The present invention generally relates to highly stable Inductor-Capacitor (LC) oscillators which utilize the LC tank temperature null phenomenon to minimize the variations of the oscillator output frequency.
Electronic clock generation classically relies on a reference oscillator based on an external crystal that is optionally multiplied and/or divided to generate the required clock. The key specifications of a clock, other than its target frequency, are frequency accuracy and stability. Frequency accuracy is the ability to maintain the target frequency across supply and temperature and is usually represented as drift from the target frequency in percent or parts per million (ppm). Long term stability, is impacted by the close-in phase noise of the oscillator. An oscillator using a high-Q element typically has a low phase noise profile, and thus good frequency stability, and is less sensitive to variations in oscillator amplifier gain, which is dependent on supply and temperature.
For example, crystal oscillators (XOs) are high-Q oscillators that provide excellent frequency stability and frequency accuracy across supply and temperature stemming from the very high quality factor (Q) of the crystal. However, not all resonators, including crystals, have satisfactory performance across temperature, thus the need for extra circuitry and techniques to decrease and/or compensate for shifts in frequency due to temperature. A temperature compensated crystal oscillator (TCXO) typically incorporates extra devices that have temperature dependence to negate the temperature dependence of the crystal. The overall outcome is an oscillation frequency with low temperature dependence.
However, the ever increasing complexity of electronic systems due to requirements of supporting multiple standards, increased functionality, higher data rates and increased memory in a smaller size and at a lower cost is pushing designers to increase the integration level through the development of Systems on Chip (SoC) in deep submicron Complimentary MOS (CMOS) technologies to benefit from the increased gate density. Reference clocks incorporating crystal oscillators have not managed to scale or integrate due to the bulky nature of crystals, thus limiting the size and cost reduction possible for electronic systems.
Recent efforts in using high-Q MEMS resonators and Film Bulk Acoustic Resonators (FBARs) have illustrated possibilities of integrating a high-Q element and Application Specific Integrated Circuits (ASIC) in the same package. However, packaging induced stress and its impact on performance still remains as a challenging obstacle, since the high-Q element may require special packages and/or calibration that are not practical for SoCs. The stress may change the temperature behavior of the resonator, possibly resulting in large frequency shifts and accelerated aging. Therefore, special assembly and packaging techniques are typically required to mitigate such effects, which increase the cost of producing such clocks. Similar problems may be encountered by any resonator that is dependent on the mechanical properties of the resonator material, which require careful design and manufacturing procedures and processes.
Design requirements for applications such as USB and SATA, which do not require superior frequency accuracy and stability, can be satisfied using oscillators with relatively low-Q elements available in a CMOS process which can have adequate phase noise profiles generating good jitter performance. Current trials include the use of ring oscillators, relaxation oscillators and LC oscillators. However, the reported frequency accuracy of these implementations suffers from large drift across supply and temperature, making them ineffective for applications requiring precise accuracy and stability. A mitigation to reduce the drift across temperature requires trimming across temperature which is neither cost effective nor practical for SoCs.
Therefore, an integrated solution that relies on existing optimized process steps in CMOS technology and that satisfies frequency stability and jitter requirements would be of great value. LC tank oscillators operating at the tank temperature-null phase to achieve highly stable output frequency have been described in U.S. Pat. No. 8,072,281, incorporated herein by reference. Techniques and circuits described herein include improvements and extensions that take advantage of the temperature-null phase.
Summary
The present invention provides a substantially temperature-independent LC-based oscillator. The oscillator includes an LC oscillator tank and frequency stabilizer circuitry coupled to the LC oscillator tank to cause the LC oscillator tank to operate at a temperature null phase generating a tank oscillation at a phase substantially equal to a temperature null phase. The temperature null phase is a phase of the LC oscillator tank at which variations in frequency of an output oscillation of the oscillator with temperature changes are reduced or minimized.
For example, the feedback loop may split the output voltage of the LC tank into two voltages having different phases, where each voltage is independently transformed into a current through programmable transconductors. The two currents may be combined to form a resultant current which is then applied to the LC tank. The phase of the resultant current is adjusted such that the LC tank operates at an impedance condition that achieves frequency stability across temperature.
The amplitude of the output voltage of the LC tank may be continuously monitored or sensed and the programmable transconductors may be adjusted substantially independently of each other. The amplitude sensing may be achieved by including an Automatic Amplitude Control (AAC) circuit in the feedback loop, in parallel with the two voltage transformation circuits.
In another aspect of the disclosure, each of the two voltages is filtered. In one aspect of the disclosure, the first voltage is low-pass filtered. In another aspect of the disclosure, the second voltage is high-pass filtered. In a further aspect of the disclosure, the path that includes the high pass filter includes a phase inversion circuit.
