PPS Signal Transfer via Broadcast Television Stations Introduction Common view (CV) techniques are often used to synchronize time and frequency standards over long distances. Using these techniques two or more laboratories receive timing signals from a single transmitter, often a satellite. One of the stations generates a time signal and acting as master compares its signal to the satellite signal and transmits the difference via some other means such as the Internet to other stations acting as slaves. Each slave station compares its signal to the satellite signal and adjusts its own signal to the same difference as the master corrected for the known propagation time. CV techniques have also been used with broadcast television stations for time transfer within the broadcast coverage area. This is how laboratories in the Washington, DC, area synchronize to the USNO standards and how the NIST radio stations WWV and WWVB synchronize to the NIST standards in Boulder, CO. It is possible to extend CV techniques to synchronize pulse-per-second (PPS) signals used for precision synchronization of networked computers. Ordinarily, PPS signals are generated by a radio clock such as a GPS receiver and connected to the computer via a serial or parallel port. Special software in the operating system kernel processes these signals to discipline the computer clock time and frequency. While GPS receivers intended for navigation are relatively inexpensive, those intended for computer timekeeping are rather more expensive and not likely to be a ubiquitous computer peripheral. Moreover, GPS receivers require rooftop antennas with clear view to the heavens, which is not always practical. In general, broadcast television signals are much stronger than GPS signals and can be received over much of the broadcast coverage area without rooftop antennas. Moreover, television receivers are much less expensive, especially small screen, black and white receivers. Thus, broadcast television signals might be an attractive time and frequency transfer mechanism for the PPS signal. Design Approach The approach is based on a technique used for time transfer between LORAN-C radionavigation stations. The horizontal sync frequency used in the NTSC television signal is not quite synchronous with the second and, in fact, the residual within the second precesses over succeeding seconds at intervals of about 3.8 s. According to the NTSC standard, the horizontal sync pulse is derived from the color burst frequency of 3.579545 MHz as the fraction 2/455 of this frequency 15734.263 Hz. A clock synchronized to this source and with 15734 ticks in the second would be 16.7718 parts per million (PPM) fast, well within the tolerance range of the nanokernel software. This is assumed the master PPS source for both the master and slaves. The remaining corrections are implemented in software. The frequency tolerance of the color burst is specified as 10 Hz or about 2.8 PPM. The tolerance of typical computer clock oscillators can be several hundred PPM. However, with the method described here, the tolerance can be no more than 63.555563 PPM with PPS comparisons at 1-s intervals. If the particular oscillator frequency exceeds this value, some means must be taken to reduce the excess, either hardware or software. The time interval between the most recent horizontal sync pulse and the 1-Hz PPS pulse represents a phase signal that can be used to discipline the local clock frequency. At each PPS epoch, the phase should increase by 16772 nanoseconds modulo 63556 nanoseconds with an uncertainty of .001 PPM. If it does not, either a pulse has been missed, a noise spike has occurred or the frequency needs to be corrected. This can be done using the same algorithms in the nanokernel and will not be discussed further here. The frequency correction can be transferred from the master to the slaves using a NTP extension field designed for this purpose. The PPS time transfer is a little more complicated. The scheme begins with the observation that the PPS precesses over the sync pulse interval at 0.263 Hz or a period of 3.8022813 s, Thus, there are a minimum of three samples per period that can be used to interpolate the exact phase of the horizontal sync interval is coincident with the PPS second. As in LORAN-C, this is called the time of coincidence (TOC) epoch. Two or more stations receiving the same television signal can record the TOC epoch independently. At designated intervals in the order of a minute they can exchange the current phase values via the Internet or some other means and compute the phase offset of each station PPS generator relative to the others. The slave stations can phase-lock to the master using conventional feedback control algorithms. Implementation A suitable device design consists of a sync separator, oscillator and time interval counter. A similar device can be used by both the master and slave stations. The sync separator generates a pulse coincident with the horizontal sync pulse from the television receiver. This pulse resets the counter, which then begins counting the oscillator pulses. The oscillator frequency must be a multiple of 1 Hz, but can be a value in the megahertz range, the higher the better. For instance, a 10-MHz oscillator will result in a PPS signal jitter no more than 100 ns. The PPS signal from a GPS radio or other precision source latches the current counter value, which is read by the associated computer using a suitable interface such as the parallel port. In the NTP protocol the client (slave) stations sends a message to the server (master) station at intervals in the order of a minute. The server includes in its reply the most recent frequency and latch values. This can be done using an extension field defined for the purpose. The client adjusts the frequency and coarse time from the reply in order to maintain accurate seconds numbering, but uses the PPS signal for the most precise adjustments. In principle, the oscillators do not have to be of precision quality and can run at a constant frequency. In the suggested design it is not necessary to slew the slave PPS signal to align with the master PPS signal, since the offset can be injected into the software using the PPSAPI application program interface. Dave sends