[time-nuts] Re: Microcontroller based frequency divider

[Tom Van Baak]

Simon,

You can probably use any microcontroller as a divider, including 
Atmel/AVR. Most modern processors have elaborate timer / counter / PWM / 
NCO features on-chip so you can avoid the archaic cycle counting 
technique used in all the PIC dividers [1].

Or you can use decade counter IC's or even discrete flip-flops if you 
don't mind the parts count. The 'HC390 dual decade counter is a long 
time favorite and you only need 3½ of them to divide 10 MHz down to 1 
Hz. Just about anything will give you sub-ns precision in the division. 
I wouldn't worry about it unless you have a critical project or have the 
specialized equipment to measure it.

The reason I picked the PIC12F675 is that it was a simple and widely 
known chip and has been in production since forever. The concept of the 
"PIC divider" goes back before the Shera GPSDO so there's been 
continuous visibility for 25 years. They are popular because they are so 
small, easy to use, with plenty to choose from [2]. As for jitter, I 
once measured TAPR/TADD-2 PIC divider jitter to be about 1 ps [3].

I see the chips are out of stock from the usual places at the moment. 
Contact me off-list if you can't find one.

/tvb

[1] http://leapsecond.com/pic/

[2] http://leapsecond.com/pic/src/

[3] http://leapsecond.com/pic/jitter/


On 4/30/2022 12:36 AM, Simon Merrett wrote:
> However, with components in short supply these days, I wondered if a similar approach could be taken on an alternate, or a few alternative, part(s). I have a Leo Bodnar GPSDO but only a 50 MHz bandwidth oscilloscope (Rigol DS1054Z), which I cannot think will be sufficient to characterise an alternative implementation.
>
> So I was wondering if there are any time nuts who might have the wherewithal to measure an alternative microcontroller frequency divider implementation. I would think a 10 MHz division to 1PPS. I'd like to know jitter and delay/offset (I'm afraid I'm still not up on all the terminology - feel free to educate me please!) to around 100 ps accuracy (I think I need nanosecond accuracy in my project). I would obviously pay for shipping the parts to be tested and the parts themselves (most likely with power and coaxial signal connectors on a PCB). Please comment, even if you aren't in a position to offer measurement.