[time-nuts] PICPET- was Affordable (cheap) COTS (etc)

[Brian Lloyd]

On Thu, Jan 23, 2014 at 2:00 PM, Tom Van Baak <tvb at leapsecond.com> wrote:

> > Has anyone else looked at the Parallax Propeller processor for timing
> > functions?
>
> Hi Brian,
>
> Oh yes. Really nice chip. But for precise timing applications I had huge
> problems with phase and temperature stability of its internal PLL. I tried
> half a dozen different boards purchased over several years. Tech support
> was not interested in someone who worried about nanoseconds.
>

Well, my disciplining code is going to run as an FLL rather than a PLL to
generate the correction for the OCXO or the Rb reference. I didn't think
that noise on the PLL multiplier to take the clock from 10MHz (using the
OCXO or Rb as the processor clock) to 80MHz would hurt when I am
accumulating clock cycles over many 10's of seconds.

OTOH, one can clock the processor directly from the reference without the
processor clock PLL and run the CPU at 10MHz. The Propeller is static CMOS
and will run at any clock speed down to DC. They spec the clock input up to
8MHz but say that it will clock just fine at 10MHz. The PLL will operate at
powers of 2 up to 2^4 (16). Most people use a 5MHz crystal and the x16 PLL
to clock the processors at 80MHz. I have been told that the Propeller will
run at 100MHz just fine.


> The architecture is really interesting, but it is such an odd chip, with
> almost zero market visibility these days, that I set aside the goal of
> using it as the basis of a general purpose 8-channel 6 ns precision
> counter. You can find various timer and counter examples at
> obex.parallax.com. If you make progress on the project, please let me
> know, ok?
>

Wilco. There are some nice things out there using the propeller.

By contrast, the PIC chips I use are fully synchronous so when you use 10
> MHz atomic references the clock/output jitter and phase stability is almost
> below what I can measure here.


Same with the Propeller. You can clock it directly from the 10MHz reference
without using the PLL and then all 8 cores are running synchronously. (They
do anyway but usually they are running at 80MHz. I need to try to see just
how fast I can run the main clock input.) I just thought that, while the
PLL would have a bit more jitter, having the extra 3 bits of resolution
would be useful in accumulating the error.


> Maybe under 2 ps. So that's why I use PIC's as the basis of all my picDIV
> and picPET projects. But I'm open to using something different in the
> future.
>

It does seem to be a very interesting processor and seems to me ideal for
performing timing functions. One can easily use it to generate pulses at
integer divisors of the system clock. At that point one would definitely
not want the jitter from the PLL.

-- 
Brian Lloyd, WB6RQN/J79BPL
706 Flightline Drive
Spring Branch, TX 78070
brian at lloyd.com
+1.916.877.5067