At 09:09 AM 7/23/2008 , Max Skop wrote:

the cheap (low cost) LCD clocks that are available with nice big digits,
temperature sensors and calendars, etc.

DDS?

At 10:09 AM 7/23/2008, Max Skop wrote...

I've thought about this too. It seems the simple way would be to clock
a PIC with the 10 MHz, then use loops to produce the 32768 Hz. It's
easier than it seems once you realize that phase noise doesn't really
matter in this application - duty cycles don't need to be exactly 50%,
etc.

The only thing that comes to mind is to divide the 10 MHz by 2500000
(5x5x100000) and then use a phase locked loop to multiply by 8192.  Of
course the PLL would best control a varactor with a clock crystal vice
controlling an RC oscillator.  Other divisor multiplier pairs are:
1250000/4096, 625000/2048, 312500/1024, 156250/512, and 78125/256.

John  WA4WDL

----- Original Message -----
From: "Max Skop" mskop@bigpond.net.au
To: time-nuts@febo.com
Sent: Wednesday, July 23, 2008 10:09 AM
Subject: [time-nuts] How to get 32.768KHz from 10MHz.

At 07:09 AM 7/23/2008, Max Skop wrote:

There's two approaches to your high level problem (driving a
clock)... one is to make 32.768 kHz, the other is to directly drive
the 1pps.  Obviously, dividing down 10M to get to 1Hz is easy.

But, for the other, you don't need to have a perfect symmetrical sine
wave.  All you need to do is make sure that there are 32,768
transitions in a second, so any sort of rough and ready divider
scheme will work.

Now.. in the lab what I did is use a HP3325B to make clocks run on Mars time

See:

http://www.techbriefs.com/content/view/2299/34/

Since the 3325 was locked to the lab's maser, it's probably the most
accurate Mars clock around (bearing in mind that Mars's rotation
isn't nearly as stable as the maser, so it's superfluous accuracy)

You can do other nifty things, by the way... you can make a clock
that displays solar elevation ("sundial time") because the rate can
be changed systematically.

I never did get around to programming a PIC to do some of this stuff,
but it would be quite straightforward.

Jim Lux

Problem is one period of 32768 is not a multiple of 100nS (one period of 10 MHz) so that won't work. Maybe there is a common denominator and it may be possible to generate an average 32768 periods over one second, even though all periods may not be equal.

Didier KO4BB

---- Mike S mikes@flatsurface.com wrote:

At 11:56 AM 7/23/2008, Didier Juges wrote...

As long as both periods are rational numbers, it doesn't matter, and it
can work. For this purpose (display for humans), it doesn't matter if
some seconds have 32768.xxx cycles and some have 32767.yyy, as long as
they average to exactly 32768.  Being off by a whole lot less than the
time it takes an LCD segment to change contrast won't hurt anything.

Max,

You are a mild case....

I also got bitten by the bug:  Now I have 4 HP10811, two rubidiums,
two cesiums,  T-Bolt  and I am seriously  considering  making my own Cesium.

A visit to Las Vegas could turn to be cheaper fun than this......

Predrag

At 16:09 23.7.2008, you wrote:

Divide 10 MHz by 250 to get 40 kHz, to be within range of the clock's
logic, then use a presettable counter loaded by the 1 PPS to count
off 32768 clocks per second.

David McGaw

At 11:30 AM 7/23/2008, you wrote:

I just prototyped this PIC algorithm and it works perfectly:
Exactly 10 MHz in gives exactly 32.768 kHz out.

There are several ways to manage the variable 38 vs. 39
instruction cycle duration of the output waveform.

The crude way, for every 65536 half-cycles, is to generate the
first 9632 using 39 instructions and the remaining 55904 using
38 instructions. This uses a total of (963239) + (5590438) =
2500000 instructions (10 MHz) to generate a 32768 Hz output
clock.

But if you note that 2500000 / 65536 = 38.14697265625 and
think about leap years, there are cute ways to distribute these
"leap cycles" more evenly throughout successive one second
periods, if that is desired.

/tvb

At 04:17 PM 7/23/2008, Tom Van Baak wrote...

TPIWWSC.

(This Post Is Worthless Without Source Code)

:-)

Start:
NOP
NOP
NOP
...
NOP
flip_bit
NOP
NOP
NOP
...
flip_bit
JUMP Start

(details omitted for clarity)

:-)

Didier KO4BB

---- Mike S mikes@flatsurface.com wrote:

At 05:42 PM 7/23/2008, Bruce Griffiths wrote...

