[time-nuts] Simple GPSDO (was: Digest, Vol 130, Issue 38)

Attila Kinali attila at kinali.ch
Wed May 27 12:40:08 EDT 2015


On Tue, 26 May 2015 16:58:00 +0000 (UTC)
Anton Moehammad via time-nuts <time-nuts at febo.com> wrote:

Could you specify a bit more clearly what kind of problem
you are trying to solve? Do you want to build a GPSDO
to stabilize your OCXO? If so, what is the main measure
you want to improve? Is it ADEV at a certain tau? Or
is it frequency accuracy?

What is the level you want to reach? Or if you do not know
this, what do you want to use the GPSDO+OCXO for?

> Now What in my mind 
> I will use Ublox 7m GPS board that can program from 0,25Mhz to 10Mhz
> (https://sites.google.com/site/g4zfqradio/u-blox_neo-6-7)
> 
> But the problem are I found someone (http://www.ra3apw.ru/proekty/ublox-
> neo-7m/) already do the spectrum analysis for this board and the result is 
> like this :
> and the last picture is far better, correct ? so what I need to do I have 2 
> plan :

Uhmm.. I don't know any russian. And i don't think many of the people here
do. Could you shortly summarize what the graphs show and to what conclusion
you came?

What I think I see there, are the spurs in the Timepulse output if
frequencies are used that are not an even divisor of the internal clock.
The internal clock runs at 48MHz, if they didnt change the general design
of the LEA modules since the LEA-6 family. Ie, a 8MHz output produces lower
spurs than a 10MHz output, because 48/8=6 while 48/10=4.8

Also keep in mind that there are two modi for the timepulse: fixed
frequency and fixed phase. Depending on which application you are
targeting, the one or the other might be the better choice.

> 1. use the board 10Mhz output than use very narrow (100Hz bandwith) Crystal 
> filter

Narrow filters induce large (for time-nuts values of large) phase shifts in
the ouput if the center frequency of the filter moves around due to
temperature changes. Hence this might not be what you want.

> 2. Use the 4Mhz output than again filter it use Crystal filter to clean it 
> than use it as reference for PLL for controling 10Mhz OCXO 

The use of a PLL is definitly the better way. If you make the loop filter
of the PLL narrow enough, then the jitter/spurs of the reference signal
will be filtered out by the PLL and the OCXO will dominate the noise for
offset frequencies larger than the loop filter bandwidth. Ie, you can
get rid of the output filter of the LEA module.

I don't know the MTI 240, but from the datasheet it looks quite decent.
So i would guess that you want to go for an as small loop filter
bandwidth as possible, probably just a couple of Hz wide, even.
How far down you can go with the bandwidth will be limited by the stability
of the PLL circuit.

> Which one you think a better way.
> last I also like to know because MTI only spec the OCXO model 240 harmonics 
> to down to -30dB (http://www.mti-milliren.com/pdfs/240.pdf) is there any 
> benefit to use a crystal filter in the output to clearer it more.

You can use filters on OCXOs to improve their output. But my understanding
is, that it is more likely to decrease the purity and stability of the
output instead. But i have to admit that this is definitly not my field
of expertise, so i leave it to others to answer.


				Attila Kinali


-- 
< _av500_> phd is easy
< _av500_> getting dsl is hard


More information about the time-nuts mailing list