[time-nuts] PLL question

Magnus Danielson magnus at rubidium.dyndns.org
Mon Aug 10 16:55:43 UTC 2009


Ulrich,

Ulrich Bangert wrote:
> Magnus,
> 
>> Indeed, but too narrow pulses may still be a problem. The 
>> benefit of a continous signal is that the disturbance can be 
>> kept at a high frequency and thus be dampend out and the 
>> channel capacity for corrections is maintained high.
> 
> I have no clue to what you are refering to. Narrow pulses? What narrow
> pulses? Continous signal? What continous signal? The AD9901 circuit uses the
> flip-flops only in case that there is still a frequency error between the
> inputs. Once this frequency error is removed by the loop the circuit works
> as if only the XOR were present. Which makes this circuit exactly an very
> intelligent and versatile steering device: Frequency comparator if necessary
> and XOR phase comparator if possible. 

I can't recall all details about all devices, OK?

I think I saw the datasheet of AD9901 a few years ago, and my attention 
was on other details.

XOR is good for continous phase-tracking if you are able to maintain 
within +/- 90 degrees. Tracking into that condition is another thing. 
Assuming good signal. XOR degrades to multiplication as noise is added, 
with the associated change in phase detector gain. But for such good 
signal/noise ratios as we have in this case, such concerns about it 
should not be excessed.

Otherwise I prefer the S/R FF, which gives a +/- 180 degrees response 
and has the phase/frequency property, so it always tracks in. You can 
build a safe S/R FF from a 74AC00. Very simple and good track-in 
properties. Combine with favorite active loop topology and you are ready 
to go.

>> You can do alot of tricks in a CPLD. Getting the basic plot 
>> of what is good or bad system design still needs to be sorted 
>> out. Not meant as criticism, but just as a kind warning.
> 
> Sure! I had better said: Due to the simple design and the very good
> performance of this phase/frequency comparator I use it as a standard in all
> my PLL designs. Since I usually do not need the high frequency capabilities
> of the original AD9901 but need only performance up to a few MHz I do not
> buy the original part from AD but have covered the circuit diagram into a
> number of cplds. These have proved to work so well in a number of PLLs for
> different purposes that I can recommend you to do the same from my very own
> experience. 

Ah, now that is a good point.

Cheers,
Magnus



More information about the time-nuts mailing list