You need to take into accout the frequency period over which you need
stability.    A locked frequency reference used for LF has very
different requirements over one used for stabilising moicrowave
sources - even though the headline stability,  typically 10^-9 may be
similar for both ends of the spectrum.

Firstly consider microwaves 10 and 24GHz in particular wher enerly
everyone uses GPS locked sources.  Short terms stability over seconds
to tens of seconds is necessary there, as otherwise SSB and CW  will
drift and.   A Simple GPSDO like the one on my website is good enough
just, but the 10 - 20Hz of frequency variation can be heard as a
steady up and down drift in tone frequency.      A better GPSDO is the
VE2ZAZ design, using teh 1 PPS to freqeuncy steer (not lock!) an
inherently good to start with OCXO.  I have two of these, one runing
24-7 as a master reference, and another set for faster lockup and
resulting lower spec for /P operation.  Both give better thatn 10^-10
over just about any measurement period I chhose to test over.

But for LF the requirements are very different.   Short term stability
over seconds is irrelevent if the signalling interval is much longer
than this, provided it remains locked at all times.   So now you can
sacrifice short term stability and hence no longer need an OCXO as the
primary source.   In fact, the 1 PPS signal from the GPS can be used
directly to steer a simple TCXO.  The note may sound horribly drifty
at HF/VHF but provided the average stays locked with a time constant
in teh tens of seconds to keep within TCXO drift. it will be fine for
LF/VLF.

So a quite simple GPSDO can be made up like this, without having to
rely on a specific GPS module with a 10kHz output.  In RadCom October
2002 I did a "GPS Locked Frequency Standard for LF Signalling "  That
did what it was supposed to do on 137kHz  for very long duration
signalling, although I wouldn't recommend doing it like that now with
its obsolete PIC and a bodged bit of software.   There is a far better
PIC with internal comparator / timer - the 16F628 and can be used in
an identical way to the VE2ZAZ design but with a TCXO and making it a
lot simpler.

G3PLX even constructed an analogue PLL using the 1 PPS signal from a
GPS module and a TCXO - not a uProc in sight!

Off air radio sources may give you semi-decent short term stability
during daytime and the middle of the night, if interference is not
present, and teh antenna tuning/ phase isn't altered, and, and, and...
Their stability over hours is of dubious value; and this is the very
requirement for LF.

So, conclusion.  I would avoid off air such as MSF and Droitwich
locked standards for very narrow band long duration working.   Go for
GPS locked stability using the 1 PPS output from any low cost GPS
module.  Either build a VE2ZAZ GPSDO, or have a go at a simple PLL
that is probably as chirpy as anythiing over several seconds, but will
remain reliably locked and stable over minutes to hours to days.

See www.g4jnt.com for a lot of the work I've done on frequency
standards and frequency locking over the years.

Andy






On 19 March 2011 17:09, James Moritz <[email protected]> wrote:
>
> Dear Ken, Wolf, LF Group,
>
> I have been playing around with the insides of a Halcyon PFS-1 
> Droitwich-derived frequency standard over the last few days, and some 
> observations might be of interest.
>
> The Receiver part of the PFS-1 is an off-air frequency standard that 
> generates 10MHz, 5MHz and 1MHz outputs from the received 198kHz carrier, and 
> is quite similar to the Quartzlock 2A device on Ebay. It has an external 
> ferrite rod antenna, a single crystal 198kHz filter, limiting amplifier and a 
> PLL which locks a 10MHz VCXO to the 198kHz signal. Most of the time it stays 
> locked, but sometimes it becomes unlocked for a few minutes at a time, for 
> reasons unknown. The 10MHz output is switched off when this happens.
>
> The rest of the PFS-1 is a DDS synthesiser that covers 0 - 16MHz in 100uHz 
> steps. It has an internal 20MHz OCXO module, which is phase locked to the 
> 10MHz reference from the receiver when that is present, and is otherwise 
> free-running, at a frequency which is calibrated via a fine-tuning trimpot. 
> The OCXO is a HCD71 module, which has specified ageing rate <1E-8 per day. 
> The OCXO module in isolation seems to do much better than that, but when my 
> PFS-1 was free-running, I found the stability was rather worse. After poking 
> around inside, I think this is partly due to the trimpot used (a 20k, 22 turn 
> component - these things are not great in my experience, not very stable and 
> with a lot of backlash in the adjustment), and partly due to the circuit 
> connected to the OCXO frequency control input, which is the PLL 
> amplifier/loop filter, with the addition of some switching that, when the 
> external reference is removed, sets the frequency control voltage using the 
> trimpot, another resistor and a 78L09 regulator IC.
>
> Bearing in mind that 1 part in 10^8 requires the 0-6V frequency control 
> voltage to be held stable to within 1mV or less, this needs quite a good, 
> stable, bias voltage and pot to achieve the specification. I have now 
> modified my unit to use the 6V output from the OCXO module and a 10k, 10 turn 
> helipot on the front panel, with a CMOS switch to select the PLL output when 
> operating. This seems to have improved the free-running stability 
> significantly.
>
> It is interesting to compare the free-running OCXO output to the 10MHz locked 
> receiver output on an oscilloscope. Once the OCXO is trimmed, the phase of 
> one waveform "walking" relative to the other shows frequency differences down 
> to the parts-per-billion level quite easily. It seems the 198kHz carrier has 
> some kind of cyclical drift in phase which occurs over a period of a minute 
> or two (the receiver PLL seems to filter out the faster phase modulation 
> data), so the direction of "walking" changes from time to time.
>
> For using the PFS-1 as a frequency source for high stability signals over 
> periods of hours, it seems to me it would be better to operate the 
> synthesiser part in "unlocked" mode, and periodically trim the OCXO frequency 
> using the receiver part. Otherwise, there are bound to be glitches in phase 
> when it switches between off-air reference and OCXO, as it does from time to 
> time. Also the observed wobbling phase of the 198kHz carrier, plus any 
> ionospheric effects around darkness hours, would be reflected in the 
> synthesiser output. The same possibly applies to the Quartzlock 2A.
>
> Cheers, Jim Moritz
> 73 de M0BMU
>