You're missing the point. The application is to drive a common, readily
available consumer clock. Simple and cheap. It can be done with a
single $1 PIC. You could spend $20 or $100 and not get better results
for the application. If you can describe a way of doing it for $0.50,
please do.

At 03:01 PM 7/23/2008, Mike S wrote:

But this is time-nuts... Any approach that doesn't have the
performance of a hydrogen maser or cryogenic sapphire resonator just
isn't good enough.  Why, we haven't even started on how to build a
radial ruling engine to make sure the clock face is precisely divided
into 60 segments to ppb accuracy.

Based on the clocks I've taken apart, dividing the 10MHz down to 1 Hz
is probably your best bet, rather than trying to hit 32768. However,
I don't know of a non-programmable single chip solution that will do
a divide by 1E7. If you want programmable chips, there's countless
ways, some more elegant than others.

Mike S,

I think you missed the point !  One does not ask the question "How does
one get a 32.768KHz signal from our 10MHz reference ?" and not expect
it to cost both in terms of money and effort.  There was no statement
of doing it cheap, re-read the original Email.

Bill....WB6BNQ

Mike S wrote:

 At 05:42 PM 7/23/2008, Bruce Griffiths wrote...

 an

 single

 You're missing the point. The application is to drive a common,
 readily
 available consumer clock. Simple and cheap. It can be done with a
 single $1 PIC. You could spend $20 or $100 and not get better
 results
 for the application. If you can describe a way of doing it for
 $0.50,
 please do.

 _______________________________________________
 time-nuts mailing list -- time-nuts@febo.com
 To unsubscribe, go to
 [1]https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
 and follow the instructions there.

References

1. https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts

A more practical offshoot of this concept is to subsample the
32 kHz oscillator at 128 Hz (ie a sampling phase detector) and use a slow
loop to tune the 32768 kHz oscillator.  The biggest problem here is that
you have to have a tunable oscillator.  Attempting to get around
this by injection locking a non-tunable oscillator is probably not
going to work very well, since the frequency tolerance is likely
to be larger than the small amount of frequency pulling you can
get with injection locking.  Basically, if the oscillator tolerance
is 0.01%, and it has a Q of >>10,000, it will be hit or miss depending
on luck.

Rick N6RK

Bruce Griffiths wrote:

Rick Karlquist wrote:

The drawback of all injection locked oscillator schemes is that the
injection locked oscillator needs to be tunable.
However a simple trimmer cap may suffice if the temperature excursion of
the oscillator isnt too large and the occasional readjustment to
compensate for aging is acceptable.

Bruce

Bruce Griffiths wrote:

The major problem with this approach is the extremely small (<<1ppm)
locking range achievable.
Thus it isnt likely to be practical with standard 32768Hz crystals, as
even a relatively small temperature change will perturb the tuning
sufficiently to make locking to the desired frequency impossible.

Bruce

Hi Mike,

Yeah, I thought I provided enough information in that posting so
that anyone that's programmed a uC before, or has seen the
free source code to the 1PPS PIC divider would get the idea.

By the way it's almost the 10th anniversary of that PIC divider:
http://www.leapsecond.com/pages/ppsdiv/

OK, you asked about the prototype for the 32 kHz algorithm. See:
http://www.leapsecond.com/tools/10m32k.c

The basic idea is as I described a few hours ago; with a correct
combination of 38 and 39 cycle instruction code paths, you get
a 32 kHz output that has some jitter in phase but has perfect
frequency accuracy from one second to infinity. This is similar
to the case of the special 100 kHz output of the PPS divider.

The cute algorithm that I prototyped is one that is inspired by the
way we handle leap years. The result is that all the "leap cycles"
are as evenly spaced throughout the second as possible, both
minimizing jitter to the extent possible with a 400 ns PIC cycle
time, and maintaining an absolute division of the 10 MHz to
32.768 kHz, as averaged over any multiple of 1 second.

There are comments in the C code. Note this is not PIC code; it
is simply my proof-of-concept (compiles under windows or unix)
for isochronous PIC assembler code that can (will?) be written.

If I know you PIC'ers with spare time out there (not me right now)
it will be done by tonight. It should be end up on the order of 40
lines of careful isochronous assembler code. I assume the AVR
would work too.

