[time-nuts] GPS module recommendation for Pi timing

[Bob kb8tq]

Hi

> On Mar 14, 2020, at 4:10 PM, Brian Lloyd <brian at lloyd.aero> wrote:
> 
> On Sat, Mar 14, 2020 at 12:52 PM Bob kb8tq <kb8tq at n1k.org> wrote:
> 
>>> I understand completely. Still, I have to come up with a
>> reasonably-priced
>>> source of time and it does seem like GPS is the right answer. I can live
>>> with units going off-line periodically due to insufficient timing
>> accuracy,
>>> but I cannot have them drifting away from correct time.
>> 
>> Only you can evaluate how important this or that aspect of the design is.
>> If *any* timing drift past 1 ms becomes an issue, some sort of TCXO based
>> GPSDO may be the real answer.
>> 
> 
> In reading the data sheet for the LEA-M8T it does have a TCXO for the
> timing outputs.

It does and it doesn’t. You can not “count on” the TCXO to do nice things
in holdover. 


> 
> 
>> With a bare crystal, you could easily have a 10’s of ppm sort of error. If
>> you do, then any outage over about 100 seconds will hit your limit.
>> 
> 
> Yes, and I just need to know that I am out-of-spec so I can turn off
> participation by that device until it has good timing again. And if I have
> to end up disciplining a local TCXO then I can do that too, and then I use
> a cheaper GPS module. Ah, design decisions. Where would we engineers be if
> we weren't making design decisions compromises. ;-)

So you need a second “something” to compare to that will let you make
this call. That or you shut down any time GPS drops out. 

> 
> 
>>> The signal will be squared and used as a clock for something else.
>>> Duty-cycle is flexible. As long as the rising edge is clean and there
>> isn't
>>> too much jitter, I can live with it. I have no erroneous notion that I am
>>> getting a clean sine wave or a symmetrical square wave. Also, this is in
>>> the "nice to have" category and I can do away with it and use a GPS
>> module
>>> that only has 1pps.
>> 
>> The output will have jitter at the same level as the time pulse. If it’s a
>> 20 ns
>> time pulse pre-sawtooth, then the 10 MHz will also hop by 20 ns.
>> 
> 
> Yes, so what happens depends on whether it is phase error or phase jitter,
> and then what the device wanting the 10MHz reference does with it. Since
> that is part of this that I don't have control over, I am going to have to
> get more information. But if that other time-pulse can be used as a 10MHz
> reference that is used to discipline another oscillator rather than being
> used directly, it should be OK. I just need more information on this part
> of it. I just don't know how clean that output from the Ublox is.

The time pulse, pps, whatever you call it is quantized to an internal free running
clock in the module. It comes out based on whatever edge happens to be
closest at the moment. Since the clock is free running you can have some
very odd behavior. 

For accurate timing or a GPSDO, there is a “sawtooth” message that tells
you the offset of the edge from reality. That message comes out once a 
second. If the edge you use comes out more often than that, you have 
no feedback about where the edge is relative to what it actually should be. 

> 
> 
>> 
>>> 
>>> So I am still hoping that someone will say, "The U-Blox LEA-M8T is a
>> pretty
>>> good choice but for what you are talking about, you might want to look at
>>> the XYZ module as well."
>> 
>> There are maybe another two dozen modules out there that might work and
>> cost the same or less. Buying a few of this and that / trying them out is
>> about
>> the only way to know if they are “good enough” for all the details of your
>> application.
>> 
>> The bigger issue is - how much can you invest to dig through a pile of
>> modules?
>> The simple answer is indeed to just pick one and accept the cost impact.
>> 
> 
> That is indeed the question. Given the experience level here I was hoping
> that others had run into similar issues and might have more information to
> share with me, or even point me to something that would works as well or
> better than the Ublox.

The gotcha is that every application is different. Without a deep dive into the 
entire design (and that’s not really what this list is for ) there is no way to start
sorting things out.  These days pretty much all the modules on the market 
(that put out a pps) will easily do microsecond level timing. 

Lots of fun !!!!

Bob


> 
> Thanks! I appreciate your input.
> 
> -- 
> 
> 
> 
> Brian Lloyd
> 706 Flightline
> Spring Branch, TX 78070
> brian at lloyd.aero
> +1.210.802-8FLY (1.210.802-8359)
> _______________________________________________
> 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.