Clock or timing signals are employed in electronic devices for many different purposes, including but not limited to clocking a microprocessor, a microcontroller, a central processing unit (CPU), a graphics processing unit (GPU), a digital signal processor (DSP), a digital logic state machine, etc., or clocking an analog-to-digital converter (ADC), a digital-to-analog converter (DAC), a sampled-data filter, a discrete-time analog signal processor, etc. The frequency of these timing signals ranges from the tens of kHz to hundreds of MHz. The power consumption requirement is in the range of one microwatt to ten milliwatts, while the typical frequency accuracy is in the range of ±10 to ±100 ppm.
Typical implementations of low-power frequency generators are classified as either harmonic oscillators or relaxation oscillators. The oscillation frequency of relaxation oscillators is set by an energy storage element (e.g. an inductor or a capacitor) and a resistor. A defining property of relaxation oscillators is that they produce waveforms with discontinuous first derivatives (i.e. pulse trains). The oscillation frequency of harmonic oscillators is set by a resonant tank, e.g. composed of two energy reservoirs, an inductive element and a capacitive element, two capacitive elements or two inductive elements. Quartz crystal oscillators are considered harmonic oscillators since a piezoelectric resonator is a resonant tank.
Quartz resonators have been the preferred resonator technology for the past 50 years for oscillators that require high accuracy and stability. Depending on their cut (i.e. crystal orientation), quartz resonators can operate on a fundamental frequency mode of oscillation, or on a harmonic or overtone mode. In addition, different quartz resonator cuts entail different stability characteristics. For example, the X-cut produces a less stable resonator than the popular AT-cut. Fundamental-mode oscillators, operating on the fundamental resonance mode of the quartz crystal, generate a fixed frequency in the range of approximately 1 MHz to approximately 50 MHz and achieve a long-term frequency accuracy better than 100 ppm. Of the various harmonic oscillators, the ones using piezoelectric resonators are known to have the best phase noise/jitter performance and frequency stability. However, such oscillators consume typically 1 mW to 5 mW (not including the power required to drive the load). Overtone-mode oscillators can generate output frequencies in the range of approximately 50 MHz to approximately 300 MHz, but also consume at least several mW.
One significant disadvantage of quartz resonators is their comparatively large size and their requirement of an expensive hermetic package (typically a leadless ceramic chip carrier or LCCC package, with a glass or metal seal). The size of such a quartz oscillator ranges from 35 mm2, which is too large for today's portable and many other applications, down to 1.9 mm2. But, as the size shrinks, cost rises and performance (such as frequency accuracy and jitter) degrades. In addition, quartz crystal resonators raise board layout considerations since they have to be placed next to the oscillator IC and also be shielded from noise and electromagnetic interference (EMI).
On miniaturized handheld devices, all large components (such as the above-described quartz resonators) need to be eliminated. However, the quartz resonators are not easily integrated with other components in a single package due to their size, their special hermetic packaging requirements and the performance issues described above. Complete chip integration is prohibitive since their size and process of manufacture are incompatible with semiconductor die fabrication.
Relaxation oscillators can achieve much lower power consumption than quartz oscillators. However, as frequency-determining devices, integrated inductors, capacitors and resistors offer very poor manufacturing accuracy and stability over temperature, which results in lower frequency stability, compared to piezoelectric resonators. Furthermore, parasitics can be significant especially for operational frequencies above approximately 500 MHz. Also, the use of resistors entails degraded phase noise and jitter because of their inherently smaller quality factor (Q). Thus, generally, relaxation oscillators tend to not be as stable as harmonic oscillators. Harmonic oscillators that do not use resistors as frequency determining elements can be more stable and feature lower jitter but require constant and stable biasing which in turn increases their power consumption. Also harmonic oscillator resonant tanks made out of inductors and capacitors require large silicon areas for operational frequencies lower than approximately 100 MHz, and this area requirement grows significantly as the resonance frequency is lowered.
