[time-nuts] 1PPS for the beginner

Tue Aug 14 14:17:39 UTC 2018

Hi

I can think of no way to get the close in phase noise of a ring oscillator down far
enough to use it as the main clock for a GPS module. It *is* reasonable that the 
L band LO in the module is a phase locked ring oscillator., 

Bob

> On Aug 14, 2018, at 9:34 AM, Dana Whitlow <k8yumdoober at gmail.com> wrote:
> 
> Thanks Bob and Azello.  Now I understand the terminology and also better the
> mechanism for the sawtooth error.
> 
> I once read that the LO, at least, in some GPS receivers, was not even
> crystal
> controlled but was rather a ring oscillator based on a string of cascaded
> logic inverters on the chip.  This always sounded improbable to me, and i've
> long wondered it this claim was really true.  Can either of you shed light
> on
> this issue?
> 
> Thanks,
> 
> Dana
> 
> 
> On Tue, Aug 14, 2018 at 8:04 AM, Bob kb8tq <kb8tq at n1k.org> wrote:
> 
>> Hi
>> 
>> A normal GPS module generates it’s outputs off of a free running internal
>> clock. Generally
>> this is a TCXO in a “timing” GPS ( = one that puts out a rational PPS).
>> This is a bit counter
>> intuitive, since you would *think* they phase lock the local source in the
>> module. They don’t
>> mainly because it makes the math easier.
>> 
>> The gotcha with a free running clock is that the device can only generate
>> an edge (like a pps)
>> when the clock edge(s) allow it to. For simplicity, lets just go with the
>> rising edge and accept that
>> there could be *double edge* designs as well.  Let’s also assume a 25 MHz
>> clock. That’s in the
>> general range of what shows up on the surplus GPSDO modules.
>> 
>> The internal fix math in the module comes up with a solution for “when
>> should I send the PPS”.
>> The clock edges are 40 ns apart. The solution says that the “right time”
>> is 10 ns after an edge.
>> The module sends out a PPS that is 10 ns early. Next second the math says
>> that the right time
>> is 30 ns after an edge. The module sends out a PPS that is 10 ns late.
>> 
>> As long as it keeps going early / late / early /late things will average
>> out. What makes it do this
>> is the local clock on the module being a bit off frequency (modulo 1 Hz).
>> As the clock drifts around
>> (and they do) you may hit a region where it is relatively stable. It will
>> then send out early /early /early….
>> ( or late / late /late…..)
>> 
>> If the local clock is a TCXO, the “stable points” are likely to also be
>> points of frequency reversal.
>> The net result is that the early / early /early never gets a corresponding
>> late /late / late to average
>> against. It will pass right through a PLL and create an offset in the
>> output.
>> 
>> I believe that NIST was the first to spot this and document it with lots
>> of plots. I could be wrong about
>> that. It was pretty much ignored in the days before SA was turned off. The
>> SA jitter masked out a lot
>> of issues. Most modern GPSDO’s use sawtooth correction messages to get
>> around the problem. There
>> may be a few still in production that don’t.
>> 
>> Without full doc’s on a GPSDO, you really don’t *know* where the PPS is
>> set to originate. It may
>> be coming from the disciplined clock on the board. It also *may* be coming
>> straight from the GPS
>> module. There are indeed units out there that will let you do it either
>> way under software control.
>> 
>> Bob
>> 
>>> On Aug 14, 2018, at 8:15 AM, Dana Whitlow <k8yumdoober at gmail.com> wrote:
>>> 
>>> Could someone please define and explain the term 'hanging bridge' in this
>>> context?
>>> 
>>> Thanks,
>>> 
>>> Dana
>>> 
>>> 
>>> On Tue, Aug 14, 2018 at 7:02 AM, Azelio Boriani <
>> azelio.boriani at gmail.com>
>>> wrote:
>>> 
>>>> Hanging bridges out of a GPSDO's PPS? Interesting... time to try to
>>>> setup a measurement and see the relation between the GPS's PPS hanging
>>>> bridges and the corresponding DO's ones.
