[time-nuts] Re: What GNSS module to buy for a good time reference?

[Bob Camp]

Hi

The way one uses a NetRS in this application is to feed it a local 
10 MHz reference. With the proper settings, the data it records is
referenced directly to that source. You can get < 100 ps resolution
and < 1x10^-14 one day frequency this way.

Other than a few devices in the $20K price range, pretty much all
GNSS gizmos have limited (+/- nanoseconds) resolution on their 
PPS outputs. The same “feed it a 10 MHz source” approach is 
how this is done on a lot of devices. 

Bob



> On Jul 3, 2023, at 2:22 AM, Ed Marciniak via time-nuts <time-nuts at lists.febo.com> wrote:
> 
> Navsparq has the NS-T but that's single band, so no further value discussing in this context.
> 
> A look at the Trimble NetRS reveals it's PPS output has 40 ns resolution and says several external factors limit accuracy to +-1 microsecond. While 40ns would come from a 25 MHz clock, it seems to be common reference frequencies are 19.2,26,38.4 MHz. Perhaps it's not exactly 40ns. Elsewhere the manual claims position 5mm+1ppm horizontal and 10mm+1ppm vertical accuracy, which probably represents a fairly decent internal receiver frequency reference. That doesn't seem directly useful for time and frequency but it's interesting for other purposes.
> 
> Ah, well...thanks for the interesting reading material the conversation generated.
> 
> 
> ________________________________
> From: John Ackermann N8UR <jra at febo.com>
> Sent: Sunday, July 2, 2023 3:00:37 PM
> To: Ed Marciniak <ed at nb0m.org>; Discussion of precise time and frequency measurement <time-nuts at lists.febo.com>
> Subject: Re: [time-nuts] Re: What GNSS module to buy for a good time reference?
> 
> I'm not aware of anything in the ZED-F9T price range that accepts an
> external reference.
> 
> The closest I know of is the Septentrio Mosaic-T, but the last I heard
> the evaluation board was >$1K and I don't think you can buy the bare
> modules in one-off quantities.  You can buy a bunch of ZED-F9Ts for
> that, but it is thoroughly modern and works really well.  Any completely
> packaged receiver with similar capabilities will cost several time that.
> 
> As Bob mentioned, there are some used dual frequency receivers that can
> do this.  The ones likely to be <$1K are the Trimble NetRS and the
> Ashtech Z12 or variants.  The NetRS works pretty well and are easy to
> configure and (relatively) easy to get raw observations out of.
> 
> The Ashtechs are really, really ancient and very proprietary, and a lot
> of them don't work anymore.  I got a couple going, and wrote some python
> to convert their serial data stream into RINEX, but I can't really
> recommend anyone go down that road these days.  The NetRS is a much less
> painful choice.
> 
> Now, for those who have are really nuts (time or otherwise), if you take
> the lid off a ZED-F9T module, you will see among the three or four ICs
> there a separate TCXO.  Don't ask me how I know this; it's not pretty.
> 
> Anyway, in theory you could remove that oscillator (I think it's ~60
> MHz) and feed in an external reference there.  I don't know anyone who's
> tried that, and I don't know if the software has the right hooks to do
> anything useful with it.
> 
> The Navspark dual-freq is designed for portable RTK applications and
> probably does pretty well at that, but from the not-that-great
> documentation, its timing performance looks to be pretty basic.  I could
> be wrong, but I don't think it supports a 0-D timing mode.
> 
> John
> ----
> 
> On 7/2/23 12:31, Ed Marciniak wrote:
>> So, what would you recommend for a lower cost module that accepts an
>> external reference and has reasonable jitter when using an external
>> reference? (I’m specifically calling that out assuming that either a
>> rubidium or excellent quartz oscillator is available because I’m not
>> necessarily concerned with jitter while using an internal reference).
>> 
>> If considering dual band receivers, is the answer different?
>> 
>> Do you have any opinions or experience to share with respect to Navsparq
>> modules? On paper, the specifications are awesome for the price, most
>> (maybe all) support binary output, are supported by open-source
>> libraries, and have rtk options amenable to using dual receivers for
>> spatial orientation. They seem almost too good to be true.
>> ------------------------------------------------------------------------
