Page 2 clearly shows ?(vertical) and ?(horizontal) to be at 95%. This is done throughout the paper.
I'd Czech in with Petr Vanicek before making such a statement.
If anyone doesn't know who that is, all your least squares arguments are from here on out should be considered null and void.
Compare that usage to that of this NGS publication on the computation of error ellipses.
http://www.ngs.noaa.gov/PUBS_LIB/AlgorithmsForConfidenceCirclesAndEllipses_TR_NOS107_CGS3.pdf
You'll note that the variable names used on the Javad controller denote something quite different. It just looks dumb. If the screen shot is changed to ?x and ?y, that actually would be much closer to standard usages. It should be a simple fix.
[sarcasm]I'll be sure to let them know that Kent McMillan said it just looks dumb.[/sarcasm]
I guess you finally read the text and came to the conclusion that you were wro... er... not right. "It just looks dumb" is probably as good of an admission as I'll get, so I'll take it.
> "It just looks dumb" is probably as good of an admission as I'll get, so I'll take it.
Well, the usage is not a common one, which is why it looks so dumb. You'll note in this paper on algorithms for computation of error ellipses, ?x and ?y are explicitly used as semi-axes of the standard error ellipse to which scale factors are applied.
http://www.mar-it.de/NavGen/final_text3.pdf
k is an variable, generally less than or equal to 4 in the NGS paper your referenced! How would making the graphic "k?" make anything more clear to the user than stating clearly "95% conf. ellipse"?!? It's like saying the error ellipse is at X confidence! What the hell is X?
I'm incredulous at myself for taking you as seriously as I have this past week.
> > The cost for the setup is as follows: Triumph LS $13,990 HPT435BT(base radio) $2,759 Triumph 2 $4,990 TOTAL=$21,739 half the cost of other brand setups.
>
> $21739.00 is not half the cost of all other brand setups.
half the cost of other brand setups
half the cost of all other brand setups
> k is an variable, generally less than or equal to 4 in the NGS paper your referenced! How would making the graphic "k?" make anything more clear to the user than stating clearly "95% conf. ellipse"?!? It's like saying the error ellipse is at X confidence! What the hell is X?
The variable k is a standard variable used in the literature to scale the axes of the standard error ellipse for different levels of statistical confidence.
If you label the semi-axes ?1 and ?2 or ?x and ?y (which is a standard usage) you're by implication stating that the values are those of a standard error ellipse.
k?x and k?y are standard indications of a standard error ellipse that has been scaled to represent an ellipse of confidence greater or lower than the standard error ellipse.
"95% confidence (k=2.447)" is explicit and competent in a way that "?1 and ?2" are not.
Last word. Triple stamped it. No erasies.
Unmanned,
Does this mean I have to start a new thread to explain how Kent's suggestion is inadequate for elevation. The vertical scalar is 1.96 not 2.4477 so now they'll have to note that k for horizontal is 2.4477 and that k for the elevation (1D) is 1.96. That simple "95% confidence" note is really looking good.
> Last word. Triple stamped it. No erasies.
Well, I'm sure we all just would prefer to see RTK controllers that didn't mislabel fundamental numbers, wouldn't we?
> Does this mean I have to start a new thread to explain how Kent's suggestion is inadequate for elevation. The vertical scalar is 1.96 not 2.4477 so now they'll have to note that k for horizontal is 2.4477 and that k for the elevation (1D) is 1.96.
That should be understood that different values of the k value apply to ?z than the bivariate ?x and ?y. I don't even recall that screen shot showing vertical uncertainty. My recollection was that the 95%-confidence error ellipses displayed were just the horizontal components.
Edit:
Yes, I see my recollection was correct:
So k=2.447
Elevation sigma is 95%
> Elevation sigma is 95%
Except the uncertainty of the height component doesn't appear on the screenshots you posted, just the horizontal uncertainties.
Edit: There it is disguised as ?h, which in normal usage means the standard error of h.
That display really needs to be rethought. A simple change to "k?h would do it", but as an example of the graphic communication of information, it's a failure. The user doesn't need to know the azimuth of an error ellipse, for example. That is junk filler.
One line with k?y k?x and k?z would be entirely too logical a solution, I suppose.
> > Last word. Triple stamped it. No erasies.
>
> Well, I'm sure we all just would prefer to see RTK controllers that didn't mislabel fundamental numbers, wouldn't we?
Nobody seems to care, other than you two. What do you care about RTK controllers anyway, Don Quixote?
::pops popcorn::
Incorrect. There are other setups for this price, some cheaper, and made with better quality and support.
Sorry, I wasn't clear. Mr. Hawley unintentionally mischaracterized the original text. The addition of the word "all" makes a big difference to the meaning. That was my actual point.
You are correct that there are systems that cost less.
James
This pretty much reflects exactly my thoughts on the subject. What I have noticed, though, is that Javad seems to be pretty much equal to the premium brands for considerably less money.
There are cheaper options out there, but not from the premium brands.
Also I think his production quality is on par with premium. His business model allows for cheaper sales of units, but isn't readily accepted by some and I completely understand that.
FWIW "premium" means less than it used to...
> Nobody seems to care, other than you two. What do you care about RTK controllers anyway, Don Quixote?
Yes, I agree with you that clearly there are plenty of RTK users who don't care about the quality of their work as long as they end up with some coordinates at the end of the field day.
I'd think it's obviously poorly designed survey gear tends to produce even lower-quality results. The way that survey accuracy is represented to the user in real time is particularly important in that respect. I wouldn't have thought that was a difficult point to understand, particularly when so few RTK users export the vectors into an adjustment by least squares as part of a subsequent QC operation.