[time-nuts] Generating a stable 26MHz and 19.2MHz from 10MHz

[Magnus Danielson]

On 07/01/2011 01:33 PM, Anders Time wrote:
> I want go generate a stable 26MHz and 19.2MHz signal with stability in
> the 1E-13 range at 1s from my 10MHz BVA.
> Have been thinking about dividing the signals down to 1MHz and 100kHz
> and PLL (50Hz BW or so) a reasonable stable 26MHz and 19.2MHz OCXO to
> that signal. But I'm worried that the division will generate noise so
> it will not be stable enough?
> Have also been thinking about using a DDS to the 26MHz and 19.2MHz and
> then pll, but the DDS subject seems to bee very difficult to get a
> grip on. Is it possible to generate a signal with that kind of
> stability in the 1s range?

There are many routes to go through.

For the PLL route you should notice that highest common frequency 
between 10 MHz and 26 MHz is 2 MHz, requiring a division of 5 and 13.

A PLL with such ratios becomes fairly simple to achieve lock and 
suitable performance. A 26 MHz with low noise should be selected. A 
PI-integrator (1 op-amp, 2 resistors and a capacitor) setup is recommended.

For the 19,2 MHz you have 400 kHz as common frequency, which would be 
another divide by 5 down from the 2 MHz. A divide by 48 is needed from 
the 19,2 MHz side. Similarly a low noise 19,2 MHz should be selected and 
a PI loop filter be selected. It should still be fairly easy to get it 
to lock properly.

Another approach would be a mixer based approach. For instance will 
synchronous dividers be able to provide an interesting solution for 10 
MHz to 19,6 MHz by recognizing that 19,6 MHz and 400 kHz both is 9,6 MHz 
away from the input, so by setting it up for a 1/25 division you also 
achieve the 19,6 MHz directly, with very good phase noise properties.

Achieving the same thing to 26 MHz becomes tricky as the synchronous 
divider techniques outputs m/n and 2-m/n the frequency of the input 
frequency. Inserting a frequency doubler provides 20 MHz and then 14 MHz 
and 26 MHz can be generated.

As for DDS it is a bit tricky to give any clear direction, it's like 
saying you should use a PLL or pen. Best result is if the DDS have a the 
DDS beat period matching the highest common frequency as presented 
above. That way annoying spurious signals can be avoided.

Cheers,
Magnus