[time-nuts] GPSDO control system

EWKehren at aol.com EWKehren at aol.com
Sun Mar 23 18:32:40 UTC 2014


Hi 
There are many issues when it comes to a GPSO. But what has to  be first 
discussed what is it one wants to accomplish. Last year when we worked  on the 
latest Shera GPSDO we always got better than 1E-11 with a unit lying on  
the bench with no enclosure or thermal management. 
Chasing elusive 1 E-13 and better, allow me to make a couple  of comments. 
In order to get there, the total system has to be under  review.  Since I 
know nothing about writing programs I leave that to  smarter people but be 
clear software and code will not do it by it self. The  most critical part is 
the thermal management of the OCXO or Rb and if analog  control is used the 
DAC.and if used its output amp. We are controlling the back  plate of the 
M100 and FRK to within 0.01 C and the front 0.1 C. The  DAC board and the 
temperature controller are on the front, Voltage regulators on  the back.
After extensive testing the LTC1655 is our preferred  choice. Take a close 
look specs are great for this application and most  important solderable. 
There are better DAC's out there  but very expensive and I am not able to 
solder. 18 bits would be nicer but 16  bits are for Rb's usable. The DAC part 
has to have its own ground plane because  ground loops can create noise and 
voltage changes it has to be tied as close and  separate to the OCXO or Rb. No 
opto Isolation necessary as long as the  controller and DAC are in the same 
box, sharing the same system ground.  Input  to the DAC can handle wide 
ground variations. Took me years to find  that out.
Absolute must  how ever is opto  isolation between GPS, controller and PC, 
again found out the hard  way. 
The other part I like to touch on is the GPS input section. I  am not a 
time nut but a frequency nut, but there has been so much talk in the  past and 
more recently about sawtooth. I am disappointed but not surprised that  no 
one has stepped up and offered a solution. The site has deteriorated to a lot 
 of talk very little action. Right now there are still affordable and 
solderable  DS1023's out there. Combined with a 12F629 or 12F1840 a sawtooth 
correction is  possible for much less than $20. I bought last year four DIP and 
10 in SOIC DS's  average price  below $ 5!  Even though I had a very bad  
experience once with a Dutch so called time nut I am willing to make the  
following offer. For the right person I make a board, PIC and DS1023 available.  
Maybe I just missed it but I do not think that there is something out there 
 readily available.
Bert Kehren
 
 
In a message dated 3/23/2014 9:02:59 A.M. Eastern Daylight Time,  
lists at rtty.us writes:

Hi

The real answer is  always “that depends”.

1)  How much does the sensitivity of your OCXO change with a change in EFC? 
1.4:1,  2:1, 4:1 …. (slope sensitivity not % linearity)

2) How quiet is your  DAC compared to your OCXO? 

3) How quiet is your reference compared to  your OCXO? 

4) How much do the DAC, reference, op-amps, resistors,  capacitors, … drift 
with time? 

5) How much test time is enough?  (hours, days, weeks ,…..)

6) How good is the survey on your GPS this  time?

7) How much does your room temperature impact your OCXO when you  do this 
or that?

8) Is your room temperature representative of the real  world? (is mine 
like yours?)

9) Do you intend this gizmo to work over a  temperature range? Did you test 
that range?

10) Are you trying for best  frequency or best time? Is your definition of 
time “GPS local  time”?

11) Are there voltage drops on your real board? Do they change  with 
anything? (or everything ?) 

12) Does your controller generate  spurs inside the control loop and 
modulate the output with them when tuned to  an offset of x.xxxxxxxxxxxxxxxxxx Hz? 

13) How do things respond to  load changes or supply voltage changes?

14) Are the parts (OCXO,  reference, dac, op amps …) responses to 
temperature, load, supply, tip,  tune,  linear / immediate or do they have artifacts 
that extend out over  longer time periods? 

This is by no means a complete list. A lot of  common GPS issues are 
notably absent.  However, I’ve seen designs fail or  fall short for problems 
related to every item on that list. Can you put this  all in a model - sure. Did 
you put all this in the model ..  ..

Bob


On Mar 23, 2014, at 5:06 AM, Bill Hawkins  <bill at iaxs.net> wrote:

> An idea is struggling to take shape  in my fevered brain.  I'd like to
> check some foundation  assumptions.
> 
> 1. The difficulty with disciplining a local  oscillator to a GPS signal
> is due to variations in the received GPS  signal and the LO.
> 
> 2. The variations occur slowly, as crystal  aging, and quickly - perhaps
> sawtooth or crystal crack propagation -  and maybe something in between.
> 
> 3. The gain of the system, in  degrees of phase angle at 10 MHz (or
> higher) per microvolt of control  signal, is fairly constant in a
> controlled environment.
>  
> 4. The power supply for the device providing the control signal  cannot
> be regulated to the accuracy required of the system, and so is  a source
> of variance. (Does anyone put the voltage reference device in  the oven
> with the crystal?)
> 
> 5. The principle source  of environmental variation is temperature.
> Humidity and barometric  pressure are not significant. This may not be
> true of the received GPS  signal due to atmospheric variations.
> 
> 6. A digital  computational device is available to calculate the control
> signal from  various measurements and previous values.
> 
> 7. There are no  supernatural forces at work, such as the experimenter
> mentally  influencing the results. :-)
> 
> That's a start . . .
>  
> Thanks for any replies.
> 
> Bill Hawkins
>  
> 
> 
> 
>  _______________________________________________
> time-nuts mailing list  -- time-nuts at febo.com
> To unsubscribe, go to  
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the  instructions  there.

_______________________________________________
time-nuts  mailing list -- time-nuts at febo.com
To unsubscribe, go to  
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the  instructions there.



More information about the Time-nuts_lists.febo.com mailing list