that is what the TS should be able to achieve.

Same with us. As we are mostly in a 3rd party QC role, you can't be bringing a knife to a gunfight, so 1" is the way we go. You can imagine my feeling one day when a guy turned up with a TDRA6000. This was closely followed by a sense of relief after I realised he couldn't even set up over a point, having relied on Spatial Analyst to do resections all the time!

Otherwise for day to day stuff we run 3"

I hate disagreeing with Conrad.
In fact, I agree with his assessment of Lieca total stations.
His testing posts were impressive and interesting. However, the sample size was one (or two?) instruments in a test not representative of either the ISO or DIN spec. Correct me if I've misremembered.

Increasingly today construction contracts include a specification of tolerance at a predetermined confidence level.

The OPs choice of instrument should meet contract specs based on accepted industry testing and practices.

His client, E&O insurer and the courts probably won't accept anecdotal evidence to the contrary be it from Conrad or myself.
Until the manufacturers and standards agencies change their testing and instrument tolerance specifications we are stuck with the current accepted methods of specifying an instruments precision.

Conrad, post: 454044, member: 6642 wrote: Iƒ??d be wary of making assumptions about surveying instruments specifications based on perfectly distributed randomness. I donƒ??t see it from modern leica instruments at least.

The worst thing to happen to statistical analysis was when Gaussian distribution started to be called Normal distribution.

Scott Zelenak, post: 454053, member: 327 wrote: I hate disagreeing with Conrad.
In fact, I agree with his assessment of Lieca total stations.
His testing posts were impressive and interesting. However, the sample size was one (or two?) instruments in a test not representative of either the ISO or DIN spec. Correct me if I've misremembered.

it's now up to 5 instruments, two 1200+, two ts15 and one ts16. all conform to the same scheme of error distributions. instruments of same spec test to within 0.3" of each other for magnitude of error in the major harmonic and similarly for the minor harmonic.

if you told me the layout of the iso test i could tell you the readings and results for these instruments ahead of time to a good degree of confidence.

i don't know the arrangement of the DIN test, but if all that is considered is differences between faces at particular points around the circle then both flavours i've tested (3" and 5") will test the same: they will be 1" total stations. that is because the instrument has diametrical angle encoders, and each direction is the mean of two circle readings 180* apart. an odd error harmonic needed to produce FL/FR differences cannot exist with an equal number of encoder read heads.

when leica had a single read head TC1010 the instrument spec was given as DIN. when leica went to diametrical read heads 1100 series the spec was ISO. i don't know if intentional, but with 2 read heads FL/FR differences should disappear for a well made instrument, and when tested in a scheme focussed on FL/FR differences a diametric instrument will test like a 1". was the iso test needed for newer generation of instruments with different error sources? without knowing the DIN test procedure i can't opine any further on that one.

If I recall correctly, Conrad was working under very well controlled conditions and found that a major source of error in his instruments was systematic, in that the error plots showed a sinusoidal dependence on the direction pointed. The manufacturer chose to correct this to a degree that depended on the instrument specification.

Under those conditions Gaussian statistics don't apply.

If random errors dominate for your instrument or for your procedures, you will see results more like the Gaussian model.