[time-nuts] Advise on building a DIY GPSDO?

EWKehren at aol.com EWKehren at aol.com
Sat Apr 9 10:07:07 EDT 2016


I do not know what U blox does but I know when we use 200 KHz out of the 1  
pps output on a $ 10 ublox 6 we consistently get better than 1 E-10 closer 
to 1  E-11 out of the Morion have the data
Bert Kehren
 
 
In a message dated 4/9/2016 10:01:05 A.M. Eastern Daylight Time,  
kb8tq at n1k.org writes:

Hi

> On Apr 8, 2016, at 9:39 PM, timenut at metachaos.net  wrote:

> 
> Hello Bob,
> 
> Friday, April 8, 2016,  6:13:07 PM, you wrote:
> 
>> Hi
> 
>> If you  start from a 24 MHz TCXO (different modules use different TCXO’
s):
>  
>> On an 8 MHz output, most of the time you divide by three.  
> 
>> On a 10 MHz output, you need to divide by 2.4. The net  result is that 
you 
>> divide by 2 sometimes and 3 other times.  
> 
>> In the 10 MHz case, there is a *lot* of energy at 12 MHz  and 8 MHz, 
along with
>> the 10 MHz output. 
> 
>> In  the 8 MHz case, most of the RF energy is at 8 MHz.
> 
>>  ====
> 
>> To correct the output by 1 ppm on the 8 MHz output,  you need to either 
drop or
>> add one pulse out of every million  pulses. Effectively you divide the 
24 MHz by
>> 2 or by 4 when you do  that. You get a bit of 12 MHz or a bit of 6 MHz 
as a result.
> If you  know you are doing a 24Mhz and a 10Mhz, why not divide the first 
by 12
>  and the second by 5 and then phase lock the resulting 2Mhz? Or divide by 
 24
> and 10, respectively and lock the 1Mhz? That way, everything is  exact.

The bigger problem is that the 24 MHz is *not* exact. It is  simply a free 
running TCXO
that happens to be in a GPS module. It has a  basic accuracy of +/- 1 ppm 
or something 
similar. It is no better or worse  than any other TCXO you could buy. 

To make it accurate they have two  choices:

1) Put a voltage control input on the TCXO and turn it into a  TCVCXO, then 
lock it up 
with a loop.

2) Let the oscillator free run  and “fix up” the output.

For a variety of reasons, none of the small  GPS modules go with option 
number 1. They 
all go with option number 2. The  24 Hz error on the (maybe)  24 MHz gets 
taken out by dropping
24 edges  every second. That’s not a lot of edges, it’s not going to turn 
the output  into absolute 
garbage you can see on a scope. It is plenty of nonsense to  mess up a 
radio or a piece of test gear. 

One easy way to look at it:  You have ~1 ppm jitter on the output (in the 
example of 1 ppm of error). A  
phase locked GPSDO with only simple filtering of a 1 pps would get you  
down to 0.01 ppm of jitter. 
A sawtooth corrected 1 pps would get you to  0.01 ppm. A good filter would 
get you to <0.00001 ppm.
Yes, I’m using a  very hand waving definition of jitter here, but it does 
illustrate the point.  You could 
look at the jitter on the pulse drop as 0.04 ppm.  

Bob

> 
> 
> Mike
> 
>> That can  be filtered out with a RF filter. The same is true with a 
(somewhat  more
>> complex) filter on the 10 MHz output.
> 
>> In  addition to the “big” RF spurs, you get a low frequency component 
to the  output
>> modulation. You are “phase hitting” the output eight times  a second. 
That gives you
>> an 8 Hz sideband along with the further  removed stuff. Since it’s not 
simple / clean
>> phase modulation,  there are more sidebands than just the few mentioned 
above. 
>  
>> What messes things up even more is that you never are quite doing  one 
ppm. You are doing
>> corrections like 0.12356 ppm this second  and 0.120201 ppm the next 
second. 
>> The pattern of pulse drop and  add is not as simple as you might hope. 
The low 
>> frequency part of  the jitter (and it will be there) is no different 
than the noise  on
>> a 1 pps output. You still need to do very long time constant  (or very 
narrow band)
>> filtering to take it out. 
>  
>> Bob
> 
>>> On Apr 8, 2016, at 7:06 AM, Herbert  Poetzl <herbert at 13thfloor.at> 
wrote:
>>> 
>>> On  Mon, Apr 04, 2016 at 06:07:54PM -0700, Alexander Pummer  wrote:
>>>> and it is relative easy to make 10MHz from 8MHz  with analog
>>>> frequency manipulation, which generates less  jitter
>>> 
>>> Could you elaborate on this a little  if time permits? 
>>> I'm more a 'digital person' but it sounds  interesting.
>>> 
>>> Thanks in  advance,
>>> Herbert
>>> 
>>>>  73
>>> 
>>>> On 4/4/2016 4:27 PM, Attila Kinali  wrote:
>>>>> On Mon, 4 Apr 2016 17:56:29  -0400
>>>>> Bob Camp <kb8tq at n1k.org>  wrote:
>>> 
>>>>>> The variable frequency  output on the uBlox (and other) GPS
>>>>>> receivers has  come up many times in the past.
>>> 
>>>>>>  If you dig into the archives you can find quite a bit  of
>>>>>> data on the (lack of) performance of the  high(er) frequency
>>>>>> outputs from the various GPS  modules. They all depend on
>>>>>> cycle add / drop at  the frequency of their free running TCXO.
>>>>>>  Regardless of the output frequency, that will put a *lot*  of
>>>>>> jitter into the output.
>>>>>  That's why you should put the output frequency of the ublox  modules
>>>>> to an integer divisor of 24MHz. Ie 8MHz works  but not 10MHz.
>>> 
>>>>>       Attila Kinali
>>>  
>>> 
>>>>  _______________________________________________
>>>> 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.
> 
>>  _______________________________________________
>> 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.
> 
> 
> 
> -- 
> Best  regards,
> Timenut                mailto:timenut at metachaos.net
> 
>  _______________________________________________
> 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 mailing list