Conference hydra::amiga_v1

    I successfully downloaded the .ARC file with Kermit (after two tries)
    and the program runs just as described in the documentation.  The
    instructions for adjusting the clock are explicit enough for the
    littlest moron, so I was able to find the correct capacitor.
    
    However, I can't seem to get the adjustment to zero.  The clock
    error ranges from plus 2 ticks in 600 to plus 4 depending on the
    capacitor setting---I can't get it below 2.  Does this mean my machine
    is bad?  I doubt I can get it replaced, since there is no "spec"
    for the clock.  Is there a more drastic adjustment that would put
    the capacator in a range that includes 0?
        John Sauter

    You may have to buy a new capacitor for it.  But, I dunno.
    
     - Chris
    

    Will this work for the A500 as well?  I haven't checked my A500
    clock for accuracy.  I just figured it wasn't very.  If I could
    adjust it to a reasonable degree of accuracy, I think I would get
    rid of that old wall clock that TICKs at 98db.   :-)
    
    
    					John

    the CBM folks on usenet said that the A500 has the same adjustment
    pot, but it is under an RF shield so you can't easily get at it.
    
    -Dave

Re: .1

Just use the program to get the error numbers as low as possible.  I never
was able to get my error numbers down to zero, but the setting I ended
up with only gains a few seconds a week.  (Lately, it seems that the
clock is getting more accurate even though I haven't touched it in months.)

Read the final section in the docs.  Getting the adjustment down to zero
may not mean that much because the short term variability of the power
line frequency is pretty large (at least here in Tyngsboro, MA, USA).

Usually when I run the program, it reports errors that indicate that
my clock should be off my more than it is.  My advice is to just get the
best adjustment you can, then wait a week to see what the effect of the
new setting is on your clock.

I would have liked the written the program to use the timer in the Amiga
that ticks based on the vertical refresh rate because that is more accurate
over short periods of time.   However, the system sets and resets that timer
constantly for its own purposes.  This means only the system knows what
the time that timer indicates.  The system service to read that timer is
not useful for my purpose because it intentionally adds fuzz to the time
so that every call returns a unique system time.  I used the line frequency
timer because the system leaves that timer alone for long periods of time
(about 10 minutes).

I'll update the docs in the "procedural" part of the docs to explain
better that adjustment that produces zero errors is pretty unattainable.

Re: .3

The program will work just fine on an A500, but you will have to drill
a hole in your A501 memory expansion/battery backed up clock in order
to get at the capacitor.  Since determining where to drill is a problem,
your memory expansion may look like Swiss cheese before you are though :-).

Lots of people complained about the clock in the A2000, but not many
have complained about the A500.  Maybe Commodore did a better job of
adjusting those clocks at the factory.

    drilling the metal shield of a pc board can cause little burs and
    flakes of metal to short out the etch and short out ic pins, causing
    possibly permanent damage to the entire computer.
    Tom

    Is it really that relevant to sync your clock to line frequency
    so accurately?  I mean, line frequency can drift over the course
    of the day.  (You are syncing to line frequency to adjust the clock,
    right?)

    Over a long period of time (days, I think) the line frequency is
    very accurate.  This is because it is synchronized to the National
    Time Service clock, which is the source for WWV.  Therefore,
    if you want your Amiga clock to be hassle-free, it is useful
    to make it track the line frequency as a first approximation,
    and then to tweak it based on observation over several weeks.
        John Sauter

Re: .9

>    Over a long period of time (days, I think) the line frequency is
>    very accurate.

I suspect that over a 24 hour period the line frequency is very accurate.


Re: .8

If you are asking, "Why the power line frequency?" then .9 answers that.
If you are asking why I care about what seems to be a small variations
in the power line frequency, consider that a error of 0.00001 is a about
a minute and a half error a week.

re: -.1
>consider that a error of 0.00001 is a about a minute and a half error a week.

Well, if you're being fussy: there's 604800 seconds per week. An error of
0.00001 would be 6.04 seconds. An error of 90 seconds (minute and a half)
would be 0.000148809. Or are you talking about a WORK week, which seems
ten times as long? :^)

Re: .11

So I was off by an order of magnitude.  Anyone who knows me is familiar
with my slight tendency for hyperbole.

    I do understand how much error can creep in over the course of a
    couple of months.  My question was more directed towards how much
    sense it might make to worry the difference between the clock and
    line frequency down to zero, when in the short term, the line frequency
    might be off itself.  Of course, if you're running it for a day
    or two, it does make quite a bit of sense.

Re: .13

>Of course, if you're running it for a day or two, it does make quite
>a bit of sense.

Unfortunately, about the longest you can run the program is about 15 minutes
before you have to reset its statistics keeping.  The line frequency timer
gets munged by the operating system in some mysterious way about every
15 minutes or so.

The program allows you to get a feel for a change in adjustment in about
a couple of minutes.  That's enough to allow you to correct really bogus
adjustments of the clock very quickly.  After that, you will have to set
the clock, wait a week, and see how far it drifted.  When I set mine, I
discovered that a week later the drift was down to a few seconds.  Over
time, I discovered that the clock, with no further adjustments, has gotten
even more accurate.  I suspect that there is some obscure (to me at least)
physical property that causes the capacitor's capacitance to change
slightly if no adjustment is made to it for awhile.  I noticed that for
the first 10 second interval after adjustment, the time results would
typically be very screwy.  Then they would settle down.

It's been a long time since I heard the details but... The accuracy of the line
frequency is very good over a long time but not necessarily over the short time.
As I recall, the power companies keep track of the deviation  during the day and
then, sometime around midnight (I may be wrong), the speed up or slow down the 
generators for a while until the error is again zero. Now, it depends on when 
you try to calibrate your clock with theirs, and how you try to make 
corrections. Depending on your luck, you might try to recalibrate when the line
frequency is right on or when it is trying to "catch-up". If you keep track of 
the error over, let's say, a week and then simply tweak the capacitor up or down
a known amount and then check it a week later, you can see if you tweaked it in 
the right direction, and how much effect the adjustment had. Then give it 
another shot, ad nauseum. Actually, How necessary is it to have an accurate 
clock on line (unless you are running some form of astronomy program or 
something)? The clocks set on the various VAX nodes I log onto are rarely closer
than a minute or two to the correct time.

Dave