The result will be a $1, 8-pin, single-chip, 10 MHz to 32 kHz, low
(not zero) jitter frequency divider, suitable for driving cheap digital
or analog clock and watch movements from a 10 MHz source.
Extra credit for the 5 MHz version.

/tvb

Quoting Tom Van Baak tvb@LeapSecond.com, on Wed 23 Jul 2008 06:27:30 PM PDT:

Or, with a bit of work, you could make a sidereal clock (handy for
astronomers) or a Mars clock (handy for a very small set of people who
want to keep Mars time).

One might also be able to do a tide clock of some sort too, but it
would probably require a variable speed.

Jim

On Wed, 23 Jul 2008 15:22:08 -0700, Jim Lux James.P.Lux@jpl.nasa.gov wrote:

Yep.  There's turd polishing and then there is time nuts turd polishing. Going
for the angstrom finish on the turd.  The latter sometimes makes me grit my
teeth....

Seems to me that all the solutions proposed so far are a bit complex, trying
to go for the 32khz frequency when that's not necessary.  The quartz analog
clockworks has a one or two winding stepper motor.  The SIMPLEST solution is
to drive those coils directly with the PIC output and scrap the rest of the
circuitry.

With some clever fiddling, one could use one of the 8 pin PICs and that WOULD
get the solution cost down to around 50 cents :-)

John

John De Armond
See my website for my current email address
http://www.neon-john.com
http://www.johndearmond.com <-- best little blog on the net!
Tellico Plains, Occupied TN
Save the whales, collect the whole set!

At 04:22 PM 7/28/2008, you wrote:

Actually, it's not even that complex... it's often an
electromagnet/solenoid driving a conventional escapement type clock
mechanism. Why use 2 coils when you don't ever need to go backwards?

One advantage of generating 32kHz (averaged over 1 second) is that
you don't have to build the power driver stage to actuate that
electromagnet.. (since it's built into the single dirt-cheap chip in
the clock in the first place)

And, one could probably figure out a way to use the original 32kHz
crystal AND run at different speeds..

FYI,
Quartz analogue clocks almost universally use a bipolar motor, a two
pole
magnetic circuit with the minimum reluctance axis displaced from the
direction
of the energised field.
When activated the 2 pole magnetic rotor aligns with the magnetic
field, when the field collapses,
the magnetic rotor moves a little towards the position of greater
self attraction, so that it is
set up for a move in the right direction when the reverse field is
applied.
To drive these motors as clock displays you either replicate the
alternate 1.5 volt 20mS pulses, or connect a capacitor,
about 10 - 100 mfd in series and drive them with a 0.5 hertz square
wave of about 1.5 volt amplitude.
The drive voltage and pulse duration should be adjusted for each type
of motor, or else
the voltage and capacitor size so that reliable stepping occurs.
Overdrive can stop some types as the rotor "poles".
just in case you were interested,
cheers, Neville Michie

On 29/07/2008, at 9:35 AM, Jim Lux wrote:

On Mon, 28 Jul 2008 16:35:53 -0700, Jim Lux James.P.Lux@jpl.nasa.gov wrote:

The clockworks that I've taken apart almost all have 2 coils.  One brand has
one. They all drive a permanent magnet rotor that turns 90 degrees on each
tick.  I'm not sure what the single coil design does to make sure the rotor
always turns the right direction.  Or maybe it doesn't matter if the rotor
turns a cam and ratchet mechanism.  I've never taken one apart far enough to
know.

No driver needed.  Each coil has about a bazillion (bazillion.000000 for time
nuts) turns of wire so fine I can't see it without my 7x OptiVisor.  I've
never bothered to measure but the resistance has to be in the hundreds of ohms
or more.  It has to be that high to get over a year's operation from an AA
battery.  Duck soup for a PIC output pin driver.

Funny how this works.  I've been thinking about this same type problem for a
few days independent of reading this list.  I'm old-fashioned and like analog
clocks much better than digital.  I also like the precision of
radio-controlled clocks.  I've bought several different WWVB analog clocks,
all of which seem to use the same cheap ChiCom movement.  They uniformly suck
(to use a technical term) at receiving WWVB where I live.  The digital
versions have no problem receiving but I don't like the looks.

What I've been thinking about is a modern version of the Simplex master/slave
clock system.  A GPS disciplined master clock sending out operating pulses to
slave clocks around my house and shop.

