[Therion] Weird handling of directions

[Tarquin Wilton-Jones]

Hi folks,

Technical problem. Perhaps something to do with PROJ. Perhaps I have my
declinations backwards?

survey testdirection -title "Test direction"
 centreline
  date 2021.08.13
  cs OSGB:ST #change this to test other grid squares
  fix startst 0.0000000000 0.0000000000 0.0000000000
  station endst "end" entrance
  data normal from to length compass clino
#for SU
#  startst endst 10 0.22666712328 0
#for ST
  startst endst 10 0.7235958904109588 0
#for SH
#  startst endst 10 1.51343150685 0
 endcentreline
endsurvey

thconfig:
source "testdirection.th"
export cave-list -surveys on -location on -o testpositions.html

Those compass bearings match the declination at those locations, on
those dates. I can see this in the Therion output pane:
geomag declinations (deg):
  2021.1.1  -0.8475
  2022.1.1  -0.6465

I have double checked them against
https://www.ngdc.noaa.gov/geomag/calculators/magcalc.shtml#declination
and I get the same numbers there (rounded to 2 dp). Therion seems to be
doing the right thing so far.

I have converted the startst location into lat,long (using both Helmert
and OSTN5+drift just in case it made a difference). Then applied a
northerly 10 metre geodesic to see where it lands (using two separate
geodesics formulae just in case it made a difference), when given that
bearing and declination.

When doing this at the SU0000000000 line (where OSGB grid and true north
line up), Therion gets the same answer as me for the endst coordinates:
SU 0.000	10.000	-0.000

When doing this at the ST0000000000 line (where OSGB grid and true north
do not line up), the geodesic calculation gives this (1 mm rounding
errors possible due to imperfections in geodesic formulae):
ST 0.276501	9.994217	0.000
(Converted using Helmert, to be compatible with Therion, but the results
are within 1 mm when converting more accurately.)
Therion gives an answer that is shifted to one side by 8 cm:
ST 0.192	9.998	0.000
This seems to cause a rotation to our entire survey. The amount of
rotation depends on the location of the cave. Where we are working, it
is a 12 cm shift within a length of just 10 metres. By the time you
reach SH0000000000 it is a 23 cm shift.

Am I wrong somewhere? Or is Therion (and perhaps Survex?) applying the
declinations in the wrong place?

If someone wants to help debug this, you can do conversion from grid to
lat/long using Helmert here (note, use SH 00000.000000 00000.000000
format with several decimal places to avoid it using the middle of a
grid square):
http://www.howtocreate.co.uk/php/gridref.php

And you can perform geodesics here:
https://geographiclib.sourceforge.io/cgi-bin/GeodSolve?type=D&input=50.7909132676dN+-3.420104865715981dE+0+10&format=g&azi2=f&unroll=r&prec=6&radius=6378137&flattening=1%2F298.257223563&option=Submit

Thanks for any help,

Tarquin