To address the deficiencies of the above-described stand-alone oscillators, several approaches have been used to extend the higher frequency stability of a certain high quartz crystal oscillator to a lower stability oscillator operating at lower power. A high-accuracy oscillator (e.g. a quartz reference oscillator) periodically corrects the frequency of a low-power, low-accuracy oscillator. The reference oscillator is typically a quartz oscillator. Since the frequency-corrected oscillator consumes a fraction of the power of the reference oscillator, significant power savings can be obtained in applications such as battery-powered horology.
A typical frequency correction for these devices utilizes a means of comparing the phase or frequency of the reference oscillator with the phase or frequency of the low-power oscillator, and either: 1) directly correcting the frequency of the low-power oscillator, or 2) correcting the output frequency by changing the division ratio of the output frequency divider.
Although the periodic frequency correction approach described above is utilized in applications such as Real-Time-Clocking (RTC) frequency generators, typically operating at 32.768 kHz, the benefits of this method diminish as the output frequency increases into the MHz range or higher. Since a quartz oscillator must be used to achieve high stability, the periodic frequency correction approach offers no cost or performance advantages. In addition, the usable frequency of typical reference AT-cut quartz oscillators is limited to less than approximately 50 MHz for a fundamental mode and approximately 150 MHz for a 3rd overtone, which introduces limitations in the power consumption, frequency division and correction mechanisms.
Furthermore, the required startup and stabilization time of MHz-range quartz oscillators is approximately in the millisecond range, which greatly increases the energy expended during the calibration process. It is thus desired to reduce this power and time as much as possible.
An additional shortcoming of the existing approaches is that the correction range is limited by the reference oscillator frequency. If several outputs are desired, output frequencies will be concentrated around a particular frequency value as allowed by minimum divider ranges and careful frequency planning.
Accordingly, new frequency generator circuit architectures are required that are compact and integratable, that can consume very low power, that provide frequency-stable operation, that are flexible, and that can be manufactured at a low cost. In addition, it is desired to integrate a high performance resonator on-chip with other circuits including a lower frequency oscillator operating at lower power.
Embodiments of the disclosure can be implemented in numerous ways, including as an apparatus and/or a system.
A detailed description of one or more embodiments of the invention is provided below along with accompanying figures that illustrate the principles of the invention. The invention is described in connection with such embodiments, but the invention is not limited to any embodiment. The scope of the invention is limited only by the claims, and the invention encompasses numerous alternatives, modifications, and equivalents. Numerous specific details are set forth in the following description in order to provide a thorough understanding of the invention. These details are provided for the purpose of example, and the invention may be practiced according to the claims without some or all of these specific details. For the purpose of clarity, technical material that is known in the technical fields related to the invention has not been described in detail so that the invention is not unnecessarily obscured.
An oscillator that comprises an integrated Bulk Acoustic Wave (BAW) resonator operating approximately in the 500 MHz-to-5 GHz range, combined with a fractional frequency divider or a direct digital synthesizer (DDS), one or more low-power oscillators, and logic circuitry to determine and apply frequency adjustment to the low-power oscillators is described herein. The adjustment can be done either via a fractional frequency divider or a direct digital synthesizer, or directly into frequency tuning elements of the low-power oscillators. Besides the low power operation, the elimination of the external quartz crystal offers significant advantages over existing solutions.
An integrated oscillator using an in-package BAW resonator, requiring no external quartz crystal, is used as a master frequency reference (or master oscillator, or reference oscillator). The oscillation frequency can be in the range of approximately 500 MHz to approximately 5000 MHz. The output of the master oscillator is processed by a direct digital synthesizer (DDS) which can include a fractional frequency divider and a phase interpolator, and delivers a reference frequency to a frequency/phase comparator, which compares it with a signal whose frequency is derived from the slave oscillator. The DDS in combination with BAW oscillators, such as demonstrated in U.S. Pat. No. 8,242,850, can be utilized. U.S. Pat. No. 8,242,850 is incorporated by reference herein. The frequency adjustment can be directly applied to the slave oscillator or indirectly applied to a secondary DDS. In one aspect, an integrated circuit device for generating an output signal includes a master oscillator circuit, a direct digital synthesizer circuit, and a slave oscillator circuit formed over an integrated circuit substrate. A reference frequency signal is generated from the master oscillator circuit by the direct digital synthesizer circuit. The output signal of the integrated circuit device is generated from the slave oscillator circuit. A frequency of the output signal of the integrated circuit device is intermittently adjusted with respect to a frequency of the reference frequency signal. The master oscillator circuit is intermittently enabled to adjust the frequency of the slave oscillator circuit output signal. The master oscillator circuit comprises a bulk acoustic wave resonator.
