GPS 18x 2.70 test (IPv6 access)
2011.10.13 Kasper Pedersen

Test of the Garmin GPS-18x with firmware version 3.70. "How precise, how accurate?"

Equipment list:

Trimble Thunderbolt
         tc=300, damp=0.700, Trimble OCXO
         Dly=-84.54: cable=52.8 ns (measured), ant=31.74 ns (specified)
         PPS=UTC, measured on face of connector
         unit is in a thermally slow environment

GPS-18x #1
         on 20cm ground plane
         cable length = 50cm (cut)
         PPS is measured after 50cm cable
         Indoor on lab desk. Ceiling has non-conductive vapor barrier

GPS-18x #2
         on 10cm steel plate
         cable length = 10m (double length)
         Has better view than #1, but also indoors

Scope
         LeCroy 9304A
         probes are matched better than 1ns on the channels used
         TIC trigger on tbolt (ch2)

50cm cable test:

channel 1: GPS-18x#1   channel 2: tbolt

18x appears to be 110ns slow compared to the tbolt.
measurement tolerance: -15ns(tbolt) + -10ns (ant) = -25ns / +15ns (tbolt) + +15ns (ant) =+30ns. So 18x is [85 ns .. 140ns] slow.
Short term jitter is 21ns rms, owing to the ~16MHz timer clock in the 18x.
During the run, the tbolt had no pps offset reading exceeding 5 ns.

10m cable test:

channel 1: GPS-18x#1   channel 2: tbolt   channel 3: GPS-18x #2 w. 10m cable.

The PPS driver in the 18x is very weak.

the 18x PPS output drives a step of only 1V into the cable. With 10m cable the start of the rise is 74ns delayed = 7.7ns/m = 42% c (a reasonable propagation velocity)
rise rime is 430ns on 10m, giving an additional 105 ns, for a total cable-induced additional delay of 179 ns. (5m: ~80ns)
The receiver delay is 126ns-0.5*7.7ns = 122 ns.

10-hour worst case test:

channel 1: GPS-18x#1   channel 2: tbolt   channel 3: GPS-18x #2 w. 10m cable.

Worst case pps timing has +/- 94 ns wander, centered on 126 ns.
94 ns is 1.5 period of the internal 16MHz clock.
+/-0.5 clock is the best the PPS output can do. The excess corresponds to 20m peak position error, which is not bad considering the receivers are indoors.

tbolt during test

To calculate correction factor for the 18x:

pps_slow = 122.2 ns + cable_length * 18.2 ns/m

The receiver delay is: 126ns-0.5*7.7ns = 122 ns.
The cable propagation delay is: cable_length * 7.7ns/m
The output RC delay is: cable_length * 10.5ns/m