[time-nuts] Frequency division by 81

[Bob kb8tq]

Hi

The gotcha is that the “HC” logic families have pretty high broadband noise floors compared to newer parts. 
At 20 log (N), divide by 81 gets you about 38 db of phase noise improvement. If you started off at -130 dbc/Hz, 
you come out at -168 dbc / Hz. That (of course) assumes you don’t hit this or that floor first. 

Agin, with limited information from the OP, it’s tough to work out just what the limits would actually be. ( what is the
starting frequency?, what phase noise offsets mater? …..)

Bob

> On Jun 18, 2020, at 3:36 PM, ed breya <eb at telight.com> wrote:
> 
> For a 10 MHz clock, 74HC would be fine. For small numbers like 81, a couple of 74HC163s would do it, and be good to go since they're synchronous anyway.
> 
> For large numbers, my go-to divider is the 74HC4040 12-bit ripple counter. It can be rigged for any fixed integer divide ratio from 3 to 4095 with a simple diode gating arrangement for feedback to the reset input. For divide by 81, three diodes (64+16+1) and a resistor comprise the feedback circuit. The bigger and more numerically complicated the divide ratio, the more diodes are needed. The straight binary 2^n ones are of course trivial. The output edge may be re-synchronized with DFFs, as Gerhard mentioned. This is advisable especially for precision time and frequency work. Also, if you use a dual FF anyway, one can be used for re-synchronizing, and the other can be used in the feedback, to improve the reset action. The reset is asynchronous, so there is a risk of race around from the outputs to the reset, but the prop delays tend to eliminate it, and I've never had a problem with it.
> 
> The 74HC4020 may also be used, depending on the divide range needed - some of its binary stage outputs aren't available, while the 4040 has all of them. The 74HC393 also can work this way, for up to 255. I've used these counters many times this way, but can't remember the exact hookup, without going back to notes on some of my projects, or figuring it out again from the datasheets. It's quite simple, but if anyone needs, I can revisit and explain it.
> 
> Ed
> 
> _______________________________________________
> time-nuts mailing list -- time-nuts at lists.febo.com
> To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
> and follow the instructions there.