In one aspect, an integrated circuit device for generating an output signal includes a master oscillator circuit and a slave oscillator circuit. The integrated circuit device is formed over an integrated circuit substrate. A reference frequency signal is generated from the master oscillator circuit. The output signal of the integrated circuit device is generated from the slave oscillator circuit. The master oscillator circuit is intermittently enabled. As used herein, “intermittently enabled,” includes both periodically and non-periodically enabled. A frequency of the output signal of the integrated circuit device is intermittently adjusted with respect to a frequency of the reference frequency signal. As used herein, “intermittently adjusted” includes both periodically adjusted and non-periodically adjusted. The master oscillator circuit includes a bulk acoustic wave resonator.
The foregoing and other objects and advantages of the invention will be appreciated more fully from the following further description thereof, with reference to the accompanying drawings.
The present invention is described in terms of several embodiments. In describing these embodiments, for the sake of clarity and compactness throughout the supporting figures, elements with identical or similar functionality have been assigned the same numbers.
In the discussion of certain mathematical relationships, certain variable values are discussed by way of illustration. However, the invention is not intended to be limited to the specific embodiments described below.
Recent advances in integrated piezoelectric resonator technology have brought about integrated Bulk Acoustic Wave (BAW) resonators. BAW resonators are fabricated on integrated circuit (IC) wafers or other wafers or substrate using monolithic deposition methods and can be quite compact. Aluminum Nitride (AlN), Zinc Oxide (ZnO) or other piezoelectric film arrangements can be deposited with electrode metals and form piezoelectric resonators that may produce frequencies in the general range of approximately 500 MHz to approximately 5000 MHz, while occupying a minimal die area, e.g. 200 μm×200 μm. Integrated BAW resonators are well known to those skilled in the art and are not described in detail herein. Exemplary integrated BAW resonators are disclosed in U.S. Pat. Nos. 8,089,195 and 8,222,795, both of which are incorporated herein by reference.
An integrated BAW resonator may be assembled with a semiconductor die in a low-cost small-size plastic IC package. In contrast, a quartz resonator must operate in a sealed cavity, which requires an expensive and larger hermetic package such as a ceramic package.
The monolithic integration of BAW resonators with silicon or other monolithic semiconductor devices can be accomplished in a variety of ways, as is well known to those skilled in the art. Such monolithically integrated assemblies provide an array of advantages such as being placed in a single package.
For example, certain Chip-Scale-Packaging (CSP) approaches are illustrated in
A Multi-Chip-Module (MCM) arrangement is depicted in
During operation of the integrated frequency generator 30, the slave oscillator circuit 35 (having relatively low power consumption and poor stability, compared to the master oscillator circuit 34) generates the output signal 51 that may be frequency-divided (see
The master oscillator circuit 34 is the main frequency reference and may be constructed as a monolithic oscillator with an output frequency in the range of approximately 100 MHz to approximately 10 GHz. The frequency of the master oscillator circuit 34 is primarily determined by the integrated BAW resonator 38. Integrated BAW resonators may be fabricated on silicon or other wafer types using monolithic fabrication methods and are much smaller than quartz resonators.
As described above, integrated BAW resonators lend themselves to complete integration either within an integrated circuit package (i.e. multi-die co-packaging) or on the same die (i.e. complete monolithic integration). This is a significant advantage over prior-art solutions that require a quartz crystal or similar resonator in an expensive package.
Another significant advantage of master oscillator circuit 34 is that it has a very fast turn-on transition time, ranging from a few microseconds to tens of microseconds, because it comprises integrated BAW resonator 38 as a resonant tank. This feature of the master oscillator circuit 34 is leveraged in the intermittent adjustment of the slave oscillator circuit 35 described below.
