|
|
| Re: Master-slave clocks explained (580171) | |||
 |
|||
| Home > SubChat | |||
|
[ Read Responses | Post a New Response | Return to the Index ] |
|
||
|
Re: Master-slave clocks explained |
|
|
Posted by RonInBayside on Mon Mar 3 19:56:39 2008, in response to Re: TA is obsessed with CBTC, and ''New'' tech for no reason., posted by Jeff H. on Mon Mar 3 19:52:17 2008. A long while back, Jersey Mike asked about synchronized clocks and "Mark" answered:There were a few different schemes, but most of them (built by International/IBM/Simplex or Standard Electric) sent out a single pulse to all slave clocks in the system once a minute. Most systems used 24V DC pulses, but anything else could have been used. The pulse triggered a solenoid in each slave clock that "unlocked" the minute hand, when the pulse ended and the solenoid released, the minute hand would advance to the next stop at the next minute position. This keeps all of the clocks moving in step, but not necessarily indicating the same time. To handle this, a correction pulse was delivered by the master clock once every hour, at the beginning of the hour, instead of a normal per-minute pulse. The correction pulse can be anything distinct from the normal pulses: for example, delivered on a separate wire, or inverted polarity. On simpler systems, the correction pulse would actuate an electromagnet that would draw the minute hand to the :00 position. This system, which would pull the minute hand forward or backward as much as 20 minutes or so, was used on Standard clocks. A more complex (and reliable?) system used by International had the slave clocks stop responding to per-minute pulses when they hit :59 and instead respond only to the correction pulse. During the 59th minute of an hour, the master clock transmits 59 normal (per-minute) pulses, to advance any clock that had fallen behind during the hour. At the last possible moment in the hour, all clocks (including those that had fallen behind) should be indicating :59 and will not be responding to the rapid pulses; when the correction pulse is delivered at the top of the hour, all clocks will advance to :00 in unison. Similarly, a clock that somehow ends up running fast will hit :59 early and stop, awaiting the correction pulse to push it to :00. (These slave clocks can be fun to watch when they're "stuck," they seemingly work very hard at the top of the hour but make no progress. A lot of noise, though.) Generally speaking, these systems synchronize the minute hand only, once per hour. If it's 3:05 when you install a clock, and you set it to 5:00, when 4:00 rolls around it'll wind up displaying 6:00. More advanced systems would also synchronize the hour hand, using a 12:00 correction pulse, similar to the :00 correction pulse. However, this degree of synchronization wasn't the huge problem it might seem to be: if a clock somehow managed to lose 60 consecutive pulses, then something else was probably wrong and the clock needed to be looked at anyway. Master clocks and repeaters often included features like "pulse counters," which would "save up" the pulses from the master clock in a power failure and deliver them to the slaves when the power returned. I don't know that any systems allowed clocks to be stepped backwards, but I'm sure it's been done. Even when the clocks would only move forward, it was easy enough to stop the system for an hour, or to set the master clock 11 hours ahead to effectively drop an hour back, to handle the fall equivalent of the dance we all should have done last night. Make the correction once at the master clock, and let it worry about sending out all those pulses. Where slave units have second hands, they're usually driven locally by the slave clock itself without any synchronization. The second hand stops when it hits :00, and restarts when the clock receives a pulse. It's not too difficult to remain accurate to a fraction of a second if you only have to do it for 60 of them before being resynchronized, even if the time base is line frequency. Even so, synchronized systems with second hands are much rarer than those without. Remarkably, just about anyone reading this, regardless of age, who attended a school with a synchronized clock system relied on time delivered by a system like the ones described. Fully digital installations are a newer phenomenon, so taking retrofits and new construction into account, this is only just now beginning to change. IBM has some information on IBM clock systems. You can also find a wealth of information on these systems from people who collect them. Here's one for Standard Electric clocks. This is a function of the master clock. On the oldest systems, the master clocks were mechanical (pendulum-and-spring movement) with an electric motor connected to wind the spring daily. In the event of a power failure, the master clock would continue to run, but the electrically operated slave clocks obviously would stop. These systems could be fitted with the "pulse counters" I described to mechanically "save" pulses that the master clock generated but could not be delivered to the slaves. Even when pulse counters were not installed, these master clocks were always equipped with a means to advance all slaves in the system. This was usually a button or switch that would send pulses to the system rapidly, rather than once per minute. On newer, fully-digital masters (yes, master clocks are still being made for pulse-driven analog systems, there's still a huge installed base), the difference between the "new" time (entered on a keypad or synchronized with an external source) and the "old" time is computed and the master sends out the proper number of pulses to catch a slow system up to time, or stops sending pulses until time has caught up with a fast system. The master, when equipped with a battery backup, can also incorporate a pulse counter, except now they're electronic instead of mechanical. And there's still a means by which the system can be advanced manually. Slave clocks are indicating instruments only. To the slaves, pulses are just pulses. They don't care if they show up on time or not, or if they're generated by a master clock or some prankster with a fistful of 9V batteries. Incidentally, it's simple to improvise a master if you want to get a few slave clocks of your own, especially if you'd be happy with a straight pulse generator and forego some of the self-correcting stuff. If you're good with some combination of electronics, computers, and embedded controllers, it shouldn't be a problem for you to put together a fancy master clock, either. Also, how do these systems support second hand operation? The systems I described usually had second-hand operation retrofitted by adding an extra wire (or two) to carry unswitched AC at line frequency (but not necessary line voltage). This drove a synchronous motor to push the second hand. Depending on the system design, the second hand would stop in place either when it hit the :00 position (once per minute) or when it hit :00 and the minute hand was parked at :59 (once per hour). It would not be restarted until unlocked by a solenoid actuated by the rising edge of a per-minute or correction (per-hour) pulse. Sometimes, the motor and gearing were designed so that it took the second hand slightly less than 60 seconds to make one full rotation; this way, there was a better guarantee that the second hand would be parked at :00 when pulse to restart it came along, and less of a need for an electromagnet to draw it into position if it were running slightly behind. This was possible because commercial power was stable enough (at least in this country) by the time these clocks appeared that, as I mentioned in my last message, it could be trusted as an "accurate enough" timebase for 60 (or 3600) seconds, as long as the second hand would be resynchronized each minute (or hour). When I say "retrofitted," I mean that, for example, IBM added clocks with sweep second hands to their catalog, not that it was common for people to replace their perfectly good 2- and 3-wire systems. You can spot these systems by looking for a minute hand that jumps in 1-minute increments. (On "normal" independent clocks, the minute hand sweeps very slowly, and when a second hand is present, it's geared directly to the second hand at the obvious ratio.) What I've described is the basic design used in most master clock systems, specifically, those manufactured by IBM/Simplex or Standard Electric Time. As usual, there were variations, but they're mostly minor details. Some systems gave hourly correction pulses at some point other than :59/:00, or they used two minutes (or more) as the correction interval instead of one. Some played twice-a-day catch-up at 6: or some other time than 12:, so as to not disrupt the workday. Some ran as plain independent clocks synchronous with line frequency for an hour, and received only hourly correction pulses to pull both the minute and second hands into position. Some played funny gimmicky games with polarity. There are probably at least as many variations as there were manufacturers. Railroads were big customers of master clock systems for more than just in-station needs. Entire roads were driven by these systems, with slave dials in every station, tower, and office, synchronized timeclocks, and more. As the first real consumers of accurate time, railroads have probably done more for timekeeping than anything before, anything since, or anything to come. Mark |
|
|
|