[time-nuts] GPSDO using 100Hz

[WarrenS]

WarrenS Email wrote:
>
> Bruce, Thanks for the feedback, Good information to know, 
> but you seemed to missed my point and question.
> in it and must use a processor?
>
> Also I should comment that on  LeapSecond.com 
> http://www.leapsecond.com/pages/gpsdo/
> you stated "where the ADEV for various GPSDO remains below 1E-11 
> over the Tau range of [0.1s, 100,000s]. This Tau has NOTHING to do with 
> the tracking time constant that the GPSDO is set to, which is usually 
> recommended to be set for a TC of several hours for optimal results.
> BTW most of what they have plotted is the results of NOT setting the GPSDO 
> tracking TC slow enough. This is why the Allan Deviation increases in the 
> tracking 
> mode at mid averaging times. The 1 second GPS tracking signal is adding 
> noise, 
> which pretty much makes my point that 1 PPS signal is not so good to use 
> if you want good fast results.
>
> Warren
>   
Analog time constants of several hours are generally considered
impractical due to the lack of suitable low noise components principally
high value resistors and capacitors.
So how do you propose to get around this with an analog control loop?

Bruce

*****************
Bruce

In answer to your question of :
"how do you propose to get around this with an analog control loop?"
Short answer is: By using the 100Hz output pulses, instead of the 1 PPS.
What I have found is that it takes about the same number of total pulses for a given accuracy
whether it is 1 Hz or 100 Hz. SO by using the 100Hz signal, the time Constant can go down
by nearly 2 decades. What I have also found is that a 100 seconds RC TC, which is very practical for analog, 
gives pretty good results AND of course this whole discussion only applies to errors caused by the 100 ns 
pulse jitter and not the 10ns GPS jitter, and it does not apply if you are using the 1Hz offset correction to get
each down to 1 ns, etc. This is for the basic GPS tracker that does not use the sawtooth correction.
The reason this works is because it is easy enough to get about a 1ns uncertainty for a 1 sec average 
using the 100 Hz instead of the nearer 100ns uncertainty of the 1 PPS. 
(Note what I am talking about is Typical performance. Worse case, a couple times a day it has some 
other  problems, that I would love to discuses with you, under an advanced topic, if we can ever get that far)

Under another subject, 
You said in one of your other post that the 100hz is not useful for an XOR phase detector because only one 
of its edges are known. 
Sorry, I thought we where above that kind of basic information exchange. I just took it 
that everyone that cares, knows the 100Hz has to be divided by 2 (using the correct edge) 
before it is feed to a simple XOR phase detector. So a 100 Hz XOR phase detector runs 
with 50 Hz square wave inputs and gives out a 100Hz, 50% duty cycle square wave at lock.
(And has a very low gain, which is why I don't use it for low frequencies)

Warren