[LEAPSECS] Bloomberg announced its smear

[Warner Losh]

On Tue, Sep 27, 2016 at 12:56 PM, Peter Vince <petervince1952 at gmail.com> wrote:
> Hi Michael,
>
> On 27 September 2016 at 19:37, Michael Shields via LEAPSECS
> <leapsecs at leapsecond.com> wrote:
>>
>> On Tue, Sep 27, 2016 at 10:27 AM, Peter Vince <petervince1952 at gmail.com>
>> wrote:
>> > I am interested that Google have chosen to linearly smear over 20 hours,
>> > thus increasing each second in that period by 13.8888... microseconds -
>> > surely such an irrational step is difficult to achieve?  Could it be
>> > that
>> > they actually step 125 microseconds every 9 seconds, or maybe 25
>> > microseconds every 1.8 seconds?
>>
>> No, there are no steps.
>
> Thanks for your reply.  If there are no steps - how is it achieved?  Do you
> have a dedicated hardware clock(s) where the frequency of the master
> oscillator is wound down a bit, and your servers synchronise to that?

All computer systems have a 'paper clock' they steer based on a
frequency input of some kind. The frequency input is calibrated
against an outside source to get an estimate of the error in the
frequency. With that you can estimate the true passage of time fairly
well. To achieve the slew, you adjust the frequency by the error you
wish to introduce. In this case 1.3888e-5, or ~14 parts per million.
After 20 yours with that frequency error, you've accumulated a second
worth of 'error' just in time to be on time again due to the
introduction of the leap second.

Warner