Re: [Gpsbabel-code] patch to increase precision in unicsv.c
Brought to you by:
robertl
From: Robert L. <rob...@gp...> - 2009-12-24 21:40:13
|
OK. I'll roll on this one. I applied it and cleaned the wreckage in the test suite's reference files. Please do a 'testo' on the next patch submission as that makes it easier for me to rubber stamp things. Thanx, RJL On Wed, Dec 2, 2009 at 5:14 PM, Hamish <ham...@ya...> wrote: > Robert wrote: > > Under what circumstances does this patch actually improve > > precision? > > after a trip into the field last week I was downloading & > converting the data and losing info in the process. Specifically > river depths from the sounder are better than 10cm while > shallow, and temperature is given better than 0.1 deg C. > This was from a consumer grade Garmin depth sounder. > > HDOP et al were just along for the ride, the idea being that > units using meters should report to the mm (to accomodate > survey grade/RTK GPSs). For lat/lon that means about %.9f btw. > > > > Do you really have a device that returns temperature to a > > thousandth of a degree > > actually I do, but that's a bit besides the point & I'm not > really asking you to support research gear (but for the cost > of 1 sigf dig that would sure be nice). > > good thermistors are pretty cheap these days, something that > gets you to better than 0.1 deg C is less than $0.50 so lots > of things have that. The idea of going to 0.001 deg instead > of 0.01 deg was to avoid introducing any FP rounding errors. > (mostly an issue for lat/lon which needs to be stored in a > double, but after tearing my hair out so many times due to > that (after the source data was gone) I try to preserve > precision+1 for everything now and let gzip worry about any > added zeros) > > > > or that returns DOP with meaningful values in the hundredths > > position? > > DOP was just to standardize at 1mm reporting. No arguement that > that will be slight overkill. I'll have a peek at some of our > Trible logs and see if I get any really tiny DOP numbers out of > that (should be good to about 2mm position in the horizontal). > I've got a feeling that the DOP there is in pseudo-meters though, > Gpsd has some code to convert, I'm not really sure how correct > their approach is though. > > > If there is actually valid information, I > > don't mind bloating the output, > > sure, I see your point. > > for bit balance, the GPX output doesn't need to go to %.6f for > those things. > > > but it's not like speed (which is almost always computed and > > not measured) > > are you sure about that? I seem to recall something recently > discussing how the speed solution was often better than fix as > it took phase differences into account. > > > from consumer grade GPS has two significant digits past the > > decimal. > > > > thanks, > Hamish > > > > > |