[time-nuts] yet another GPSDO design, or so

Stanley Reynolds stanley_reynolds at yahoo.com
Sun Jun 27 13:04:46 UTC 2010

I have been thinking about a faster counter also but the Shera board was depending on the jitter in the 24 Mhz clock to average out the +- count. The faster clock would reduce the need for this but without the right amount of jitter we lose the benefit of this average.


----- Original Message ----
From: "EWKehren at aol.com" <EWKehren at aol.com>
To: time-nuts at febo.com
Sent: Sat, June 26, 2010 8:14:02 PM
Subject: Re: [time-nuts] yet another GPSDO design, or so

What you want is basically a Shera Board. That design has been around for  
quite some time and has served me very well. I have a total of six running  
including two controlling Rubidium. There are in my opinion a couple of  
problems: not every 4066 works on the design the 18 bit D/A is very hard to 
find  and now expensive and the single step of the D/A is intended for a 1.7 
E-13  frequency step. I have build a input section that counts 100 MHz in 
stead  of  24 MHZ making the unit create steps of 4.3 E-14 which works better on 
my Rubidium's and Datum FTS 1000.  Also it eliminates the 4066's. Since I  
do not know how to write code that was my solution. I have also designed a 
later  version Shera, with less IC's and a low cost dual D/A but I do not 
have the  programming skill. 
If you contact me directly I will send you a copy of the QST Shera article, 
my design and the D/A data sheet.. I am sure you can replace the PIC with 
an  Atmel device.

In a message dated 6/26/2010 1:16:09 P.M. Eastern Daylight Time,  
attila at kinali.ch writes:


I recently had a look at the data sheet of the LEA6-T  GPS module
from ublox, which now features a second time pulse output  that
is capable of delivering a 10MHz signal, synchronized to  GPS.

After thinking quite some time quite some time about  building
my own GPSDO and struggling with the question how to  synchronize
a 10MHz signal to a 1Hz signal that has some substantial  phase
noise, the new LEA6-T module seems like to make things a  lot
easier. Although the LEA6 specs do not say anything about how
the  timepulse output is generated or how it is synchronized
to GPS, i assume  that it will either have some jumps or phase/frequency
noise due to  oszillator and synchronization imperfections.

But, it should be  possible to use the LEA6-T together with
some OCXO and a PLL setup to  stabilize the OCXO to get a high
quality frequency  standard.

Unfortunately, my knowledge in that field is rather limited,  thus
before starting to make wrong design decisions i'd like to ask
for  some advice here.

My basic idea is to feed the 10MHz output of the  LEA6-T and
the 10MHz OCXO into a current output PFD, do some  low-order
filtering of the output signal. Feed that into an ADC which
is  read by a uC which in turn controls an DAC that sets over
some amplifier  stage the EFC input of the OCXO.

As PFD i thought about using a ADF4002  from Analog, which
is actually an PLL, but allows to bypass the input  dividers,
so that it can be used as pure current output PFD.

I'm not  yet sure what kind of output filter i want to use.
I probably have to add  at least one low noise opamp there,
to isolate the PFD output/filter from  the ADC. I'm also
not sure what filter frequency i should use here. It  will
have to be below 10MHz for sure, probably in the lower 
kHz range,  but how low is the question. The lower the easier
gets the ADC stage and  the less work has to be done in the uC,
but using a low frequency filter  either means using an active
filter (noise) or high value R or L (again  noise, especially
the L might couple in 50Hz noise from the enviroment or  show
microphone effects).

The ADC will be either a low-noise 16bit  type or a 24bit
type. This will largely depend on the sample rate to  be
used and the availabilty of the ADCs. Any good advices
on what to use  here? Should there be some form of signal
conditioning done? If, what form  of conditioning would
you advise me to use?

As a uC i thought about  using a AT91SAM7 variant from Atmel.
I know these beasts (and their bugs)  pretty well by now
and already have some code ready for those.
I thought  about clocking the uC with a 40MHz crystal that
is synchronized to the  10MHz OCXO using a PLL. This would
allow me to generate quite  precise+accurate digital signals.
Unfortunately, there doesnt seem to be  VCXOs at 40MHz available
so that means that i'd have to build one by  hand.

The loopfilter is going to end up in the uC as it is easier
to  build such low frequency filters digitally than in analog.
I havent put  much thought into how that filter should look
like, as this can be easily  changed later.

The DAC will probably be a 16bit type (there does not  seem
any higher resolution DAC with sane specs and still  reasonable
availability). The amplifier for the DAC output will be a  two
stage amplifier. One stage that adds an (adjustable) offset
and one  stage that adds the (again adjustable) amplification.
This approach is  choosen as the needed EFC range will probably
much lower than the full  range. Hence the resolution of the
DAC can be enhanced by producing only  values within that range.
The disadvantage here is that it requires  calibration.

A rough guestimate is that the whole thing will probably  cost
less than 500CHF (including PCB production, but excluding  OCXO).
Yes, i know, i could get a Rb frequency standard for that  money
on ebay. But where is the fun in that? ;-)

Beside whether this  setup makes sense, the two biggest questions
i have are, what OCXO to use.  Are the ISOTEMP 134-10 that are
available on ebay "good enough" for such an  application?
Or shall i look for something better/different?

And the  other is, how do i amplify and distribute the 10MHz
signal i get out of the  OCXO to be used by other devices
with minimal phase  noise?

Thanks for your help

Attila Kinali

If you want to walk fast, walk  alone.
If you want to walk far, walk together.
-- African  proverb

time-nuts  mailing list -- time-nuts at febo.com
To unsubscribe, go to  
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