[time-nuts] Setting up PRS10 with M12+T GPS

[Magnus Danielson]

Hi Björn,

On 05/02/2012 11:50 PM, bg at lysator.liu.se wrote:
>>> The phase of the PPSOUT and the generated 10 MHz will probably not align
>> up,>something which is critical to some applications.
>> Yes, for example the DVB-T SFN requires the PPS and 10MHz phases to be
>> aligned, better, not to move. This requires to slightly alter the 10MHz to
>> keep the PPS aligned.
>>
>> Let me ask Bjorn: did you succeed in aligning the two PRS10?
>
> Somewhat... I had a "stable" offset between the 10MHz PRS10 signals of
> almost 8ns, moving +-2.5ns over some hours.

Recall that input resolution is about 0,2 ns and output resolution is 
about 0,5 ns and that 1 ns seems to be around where input and output 
precision is really given... so some of that may be traceable to those 
aspects.

> Moreover,
>> what
>> do you mean by "each other", a scheme like this:
>>
>>
>>
>> |-------------------------------------------------------------------|
>> |                                                                   |
>> |       PRS10A                        PRS10B           |
>> |--->PPS IN PPS OUT------>PPS IN PPS OUT--|
>>
>>
>> hoping that the font doesn't mess up the ASCII schematic...
>
> Almost, I also let the PRS10A be steered by a Tbolt.
>
> PLL settings 8 on PRS10A, and 0 on PRS10B, which should make PRS10B quick
> enough to follow any wiggle the PRS10A do slowly following its Tbolt
> supplied PPS_in.
>
> I had
>
>      locked Tbolt 1PPS going into PPS_IN of PRS10A
>      PPS Out of PRS10A going into PPS_IN of PRS10B.
>
> I watched/measured
>      1) PPS out from Tbolt on ch 1  (yellow)
>      2) 10MHz out from PRS10B on ch 2 (green)
>      3) 10MHz out from PRS10A on ch 3 (blue)
>      4) 1PPS out from PRS10 B on ch 4 (red)
>
> see attached screendump from a 500MHz Scope.  I did not have enough time
> to completely remove offsets between 10MHz signals from PRS10A&  B. More
> tuning should have been done on the PRS10A wrt the Tbolt PPS also.
>
> Any comments?

The timing offsets of both PRS-10s may need a bit adjustment to align up 
better. The thunderbolt PPS noise is obviously not wonderful. The higher 
noise on the PRS-10B 10 MHz (green) compared to the PRS-10B PPS (red) is 
interesting. This may be due to the PPS output delay "hiding" the error, 
something which you can grasp from the last sentence of page 33 in the 
manual.

Thus, the PPS output delay is used to "cover up" the time-error that the 
PLL loop provides. It looks like it is better locked-up then it actually 
is, so looking at the loop state will be necessary to really know.

Cheers,
Magnus