[time-nuts] 10 MHz -> 16 MHz clock multiplier

[Bill Fuqua]

    If you divide by 5 the phase noise sideband amplitude (voltage) will be 
divided by 5.
That is a reduction of 14db for all phase noise sideband frequencies . Then 
when you multiply that by 8 you will add the phase noise
sidebands will be multiplied by 8 or 3x6db or 18 db. So the end result will 
be a
factor (if I did my math right in my head, which is getting more difficult 
these days) 4db increase.
   The crystal filter will reduce the phase noise sidebands to some degree. 
That is to say
if you had a perfect filter you could pass only the carrier with out the 
sidebands and thus
no phase noise. However, if your filter has 1kHz bandwidth you would only 
eliminate the
sidebands beyond 500 Hz on either side of the carrier. So the answer is 
sort of yes.
   When I read the email about multiply by 16 and divide  by 10 it occurred 
to me that
it would be easier to divide and then multiply and then I began to brain 
storm, which
is hard to do when you stay up too late.
   If you chose to use a crystal ladder you need to use 16 MHz parallel 
resonant
crystals since the series resonance will be slightly less than 16 MHz. These
crystals are rather cheap. If you want to use a simple high Q (narrow 
bandwidth)
phasing type filter you need to use a crystal with a 16MHz series resonance and
use a termination resistance greater than the crystals series resistance. 
You can
adjust the crystal filters bandwidth by changing the termination 
resistance. This type of
filter was mostly  used in early vacuum tube receivers. They usually shot 
for a minimum
bandwidth of 500 Hz or so. you need to adjust the "phasing capacitor" so it 
equals the
crystals parallel capacitance to minimize feed thru.
   Experimenting with these filters is a lot of fun. I have made lots of 
crystal filters.
I even have a digital crystal impedance meter so I can compare crystals. If
reducing phase sidebands in not a goal all you need is a filter that will 
eliminate
all the other 2 MHz harmonics.

73
Bill wa4lav


At 05:37 PM 1/4/2013 +0000, time-nuts-request at febo.com wrote:
>THis is exactly what they are talking about the 74HC390 can do over
>50MHz and costs abut 30 cents.  You don't need ECL or anything so
>exotic the 30 cent part will work.   Set it for divide by 5.  I guess
>this is imperfect enough that there is some fourth harmonic content in
>the 2MHz square wave, then you select that with a narrow band filter
>and amplify it to whatever you need.   A smart design might try and
>add fourth harmonics be using a slightly not-symetric 2MHz square wave
>
>My question is about the phase noise of the final 16MHz signal.  Do
>crystal filters "clean up" the signal.  It seems that after several
>16MHz crystals in series the output should look a lot like an XO.