[hpsdr] Trade-offs in loop filter design

[Chris Bartram]

Bob 

> The other side of this is accuracy, as in frequency accuracy.  The nice
> thing about the 2nd order PLL circuits is they can help achieve accuracy
> and phase stability.  But outside of the "noise bandwidth" of the phase
> locked loop filter,  you have approximately the same noise as before but
> now it is centered on frequency!   Increasing the bandwidth of the PLL
> filter, will allow more wandering in the frequency that we desire.  So
> it is clear there is a trade off.
>
> The trade should probably occur "at the spot" of the cross over in terms
> of crystal phase stability and locked longer term stability (which will
> translate into frequency accuracy) in the typical alan deviation plots
> of crystal oscillators and free running results from a PLL based
> oscillator. This will be a neat little analysis to do and the design
> should be interesting to achieve it.

It probably says a lot about my inability to express myself adequately using 
algebra, but I've always found it simpler to do the _final_ optimisation of 
PLL loop filters empirically. There's a complex trade-off between controlled 
oscillator noise and reference noise characteristics which is mediated by the 
loop. 

The old texts about PLLs used to talk about a 'phase noise pedestal' which was 
really a statement which said 'OK we can clean-up the carrier so far, but 
when we run out of loop gain, the pesky VCO will still dominate the far-out 
noise'. That's not something we can do much about unless we concentrate hard 
on cleaning up the VCO. 

The work I've recently been doing for uWSDR has been to optimise the VCXO we 
intend to use as a phase-locked clock in LO DDSes. By using a good VHF VCXO 
we can get very close to - or just conceivably better - the performance of 
the reference at VHF, obviously allowing for the degradation caused by 
frequency multiplication.

A similar set of trade-offs occurs when we use a second loop to multiply the 
VHF oscillator to the microwave region... Ulrich's comments about VCOs (in 
particular multiple resonator VCOs - I've read the papers..!) are understood 
and appreciated! 

In all cases, a lot of attention needs to be given to the phase detector and 
to the way in which the frequency translation is achieved. Sampling phase 
detectors are a better solution than phase/frequency detectors and frequency 
dividers.

Vy 73

Chris
GW4DGU