This invention relates to the synchronization and clocking of wireless devices such as base stations and access points including but not limited to CDMA and Wideband CDMA (UMTS and CDMA2000) femtocells and picocells and WiMax and LTE access points.
GSM is an asynchronous digital cellular telephone technology and hence GSM base stations do not require synchronization unless they are used for positioning of cellular telephone handsets. However, CDMA base stations have always required synchronization to facilitate hand-off of handsets between base stations. Synchronization allows the handset to switch base stations during a call while maintaining track of the spreading code. If the handset is too far out of synch with the spreading code after handoff then it will take too long to reacquire code lock and the call will be lost. In addition, Wideband CDMA (CDMAOne, CDMA2000 and UMTS) base stations and WiMax and LTE broadband access points all require synchronization and syntonisation (frequency control) to relatively stringent requirements.
CDMA base stations have more demanding synchronization requirements and have always employed GPS for synchronization. Typically, the GPS antenna is deployed on a mast external to the building housing the base station and is placed so as to provide an excellent view of the sky.
Conventional base stations use T1 links (or equivalent) for the back-haul and hence employ Synchronous Digital Hierarchy (SDH) globally or Synchronous Optical Networking (SONET) in North America as the transport protocol. This provides a mechanism for syntonisation and synchronization.
The main base station processor communicates with the BSCP via a serial port and the BSCP controls and monitors the status of the GPS receiver via another serial port. In a CDMA or WiMax base station the GPS would be the primary reference but in a UMTS base station the 8 KHz_In and 1PPS_In would be the primary references derived from the SDH back-haul link. In each case, the BSCP would control the switchover to the secondary reference via controls to the PLL circuit. When both GPS and external references are in holdover the BSCP would direct the PLL to hold the control voltage to the OCXO.
In a CDMA base station the timing would be required to holdover to within 8 μs for 24 hours. In order to meet this requirement, a very expensive double-oven OCXO would be incorporated.
Typically, a base station would incorporate 2 complete base station clocks in a redundant configuration in order to allow for hardware failure of the clocks themselves.
Such a GPS deployment as is used in CDMA base stations is not economically viable for femto-cells or, to some extent, pico-cells. In these applications, the GPS antennas are embedded in the product either within the plastic enclosure or attached rigidly to the enclosure. Hence, the embedded GPS receiver must operate indoors, generally in an assisted fashion. No redundancy is incorporated.
Various synchronization techniques have been used by UMTS femto-cell developers including:
However, femto-cells use the subscriber's broadband internet connection for the back-haul. A tunnel or Virtual Private Network (VPN) connection is established between the femto-cell and the network core. Since the open Internet, with its highly variable latencies, is used for the back-haul, use of the back-haul network for synchronization is extremely problematic. Furthermore, since one of the prime drivers for deploying femto-cells in USA is to overcome the poor indoor cellular coverage typical of US networks, it is not possible to rely on the availability of adjacent macrocell transmissions for synchronization either.
Synchronization via Indoor Assisted GPS is much more accurate than synchronization via the internet and, arguably, is more dependable than synchronization to adjacent cell transmissions. Furthermore, since GPS can provide highly stable frequency as well, it is possible to reduce costs in the femto-cell by causing the GPS receiver to discipline its own oscillator and to supply the clock signal from that oscillator as the master clock or system clock for the femto-cell. This has led to the development of highly integrated low cost GPS Clock Modules (GCMs) for integration within femto-cells to perform the functions described above.
In a small percentage of indoor locations, periods of holdover may be experienced where the GPS signals fade to the point where the synchronization specification may not be maintained. This invention incorporates several schemes that overcome this difficulty by falling back on any of the above alternative synchronization schemes (including SDH/SONET in the case of base station applications) during GPS holdover. It provides a low cost GCM that with enhanced interface capabilities and firmware can be used to do this at minimal additional cost. It also describes how such an enhanced GCM could be used as the core of a very low cost macro base station clock.
The goal of the invention is to facilitate a scheme to back up GPS synchronization and clocking for femto-cells. The mechanism for doing this must be readily integrated into a low cost GPS Clock Module (GCM) as would be suitable for integration into a femtocell. It must also avoid phase discontinuities in the master clock supplied by the GPS receiver circuit. This implies that any switching between synchronization modes must occur on the control side of the oscillator rather than at the outputs of two oscillators.
In
In GCM solutions designed for use in femto-cells, the 1PPS alignment is maintained through the oscillator discipline and not by choosing the nearest clock edge as is conventionally done in GPS receivers. The reason for this is to ensure that the number of clock cycles between 1 PPS pulses is always exactly the same. It also has the effect that the clock frequency error is strictly proportional to the change in synchronization error between any two pulses.
Note that the 1 PPS is coherent with SCIk and hence SCIk may be used within the synch discriminator to precisely interpolate between the 1 PPS and an external reference event.
