[time-nuts] 1PPS to 32.768 khz

Bob Camp kb8tq at n1k.org
Thu Oct 20 07:08:31 EDT 2016


One problem with a PLL and a 1 Hz input are the values of components
you get in the loop. The other issue is the cost of the VCXO that will get 
you to 32,768 KHz. The PLL as described by the OP would need the 1 pps
divided by 2 with a lot of PLL chips. You now are locking 32768 to 0.5 Hz. 

It gets complicated pretty fast.


> On Oct 19, 2016, at 11:07 PM, David <davidwhess at gmail.com> wrote:
> I was thinking 32.768kHz VCXO and phase detector to make a simple
> analog PLL.  I found a datasheet for a suitable VCXO and assuming a
> total error of 20ppm, it would only need to be divided by 2 to prevent
> locking to the wrong frequency making an analog PLL pretty simple.
> Safer to divide by 4 or 16 of course.
> But I wonder if a microcontroller using a 32.768kHz external clock
> could be the phase detector itself.  Strobe the microcontroller ADCS
> from the 1 PPS to sample the sine 32.768kHz clock, simmer, and serve.
> Or just count cycles to make a FLL.
> The most annoying thing about using a varactor for tuning is that they
> are not amendable to low drive voltages.
> On Wed, 19 Oct 2016 18:14:32 -0400, you wrote:
>> Hi
>> As has been already mentioned, a lot depends on what you have. The drop dead cheapest way to do it:
>> Start with an MCU with an internal oscillator. There are *lots* to pick between. Which sort really does not matter. 
>> For example, I’ll use one that starts at 4 MHz. 
>> Divide the 4 MHz down to 32,768, or as close as you can get. The common clocks aren’t going to divide
>> straight to 32768 so you will need to do two divides. One will be a bit fast, the other a bit slow. You flip between the 
>> two in a fixed pattern to get the result to average out correctly. 
>> Next take the same 4 MHz and run a counter /timer off of it. Capture the pps edge with the timer. It will drift
>> a bit since the clock in the MCU is not perfect. Based on the drift, modify your dividers to correct the outcome. 
>> If you are “good” to 1 part in 32,768 in each second, that’s close enough for a wall clock. You will have no net
>> error long term if you do it right. 
>> That’s all a sub $1 solution
> .
>> Bob
> _______________________________________________
> time-nuts mailing list -- time-nuts at febo.com
> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.

More information about the time-nuts mailing list