Are you sure that both you and the GPS are using the same sea level
reference?  The GPS is likely using WGS-84.  How did you manually measure
the antenna hight?  Note that almost all topographic maps in the US were
made long before 1984

On Sun, Jun 26, 2016 at 6:02 AM, Mark Barettella via time-nuts <
time-nuts@febo.com> wrote:

--

Chris Albertson
Redondo Beach, California

Hi

What are you using to determine the “correct” altitude of the antenna? In a lot of cases, the GPS
is more correct (relative to it’s specified ellipsoid) than the reference being used.

Bob

Yo Mark!

On Sun, 26 Jun 2016 09:02:55 -0400
Mark Barettella via time-nuts time-nuts@febo.com wrote:

Others have covered some obvious details.  Different ellipsoids,
long term surveying, etc.

Depends on how accurate you need.  I'll assume your estimate is perfect,
which is that your GPS is reading off by 22 meters.

The speed of light is  299,792 kilometers/second.

So to travel 22 meters is about .000,000,073 Seconds.  Or 73 nanoSecond.

All else being equal, does a constant 73 nanoSec matter to you?

For comparision, a Trimble RES SMT 360 only promises 15 nanoSec (1 sigma).

If all you want is a stable frequency from your gpsdo then the offset
is not relevant.

RGDS
GARY

Gary E. Miller Rellim 109 NW Wilmington Ave., Suite E, Bend, OR 97703
gem@rellim.com  Tel:+1 541 382 8588

Poor coordinates for the antenna couple into the apparent distance to
a satellite in such a way that the error in the distance tracks a
parabola-shape as the satellite rises and descends, with a
corresponding error in the apparent satellite time. At any instant,
the combination of N satellites then leads to an apparent GPS time
with a mean offset roughly equal to position error, and peak to peak
scatter in the satellite times  equal to the position error as well.

The offset isn't a problem, unless you care about time of day. How the
noise feeds into your GPSDO depends on the time constant of the servo.

I tried a simple numerical experiment, where I post-processed raw code
measurements from a timing receiver for different antenna heights,
comparing the results with those for its true position, known with an
accuracy of a few cm. The output of this calculation is a CGGTTS
time-transfer file, where the satellite time is averaged over 780 s.
For a 15 m error in height, at my latitude (30 deg) the offset is
about 28 ns and the peak to peak scatter is about 50 ns. If you then
try to do what the GPS receiver does, ie take an average of the
satellites, you get a peak to peak noise of about 10 ns.

I plotted the TOTDEV of this noise.

As you can see, at about 1000 s, the instability is about 10^-12,
averaging down to about 10^-13 at one day.
This does not  compromise a GPSDO's frequency stability at these
averaging times. At shorter averaging times there may be a problem
though, depending on the time constant of the GPSDO, and the stability
of its oscillator.

Cheers
Michael

On Sun, Jun 26, 2016 at 11:02 PM, Mark Barettella via time-nuts
time-nuts@febo.com wrote:

Hi

If your time is off because your antenna location is off and you are running a GPSDO …
The error does matter for frequency.

1. Start with a group of sats in the sky all in the direction of the  error vector. Your
   GPSDO will get a time solution that is 72 ns early.

2. Move the group of sats so they are at right angles to the the error vector. Your
   GPSDO will get a time solution that is correct (the error does not impact the solution).

3. Move the sats so they are in the exact opposite direction of the error vector. Your
   GPSDO now gets a time solution that is 72 ns late.

As the sats move in the sky, the solution “swings” back and forth between +72 and -72 ns.
It gets a bunch of zeros in there as well. How often you get a “all bad minus” versus “all bad
plus” depends a lot on your antenna location. Even if it is simply a 2/3 of the sat’s over there
and 2/3 over here …. there is still a significant swing.

A swing in time will give you a delta frequency. If you otherwise would get 1x10^-11 delta F out
of your GPSDO, That will double if the +72 to -72 ns swing occurs in 4 hours.

Bob

Hi Gary,

I want to echo what Bob just wrote. People get carried away with "a nanosecond is a foot" and think it applies 100% to GPS timing and position, or in this case, elevation errors.

Equating 22 m with 73 ns, or equating 1 foot with 1 ns is only true in the impossibly rare case of one satellite directly above you. In reality, 1) most of the time the SV are further down and so the error is reduced by sin(angle). And, 2) the GPS timing solution is typically based on lots of satellites, not just one, and so the effects of position error is further reduced by the ensemble mean.

An accurate position is desirable. No question about that. This note is just a plea not to apply the speed-of-light number or the "nanosecond a foot" rule-of-thumb out of context.

/tvb

----- Original Message -----
From: "Gary E. Miller" gem@rellim.com
To: "Mark Barettella via time-nuts" time-nuts@febo.com
Sent: Monday, June 27, 2016 12:31 AM
Subject: Re: [time-nuts] Impact of GPS antenna height measurments

Yo Mark!

On Sun, 26 Jun 2016 09:02:55 -0400
Mark Barettella via time-nuts time-nuts@febo.com wrote:

Others have covered some obvious details.  Different ellipsoids,
long term surveying, etc.

Depends on how accurate you need.  I'll assume your estimate is perfect,
which is that your GPS is reading off by 22 meters.

The speed of light is  299,792 kilometers/second.

All else being equal, does a constant 73 nanoSec matter to you?

For comparision, a Trimble RES SMT 360 only promises 15 nanoSec (1 sigma).

If all you want is a stable frequency from your gpsdo then the offset
is not relevant.

RGDS
GARY

Gary E. Miller Rellim 109 NW Wilmington Ave., Suite E, Bend, OR 97703
gem@rellim.com  Tel:+1 541 382 8588

Yo Tom!

On Mon, 27 Jun 2016 15:49:00 -0700
"Tom Van Baak" tvb@LeapSecond.com wrote:

I should have mentioned that I made some simplifying assumptions:
The cow is sperical, its albedo is 1.0, and she is in a vacuum.  :-)

I could nit-pick your nit-picky analysis, but at least I came up with a
real number.  What is your real number?  I really would like to hear a
fully derived answer, and see how well my back of the hand calcuation
compares.

As a time-nut, you are right, but since I am an engineer, if I get it
within a factor of 4 I'm happy.  :-)

Almost all people when they hear 73 nanoSec, which is hardly more than 2
sigma of a good GPS, they say "Never Mind".

RGDS
GARY

Gary E. Miller Rellim 109 NW Wilmington Ave., Suite E, Bend, OR 97703
gem@rellim.com  Tel:+1 541 382 8588

Tom Van Baak said:

"This note is just a plea not to apply the speed-of-light number or
the "nanosecond a foot" rule-of-thumb out of context."

It works reasonably well as a rule of thumb. It's an upper limit  but
if you wanted to refine it a bit, divide by two. The average value of
sin(x) on [0,pi/2] is 2/pi ie about 1/2.

This agrees quite well with the results of the post-processing I
described earlier.

Cheers
Michael

On Tue, Jun 28, 2016 at 8:49 AM, Tom Van Baak tvb@leapsecond.com wrote: