The present application claims priority from European patent application EP06120455 filed on Sep. 11, 2006.
The current invention relates to a low-power management system addressing the problem of providing high stability frequency references despite temperature variations. This invention is meant to be integrated in GPS (Global Positioning System) devices which requires both clock precision for better performance and a long lifetime for commodity and ease of use in combination with other applications (e.g. so-called Assisted-GPS applications).
More specifically, the described invention relates to a method for temperature compensation in a clock oscillator using quartz crystals, which integrates dual crystal oscillators. The minimal power consumption is achieved through an efficient use of a processor in charge of the synchronisation of the two oscillators.
For devices requiring a highly accurate clock precision like GPS devices or satellite radiolocalization devices in general, effective temperature compensation must be achieved. To this end, TCXO (Temperature compensated Crystal Oscillator) are often employed. The cost and power consumption of available TCXO devices, however, are high. Thus, there is a need for an alternative way of providing a highly accurate clock which is less expensive and has lower power demands.
Devices providing an accurate time base are known, which use two thermally linked quartz crystals having different temperature coefficients. In such devices one crystal, cut in a manner that minimizes the temperature coefficient, is usually used as the “reference” oscillator while the other crystal is regarded as a “temperature oscillator”, in that it has a temperature coefficient which is quite linear with the temperature. In this way, the difference of the frequency generated by the two crystals univocally determines the common crystals' temperature, and can be used to correct the reference frequency, according to a known correction function.
An example of such quartz oscillators is known as CDXO (Calibrated Dual Crystal Oscillator). The appeal of such device reside in the fact that they comprise a memory which is factory programmed with an individual calibration function, coded for example as coefficients of a high order polynomial function, expressing the deviation of the “reference” oscillator as a function of the frequency difference between the two oscillators.
Usually the compensation for frequency deviation according to temperature is carried out by a Digital Signal Processor (DSP) that measures the frequency deviation of the “reference” oscillator based on the calibration function and on the inputs of the second oscillator. It communicates this frequency deviation to a numerically controlled oscillator (NCO) that numerically generates the reference frequency, for example 1 KHz for a 1 mS precise clock.
The correction function is, by nature, highly non-linear. These systems, therefore, may be difficult to integrate in portable GPS receiver, that require a very precise correction, in real-time, over a large temperature span, and in which the computing power is limited.
It is an object of the present invention to provide minimal processing power consumption while performing the compensation for the reference oscillator frequency drift depending on the temperature.
It is a further object of the invention to provide an optimised computer program carrying out the steps of the frequency compensation method.
The above objects are attained by employing an architecture using no temperature, but solely a frequency related calibration function that is stored in a memory means and assessed with a stepwise linear estimation in order to correct the reference oscillator frequency according to the temperature variation. This requires fewer cycles of the Digital Signal Processor (DSP) in charge of determining the temperature dependent frequency drift; and hence allows for a lower processing power to carry out the correction. This is a very useful feature for GPS devices that precisely need low power management characteristics.
The invention will be better understood with the drawing illustrated in
With reference to
The resonators 14 and 16 are contained and thermally linked in the same device 10. It can then be assumed that both crystal have a common temperature. The device 10 contains also a memory area 12, for example a EEPROM or a ROM, programmed with a correction function, for example the coefficients of a 5th order polynomial expressing the frequency deviation of the reference oscillator 14, as a function of the frequency difference between the reference oscillator 14 and the temperature sensitive oscillator 16. The calibration coefficients are stored in the device by the manufacturer. The content of the memory 12 is readable through an appropriate bus 13, for example a serial bus.
Optionally, if the memory area 12 should be user-writable (for example a flash memory), the calibration coefficients may be updated by the host system 40, for example to accommodate aging of the quartz crystals. This may be useful in radiolocalization devices in which the temperature calibration of the resonators 14 and 16 can be precisely verified, by comparison with the GPS time reference.
The two outputs of the two oscillators 14 and 16 are compared in the frequency comparator 80, in order to obtain a beat frequency which is temperature dependent, and whose characteristics with respect to temperature are stored in the memory means 12.
The radiolocalization receiver also comprises a GPS core, for performing known radiolocalization function together with a digital processor 40. According to the described example, the same digital processor 40 is also used to obtain a precise frequency reference from the dual crystal device 10. It will be understood, however, that the present invention also includes the case in which the frequency reference is provided by a separate processor, or by a dedicated circuit.
The signal generated by the reference oscillator 14 is used as time base for the numerically controlled oscillator 60 which delivers at its output a clock signal 65. An increment value 62, provided by the DSP unit 40 has to be set in order to have the desired frequency of the clock signal, typically 1 kHz in the case of a GPS unit.
The clock signal 65 is also applied to the frequency comparison circuit 80, which generates a control signal 85 indicative of the relationship between the frequencies of the two oscillators 14 and 16. It will be understood that the frequency comparison circuit 80 could operate in several ways, within the framework of the invention. For example the comparison circuit 80 may include a mixer and possibly a low-pass filter, in order to extract a beat frequency equal to the difference of the frequencies 14 and 16, and a counter, to count the beats of the two oscillators in one millisecond: In alternative, the comparison circuit 80 may simply count the output of the temperature oscillator 16 in a time interval determined by the clock signal 65, from which the beat frequency fan be derived by software in the DSP 40. We will assume, to fix the ideas, that the control signal 85 expresses the beat frequency between the reference oscillator 14 and the temperature oscillator 16, it being understood that the invention also comprises other variants in which the control signal carries, say, the ratio between the same frequencies.
