@PaulZC that all looks correct. Just watch the signs, the orthometric height is the ellipsoidal height less the geoid height. In many areas (including all 48 US states) the geoid height is a negative number, so it is effectively addition, and I confuse this often.
H = h - N where H is the orthometric height, h is the ellipsoidal height, and N is the geoid height
Here’s a good diagram: https://www.researchgate.net/figure/Rel … _273133676
Regarding applying the geoid model of your choice, you don’t really need to know what internal geoid model the ZED-F9P in the SparkFun product is using. Just back out whatever the internal geoid model is by using the separation provided in the $GGA message and you’ll have the ellipsoidal height. Then apply your own geoid model.
The ellipsoidal height is: h = H + N
Below is a $GGA snip from a project of mine in Virginia, USA.
$GNGGA,184252.00,3811.5564944,N,07936.8396046,W,2,12,99.99,650.856,M,-33.810,M,0131*4A
H=650.856m, and N=-33.810m. That H and N are dependent on the geoid model built into the Z9P).
h=617.046 (This is the ellipsoidal height independent of the geoid model)
The geoid model I use specifies N to be -31.480 at that point, so my orthometric height is H = h - N = 617.046 - (-31.480) = 648.526m
BUT the original question asked about DATUM. That’s a different issue then geoid models and orthometric elevations.
If you’re using RTK, the horizontal and vertical datum will depend on the datum of the RTK base or the network reference station. Most network reference stations in the US are in NAD_83(2011) so the output coordinates when using those stations are in that NAD_83(2011) horizontal datum and the NAVD88 vertical datum. The ellipsoidal height and orthometric heights will be referenced to these datums.
Tony