On Wed, 23 Apr 2014 15:07:52 +0200
Antonio P. [email protected] wrote:
Using an external 10 MHz reference clock, a square wave will offer the
best phase noise performance, but a sinusoid is acceptable.
The difference between the phase noise of a square wave and a sinus
input is negligible in a radio application. Your system will be
by the noise of your input. Unless you are doing detection with very
long integration time (in the order of seconds and longer). And even
there i’m not sure whether it’s not still the input noise that’s
the performance instead of your reference clock noise. And in that case
you care more about close in phase noise which is dominated by
mixing of low frequency noise sources into your signal, than what we
commonly refere to as jitter.
To get that down you want to:
use sinusoidal input and not square wave
Square wave has high frequency components, depending on the squaring
buffer you will have considerable spectral components between 100MHz
and 1GHz… or even more. Cables have worse and worse characteristics
the higher you go. Also small impedance mismatches will lead to
more severe reflections the higer the frequency is. Also as the
temperature changes, you will have changing phase shifts for
frequency components (cables are non-linear in that regard).
Change the clock multiplexer (SY89545) from a LVDS->LVDS type to a
pure analog part. Or rather, remove it completely and go directly
onto the AD9510
But that’s only the case if and only if you care about long integration
times and the 10th decimal point of performance.
For all else. Using a good quartz oscillator is more than good enough
and you dont need to care about anything but having approximately the
right signal levels.
The trouble with you, Shev, is you don’t say anything until you’ve saved
up a whole truckload of damned heavy brick arguments and then you dump
them all out and never look at the bleeding body mangled beneath the
– Tirin, The Dispossessed, U. Le Guin