>> *From:* John Ackermann N8UR via time-nuts <time-nuts at lists.febo.com>
>> *Sent:* Sunday, July 2, 2023 10:41:37 AM
>> *To:* time-nuts at lists.febo.com <time-nuts at lists.febo.com>
>> *Cc:* John Ackermann N8UR <jra at febo.com>
>> *Subject:* [time-nuts] Re: What GNSS module to buy for a good time
>> reference?
>> Hi Erik --
>> 
>> The ZED-F9T will give you about the best (lowest jitter) PPS accuracy of
>> any of the modestly priced modules, but I don't think post-processing
>> the results will help you with timing.
>> 
>> You can use the u-blox RAWX message to output raw data which can then be
>> converted to RINEX format by one of several means, and then send the
>> RINEX file for processing (or process it yourself with something like
>> RTKLIB).  The results will give you highly accurate position
>> information, usually better with longer observation times.
>> 
>> But the ZED-F9T uses a free-running TCXO for its clock, and can't accept
>> an external reference source.  As a result, the clock data returned from
>> the post processing service is pretty much meaningless because of the
>> TCXO inaccuracy and instability.
>> 
>> By contrast, the clock results from receivers that are locked to a
>> quality external reference can be used to determine both time and
>> frequency offset and stability down to parts in 10e-15 over long intervals.
>> 
>> I am not sure which post-processing site is best for Europe.  I don't
>> think that NRCan has any boundary restrictions (I use them in the U.S.)
>> but I don't know if their algorithms cross continents.
>> 
>> In a roundabout way, you can improve your PPS performance by getting a
>> good post-processed receiver position, and using that as the ZED-F9T's
>> fixed location mode position.  The closer your stated position is to
>> actual, the better the PPS results (though a few centimeters won't make
>> much difference).  But that's the only timing benefit post-processing
>> provides for the ZED-F9T.
>> 
>> John
>> ----
>> 
>> On 7/2/23 11:09, Erik Kaashoek via time-nuts wrote:
>>> As I needed the accurate PPS in the coming weeks I decided to go for the
>>> ZED-F9T L1/L2 and if all is well it will arrive tomorrow.
>>> Now I need to understand how all this "postprocessing" works.
>>> Some questions:
>>> 
>>> 1: The ublox tools show ability to output basic or full raw data and there
>>> are various scripts  to convert ublox data to Rinex and there are some
>>> websites listing the commands required to get the required output but is
>>> there a dummy's guide somewhere on how to get the RINEX data from the
>>> ZED-F9T in the correct version/format for the postprocessing?
>>> 
>>> 2: NRCan seems to process only for Canada (according to their website,
>>> correct???). Auspos is listed as processing for the whole world. There are
>>> some others. What would be a recommended service? I'm located in Europe.
>>> 
>>> Erik.
>>> 
>>> 
>>> Op wo 21 jun 2023 om 19:34 schreef Bob Camp via time-nuts <
>>> time-nuts at lists.febo.com>:
>>> 
>>>> Hi
>>>> 
>>>> The “big deal” with no L2 is fairly simple:
>>>> 
>>>> One of the great things about dual freq modules is the ability to send off
>>>> data
>>>> and very quickly get back a corrected version. If you use NRCan, this
>>>> includes
>>>> clock corrections. They are good to the ~ 0.1 ns level. You can get to <
>>>> 1x10^-14
>>>> in less than a day.
>>>> 
>>>> The problem is, the free correction services (at the moment) are L1 / L2
>>>> based.
>>>> For whatever reason, they don’t (yet) understand L5. That may change, or it
>>>> may not change. Right now, it’s the way it is ….
>>>> 
>>>> If you want to go crazy, the Mosaic-T is the best of the best in terms of
>>>> GNSS
>>>> modules at the moment. They apparently are well aware of this and charge
>>>> accordingly. You can buy a *lot* of F9T’s for the price of one Mosaic-T.
>>>> 
>>>> Bob
>>>> 
>>>> 
>>> _______________________________________________
>>> time-nuts mailing list -- time-nuts at lists.febo.com
>>> To unsubscribe send an email to time-nuts-leave at lists.febo.com
>> _______________________________________________
>> time-nuts mailing list -- time-nuts at lists.febo.com
>> To unsubscribe send an email to time-nuts-leave at lists.febo.com
> _______________________________________________
> time-nuts mailing list -- time-nuts at lists.febo.com
> To unsubscribe send an email to time-nuts-leave at lists.febo.com