Al
I like the truetime products. In general easy to understand and last a long
time.
But there never seemed to be that many. Sure they were used in broadcasting
and maybe power. But the others like the 3801 and tbolt were used in telco
and mobile apps so there were 10,000s turned out and thats why we get them
for cheap. I simply never see the truetime dc60 or gps units around. Though
I have my stock of dc468 sat clocks. :-) Working. I hacked a goes sat
replacement 3-4 years ago.
That said some of the older gps technology is a bit slippery on exactly how
good they are.
So for perhaps amateur purposes they are totally fine but when you start
comparing to a Tbolt or 3801 various behaviors apear.
Odetics GPStars as an example slip cycles on purpose. Its a mode you can
set and by default is how they are set.
For what they were intended for they are perfect. But at least 1 X10 poorer
then other devices. Its not at all broken. It was a general time piece for
radio networks. Give or take 500 ms.
Regards
Paul
WB8TSL

On Tue, Dec 4, 2012 at 9:58 AM, Al Wolfe alw.k9si@gmail.com wrote:

Hi

I would guess that HP/Agilent/Symmetricom and Trimble made 100X more GPSDO's than the next five people in the business combined over the 1995 to 2005 period.

Bob

On Dec 4, 2012, at 10:26 AM, paul swed paulswedb@gmail.com wrote:

With the price of T-Bolts now higher, does it make sense to build your own
GPSDO?

What is the simplest phase detecter that could work?  I think only that,
and then a duouble oven crystal from eBay, a GPS and and Arduido.

Yes the Aruino is expensive compared to a bare uP chip but using one, I
thin you could build a GPSDO without a PCB and the Arduino's USB connection
could be usful for power and logging/control.

If ther phase detector where simple enough it could be build on a prototype
board the fits on top of the Arduino.

There are some other designs but because programming a uP and making a PCB
seem to be rare skills that job tends to fall on one person.  Anyone can
program an Arduino and with out need of a PCB the entire design could be
puted on a web page and the replicated with common parts.

On Tue, Dec 4, 2012 at 4:01 PM, Bob Camp lists@rtty.us wrote:

--

Chris Albertson
Redondo Beach, California

Hello, I've written few messages on this mailing
list, I'm an absolute beginner on timing science.

I've never introduced myself, so this is a sort
of introduction...

Although this has nothing to do with
arduino, reading the post by Chris Albertson
about an arduino-based GPSDO, I will share
what I'm thinking to do.

Now I'm slowly documenting myself, the
argument is so vast and over my head that
I need time to start understanding, the
luck is that this mailing list (their memebers)
is... powerful :)

I'm seriously thinking to attempt a gpsdo.
It's mainly to learn something new.
For some reason I collected some Rb oscillators,
and I'd like to have a 10MHz absolute reference,
so I will try to discipline one of the Rb, and
later maybe an OCXO.

The project will proceed slowly and there is
some probability (small, but not null) that
it will be abandoned, because of time problems
of the author (could be a paradox?).

The platform I will try to use is the STM32F103
microcontroller, the reasons are:

• Mainly I have some devboards I used to develop
  some hw in the past, that now sit unused,
  and since I've used it in the past I have some
  experiece with it.
• It contains extensive timing hw, it can measure
  external pulses with internal 72MHz Ck as timebase,
  downside is the 72MHz is internally generated by a PLL
  from a lower frequency.
• It has two 12bit DAC, and external reference
  options.
• It has good and fast (12bit 1uS) ADC useful if one would
  try to build a time to amplitude interpolator (is
  that name right?).
• DMA facility to collect measurements without
  messing with nested interrupts.
• At least two RS232 interfaces, easy to interface
  to a PC.

For a beginner probably the learning curve is steeper than
an arduino, but I never used arduino, so cannot compare.
But for who is not accustomed to program an arm
micro, I think it will not be too difficult to try:

• devboards are CHEAP (ST and others seem to sell
  them at less than the IC cost alone)
• Programming the flash is easy, the uc contains a
  bootloader, you only need an usb-to serial adapter
• Once developed a working example, one can mod
  the progam easily, it will be written in C language.
• There are free and working toolchains for these
  devices.

For now my plans are rather nebulous, but roughly:

• I will start trying to check the jitter of the GPS I
  will use (hope I receive it soon), with a counter and a PC.
• I will start building a time interval counter with
  the stm32, that will use an external 10MHz as timebase
  to measure the PPS of the gps, this will make me
  getting familiar with internal micro timers, they have
  some million possible configurations.
  I will try to use the internal PLL clock as a
  digital interpolator, to try to reach better than 20nS
  resolution.
• start to build some form of simple disciplining...
  dont know yet how.

Fabio.

Chris Albertson albertson.chris@gmail.com ha scritto:

This message was sent using IMP, the Internet Messaging Program.

On 12/05/2012 08:03 AM, Fabio Eboli wrote:

Coincidentally, my previous time-nut project was built around the same
chip. I built a simple GPSDO using a STM32F103C with a bit of support
circuitry, using the timer in "input capture" mode to timestamp pulses
and act as a coarse time-to-digital converter. I got a simple PLL
control algorithm working but haven't yet refined it so it tracks rather
poorly. My intent was to adopt some of the self-tuning attributes of
NTPns, which I will likely revisit for the next project.

