Searching \ for 'Timing Questions' in subject line. ()
Make payments with PayPal - it's fast, free and secure! Help us get a faster server
FAQ page: techref.massmind.org/techref/index.htm?key=timing+questions
Search entire site for: 'Timing Questions'.

Truncated match.
PICList Thread
'Timing Questions'
1998\02\08@225133 by Bob Nelson

flavicon
face
If you were to take any 4 of the shelf  20Mhz  16C54 PIC chips, put them in
different locations using good quality Xtals and battery power but the
ambient temperature varied by 50 degrees F,  run identical code in them for a
timing function,  would they all have the same accuracy?

If so,what is the shortest time span that one could measure?

I have an idea for a project but need to know if it is feasable.

Bob

1998\02\09@182855 by Steve Baldwin

flavicon
face
> If you were to take any 4 of the shelf  20Mhz  16C54 PIC chips, put them
in
> different locations using good quality Xtals and battery power but the
> ambient temperature varied by 50 degrees F,  run identical code in them
for a
> timing function,  would they all have the same accuracy?

The PIC itself wouldn't have any influence once the system was up and
running (unless you had one of Einstein's experiments in mind). Everything
internally is governed and in sync with the clock. You'd  have to ensure
that the edge of what you are timing was sufficiently fast and stable
enough to make your measurement worthwhile.

Your biggest variation would occur at startup, when the oscillator starts
and how you generate the reset. Depending on your application, you may be
able to synchronise the devices after they have been started and avoid that
problem.

For a standard AT cut crystal, you would expect an initial tolerance of +/-
50 parts per million and then a temperature tolerance of +/- 10ppm and a
similar amount for aging. So worst case extremes would amount to 0.014% or
140us per second (12 seconds/day).

You could calibrate out the initial tolerance and if you can measure the
ambient temperature to say 5 degrees, you could make some correction in
software. There's not much you can do about aging.

You could improve that by going to an ovenized crystal oscillator and get
+/-0.1ppm but at the expense of $$$ and 300-400mA supply current.
You may also want to consider GPS as a stable and widespread time source.

Steve.

1998\02\10@031517 by Russell McMahon

picon face
Observation reveals that, if you trim a digital wrist watch
crystal oscillator carefully you can get long term
accuracies of well under 1 second per day drift. This is a
low cost 32 KHz crystal which hasn't been chosen for its
precision! This is a lot better than the specifications for
the crystal that the manufacturer would claim. Importantly,
even if it drifts, my experience is that it drifts
consistently and in the same direction. Of course, in a
professional design you are obliged to do a worst case
calculation based on the manufacturer's worst case specs
:-). The watch has the advantage of being attached to a very
good "oven" (your wrist, maintained at about 98 degrees F).
An article for a "micropower crystal oven" appears in
Electronics and Wireless World, August 1997 page 660. The
circuit uses 2 '555 oscillators which apply a variable mark
space signal to a heater resistor with the overall mark
space being controlled by feedback from an NTC thermistor
located with the crystal in the "oven" compartment. Power
consumption is not specified but can be minimised by running
the oven at "just" higher than the highest external
temperature liable to be experienced. More insulation will
require less wattage (and produce a slower thermal
response).


{Original Message removed}

1998\02\10@032728 by William Chops Westfield

face picon face
   Observation reveals that, if you trim a digital wrist watch
   crystal oscillator carefully you can get long term accuracies of
   well under 1 second per day drift. This is a low cost 32 KHz
   crystal which hasn't been chosen for its precision!

It's also used in an environment that maintains a relatively
constant temperature...

BillW

1998\02\10@040435 by Scott Newell

flavicon
face
>>    Observation reveals that, if you trim a digital wrist watch
>>    crystal oscillator carefully you can get long term accuracies of
>>    well under 1 second per day drift. This is a low cost 32 KHz
>>    crystal which hasn't been chosen for its precision!
>>
>It's also used in an environment that maintains a relatively
>constant temperature...

I've been reading up on timekeeping recently, and ran across a bunch of
interesting information.  Here's a couple of links worth checking out.

Best of all is the Hewlett Packard application
note on the science of timekeeping.  It includes a
description of an experiment performed on some really
cheap digital stopwatches over a period of 3-4 months.
They decribe how you can model the errors in the
frequency and aging to correct the indicated time.

It's available on-line at:
http://www.tmo.hp.com/tmo/Notes/English/5965-7984E.html


The HP app note references a patent on a 'smart' watch
that updates an equation containing drift and aging
correction terms.  It looks like it would make a fun
little project for a PIC, lcd, and oven oscillator.

It's also available on-line (as most recent patents
are) at:
http://www.patents.ibm.com/details?patent_number=5%2C274%2C545


There are several other HP app notes on subjects such
as electronic counters, GPS time standards, crystal
oscillator design, etc.  Many of these would be relevant
to any discussion of high-accuracy oscillators and
timepieces.


later,
newell

1998\02\11@054229 by Harold Hallikainen

picon face
       How about doing the correction in software based on measured
temperature and a look-up table?  This seems a lot less power hungry than
a crystal oven.

Harold



On Tue, 10 Feb 1998 20:52:23 +1300 Russell McMahon <spam_OUTapptechTakeThisOuTspamclear.net.nz>
writes:
{Quote hidden}

_____________________________________________________________________
You don't need to buy Internet access to use free Internet e-mail.
Get completely free e-mail from Juno at http://www.juno.com
Or call Juno at (800) 654-JUNO [654-5866]

More... (looser matching)
- Last day of these posts
- In 1998 , 1999 only
- Today
- New search...