Another advantage of the integrated reference frequency generators disclosed herein is that they can be integrated monolithically or co-packaged with sensors that comprise a BAW resonator. Example devices that can be integrated with the integrated reference frequency generator include pressure sensors, acoustic microphones, acceleration or vibration sensors, physical, chemical and biological sensors, as well as electronic devices that comprise a BAW resonator, such as RF filters, RF transmitters, RF receivers and RF other circuits, etc.
The output frequency of the slave oscillator circuit 35 is in the range of approximately 1 kHz to a few hundred kHz (e.g. for applications real-time clock applications) and from approximately 1 MHz to a few hundred of MHz (e.g. for CPU clocking applications). Since the slave oscillator circuit 35 operates continuously, the slave oscillator circuit 35 is preferably designed to have a low power consumption. As a consequence, the slave oscillator circuit 35 features poor absolute and temperature-dependent frequency stability when it operates in free-run mode. To correct this excessive frequency drift, the slave oscillator circuit 35 is periodically adjusted by tuning one or more of its frequency determining elements, based on the frequency adjustment signal 39. The frequency adjustment signal 39 is determined based on a comparison of the slave oscillator output signal 51 with the master reference frequency signal 40, as described below.
The master oscillator circuit 34 features superior frequency stability compared to the slave oscillator circuit 35. The DDS-M circuit 32, e.g. similar to the one described in U.S. Pat. No. 8,242,850, processes the master oscillator output signal 50 to synthesize a master reference frequency signal 40 that is output from the DDS-M circuit 32. Depending on requirements, the DDS-M circuit 32 may divide the frequency of the master oscillator circuit 34 output signal 50 by fractional division techniques or integer division techniques. Optionally, the DDS-M circuit 32 may use phase interpolation techniques to achieve a lower output jitter. The master reference frequency signal 40 and the slave oscillator output signal 51 are applied to the frequency adjustment circuit 33 which produces the frequency adjustment signal 39. The frequency adjustment signal is applied directly to the slave oscillator circuit 35 and changes the frequency or phase of the slave oscillator output signal 51.
The temperature compensator circuit 36 senses the operating temperature of the slave oscillator circuit 35, and may independently adjust the slave oscillator frequency by adjusting the value(s) of the slave oscillator's frequency determining element(s). The operation of the temperature compensator circuit 36 may be dependent of or independent of the scheduler/controller circuit 37.
Note that, for the sake of conciseness, in parts of this document where clarity is not affected, a signal and the frequency of that signal are referred to interchangeably.
The output signal 52 of the DDS-S circuit 45 is compared in the frequency adjustment circuit 33 with the master reference frequency signal 40 output from the DDS-M circuit 32. The output frequency adjustment is accomplished by changing the divisor (or division ratio or modulus) of the DDS-S circuit 45, rather than by directly altering the frequency of the slave oscillator circuit 35 as described with reference to
The embodiment illustrated in
The master oscillator circuit 34 operates in the approximate range of approximately 500 MHz to approximately 5000 MHz and consumes more power than the slave oscillator circuit 35. On the other hand, the frequency stability of the master oscillator circuit 34 is much better than that of the slave oscillator circuit 35. Thus, operating the master oscillator circuit 34 intermittently conserves power and with careful design can ensure that the slave oscillator frequency stability remains within specified bounds.
The intermittent operation of the master oscillator circuit 34, the DDS-M circuit 32 and the frequency adjustment circuit 33 results in lower power consumption over the case where the adjustment is continuously applied to the slave oscillator circuit 35. The total average power is given as follows:
P
0avg
=P
SO
+P
CORR Eq. 1
where PSO is the average power of slave oscillator circuit 35, and the scheduler/controller circuit 37, and PCORR is the average power of the adjustment circuitry, including the master oscillator circuit 34, the DDS-M circuit 32, and the frequency adjustment circuit 33. As mentioned above, the master oscillator circuit 34, which includes integrated BAW resonator 38, has a rapid turn-on time which may significantly reduce the additional average power consumed by the adjustment circuitry.