Some more details about what was on the board:

• A NC7WZ14 CMOS inverter to square up the sine wave from the OCXO,
  which then feeds...
• A PIC12F1501 as a programmable divider, using TVB's picDIV code
  lightly modified to work on that particular chip
• The STM32F103 itself, which compares pulses from the divider to pulses
  from the GPS receiver and makes adjustments via...
• A slow 16-bit DAC constructed from a PWM output on the STM32, a
  two-pole RC filter, a buffer op-amp, and a third RC pole. This drives
  the OCXO's frequency control. The PWM is also tweaked over 16
  consecutive periods to add 4 more bits of precision, a sort of crude
  pulse-density modulation.
• There's also an op-amp to buffer the 10MHz sine wave for 50 ohm
  output, and a digital buffer for a 50 ohm PPS output from the divider

Here are the design documents, if you're curious:
http://hg.partiallystapled.com/circuits/serafine/raw-file/d75ab09ca163/out/production.PDF

The precise parts of course are not important, it's just an example of
things I chose to get the job done. The general shape of it is the same
as many, if not all, other GPSDOs out there. I'm reasonably happy with
the hardware as a GPSDO experimentation platform (but not looking to
sell anything at this time).

The current project, as I've mentioned before, is a self-contained
GPS-to-NTP server based on STM32F107, which has built-in ethernet but is
otherwise very similar to the F103. The finished board won't be nearly
precise enough to compete with a "real" GPSDO as it is based on a small
on-board VCTCXO but should shore up the algorithms enough for me to
revisit the GPSDO again.

-- m. tharp

Hello Michael!

Michael Tharp gxti@partiallystapled.com ha scritto:

...

...

Thank you very much, I will study it with interest,
it will be very helpul to see what you have done.
Can I ask you more details? I didnt's understand
how you are using the timers: are you timestamping
each pps transistion using the internal clock?
Are you using the pll to obtain 72MHz (x9) for the clock?

Good, to be clear my project will not be commercial
in any way, only an amateur attempt, documented as far
as my time permits. And my skills are not enough for
anything that can be sold in this field :)

This is another advantage of the STM32 (or other manufacturers
cortex arm micros), one can easily port the project up to higher
specs devices; for example ST sells the stm32F4 discovery for
around 20eu that is powered by an stm32F407 that is cortex M4F
device full of ram and flash and communication ports, and that
support FPU etc, it's a 168MHz device, but I'havent checked
it's timers capability.

I will stick with the 103 for now, my goals are the basic ones:
building a counter and figuring out how to discipline a Rb,
the communication/logging/pc support software is really a big
work to do and i'm not planning to go there for some time :)

Fabio.

This message was sent using IMP, the Internet Messaging Program.

On 12/6/2012 4:26 AM, Fabio Eboli wrote:

Yes, the crystal oscillator is multiplied up to 72MHz which then drives
the timer. Even though the particular timer peripheral I chose happens
to be on the APB1 bus which is restricted to 36MHz, the timer itself is
still fed with the 72MHz clock. Both PPS signals (generated from OCXO
and received from GPS) are then independently timestamped. The
timestamps are extended to 64 bits by adding the value captured from the
IC to an "epoch" variable that is incremented every time the timer
itself rolls over. This works fairly well but my implementation is
slightly buggy, occasionally the timestamps will be off by an epoch
(plus or minus 65536 ticks) but such a large deviation is easily
detectable and is discarded. The timestamps are subtracted to get phase
difference which is then fed into the proportional-integral controller
which seeks to zero the phase difference, with two different "speeds"
for early startup and later settling once the oscillations dampen. This
last part is the bit that needs major work since the phase difference
continues oscillating by up to 5 "ticks" (72MHz periods) and sometimes
has excursions to 10 or 15 before it settles down again. NTPns seems to
be self-tuning which could help a great deal. The coefficients I'm using
are experimentally determined which is probably why the settling isn't
very good. There's also the problem of not currently having a TIC or
similar equipment for quantifying the performance of the system as a
whole, I should buy or build one sooner rather than later.

Michael Tharp gxti@partiallystapled.com ha scritto:

Ok, now I understand.
How is faring the pll clock on 1s intervals? The datasheet reports
300pS jitter, that should be good.
My plan for now is this:
(sorry if my terms are not exact), i will start with a Rb instead
of an OCXO. I want to try to measure the ratio between the GPS and
the Rb, to do so I will use the Rb 10MHz clock a timebase for the pps.
The first counter will be setup as input capture with external clock,
the clock is the Rb 10MHz and the captured will be the rising edges
of the GPS.
This will give a count of around 10^7. Then I will use a second
counter with internal clock that will start on the pps edge and stop
on the Rb edge, the result will go from zero to 100nS/13.9ns=7
This count will give some interpolation.
As far as I understood all this could be done in the internal hw, but
it remains to be seen if there are problems with the actual hardware
(bugs, silicon errata etc).
The external hw for now will be a normal XO for the micro, and some
logic to square the inputs, perhaps I will setup also a FF
to have a pulse that starts with the PPS and stops with the
first 10MHz edge, this will be useful for testing with the
real counter and to eventually implement an analogue TAC.

Just now I was simulating on ltspice some TAC interpolators,
if the previous system will go I'm thinking to add an analog
interpolator that will convert the 0 to 100nS edge difference
between the Rb clock and the pps in analog value, to increment
the single shot resolution. I was also thinking that for the
disciplining I dont need to have the interpolator giving the
exact time interval, but I need it only be sensitive and stable
enough.

The control loop is entirely another problem :)

Ok now I need to start burning some flash to check if
the thing can go somewhere.

Fabio.

This message was sent using IMP, the Internet Messaging Program.

Hi

I believe your second counter (the one that starts on one edge and stops on the other) will have trouble. It is unlikely you will get your 14 ns resolution out of it.

The analog TDC is a fine idea. Cheap, easy, and requires a few parts that are not built into the micro controller. The simple answer is to use a cheap CPLD (possibly sub $2) for a bunch of this stuff and move on.

Bob

On Dec 7, 2012, at 4:26 AM, Fabio Eboli fabioeb@quipo.it wrote: