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

Attila Kinali attila at kinali.ch
Sat Apr 9 11:41:47 UTC 2016


On Fri, 8 Apr 2016 21:31:31 -0500
"Graham / KE9H" <ke9h.graham at gmail.com> wrote:

> The lowest jitter way to do this kind of conversion is to multiply the
> signal up to some common multiple frequency, then divide it back down to
> where you want to be.  For instance, with 8 or 24 MHz, multiply up to 240
> MHz, then divide by 24 to get 10 MHz.

In this case, I would rather recommend against multiplying up first.
The added complexity (which is quite a bit, compared to a simple digital
divider) is managable, but there is an issue with intermodulation products
that is not so easy to deal with:

The input signal, which comes out of a digital "synthesis" chain within
the GPS module, is anything but clean. There are lots of spurs close to
the carrier. Due to this, the multiplied signal will contain lots of
intermodulation products, which will degrade the signal considerably.
And these intermodulation products will not magically vanish once you
divide down again. It would be better to multiply the 10MHz signal up
to 60MHz and divide down to 12MHz instead.

Additionally to this, keep in mind that division will decrease the noise
floor with only 10*log(N) when done using digital division instead of
the 20*log(N) when done analog (using regenerative dividers, lamda dividers
and such). Ie, multiplying a signal by 10 and then dividing it down by 10 
using digital logic will increase the noise floor by 10dB. (see [1])

Given all this, the gain in performance using a multiply-divide chain
for the PLL instead of a divide-divide chain, is irelevant considering
how dirty the GPS module provided reference frequency is.

			Attila Kinali


[1] "The sampling theorem in Pi and Lambda dividers", 
by Calosso, Rubiola, 2013, 
http://rubiola.org/pdf-articles/conference/2013-ifcs-Frequency-dividers.pdf
slides: http://rubiola.org/pdf-slides/2013C-IFCS--Dividers.pdf

-- 
Reading can seriously damage your ignorance.
		-- unknown



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