[time-nuts] Allan Deviation -> continuing saga...

[Tim Shoppa]

John Ackermann N8UR <jra at febo.com> wrote:
> Didier Juges said the following on 10/28/2006 12:18 AM:
>
> > The clock shaper uses a commercial 10.7 MHz IF transformer, readjusted 
> > down to 10 MHz (see schematic at 
> > http://www.ko4bb.com/ham_radio/Manuals/2_GPS_Stuff/Clock_Shaper.jpg).
> > It provides light loading to the OCXO through the 390 ohm resistor, yet 
> > provides about 2V p-p to the comparator, to get it out of the noise. I 
> > do not know the temperature stability of the transformer, or the loaded 
> > Q, I will check that later. I realize this is probably not an optimum 
> > design, but it's what I could do with what I have on hand. I am not 
> > using the fine tooth comb and tweezers yet, I am still with the rake and 
> > the fork lift :-)
>
> To minimize jitter and tempco, you probably ought to stay away from a
> comparator if you can.  I've had good success using the input circuit
> from Brooks Shera's GPSDO which is a 74HC4046 PLL, but using only the
> input circuit and not the PLL section itself.  Alternatively, you might
> get away with a simple capacitor followed by 50/50 voltage divider to
> provide 1/2 Vcc bias going to the input of the divider chip.

The 74HC4046 input circuit is really nice for dealing with a
wide range of levels up to the 20MHz range. Others here will
pooh-pooh the phase comparators on the HC4046 but 
they're just SO convenient!

But by far my favorite as of late are differential
line receiver chips. Different families have different thresholds
and hysteresis points, the ones I use most commonly are RS-423 (for
slowish large-swing signals) and the LV diff SCSI chips (almost
only available in surface-mount but good for fast not-so-big-swing
levels.) And a 74(whatever logic family but usually HC as of late)14
will do great too if you have a few volts P-P.

I would be very reluctant about feeding a 10MHz sine wave into a
high-speed (especially 74F series) counter without conditioning.
It's not jitter but double-counting etc. that will get you there
due to the slow rise time. Even nice sharp LSTTL pulses into 74F
counters are sometimes "too slow", and a 10MHz sine will by positively
glacial in comparison.

Tim.