Re: [rsgb_lf_group] Re:: Fwd: LF: Frequency References GPSDOs etc. (was EbNaut etc...)

[Andy Talbot <andy.g4jnt@gmail.com>]

What is needed is a design for a homebrew 1 PPS PHASE LOCKED source.    Both the Brooks Shera  and VE2ZAZ designs are frequency locked loops - although it could be argued that frequency and phase are the same things, and it is the order of the PLL that is different.  But whatever...  they give noticeable errors short to medoium term

Commercially quite low cost GPSDOs, like one I have from [what used to be] Connor Winfield give results that when averaged over a few minutes are quite accpetable in terms of absolute accuracy, and are in the sub-PPB region over shorter periods.    The paper on my web site,  URL given in a previous post, shows the CWin  GPSDO short term stability.

Those low end units  use what would appear to be phase locking, and I see no reason why a simple edge triggered phase detector at 1PPS, using high speed logic couldn't be made to work.  Problems ion the past have centred round phase detectors being not very useful when asked to deal with difference in tens of ns, but these days, one made using VHF logic or a gate array ought to be able to manage.   I don't think an XOR phase detector is feasible, the filtering to get rid  of the 2Hz residue, even at very long time constants, would be too much.    Getting a real long time constant could be a bit of a headache without going into DSP, but good quality non electrolytics are around for several uF, and if we're not worried by size, several in || to say 50uFm with 20M resistors around CMOS opamps ought to be good enough - that's up to 20 minutes time constant and in the realm of needing an OCXO as the 10MHz source.  Construction would need care to avaoid leakage though.    My MSF reference way-back-when used CR values of this sort of value, and it did mostly work. Phase changes of RF due to  proximity effects around he ferrite rod were its major downfall.

  And all the rest is just basic division

On 9 December 2015 at 02:54, g8lco10@yahoo.co.uk [rsgb_lf_group] <rsgb_lf_group@yahoogroups.co.uk> wrote:

 

It is a not insignificant fact that Caesium Clocks are very expensive and have finite lives as their caesium "fill" moves from the source end to the ioniser. Current HP 5071's have a 7 year lifetime on the High Performance option, the replacement Physics package was over $20,000 some years back. So that's a cost of ownership of  around £2,000 per year after laying out  the cost of a cheap house. Even if you go for the lower performance option involving a reduced beam current and poorer stability you still have the initial cost and >£600 per annum cost of replacing the Cs.

The earlier 5061's  that sometimes pop up surplus had around 5-6 year run lives so could require a very expensive repair. So the Cs clocks in Ham ownership are few and far between. An "unused " one is not a good idea either, the physics pack needs to be in standby not off !

 The notion that a Cs clock is primary is interesting.  Although  time is now defined as a number of cycles of a Cs state transition frequency that applies to very long tubes run as National Standards and given a National Standards level of support, a rack mount short tube IS TWEAKED onto frequency by adjusting the C magnet field. The tweak is very fine however it is finite, the Cs resonance is quite wide- around 1KHz- and is biased by an axial magnetic field. Individual rack mount clocks do show some variation, for critical applications people run groups of clocks and average the outputs ( e.g. the working GPS master standard)

It seems improbable that Cs will be "primary" much longer, H has better short term noise whilst laser cooled clocks go way beyond the Cs limits by greatly reducing thermal noise however they won't be cheap.The question of which will win is complicated because the different systems keep leapfrogging in performance as new refinements are developed.

Rb and Cs clocks are atomic resonators, they work by amplitude locking the synth harmonics of a good xtal oscillator to the resonance so the noise performance of the clock is that of the crystal oscillator at around 1S averaging time.  You cannot  increase the loop bandwidth much because of modulation leakage and bode stability. So a very good GPS corrected crystal osc is about as good as an atomic standard for some uses in ham radio in terms of short term noise.

Good GPS based oscillators use Rb or high grade double oven crystal oscillators that run with long integrating times. They offer hams tracability of frequency and very low noise for very little cost, a second hand Trimble Thunderbolt or a Rapco Rb cost £100-300 and deliver a very clean signal.

There is the very low cost route of GPS Rx and huf and puf stabilisation of a simple temp controlled oscillator. For the very low cost this does give some degree of frequency certainty within a few Hz. However counting the oscillator using the 1pps as a gate produces much lower stability as the crystal hunts around the mean fx. To get closer to the quality of the GPS signal a high grade crystal oscillator and integrated phase measurements are needed but that provides a reference ideal for LF and MF signalling at very narrow bandwidths.

73's
Alan G8LCO

__._,_.___

---

Posted by: g8lco10@yahoo.co.uk

---

Reply via web post

•

Reply to sender

•

Reply to group

•

Start a new topic

•

Messages in this topic (2)

Visit Your Group

• New Members 3

• Privacy • Unsubscribe • Terms of Use

.

__,_._,___