NEW ZEALAND:   Talking Clock on  0900 45678
The Measurement Standards Laboratory of New Zealand  advises the following about the 'Talking Clock':
"MSL operates a talking clock that is available to the New Zealand public via the Telecom NZ 0900 service. On dialling 0900 45-678, the user receives a voice announcement of the time followed by three "pips" of 1000 Hz tone. The start of the third pip is the exact time. The announcement and pips are repeated at 10 second intervals to give six time announcements in all, ensuring that the full minute is always included."

Spot checks by NZ Section members (some years ago) indicated that there were 'some concerns' over the actual accuracy, but that this time signal is probably within 0.1 second of the correct time.  (We should check with the NZ MSL whether this service is "audited", i.e. whether it is cross-checked against a time standard.)
 

AUSTRALIA:   "Talking Clock" on  1194 (about 40 cents per call)  and commercial services on 1900 912 000 and 1900 911 481 (much higher cost) only
Since early 2006 the services on on   1194,   1900 912 000  and  1900 911 481  are derived from a dedicated  time server  operated by a private company (separate from the main telephone carriers), with reference to UTC (AUS)) as maintained by the  National Measurement Institute  (formerly the CSIRO National Measurement Laboratory)  -  use the 'Capabilities' tab to see their section on   Time and Frequency Capabilities     To the best of our (April 2006) information, this time signal is stated to be accurate to 1 millisecond (0.001 second) when the time pips enter the public switched telephone network at the source in Melbourne.  When propagated over landlines, any delay within Australia should be "small" for most users.  

However, all telephone systems worldwide now use transmission technologies involving numerous computer conversion stages, repeaters, etc.  where telephone signals are "packed" and "unpacked" many times.  Therefore it is "conceivable" that delays could be introduced, but these are difficult to quantify.  Also see the separate note below on the use of  mobile ("cell") phones.  A "typical maximum" delay of 30 milliseconds is quoted on the web site of  NIST in the USA  and the same value was quoted by Telstra in Australia in 2003.

NOTE:  the third commercial service on  1900 912 902  is *NOT* in the above group of '1900' numbers.  The third service is operated by a different company which uses a different,  as yet unknown source of time.   Please do not use this as a reference, unless you have better information on its accuracy.

The source of the Telecom/Telstra time was a dedicated group within the Telstra Research Laboratories (since closed down in March 2006).  The time reference was of a very high (internationally recognised) standard.  When dialled over the telephone, the time signal ran for about 5 minutes, after which the call was disconnected.  The hour and minute spoken was the local State zone time, generated accurately in each State capital city, but the start of each minute and second marker were accurate UTC.  Telstra advised in those early days:  "in all Australian States local time on 1194 should be within 30 ms for callers in every part of the State they are making the call from.  In practice this is typically better than 10 ms".  Until December 2002, the 1194 service had indeed been very reliable and accurate.

From January 2003 to March 2006 there was a period of "uncertainty" during which the 1194 service was provided by a third party and the service was inconsistent.  There were occasions when the spoken time was pretty good (i.e. within 0.1 sec), but other occasions when the systems drifted away by anything up to 5 or 10 or 20 seconds.  This lack of reliability made the service unusable.  At that time, we lodged a number of complaints about this with the Telephone Information Services Standards Council (TISSC), the industry regulating body, but this organisation was unwilling to take action  -  their web site is:   http://www.190complaints.com.au/    Typically, when complaints are made, the TISSC will offer to refund the cost of the relevant (faulty service) phone call, but  the TISSC will not take action to enforce correct time announcement  by the service providers.   This is a deplorable situation, which regrettably also appears to be the case in the USA.

The good news is that from early 2006 we have been advised that the time spoken on the  1194  service is again correct, as outlined above. 

CAUTION -  Mobile ('Cell') Telephone Networks introduce a further delay
Accessing a telephone time service through a mobile telephone causes an additional small processing delay.  This can be noticed by comparing the audio from a fixed landline phone to e.g.  1194  with the audio from a mobile handset dialing the same number  -  simply hold one to each ear.   Sometimes the two signals match, sometimes they are very noticeably different.  We have found that with the Australian GSM system the mobile signal arrives a fraction of a second late and we do not know if this delay is a fixed value - users should definitely be aware of this.

The National Institute of Standards and Technology (NIST) in the USA also runs a telephone service for users in the US, and the NIST web site also alerts the user to significant propagation delays in the mobile telephone network as advised on their web site:   http://www.boulder.nist.gov/timefreq/stations/sig.html
 

**  Firstly, your network needs to support auto-update of time (and date).  Many (but not all) GSM networks do this and most 3G (including UMTS) can do this.  **  Secondly, you need a handset that has the auto-update capability and this needs to be enabled (check the Setup menu).  Most of the early (1990s) handsets did not have a clock, full stop.  Later they would come standard with a clock display derived from the local crystal oscillator.  Today nearly all newer handsets are capable of periodically auto-updating their local clock.  However, it is not easy to know this before actually using a (network + handset) system.  It is rarely mentioned in the published specifications and most sales staff have no detailed knowledge of this.

Sadly, from a timing point of view, from trials with a few different handsets it is found that with certain handsets the minute changeover is often (but not always) within a few tenths of a second, but this cannot always be guaranteed;  some time later the minute changeover may be a whole seconds early or late.   On other handsets the changeover can be off (early or more often late) by as many as 40 seconds.  Presumably (?) this must be related to the other tasks the phone processor is working on.  This accuracy may be acceptable for most home or office use, but not for scientific work. 

Therefore NOTE that if you were to consider using a mobile (cellphone) handset for any kind of timing purposes, it is essential that you repeatedly check YOUR  (handset and network combination) against an independent source of accurate time.

Auto-update, i.e. if your network sends time and date to handsets:   Please check the setup ("time settings" or similar) of your phone and check for an "automatic time update" function.  If this function is enabled, the time and date display on such a mobile phone is sourced from the base station.  You do not need to "set" the time manually; just switch it on and wait for the first full minute changeover.  However, for the accuracy of the minute changeover, see above.

Quartz clucks:  if your network or phone does not support time broadcast or if you have turned this function off, then your phone will continue to use a local clock with a display derived from its own quartz crystal oscillator, which needs to be set by the user and checked from time to time.  The accuracy, long term drift, temperature dependence of this clock are like that of any quartz watch or clock; most are pretty good and some are badly adjusted.  Once set accurately they will provide excellent time information for a period of "some hours" or more, depending on the quality of the internal clock.

This means that if you are lucky and have a handset with a well-adjusted internal (quartz) clock, and you manually set this to a standard time source, then for some time (some hours?) the time displayed can in fact be more accurate than the time derived from a 'typical' auto-update handset.  Life was not meant to be easy.