[time-nuts] GPSDO Component Selection

[Magnus Danielson]

On 09/09/2012 11:05 PM, Chris Albertson wrote:
> On Sun, Sep 9, 2012 at 1:14 PM,<bg at lysator.liu.se>  wrote:
>
>> True for a cheap oem navigation receiver. Not true for a geodetic quality
>> receiver, who usually have some options (external frequency input, PPS_in)
>> to make them the best timing receivers available. However they are much
>> more expensive than the typical single frequency timing reciver.
>
> I looked at every link and can't see where they give a timing accuracy
> spec on the PPS with respect to UTC.   Possition accurracy is very
> good and we might assume the timing is as good.  But they don't say it
> is.  What's interesting is these GPSes will accept an accurate clock
> input in order to give better location data.   That is the opposite of
> a timing GPS where you tell it accurate location data so that it can
> get better timing.   Cutting down the unknown in one lets you do
> better in the other.   I assume these all cost well over $50.  You can
> get a pretty good timing GPS for $30 and it WILL have the PPS error
> specified.
>
> To the OP.  None of this matters a lot because PPS is a standard input
> signal.  It is easy to swap out a GPS receiver later.  Same with the
> OCXO.  From a control point of view they are all pretty much the same.
>   You can swap them out later

There can be uncompensated delays, that always needs to be calibrated 
out. If you go looking around, you will find that they have gone through 
fair amount of effort to characterize the delay and it's temperature 
dependence among other things.

Also, the PPS generation and it's precision isn't a good quality measure 
here, as the PPS is typically generated out of the internal clock and 
stepped in that clock. For the higher precision you need to use the data 
alongside it.

For most of these, you do not provide external time but a better 
reference clock for a stable frequency, which will increase the quality 
of both time and position results. There are receivers in which you 
provide time as in 5/10 MHz and PPS, but then you get the time 
difference and position out of the receivers.

So I agree with Björn here, but it may be beside the point unless you 
are eager to go over the steep step and into really good receivers.

There even are rubidiums which learns of time and frequency errors out 
of GPS receiver messages.

Cheers,
Magnus