Battery-operated and real-time clock (RTC) devices may require frequency generation at extremely low power (e.g. about 25 microwatts) while maintaining high frequency accuracy (e.g. 10 ppm). For example, a relaxation oscillator used as the slave oscillator circuit 35 operates at the typical RTC frequency of 32.768 kHz with a combination of a resistor and a capacitor as frequency-determining elements. Such a slave oscillator draws 5 μA from a 2.7V supply, i.e. consumes PSO=13.5 uW, but does not have the required frequency stability and requires periodic adjustment. In addition, a master oscillator circuit 34 that includes a 2 GHz integrated BAW resonator 38 may operate at 2.7V and draw 2 mA, i.e. consume 5.4 mW, when enabled. To produce a reference signal at 32.768 kHz, the DDS-M circuit 32 divides the frequency of the 2 GHz signal by 61035.15625. When enabled, the DDS-M circuit 32 also draws 2.4 mA at 2.7V, i.e. consumes 6.5 mW. The frequency adjustment circuit 33 consumes an additional 0.1 mW when enabled. The operational duty cycle of the frequency adjustment circuit 33 along with the master oscillator circuit 34 and the DDS-M circuit 32 determines the additional power consumption above the baseline of 13.5 μW. So if, for example, the adjustment happens once a second and is completed within 1 msec, then the duty cycle of 0.1% results in an additional PCORR=(5.4 mW+6.5 mW+0.1 mW)×0.001=12 μW. This results in a total average power consumption of P0avg=13.5 μW+12 μW=25.5 μW.
An additional benefit of the intermittent operation of the adjustment circuitry is that the aging of the master oscillator circuit 34 and the integrated BAW resonator 38 is slowed down because these devices are only enabled during a fraction of the operating time of integrated reference frequency generator 30. Such low duty-cycle operation not only enhances the frequency stability of the master oscillator circuit 34, but also can significantly extend the serviceability of the integrated reference frequency generator 30.
Another significant advantage of the embodiments illustrated in
ΔfOUT(t)=ΔfOUT/Δt*[t−tadj] Eq. 2
where ΔfOUT/Δt is the frequency drift slope (assuming linear deviation with time), t is a given time instant, and tadj is the time instant of the last adjustment of the slave oscillation frequency. The estimated instantaneous frequency deviation ΔfOUT(t) may then be converted to the frequency adjustment signal 39, which is applied directly to the slave oscillator circuit 35.
In the embodiment shown in
Based on the above relationship, for a given maximum known frequency drift slope and a desired maximum frequency deviation requirement, the frequency adjustment interval Δtadj is given as follows:
Δtadj≦1/max[ΔfOUT/Δt]*ΔfOUTmax Eq. 3
where max [ΔfOUT/Δt] is the maximum expected frequency drift slope of the unadjusted slave oscillator circuit 35, and ΔfOUTmax is the specified maximum frequency deviation.
The scheduler/controller circuit 37 determines the time instants at which adjustments are performed by controlling the operation of the master oscillator circuit 34, the DDS-M circuit 32 and the frequency adjustment circuit 33. There are several possible regimes under which the scheduler/controller circuit 37 can operate. For example, the frequency adjustment process can be triggered by: i) a change of operating or ambient temperature, ii) a predetermined elapsed time, iii) user control, or iv) a combination of any of the above and other criteria.
In one embodiment, given a maximum possible frequency deviation slope and a maximum frequency deviation specification, the scheduler/controller circuit 37 is triggered by a timer 96 (shown in
N≦(fOUT+ΔfOUTmax)*ΔfOUTmax/max[ΔfOUT/Δt] Eq. 4
Variations of this embodiment are illustrated in
In another embodiment described with reference to
As stated above, the free-running slave oscillator instantaneous frequency 80 drifts linearly over time. The output instantaneous frequency 81 is periodically adjusted to a nominal output frequency at times designated by pulses in the frequency adjustment enable signal 82. The output instantaneous frequency 81 is adjusted according to the observed difference between the frequency of the master reference frequency signal 40 and the frequency of the output signal 52 of the DDS-S circuit 45 by the frequency adjustment circuit 33 via the DDS-S circuit 45. Immediately after power-up, while the drift compensation is not active, the instantaneous output frequency 81a follows the free-running slave oscillator instantaneous frequency 80. Then, the first frequency adjustment is performed, and the instantaneous output frequency 81 is adjusted to its nominal value. Then, the free-running slave oscillator instantaneous frequency 80 continues to drift linearly, and the output instantaneous frequency 81b drifts again from the nominal frequency value. The frequency adjustment process is repeated periodically. The maximum frequency deviation is determined by the adjustment interval and the slave oscillator circuit 35 output signal 51 frequency drift slope.