>>>> On Tue, Aug 14, 2018 at 9:55 AM Mike Cook <michael.cook at sfr.fr> wrote:
>>>>> 
>>>>> 
>>>>>> Le 14 août 2018 à 09:29, Mike Cook <michael.cook at sfr.fr> a écrit :
>>>>>> 
>>>>>> Sorry about the previous blank mail. Finger jitter.
>>>>>> 
>>>>>> 
>>>>>> 
>>>>>>> Le 14 août 2018 à 04:29, Chris Caudle <chris at chriscaudle.org> a
>>>> écrit :
>>>>>>> 
>>>>>>> On Mon, August 13, 2018 9:16 pm, Chris Burford wrote:
>>>>>>>> I have a (generic?) GPSDO which contains an Oscilloquartz STAR 4+
>>>> OCXO
>>>>>>>> that I am using to steer a PRS10 RFS. I'm a little confused on where
>>>> the
>>>>>>>> 1PPS is coming from with respect to the GPSDO.
>>>>>> 
>>>>>> As Chris points out the 1PPS from a GPSDO will « generally » be
>>>> derived from the primary frequency and can show better performance than
>>>> directly from a GPS receiver.
>>>>>> However this is becoming less and less true.
>>>>>> If you look at the Oscilloquarz blurb for the Star 4+ ( I found some
>>>> here <http://pdf.directindustry.com/pdf/oscilloquartz-sa/
>>>> star3-4/62169-330779.html#search-en-oscilloquartz-star-4> ) , you will
>>>> see that the phase stability (jitter) on the 1PPS output is +/- 30ns
>> when
>>>> locked to GPS, an it has a timing grade GPS receiver. This is not as
>> good
>>>> as other GPS modules now. 15ns is normal, with some less than half that.
>>>>>> The PRS10 has outstanding PLL control already. The SRS product doc
>>>> gives +/- 10ns accuracy with +/-1ns resolution.
>>>>>> I don’t think that you are buying much with disciplining the PRS10
>>>> with a GPSDO 1PPS. Do you have any TIC measurements in this config to
>>>> compare with a direct GPS 1PPS feed?
>>>>>> 
>>>>> 
>>>>> I forgot to mention one other thing which may be of interest to some.
>>>> The 1PPS wave form output from the PRS10 is pretty mediocre. I put the
>>>> details in another post here sometime back.
>>>>> The Star4 spec is +/- 10ns, something I can only get from my PRS10s
>> with
>>>> a 74HC7001 shaper.
>>>>> 
>>>>> 
>>>>>> 
>>>>>>> 
>>>>>>> A GPS disciplined oscillator contains a GPS receiver which outputs
>>>> 1PPS
>>>>>>> based on receiving the GPS signals and calculating the position  +
>>>> time
>>>>>>> equation. That PPS signal is noisy in time, it jitters around
>>>> relative to
>>>>>>> the ideal 1 second period.  The GPSDO implements a long time constant
>>>> PLL
>>>>>>> to synchronize the output of the OCXO to the long term average
>>>> frequency
>>>>>>> and phase of the GPS PPS, so what you see externally is 10MHz
>> directly
>>>>>>> from the OCXO, 1 Hz (PPS) which is divided down from the 10MHz OCXO,
>>>> and
>>>>>>> those are controlled by a PLL so that long term the phase of the PPS
>>>>>>> divided down from the OCXO follows the PPS calculated by the GPS
>>>> receiver,
>>>>>>> but with lower jitter.
>>>>>>> 
>>>>>>> --
>>>>>>> Chris Caudle
>>>>>>> 
>>>>>>> 
>>>>>>> 
>>>>>>> _______________________________________________
>>>>>>> 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.
>>>>>> 
>>>>>> Don’t worry about how powerful the machines are. Worry about who the
>>>> machines are giving  power to.
>>>>>> 
>>>>>> 
>>>>>> _______________________________________________
>>>>>> 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.
>>>>> 
>>>>> Don’t worry about how powerful the machines are. Worry about who the
>>>> machines are giving  power to.
>>>>> 
>>>>> 
>>>>> _______________________________________________
>>>>> 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.
>>>> 
>>>> _______________________________________________
>>>> 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.
>>>> 
>>> _______________________________________________
>>> 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.
>> 
>> 
>> _______________________________________________
>> 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.
>> 
> _______________________________________________
> 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.