[time-nuts] Advantages of GNSS ???

[Bob kb8tq]

Hi

Whatever they are doing to create an output is being done modulo the internal TCXO period. 
That gets you down to a +/- 4 ns error on either the F9T or the F9P.  They can do that however 
often they like, it’s still a very coarse correction as far as I’m concerned. 

What matters is how often they come up with the data to tell you how far off from reality that magic
edge *is* in reality. They do that to well under a nanosecond, both on the F9T and the F9P. That 
update info only is calculated / supplied once a second. This information *plus* the edge is (at lest to 
me) the only source of accurate time out of the device. It’s at least 10X better than what you get 
without the correction 

If indeed you *do* configure to let them modify the TCXO <-> pulse output at a greater rate than 
once a second, the connection between the sawtooth correction data and the actual pulse becomes
a bit obscure. Only one edge has data for it. The rest have an offset of “who knows what”.  Having a 
bunch of “unknown” edges in with one accurate one seems like a route to confusion. 

This all gets right back to the recommendation that I quoted about running the pulse rate at 1 pps
for precision use. There are similar notes in similar documents going back at least to the LEA-5T
era of modules. Are any of them 100% clear - nope. You have to either call up and ask “what does 
this really mean?” or guess a bit.

Is this a knock of some sort on uBlox? Certainly not. No document ever written is ever going to be
perfect. They are not the only outfit doing this (one second calculation of error). They aren’t any
worse at explaining it than the other guys. They all do a pretty good job of describing what’s going
on face to face. 

Bob

> On Jul 8, 2019, at 9:51 AM, Leo Bodnar <leo at leobodnar.com> wrote:
> 
> Bob, what are you calling "time correction"?
> 
> You are now quoting F9T which is not the product original statement related to (F9P.)
> 
> If you refer to internal Ublox adjustment of instantaneous timepulse train then it is performed at navigation rate - up to 20Hz on F9P.
> F9P's rate of TP can be set from 0.25Hz to 10MHz according to the datasheet but in reality is usually wider.
> 
> TP edge quantisation is a separate issue to TP base rate adjustment and it is done at TP rate.
> 
> Leo
> 
> 
>> Subject: Re: [time-nuts] Advantages of GNSS ???
>> They go absolutely crazy updating nav?.. but they only come up with the timing correction once a second.
>> 
>> From the latest version of the F9T Manual UBX-19005590 - R02  on page 43: 
>> The recommended configuration when using the UBX-TIM-TP message is to set both the measurement rate (CFG-RATE-MEAS) and the time pulse frequency (CFG-TP-*) to 1Hz.
>> 
>> Since the rate of UBX-TIM-TP is bound to 1 Hz, more than one UBX-TIM-TP message can appear between two pulses if the time pulse frequency is set lower than 1 Hz. In this case all UBX-TIM-TP messages in between a time pulse T1 and T2 belong to T2 and the last UBX- TIM-TP before T2 reports the most accurate quantization error. In general, if the time pulse rate is not configured to 1 Hz, there will not be a single UBX-TIM-TP message for each time pulse. 
>> 
>> Sorry if the quote comes through a bit garbled. Sometimes this cut and paste stuff does not quite do the trick. 
>> Indeed it?s not 100% clear what they are doing from the docs. They say this and that here and there. The 
>> bottom line is still that you can only trust the sawtooth offset data to be correct at the one second point. 
>> 
>> Bob
> 
> 
> _______________________________________________
> time-nuts mailing list -- time-nuts at lists.febo.com
> To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
> and follow the instructions there.