I thought about wireless, including synthesizing my own WWVB signal but I know
that I'll not get enough round tuits to do that.  What I'm working toward is
just about what I described above, except that the master clock will drive 4
conductor telephone station wire and the slave clocks will contain no
electronics.  Only the clockwork and the coils.  All the clocks will be wired
in parallel.

This is an open-loop system that assumes all the clocks are in the same
mechanical position when the master is activated.  Perfectly acceptable, given
the relatively few number of clocks and the small area involved.

This architecture should give me what I want - REALLY simple, no electronics
in the individual clocks, "atomic" accuracy, automatic DST correction and
perfect synchronism.

Comments?

John

John De Armond
See my website for my current email address
http://www.neon-john.com
http://www.johndearmond.com <-- best little blog on the net!
Tellico Plains, Occupied TN
No one can be right all of the time, but it helps to be right most of the time. -Robert Half

At 05:04 PM 7/28/2008, you wrote:

The ones I have here have a single solenoid with a sort of wishbone
shaped thing around it. A curved metal rod with a thicker center
portion is connected between the arms of the wishbone.  The rod goes
through the center of the solenoid, and it looks like when the
current pulse is applied to the solenoid, the armature is pulled through.

Imagine a sort of pendulum with the solenoid at the bottom.

In any case, faking the 32.768 is going to be easiest, because the
clock mfr has already fooled with figuring out the right voltages,
currents, and pulse widths.

Traditionally, such systems had all the slave coils in series.

I suspect you will find that the better way; less room for weirdness.

Otherwise, I like it!

Tom Frank

Max Skop wrote:

I would consider locking a 8,192 MHz crystal to the 10 MHz. The
comparator frequency of 16 kHz using a division of 625 from the 10 MHz
and a 512 from the 8,192 MHz oscillator. The 32,768 kHz output can be
generated by first dividing by 125 and then a division by 2 for propper
50% pulse ratio.

Not too complex and the comparator frequency is high enought for a nice
filter should be able to do it. I would use a SR flip-flop for phase
detector. For this kind of work they work very well IMHO.

Cheers,
Magnus

Hi John:

All the quartz clocks I've taken apart have a single coil, for example see:
http://www.prc68.com/I/QuartzClk.shtml

The drive to that coil is bi-polar and is from a single AA cell, so the driver
IC must use an H-bridge.  If you want to drive a clock like this at 1 PPS from
a 10 MHz source you will need to use some kind of conversion circuit.  That's
why it's easier to generate the 32768 Hz signal and just drive the xtal pins on
the  clock's IC.

Have Fun,

Brooke Clarke
http://www.prc68.com/P/Prod.html  Products I make and sell
http://www.prc68.com/Alpha.shtml  All my web pages listed based on html name
http://www.PRC68.com
http://www.precisionclock.com
http://www.prc68.com/I/WebCam2.shtml 24/7 Sky-Weather-Astronomy Web Cam

Neon John wrote:

Hi Brooke,

The drive for the coils is really not very critical.  All you need is
a cmos inverter and a couple of capacitors:

~1PPS--[>o---+----[>o-------|(-----+
~            |                    |
~            |                  MOTOR
~            |                    |
~            +--------------|(-----+
~

Yes, it is just that easy.

-Chuck Harris

Brooke Clarke wrote:

Chuck Harris wrote:

Actually, when I spend just another minute thinking about it, it is even
easier:

~1PPS-----[>o-----|(------+
~                        |
~                      MOTOR
~                        |
~                        ///

-Chuck Harris

Brooke,

Here's what I did with a PIC (or Basic Stamp, etc.):

Current limit the coil to two pins of your microcontroller and
configure both as inputs (high Z).

Then for about 50 ms each second make the pins output; for
even seconds set the pins to 0 and 1; odd seconds 1 and 0.

/tvb

----- Original Message -----
From: "Brooke Clarke" brooke@pacific.net
To: "Discussion of precise time and frequency measurement" time-nuts@febo.com
Sent: Thursday, September 04, 2008 12:25 PM
Subject: Re: [time-nuts] How to get 32.768KHz from 10MHz.

For those still interested in the 10 MHz to 32 KHz divider issue, I
offer the PIC source to a tested divider that runs in an 8-pin DIP
package device:

http://www.ni6e.com/time/10M32K.html

Regards,
Peter

Tom Van Baak wrote: