[time-nuts] Enhancing the TIC232 code for better data

[Richard H McCorkle]

The reason the Fairchild HC4046 doesn't work is it doesn't have 
a S/R F/F output on pin 15 and the other devices do. 
What I found in testing the different HC4046's with an MTI-260 
(very low phase noise oscillator) GPSDO controller I designed  
was the Phillips part introduced the least phase noise in the 
phase plots and the TI HC4046 introduced more. (I also tried 
a discrete HC74 phase detector and it was even worse!) I wasn't 
so worried about the delay in the phase detector as I was in the 
change of delay over temperature and the short-term noise 
introduced. The Phillips HC4046 performed best in both these 
areas, so now  I use it as first choise for low phase noise designs. 
I'm sure there are more modern devices out today, but I tend to 
go with devices I have experience with.

The additional gating on the S/R F/F insures the proper start 
and stop action so you can't get a phase reversal. They insure 
you can't input a start if not already stopped, and you can't 
input a stop without a start first. 

One variant of the averaging design that I am using to control 
an LPROII Rb oscillator uses the 4046 to gate a 50 MHz 
clock into TMR0 on a 16F873A using an HC00 external 
to the PIC to gate the clock. I average 100 seconds of 50 
MHz phase samples to determine how much to change the 
EFC to the LPRO. Reading the prescaler is a pain, but can 
be done (see AN592) and having the 50M clock gives 
excellent resolution and makes the work to read the 
prescaler worth it in controlling an Rb.

Once you start using Allan deviation software to determine 
stability you start to see the stability over different time
intervals. If you use a short loop time the short term 
stability becomes worse and the long term stability 
improves, (You are always in tight phase lock, but you are
constantly changing things) Systems designed for time
keeping tend to have short loop times to maintain good 
long term stability, but generally have a noisier output.
Using a longer loop time can improve short-term stability, 
but long-term stability suffers. Systems designed as a
low phase noise frequency sources use long loop times 
to minimize the short term changes and tend to have less
noise in the output, but are not as good for timekeeping.
(The phase wanders around more but the short-term 
changes aren't as much.) The best overall stability is 
achieved when you can get the chart as close to horizontal 
as possible. (You make your changes just in time and just
enough to keep the phase from wandering too far.) 
The AD chart can also show temperature effects 
graphically as they happen over 1/2 day intervals, so data 
at about 43.2K seconds averaging time shows off the daily 
temperature effects in the data. Once you start using the 
AD chart you begin to see when a loop time is set short 
(high at start of graph and typical for timekeeping) or set 
long (high at end of graph and typical for low phase noise 
frequency sources) and when you have the loop time 
between the two at best overall stability. (about horizontal)  

Enjoy!
Richard

----- Original Message ----- 
From: "Brooke Clarke" <brooke at pacific.net>
To: "Richard H McCorkle" <mccorkle at ptialaska.net>
Sent: Sunday, August 13, 2006 9:02 AM
Subject: Re: Enhancing the TIC232 code for better data


> Hi Richard:
> 
> I really like your idea of averaging readings to improve resolution but 
> there's a couple of things about the TIFC232 that have been bothering 
> me.  Maybe you can shed some light on them.
> 
> (1) You mention that the brand of 4046 matters and even that there's a 
> pecking order among 3 brands.  Since the 4046 is only being used as a 
> phase detector, very much like if not exactly like a set-reset 
> flip-flop, I'm trying to understand what it is about the different 
> brands or what specification would cause a difference in results.  Have 
> you done any testing of the 4046 phase detector parameters on a stand 
> alone basis?  Do you have a scope that can make sub nanosecond 
> propagation time measurements?  The PC3 output of the TI 4046 has "and" 
> gates on the Set and on the Reset inputs but I don't understand why. 
> 
> The propagation delay for the TI 74HC4046 PC3 output is specified at 49 
> ns at 4.5 Volts (42 ns @ 6 V).  Note that this changes to 74 ns over 
> temperature.  Do you have data sheets for the other brands? 
> 
> (2) You mention that it's better to zero Timer 1 after each measurement 
> rather than let the timer act as an accumulator.  Although I have 
> downloaded both the German and AVAR software packages, so far haven't 
> set them up and learned how to use them.  I'm beginning to get a feel 
> for the variations in time domain measurements but am far from being 
> able to actually use Allan plots.  If you posted the plots there may be 
> someone on the Time Nuts mailing list that could help.
> 
> Thinking about the above it occurs to me that it might be better to have 
> the phase detector and the gate for the 16 Mhz oscillator all be 
> external to the PIC.  Maybe the phase detector and the gate need to have 
> propagation times that are fast compared to a nano second?
> 
> Have Fun,
> 
> Brooke