The synchronization reference may be derived using any available Non-GPS synchronization scheme and the synch discriminator circuit may be implemented using a combination of hardware and firmware. The synch error passed to the GCM should also include an estimate of the first order error statistics (e.g. standard deviation) in the error measurement. Alternatively these statistics may be known a-priori and stored in the GCM.
Note that the synch discriminator may operate continuously or just during GPS holdover periods. If it operates continuously, the GCM may ignore the error measurements when locked to GPS or may use them in combination with its internal GPS measurements.
The core idea is to connect an external reference at, say, 1 Hz, to a GCM I/O with timing functionality (as illustrated by the 1 Hz Ref I/O in the figure). This I/O can be used by the firmware to estimate very precisely, the period of the external frequency reference. Then, during holdover, this precisely calibrated period can be used by the GCM to maintain time and frequency. This would allow an OCXO, as in the figure, to be used to substantially enhance the holdover performance of the GCM itself. Alternatively, it would allow a frequency reference derived from, say, SDH (typically 8 kHz) or NTP to be used as a backup during holdover.
Enhancement 2 involves modifying the GCM firmware to support interfacing an external temperature sensor to an Analog Measuring Unit (AMU) input (small dots in the figure). The firmware would then be enhanced to temperature compensate the external oscillator using the same algorithms as the firmware uses to temperature compensate the GCM's own internal Temperature Compensated Crystal Oscillator (TCXO).
This would allow the holdover performance of the GCM to be improved substantially using a lower cost OCXO or even a more expensive TCXO than that used within the GCM itself.
Enhancement 3 would add the ability to discipline the external oscillator while the GCM is locked to GPS. This would allow the external oscillator to be used as a system clock in its own right. This would allow the GCM to be used to satisfy requirements for system clock frequencies that it did not directly support as well as providing enhanced holdover performance based on the stability of the external oscillator.
In principle this could be done using a Pulse Width Modulator (PWM) output of the GCM together with a few passive external components. Alternatively, it could involve the use of an internal Digital-to-Analog-Convertor (DAC) or an external DAC which would interface to the GCM most conveniently using a SPI port (fat square dots in the figure).
Finally, Enhancement 4 involves support for a 1 PPS timing signal interfaced to a second timing input (large dots in the figure). This would be used by the firmware to control the absolute time of the 1 PPS output during holdover or, in fact, as a primary timing reference, depending on the application requirements. In addition, provision would be made to switch between an 8 kHz reference input and the external oscillator under firmware control (via the second large dots I/O in the figure) to meet application requirements.
With some subset of the four enhancements described, a very low cost
GCM can be adapted to meet more demanding holdover requirements by utilizing external syntonisation and synchronization sources. With all 4 enhancements, a highly integrated GCM such as that used in a femto-cell could be adapted to meet all the requirements of a full macro base station clock very cost-effectively compared to conventional approaches.
As indicated in the diagram, master clock frequency errors are deduced from the input sequence of synch errors and both estimates (synch and frequency) are processed by the External Synch Controller. Note that this is feasible because the synchronization is strictly maintained through the oscillator discipline as discussed earlier. Because of that, the changes in synchronization error between any two points in time are proportional to the average clock frequency error between those two points in time.
In the preferred embodiment, either the External Synch Controller is or both synch controllers are supplied with oscillator crystal temperature measurements. These are used to improve oscillator discipline.
The use of temperature permits the oscillator drift to be compensated for by the controller especially during holdover when less precise frequency and synchronization error estimates are available. In the event that even these external measurements are not available, then the use of temperature measurements can extend the holdover period during which the GCM can maintain both the frequency and synchronization within specification.
The GPS time filter is a sequential Kalman filter that processes Time-Of-Transmission (TOT) measurements and Doppler frequency measurements from the available satellite signals to obtain estimates of absolute time error (with respect to GPS time) and master clock frequency error.
One characteristic of IP-based synchronization is that it suffers from time-varying biases. During GPS lock it is possible to estimate and track the time varying bias of the external synch error estimates.
During GPS holdover, the filter would maintain the bias and it would be eliminated from the time error estimate passed to the controller. In this way, improved accuracy would be maintained during all but prolonged holdover periods.
Alternative forms of algorithms can also be envisaged, including a separate bias tracker, estimators other than a Kalman filter or more sophisticated Kalman filters that track the bias drift rate as well as the bias itself.
Although the invention has been discussed in terms of specific embodiments, persons of skill in this art will see its full utility. Accordingly the invention is intended to cover what is described in the following claims:
Filing Document | Filing Date | Country | Kind | 371c Date |
---|---|---|---|---|
PCT/US09/42580 | 5/1/2009 | WO | 00 | 1/19/2011 |
Number | Date | Country | |
---|---|---|---|
61049497 | May 2008 | US |