Re: CLKADJ?

On Sun, 11 Jun 2006 14:37:11 -0500, Heiko Arnemann wrote:

When using my time keeper program, it is required
to make all adjustments with special programs,
in order to record the adjustments.

Any adjustment by using built in tools are not recorded and so,
further required adjustment would be calculated with an error...

These points are exactly what I meant;
so long as the built-in functions remain available,
they can bypass any suggested alternative system
(including any "daily alarm" system, but in the latter case,
corrections are more frequent
and maximum uncorrected deviations are less).

We have had philosophical differences like this before,
in connection with "hh.mmss rounding" -- you did not want
to deal with with any possibility that the user would do anything
that wouldn't work, while I tended to assume that any possible
input might be given, and that we want to cover all cases
as best we can, to avoid giving users any way
to make the system fail and get incorrect results.

We don't want the airbag to become necessary,
but we provide one anyway, just in case of really bad driving :)

Scott reported about an hp 49g+ which was
running a daily alarm in combination with a time keeping
program. On weekends, when this calc has not been touched
the calculator misses alarms and additional to that it
misadjusted the clock. He was using my time keeper and
my timekeeper simlply needs one call << AAD >>
to get an Automatic ADjustment. When doing this
call later it makes a proper job on HP 49G, but
not on hp 49g+

Unless we have some other clock source,
no program can restore a broken (irregular) clock
back to the correct time.

your proposal works perfect with HP 49G but not for the 49g+

*Neither* proposal can correct for unexpected changes
within the internal clock; my comparison between the schemes
is based only on the ability of users to bypass the schemes
and use internal functions directly, and I like more frequent
adjustments (via alarms) only because this scheme is automatic,
and limits the difference between real time and calc clock time
to a lower value.

ROM 2.01

Aren't later versions now available?
Have later versions fixed any clock bugs?

and it is very hard to find out, what is messing up the
clock. That is one of the causes, I have decided to avoid alarms,
while testing the "time management" [steady drift] of the 49g+

If the internal clock goes awry due to alarms alone,
then it would be good to avoid any alarms at all,
and use a system that isn't alarm-based.

If alarm-related clock issues are solved in current or subsequent
ROM versions, then I'd recommend alarm-based adjustment over
a system requiring the user to remember to manually adjust,
or a system that allows long drifting between adjustments,
mainly because of all the opportunities to neglect adjustment,
and to use internal functions directly meanwhile
(including even the built-in on-screen display).

[r->] [OFF]
.

Relevant Pages

  • Re: Stolen and degraded time and schedulers
    ... If you want accurate time accounting, don't use the TSC. ... The adjustments that I spoke of above are working regardless of ntp .. ... as does interrupt latency since the clock is essentially ... CPU does work as a timebase, then using the same warping mechanism would ...
    (Linux-Kernel)
  • Re: CLOCK_MONOTONIC and
    ... reference clock. ... all timeout values will occur at once then. ... so the NTPD way of small adjustments won't ...
    (comp.unix.programmer)
  • Re: [PATCH 2/5] introduce CLOCK_MONOTONIC_RAW
    ... the raw clock doesn't need any adjustments, ... (Some day we may want to have multiple active clocks for various purposes ... I think that crufts up the clocksource structure (which is ...
    (Linux-Kernel)
  • Re: Why should I (not) use an internal oscillator for 8-bit micros
    ... > with the incoming data as it doesn't have a clock to sync to. ... > This is always presuming that the receiving UART has ... 2.5% match between the sender and the receiver. ... of other schemes for adjustments - CAN controllers adjust their sub-sampling ...
    (comp.arch.embedded)
  • Re: Never on a Saturday...Alarms, if you dont want to read about alarms
    ... >>> be in multiples of clicks. ... > executing 8129 CLKADJ or adds some unintended adjustments. ... I really do think the alarms mess with the clock. ... After a weeken of idle time, ...
    (comp.sys.hp48)