RPI 2 and 3 with USB GPS and PPS Input

Trying to get this to work reasonably.... I can use ntp with a USB GPS 
fine; I do not want to use the serial I/O for it (it has serial I/O pins 
that are 3.3v compliant) because then if its connected, unless I 
manually configure a different uBoot, the box won't come up if the GPS 
is talking (and it always is, of course.)

I have the gpio-pps overlay in with the .ko loaded and the device node 
is created on boot.  Configuring it in /etc/ntp.conf as:

server 127.127.20.0 prefer
server 127.127.22.0 minpoll 4 maxpoll 4
fudge 127.127.20.0 time2 +0.066

comes up with both the GPS and PPS initially, and appears to be ok:

ntpq> peer
      remote           refid      st t when poll reach   delay offset  
jitter
==============================================================================
*GPS_NMEA(0)     .GPS.            0 l   14   64  377    0.000 +2.243  11.107
oPPS(0)          .PPS.            0 l    6   16  377    0.000 +1.723   1.514

But.... after some period of time ntp declares BOTH to be "false 
tickers" and drops them.  Once it does it will not consider them "ok" 
again even if I wait an hour or more unless I shut down and restart ntpd.

ntpq> peer
      remote           refid      st t when poll reach   delay offset  
jitter
==============================================================================
xGPS_NMEA(0)     .GPS.            0 l   46   64  377    0.000 +6.181   0.359
xPPS(0)          .PPS.            0 l    6   16  377    0.000 +2.467   0.128

In ntpd's usual way it is less than helpful (that is, no log messages of 
information at all) about /why /it did that. I'm assuming it sees 
something discordant between the PPS signal and the NMEA output that 
causes it to declare both "bad"; unfortunately it looks like gpsd 
doesn't know how to use the gpiopps kernel module at all nor the 
/dev/pps0 device node which bites as otherwise I suspect I could at 
least figure out why its upset.

This is consistent on both a Pi2 and Pi3 on 13.2-STABLE and 13.3-STABLE 
respectively (the latter built this afternoon) and I moved the PPS pin 
from the "built in" one that was concordant with PCM use to another with 
no change in behavior.

Has anyone had any luck with this?  I can run without the PPS if I'm 
willing to live with that and it does appear to be stable.

Thanks in advance.

-- 
Karl Denninger
karl@denninger.net
/The Market Ticker/
/[S/MIME encrypted email preferred]/