Based on the control signal 85 the DSP 40, which as read beforehand the calibration coefficients from the memory 12 compute, periodically and in real-time, the increment value 62 for the NCO 60, in order to obtain a clock signal 65 having precisely the desired frequency, for all possible temperatures of the device 10.
Such a dual crystal system can provide a very fine resolution while employing an only frequency-related calibration function. Yet in order to achieve this with very frequent sampling times, the digital processor 40 can be quickly overloaded. It is therefore preferable to compute the instantaneous correction coefficients as an interpolation of the calibration polynomial. To this effect, for example, the DSP divides the expected range of the control signal 85 in a finite number of steps and fills in advance an interpolation table with the appropriate values to quickly calculate the interpolation function for each value of the control signal 85. The interpolation technique may be varied, according to the need. Preferably a piecewise linear interpolation is used.
In a preferred embodiment, a software program will carry out the described frequency control method with a scaling function, so that integers and no floating numbers will be manipulated, depending on the size of the registers used, in order to comply with the low power management policy of the invention in order to fit in with the requirements of GPS devices which are precisely characterised by low power characteristics.
In an alternative embodiment of the invention whose flowchart is represented in
In step 103 the DSP reads the beat frequency or the control signal 85 from the comparison unit 80, and in step 104, the DSP 40 updates the increment value 62 for the NCO 60 based on the control signal 85 and on the values pre-calculated in step 102. Steps 103 and 104 repeat as long as needed, for example according to a periodic interrupt of the DSP 40.
We will now work out, in further mathematical detail, an example of the present invention.
To obtain an accurate indication of the deviation of the Fref frequency generated from the reference oscillator 14 from its nominal value, a 5th order polynomial is applied:
ΔFref=C0+C1·ΔFbeat+C2·ΔFbeat2+C3·ΔFbeat3+C4·ΔFbeat4+C5·ΔFbeat5 (1)
Cn are the coefficients of the polynomial calibration function obtained from the EEPROM 12, opportunely adjusted and scaled, if needed. ΔFbeat is the deviation of the beat frequency 85 from the expected beat frequency FBeat
ΔFbeat=(Fbeat−FBeat
Gatems is the period in ms over which the measurements are made.
F
Beat
nom
=F
TempCount
nom
−F
RefCount
nom (3)
FTempCount
FXTemp is the nominal frequency of the XTemp oscillator (in Hz) at 25° C. Foffset is an offset frequency adjustment to XTemp obtained from the data in EEPROM 12. GateCount is the number of counts of XRef in the measurement period.
GateCount=Gatems·FXRef/1000 (5)
FXRef is the nominal frequency of the XRef oscillator (in Hz) at 25° C., but note that the term FXRef/1000 is actually the factor applied to obtain the 1 mS clock signal.
FRefCount
Hence restating equation 3
F
Beat
nom=GateCount·(FXTemp+Foffset)/FXRef−GateCount (7)
There are two options for FBeat depending whether the system is configured to measure a Beat frequency-count, or to measure the count from XTemp.
If the Beat frequency is being counted, then
Fbeat=MeasuredCount (8)
Otherwise, if the count from Xtemp is measured, then
F
beat=MeasuredCount−GateCount (9)
In the presented example, The NCO uses a 39-bit counter and a 28-bit increment register. It is clear, however, that other arrangements are possible.
F
req=(Incr·Fref)/239 (10)
Freq is the desired frequency
Fref is the input frequency
Incr is the Increment value.
By rearranging we obtain
Incr=(Freq·239)/Fref (11)
Initially, and by default the NCO will be programmed with an nominal Incrnom value obtained from the nominal frequency of XRef (XRef
When the control loop is active, periodically the system provides a frequency error for XRef. The NCO 60 must be adjusted to correct for any drift in the frequency of XRef. This will be done by applying Equation 11 where Fref is the nominal frequency of XRef, plus the change in frequency ΔFref obtained from Equation 1. Thus:
F
ref
=X
Ref
nom
+ΔF
ref (12)
To reduce the scale of the calculation required to obtain the updated Incr value, it is possible to scale the Incr value based on the ΔFRef value. Hence
Incr=Incrnom+(ΔFRef·Slope256/XRef
Incrnom is the Incr value calculated using the nominal frequency of XRef.
Slope256 is a pre-calculated Slope factor across the range of variation in XRef, scaled by 256.
Slope256 is calculated by the following equation:
Slope256=256·(IncrXrefmax−IncrXrefmin)/(Xref
Where
IncrXrefmax is the Incr value calculated using Xref
IncrXrefmin is the Incr value calculated using Xref
Xref
Xref
Number | Date | Country | Kind |
---|---|---|---|
06 120 455.8 | Sep 2006 | EP | regional |