[time-nuts] 30 MHz freq adjust by Hz

[glen english LIST]

Hello group

I have an idea that might work, and I wanted to discuss  with likeminded 
that might already have experience with the problem.

Shifting a fixed oscillator a few Hz using a image reject mixer.

background : From time to time I (and others) make lock boards for ham 
gear, pulling the internal VCXO (vary from 11 to 55 MHz ish - which are 
out by a few Hz ) in against a 10 MHz input. Frequency accuracy is 
required for narrow band modes, and low phase noise 10kHz-200kHz is 
required as not to desense your ham neighbours.

I use fast LVDS diff receivers to square stuff up and ADF4157 high res 
fract and about 10Hz BW. That's all fine.  ....That aside, there are a 
bunch of radio that have only XOs, no control facility. Varying the 
supply voltage as a means of control is one way, but that doesnt work 
for the ovenized nor internal regulator types.

Other people to solve this problem by applying an external oscillator 
that has been disclipined - usually like crappy SiLabs spury synthesiser 
chips that can produce the oddball reference frequencies like 
31.28234MHz ! Actually they are pretty good for what they are, but they 
are certainly not as clean close in , and particularly poor in spurs far 
away. they are what they are.

Of course the great way is a DDS , and run something like a 400-700 MHz 
VCO/SAW/BAW clock. One needs to go that high to get decent oscillator Q 
, and of course the DDS needs the high clock. The clock is of course  
pulled to the 10 MHz with something like a ADF4002 etc integer synth 
running high BW like 200 kHz to kill to close in VCO noise.  But that's 
alot of stuff

*** I thought in the shower this morning of inserted a block, and 
shifting the internal radio oscillator (running at say 31.28234MHz the 
few Hertz either side it needs to move.)

- by using an image reject (full complex)  mixer with a +/- 5 Hz 
oscillator applied.

- by using a analog or digital variable delay line to remove or insert 
delay as to strecth or contract the period . almost like a phase 
modulator, but I think it is going to wrap and cause trouble. Hmm if I 
play with the high and low period separately, I might be able to fix it 
when it wraps. But that technique will likely insert noise for any soet 
of small easy implementation.

- alias to close to baseband using another oscillator (fixed) and then 
alias back up. Aliasing technique are  very cheap and useful in DSP. Hmm 
that might be soemthing I do in DSP for other signal processing tricks, 
but not on a small board .

The  cpx mixer is the 1st thought :

Perhaps a commutating HC-CMOS switch quadrature DBM (like HC4316) with 
the complex LO +/- 5 Hz coming from something I can dream up.

For a single frequency, I would be able to get the quadrature matching 
at least -60 over temperature  in my experience with something like this..

That would be the unwanted sideband down that far. Of course with square 
wave drive, the mixer will be sensitive to the harmonic series, but the 
input is squeaky clean, so that's no issue. ALTHOUGH hmm the close in 
noise say + /- 10 Hz would get a say as it would be aliased in, but the 
10Hz noise on those XOs is usually prety good, and for this purpose, it 
is the noise at offsets  of 10kHz up to 300kHz that are the most 
important.  Control bandwidth only has to track thermal drift in the 
radio, have be fractions of a Hz, so the system could spend some time 
calculating and generating the LO.

Anyone tried this (IE shifting the frequency of the source)  ? Comments ?

Glen

(VK1XX, AI6UM)