[time-nuts] PICTIC II questions

[Richard H McCorkle]

Fellow Time Nuts,
The PICTIC II is a spin-off of a low cost front end developed for a
GPSDO to compare the oscillator to GPS 1PPS over very short (< 20us)
time intervals. In that application all that was required was 1ns
TIC resolution to match the resolution of the GPS sawtooth correction
to get the best accuracy possible from the GPS receiver. While the
interpolator can provide gains in the 400-800 range the data span
with a 10 MHz timebase is typically reduced to 100 counts so
variations up to +/- 7 counts in the interpolation hardware over
temperature have minimal effect on the 1ns data accuracy.
   The interpolator hardware was much better than required for that
application so the PICTIC II was developed as a learning platform to
get others playing with interpolation and improving the design, not
as a piece of precision test equipment. If you want to measure the
long-term performance of a GPSDO against GPS 1PPS and not tie up your
SR620 for months doing so then the PICTIC II is just the ticket. If
you think you can measure long intervals with an XO timebase or
achieve the 25ps accuracy of a $5000 SR620 using a $50 PICTIC II
then you are being totally unrealistic.
   Other factors besides resolution need to be considered when deciding
how to populate the PICTIC II depending on how it will be used. An XO
timebase may be fine measuring intervals of a few microseconds as
originally intended for a GPS monitor, but as the measurement interval
increases so do the errors from the timebase. A high stability timebase
is required if you want to measure longer intervals accurately and most
of us have a high stability 10 MHz source available. The timebase rate
will determine the digital portion of the count and reducing the
interpolation required for a given resolution will improve the accuracy.
There is a point where the system cost increases faster as the timebase
rate increases and using a high stability 1 GHz timebase for a 1ns
resolution time interval counter is not low cost solution. That is why
most commercial time interval counters use some form of interpolation.
The PICTIC II demonstrates how to achieve 1ns resolution with a 10 MHz
timebase (XO, OCXO, Rb, Cs) while keeping the size and cost to a minimum.
   While using a faster timebase or higher interpolator gain increases
the resolution that doesn’t imply the accuracy will also increase. The
PICTIC II uses CMOS logic with propagation delays that vary with
temperature much more than the ECL logic used in a commercial counter
like the SR620, severely affecting the accuracy below about 250ps. The
interpolator was modeled after the SR620 design but simplified to use
the least amount of hardware possible to reduce the size and cost. As
the timebase rate is increased a smaller cap is used so stray
capacitance and the capacitance of the switching devices have a larger
effect on the charge linearity. The PICTIC II uses software calibration
methods that are not as precise as those in a commercial counter so the
accuracy is not specified other than to say it works well for GPS
monitoring applications at 1ns resolution with a 10 MHz timebase once
set up properly. If you want to log GPS data over months at a time then
a $50 PICTIC II should be sufficient for purpose. But if you want lab
grade accuracy over long time intervals with 25ps resolution then by
all means use a lab grade commercial counter like the SR620 and not a
PICTIC II!

Richard



>> I know the resolution can be enhanced by swapping out some parts, but it's
>> been too long to recall the details.
>
> On the schematic is a list of alternate XOs. For 50 MHz, the 1000 pf
> cap is changed to 180 pf. This is stated to give 25 ps maximum
> resolution. If this simple change is all that is required, I can
> easily substitute the parts to obtain better resolution. Has anyone
> made these changes and used a PICTIC II with a GPSDO and rubidium as
> inputs? Is it good enough?
>
> Joe Gray
> W5JG
>
> _______________________________________________
> 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.
>