The instantaneous supply current drawn by the integrated reference frequency generator 30 is also graphed in
As stated above, the free-running slave oscillator instantaneous frequency 91 drifts in a random walk fashion over time. The frequency adjustment process is periodic with an adjustment interval designated by pulses in the frequency adjustment enable signal 82. Immediately after power-up, while the drift compensation is not active, the instantaneous output frequency 83a follows the free-running slave oscillator instantaneous frequency 91. Then, the first frequency adjustment is performed, and the output instantaneous frequency 83 is adjusted to the nominal output frequency. Then, the instantaneous output frequency 83b drifts again in a random walk fashion until the adjustment process is repeated.
The supply current plots show that the average current draw 86 (Iave) is slightly higher than the slave oscillator supply current 84 (IFR) but much lower than the peak of current 85 (Iadj) drawn during the frequency adjustment process.
According to an aspect of the invention illustrated on
As stated above, the free-running slave oscillator instantaneous frequency 80 drifts linearly over time. The frequency adjustment is periodic with an adjustment interval designated by pulses in the frequency adjustment enable signal 82. The difference between this scheme and the schemes illustrated in
At power up, as the drift compensation is not active, the output instantaneous frequency 87a follows the free-running slave oscillator instantaneous frequency 80. Then, the first frequency adjustment is performed, and the output instantaneous frequency 87 is adjusted to its nominal value and the computed forward frequency drift slope estimate 88 is forward-applied via a frequency adjustment mechanism which can be implemented, for example, in the frequency adjustment circuit 33 or the temperature compensator circuit 36 or both. Then, the instantaneous output frequency 87b drifts again in linear fashion until the frequency adjustment process is repeated. The power consumption is graphed as in
The forward frequency adjustment method can achieve better frequency stability over short adjustment intervals compared to simpler adjustment methods.
At power up, the output instantaneous frequency 89 follows the free-running slave oscillator instantaneous frequency as the drift compensation is not active. Then, the first frequency adjustment is performed, and the output instantaneous frequency 89 is adjusted to its nominal value and the forward frequency drift slope estimate 88 is computed (in this example, the value of the forward frequency drift slope estimate computed during the first frequency adjustment is 88b) and is used to forward adjust the output instantaneous frequency 89 until the next adjustment operation. Then, the instantaneous output frequency 89b drifts again in a random walk fashion until the frequency adjustment process is repeated. For comparison, purpose, the third graph from the top shows the output instantaneous frequency 89 obtained with feed-forward drift adjustment and the output instantaneous frequency 83 (shown in dashed lines) obtained without feed-forward drift adjustment as illustrated in
Referring back to
To facilitate a detailed discussion of the frequency adjustment capabilities of the invention, the embodiments of
The frequency of the output signal 50 of the master oscillator circuit 34 and the frequency of the master reference frequency signal 40 are measured, and calibrated if needed, during or after manufacturing, and the relevant parameters are stored in the memory 102 of the DDS-M circuit 32. The relationship of the master frequency drift to the ambient temperature variation can be captured and modeled by a piece-wise linear function or other similar function of low circuit complexity, or any other function. The temperature compensator circuit 31 is enabled by the scheduler/controller circuit 37 during the frequency adjustment window and outputs a digitized frequency adjustment value to the DDS-M circuit 32 so that the frequency of the master reference frequency signal 40 is intermittently adjusted.
According to a different aspect of the invention, the temperature compensator circuit 31 is operated intermittently and enables the scheduler/controller circuit 37 when the measured temperature exceeds a previously measured temperature by a certain value that requires immediate frequency adjustment.
