[FOSS-GPS] gap in base observations unsettles RTKPOST indefinitely

[Michele Bavaro]

Hello,
I am doing Kinematic between base (LLH = 44.434828231 12.300198659 45.9096) and rover.The rover has 4Hz.The base has normally 0.2 Hz rate, but across midnight experiences a gap of about 90 seconds.So the age of differential correctly grows and solution goes from 1 (fix) to 5 (single).However when the base eventually comes back, a perfectly good set of observations pushes the EKF so far that AR does not recover anymore.So far I am solving this by adding 
            memset(rtk->x, 0, rtk->nx*sizeof(double)); /** RESET FLOAT STATES?? */
in rtkpos.c after 
            errmsg(rtk,"age of differential error (age=%.1f)\n",rtk->sol.age);
However, my fix looks dirty and the problem should not manifest in the first place?So far I am guessing the problem is here:
        if ((nv=ddres(rtk,nav,dt,xp,Pp,sat,y,e,azel,iu,ir,ns,v,H,R,vflg))<1) {
inside relpos() in rtkpos.c. My feeling is that the age of correction when the base wakes up is said to be 0.2 seconds when in reality more than 90 have passed and so the partial derivatives go off scale?Such behavior, if confirmed, may impair any real-life situation in which base and rover talk unreliably between each other...

All data are here:https://www.dropbox.com/s/igwt9jsh8pco1bs/160531n-aann.zip?dl=0

Can anybody comment on this and propose a cleaner fix?

Cheers,Michele
P.S.: This happens with both 2.4.2b11 and 2.4.3b12
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/foss-gps/attachments/20160622/665b2b62/attachment.html>