[time-nuts] Ublox neo-7M GPS
ed_palmer at sasktel.net
Tue Aug 19 19:11:24 EDT 2014
Does anyone have a neo-7M and an HP 5371A or a 5372A Analyzer? Use the
Histogram Time Interval function to measure a block of samples. That
will show the length of the samples with a resolution of 200 ps. That's
what I did a couple of years ago when I analyzed the Navsync CW-12 with
the old and new firmware.
Remember the explanation of a GPSDO's Adev curve. At low values of Tau,
the value is determined by the oscillator (whether OCXO or TCXO). At
high values of Tau, the value is determined by the GPS system.
I think of 'The GPS Line'. It's a line on the Adev graph that passes
through 1e-10 @ 100 sec. with a slope of -1. Use a *really* fat pencil
when you draw the line! Every GPSDO follows that line - nothing exists
to the right of it. The oscillator determines where the curve for that
particular GPSDO is on the left side of the line. When the oscillator
performance hits the GPS Line, the graph turns down and to the right and
follows the line.
Since an NCO (Navsync, Ublox, whatever) has no internal oscillator, it
just follows the GPS Line. That means that at a Tau of 1 sec. the Adev
can't be any better than 1e-8. A low clock speed could make it worse
due to limited resolution on the step size. Said's GPSTCXO has a nice
TCXO oscillator which gives an Adev two orders of magnitude better than
that at 1 second, but that difference disappears at 100 sec. Most
GPSDO's use an OCXO which give even better performance at 1 sec. but
eventually, the GPS line corrals everyone and imposes similar performance.
For any particular application, the user has to decide what level of
performance is necessary. If an NCO is good enough with it's
cycle-to-cycle anomalies and limited low Tau performance, use it. If
not, move up to a real GPSDO.
On 8/19/2014 3:23 PM, SAIDJACK at aol.com wrote:
> Hi Tom,
> last time I looked at these I tried figuring out what they were doing. It
> is very hard to get measurement data, our TSC did not converge on their
> signal, and looking at the output on a scope revealed only a bunch of crazy
> random phase jumps. I guess one could use a counter to measure how many time
> pulses are being sent in x seconds with x being a large number, or divide
> the output by 10 million and see how the pulse moves back and forth compared
> to the 1PPS UTC output..
> Since I don't know the exact algorithm being used, I said
> "adds/drops/extends/retards" in my previous email. I did not mean to imply that the unit is
> doing all or any of those items. But that is exactly part of the problem
> isn't it, there is no clear description of what exactly is happening in the
> uBlox documents or the CW docs for that matter that I could find.
> I for one would not use that output to drive a processor or other digital
> device directly, who knows what happens if the processor sees a 100ns, then
> a 110ns, and then an 70ns pulse if it is only rated at 10MHz and 100ns
> pulse-width +/- a couple percent for example.. Without knowing the exact
> minimum phase time period specification that could come out of one of these
> NCO's, one should not properly use that signal in a digital design.
> My initial concern was that this is time-nuts, and we should call a GPSDO a
> GPSDO, and an NCO an NCO in my opinion. Nothing wrong with one or the
> other, but they sure are not the same thing - by 6 or more orders of magnitude
> in phase stability. We usually are concerned here about parts per trillion
> stability and accuracy, and now we are mixing things up that are millions
> of times worse than one another..
> In a message dated 8/19/2014 13:08:52 Pacific Daylight Time,
> tvb at leapsecond.com writes:
> Hal, as long as you maintain long-term phase lock it's a disciplined
> oscillator. So, yes, a carrier tracking WWVB receiver with sufficiently stable
> flywheel LO is a WWVBDO.
> Said, too-short or too-long 100 ns cycles is one thing. Still ok for many
> applications. But tell me more about extra or missing pulses in the
> ublox-7. That sounds like a show stopper to me.
> /tvb (i5s)
>> On Aug 19, 2014, at 2:05 PM, Hal Murray <hmurray at megapathdsl.net> wrote:
>> SAIDJACK at aol.com said:
>>> its not a GPSDO though, not even a simple one :)
>>> It does not discipline an oscillator. It generates the output by
>>> mathematically calculating how many phases it has to add/drop in a second, then digitally adds/drops/extends/retards the phase of the output clock to achieve an average of number of desired clock cycles.
>> Is there something about the term GPSDO that says I have to do the "D" in the analog domain rather than the digital domain?
>> I agree that current technology doesn't give results that are useful for many applications that currently use GPSDOs. What if the clock ran at a GHz? 10 GHz? Sure, it would have spurs, but would it be useful for some applications?
>> Is a GPSDO still a GPSDO if the D/A driving the VCXO only has a few bits?
>> How many bits does it need to be a real GPSDO?
>> Is a battery powered wall clock listening to WWVB at 2 AM a WWVDO? It's got a pretty good ADEV if you go out far enough.
>> These are my opinions. I hate spam.
More information about the time-nuts