The scheduler/controller circuit 37 controls the operation of the master oscillator circuit 34, the temperature compensator circuit 31 and the frequency adjustment circuit 33. The timer 96, which may be implemented with a large-ratio frequency divider, is clocked by the slave oscillator output signal 51 and signals the scheduler/controller circuit 37. The scheduler/controller circuit 37 periodically enables the frequency adjustment circuit 33 along with the master oscillator circuit 34 and the DDS-M circuit 32. The scheduler/controller circuit 37 also enables the temperature compensator circuit 31 in the master oscillator path to introduce additional frequency adjustment to the DDS-M circuit 32 and the temperature compensator circuit 36 in the slave oscillator path. The scheduler/controller circuit 37 enables the frequency adjustment process (implemented by the frequency adjustment circuit 33 and other circuits as required) intermittently, periodically or not periodically, with a schedule that can be deterministic or pseudo-random or random, or alternate between deterministic, pseudo-random and random periods of operation. The scheduler/controller circuit 37 can be triggered in various ways, e.g. through the timer circuit 96 driven by the slave oscillator circuit 35 output signal 51, a threshold-crossing of the temperature compensator circuit 31, an external stimulus signal, etc.
The frequency of the slave oscillator output signal 51 may be temperature compensated in multiple ways. In one embodiment, the temperature compensator circuit 36 directly applies a frequency adjustment signal to a tuning element 95 of the slave oscillator circuit 35. In the embodiments shown in
The DDS-S circuit 45 in the embodiment presented in
In addition, the frequency modulation circuit 109 can alter the output signal 53 using dithering and related techniques, for example, to suppress spurious tones that may otherwise be generated by the DDS-S circuit 45. Dithering techniques include continuous-time and discrete-time modulation by a random or a pseudo-random signal, temporally shifting the edges of clock signals with random or pseudo-random jitter, etc.
Further, the frequency modulation circuit 109 can adjust the output signal 53 to correct deterministic errors that affect output signal 53. For example, the frequency modulation circuit 109 can measure the value of the power supply voltage applied to the integrated reference frequency generator 30 (or applied to some of its components), and it can adjust the output signal 53 to cancel frequency errors induced by variations of the supply voltage, based on a model or describing function of the sensitivity of the frequency of the output signal 53 to supply voltage variations.
An additional distinction in the embodiments of
In the embodiment of
In the embodiments of
In the embodiment of
In the embodiment of
In the embodiments described above, the rate translation circuit 99 may implement linear translation, or alternate linear or nonlinear functions, that facilitate the adjustment and control of the output frequency of the integrated reference frequency generator 30. Optionally, for a higher comparison rate, the frequency divider circuit 97 may be bypassed.
Referring to
The configuration of the integrated reference frequency generator 30 of
The master oscillator circuit 34 comprises the integrated BAW resonator 38 and drives the DDS-M circuit 32 which in turn outputs the master reference frequency signal 40 that is input to the frequency adjustment circuit 33. This reference frequency chain is controlled by the scheduler/controller circuit 37 which ensures that each of the slave oscillator circuits 35a, 35b, 35n is adjusted at the appropriate time interval. The frequency adjustment process is the same as described in the embodiment of
The temperature-dependent frequency drift of the master oscillator circuit 34 output signal 50 can be adjusted indirectly by the temperature compensator circuit 31 that periodically adjusts the divisor of the DDS-M circuit 32. The temperature-dependent frequency drift of the slave oscillator circuits 35a, 35b, 35n output signals 51a, 51b, 51n can be compensated through direct frequency adjustment by the temperature compensator circuits 36a, 36b, 36n. The advantage of having multiple slave oscillator circuits 35a, 35b, 35n is that a number of unrelated output frequencies can be generated with low power consumption, high frequency stability characteristics and a completely integrated reference frequency source.
The intermediate master reference frequency signals 40a, 40b, 40n are processed by a reference frequency selector 160 to generate the master reference frequency signal 40. For example, the reference frequency selector 160 may compare all intermediate master reference frequency signals 40a, 40b, 40n against each other to determine which intermediate master reference frequency has drifted the least, e.g. due to aging or other reasons. Once this is determined, the selected intermediate master reference signal is used as the master reference frequency signal 40 and output to the slave oscillator adjustment circuit, as described above.
The circuit architecture described with reference to
In an alternate embodiment related to the one illustrated on
As these and other variations and combinations of the features described above can be utilized, the foregoing description of the preferred embodiments should be taken as illustrating, rather than limiting, the scope of the invention.