[time-nuts] quick and very dirty phase comparator

Thu Jun 5 00:23:51 EDT 2008

Bill,

yes, I have been thinking exactly this way too. The easist way would be
to divide the reference by 2 giving 5 Mhz and then use an odd integer
multiplier in the DDS. However, I believe that the AD9851 is limited
concerning its multiplier range, so it might be necessary to use an
AD9852/54 which again is a complication (that I don't like) of an
otherwise "easy" project.

73 Ulrich

> -----Ursprungliche Nachricht-----
> Von: time-nuts-bounces at febo.com 
> [mailto:time-nuts-bounces at febo.com] Im Auftrag von WB6BNQ
> Gesendet: Donnerstag, 5. Juni 2008 00:41
> An: Discussion of precise time and frequency measurement
> Betreff: Re: [time-nuts] quick and very dirty phase comparator
> 
> 
> Ulrich,
> 
> One of things I noticed when playing with that DDS design 
> tool is you can achieve the same output frequency using a 
> different clock frequency, thus removing the spurs.  THe 
> trick is to look at clock frequencies that you can lock to 
> your house reference for stability.  Also, it might be 
> possible to use cascading DDS chips, one providing the clock 
> for the other.
> 
> What are the thoughts on such an approach ?
> 
> Bill....WB6BNQ
> 
> 
> Ulrich Bangert wrote:
> 
> > Bruce,
> >
> > > Even a DDS followed by a PLL cleanup loop (10811 plus analog PD 
> > > etc.) should work well although with a binary tuning word 
> obtaining 
> > > an exact 10.00001MHz (or alternatively 9.99999 MHz) output isnt
> > > possible. A DDS has some advantages over a synthesizer using
> > > dividers in that
> > > additional noise isnt aliased into the output.
> >
> > Since I am well familiar with the Analog Devices DDS circuits, this 
> > has been my very first idea. The most simple one for that purpose 
> > would be a AD9851 (180 MHz, 32 Bit, built in clock multiplier). But 
> > when I used the DDS design tool available on the AD web pages I 
> > received a big warning saying that using a "clock X multiplier" 
> > frequency that is a near integer of the output frequency generates 
> > lots of unwanted spurs. Which was new to me since I do so 
> in my GPSDO 
> > but should they not know better? This is why I dropped the 
> thoughts on 
> > DDS.
> >
> > Best regards
> > Ulrich Bangert
> >
> > > -----Ursprungliche Nachricht-----
> > > Von: time-nuts-bounces at febo.com 
> [mailto:time-nuts-bounces at febo.com] 
> > > Im Auftrag von Bruce Griffiths
> > > Gesendet: Mittwoch, 4. Juni 2008 10:44
> > > An: Discussion of precise time and frequency measurement
> > > Betreff: Re: [time-nuts] quick and very dirty phase comparator
> > >
> > >
> > > Ulrich Bangert wrote:
> > > > Bruce,
> > > >
> > > > thank you for correcting me. Here I have clearly fooled myself. 
> > > > However your posting originated some new ideas: With the
> > > GCD becoming
> > > > THAT low an analogue phase lock to a 10 MHz reference 
> will not be 
> > > > easy. But if we stop to think about phase locked VCXOs 
> we need not 
> > > > bother anymore about odd exotic xtal frequencies
> > > Yes you would need a VCXO with low close in phase noise for the 
> > > 17.73447MHz source. That crystal frequency isnt too exotic as RS 
> > > components have suitable crystals, if you build your own VCXO.
> > > >  at all that may generate us a GCD of 10.
> > > > Instead we are free to choose for example 10000010 Hz for the 
> > > > controller's frequency. Which brings us back to a 
> construction of 
> > > > a good offset generator.
> > > >
> > > > Until now I have believed that a good (low phase noise, high
> > > > stability) offset generator would involve
> > > >
> > > > a) a number of single sideband mixers (as described in
> > > > www.horology.jpl.nasa.gov/papers/fssa.pdf)
> > > >
> > > > or
> > > >
> > > > b) the well known offset synthesizer circuitry as described by 
> > > > Rick Karlquist.
> > > >
> > > > I am sure that both ideas work excellent, although I am
> > > unsure whether
> > > > a) can generate an 10 Hz offset. However, both methods involve 
> > > > circuitry that I would not call exactly "quick and dirty" and 
> > > > their use would overstress the try to make something really
> > > simple. On a new
> > > > internet search for "offset generator" I came over this one:
> > > >
> > > >
> > > 10Hz offset by method a is trivial (9.99999MHz is just as 
> useful as
> > > 10.00001MHz):
> > >
> > > 1) Use a LSB mixer to mix 10MHz with 10MHz/1000 to 
> generate 9.99MHz
> > >
> > > 2) Bandpass filter this and then use a USB mixer to mix 
> 9.99MHz with 
> > > 9.99MHz/1000 to generate 9.99999MHz.
> > >
> > > 3) Use a PLL to phase lock a low noise VCXO (spare 10811A 
> or similar 
> > > detuned mechanically by 10Hz??) to the 9.99999MHz output 
> to remove 
> > > spurs etc.
> > >
> > > Even a DDS followed by a PLL cleanup loop (10811 plus analog PD 
> > > etc.) should work well although with a binary tuning word 
> obtaining 
> > > an exact 10.00001MHz (or alternatively 9.99999 MHz) output isnt
> > > possible. A DDS has some advantages over a synthesizer using
> > > dividers in that
> > > additional noise isnt aliased into the output.
> > > >
> > > www.diva-portal.org/diva/getDocument?>
> > urn_nbn_se_liu_diva-1838-1__fullt
> > > > ex
> > > > t.pdf
> > > >
> > > > What do you think about that topology? Let the "IF in" be
> > > the needed
> > > > offset and the "offset OSC" be our 10 MHz reference. Would that 
> > > > not make an really easy way to generate an precise offset with
> > > the wanted
> > > > features?
> > > >
> > > >
> > > Not much (can work well with 20kHz offset but not with 
> 10Hz offset), 
> > > direct generation mixing 10Hz with 10MHz like this 
> requires a rather 
> > > low PLL bandwidth.
> > > Also filtering out the unwanted sideband may be problematic.
> > >
> > > > Best regards
> > > > Ulrich Bangert
> > > >
> > > >
> > > Bruce
> > >
> > >
> > > _______________________________________________
> > > 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.
> 