Description
Referring now to
The implementation of an ideal pure inductor or capacitor is usually not possible due to the physical limitations of having a finite quality factor Q. Integrated inductors in CMOS technology to date have low Q factors when compared to MEMS resonators and crystals. Sources of losses in an inductor include the inductor metal ohmic losses rL and substrate resistive losses rSUB. Both of these losses are usually temperature dependent, and therefore, the overall impedance and Q of the inductor is temperature dependent.
The integrated capacitive part of the tank also suffers from a finite temperature dependent Q as well as temperature dependence of the capacitance value. As a result, the physical implementation of an integrated LC-tank will dictate a strong temperature dependence of the impedance and Q factor of the tank, which results in a temperature dependent tank resonance frequency.
An oscillator built using an LC oscillator tank 10 typically includes an amplifier responsible for overcoming the tank losses. For the oscillator to have sustained oscillations, the Barkhausen criterion requires an open loop gain greater than 1 and phase equal to zero. Assuming that the amplifier contributes a zero phase, then for oscillation to occur, the LC oscillator tank impedance ZTank must have a zero phase. The phase condition is used to derive the oscillation frequency ωosc as follows:
An oscillation condition of φTank=0 results in:
From the above equations 1-3, it can be seen that the oscillation frequency is temperature dependent if rL is temperature dependent. A linear variation of rL with temperature results in an almost linear variation of the oscillation frequency. In addition, any temperature variation in C would strongly contribute to the temperature dependence.
This is graphically shown in
rL=ro(1+α(T−T0)) Equation 4
where α is a temperature coefficient of rL.
It is to be noted that the oscillation frequency is determined using the intersection of φTank=0 with the phase plots. The corresponding oscillation frequency across temperature is plotted in
Examining the phase plots again in
When the intersections occur at the same phase, a temperature insensitive tank operating point is created, and the tank is said to be operating at a temperature “null” with a phase φNull. The ideal temperature null phase occurs when the phase plots across temperature intersect at precisely the same phase. Oscillation with a phase across the tank ideally equal to φNull results in an oscillation frequency with zero deviation across temperature.
More realistic tanks exhibit a temperature null with small frequency deviations across temperature. This is illustrated graphically in
A Global Temperature Null (GNull) can be defined as a phase operating point φGNull that results in a minimum frequency deviation f across a temperature range T with a very small or zero change in oscillation frequency over temperature
at the center of the temperature range T0. A measure of the quality of the temperature null is the oscillation frequency deviation across temperature. A Figure of Merit (FOM) of the tank temperature null may be defined as:
where fT0 is the oscillation frequency at T0. The smaller the value of the FOM, the better the null quality is with the perfect null occurring at FOM=0.
A Local Temperature Null (LNull) can be defined as a phase operating point φLNull with
Alternatively, LNull can be defined at temperature T as the intersection of the phase plots of temperatures (T+δ) and (T−δ) where δ is infinitesimally small.
The GNull oscillation frequency ωGNull around temperature T0 may be derived by finding the intersection of two phase curves at temperatures T0+ΔT and T0−ΔT. For an LC oscillator tank with a linear temperature dependence of rL the phase and frequency at the GNull are as follows:
In order to force the tank to oscillate at a non-zero phase while satisfying the Barkhausen criterion, the active circuitry associated with the tank must provide the same phase as the tank but with an opposite sign. In other words, if the tank phase is −φ, then the active circuitry must provide a phase φ. One aspect of the disclosure of a temperature-stable oscillator is shown in
In another aspect of the disclosure, the phase φ introduced by the phase shifter block 210 is programmed by the digital word φControl so as to be as close as possible to the tank temperature null phase φGNULL.
Another aspect of the disclosure is shown in
In order to understand how the temperature-null phase φGNULL is obtained from the apparatus in
VT=Vo<0 Equation 7A
VH=AHVo<−θH Equation 7B
VL=ALVo<−θL Equation 7C
IH=−gmhAHVo<−θH Equation 7D
IL=−gmlALVo<−θL Equation 7E
IT=Io<φ Equation 7F
By direct derivation, the phase shift φ across the tank impedance (ZTank) between IT and VT is given by:
Hence, the phase φGNULL can be obtained accurately by adjusting the ratio between the two programmable transconductors 220a and 220b through the digital control words denoted in
In another aspect of the disclosure the phase shifter 210a of
As described in M. E. Van Valkenburg, “Analog Filter Design,” (Oxford University Press, 1995) in reference to
Here R and C are the resistance and the capacitance, respectively, used in both the LPF and HPF. Furthermore, ω is the angular frequency at which the LPF and HPF are operating. For the specific case of the oscillator, ω is the oscillation frequency.
Finally, the phase shift φ across the tank impedance (ZTank) between IT and VT is given by:
Hence, the phase φGNULL can be obtained accurately by adjusting the ratio between the two programmable transconductors gml and gmh through the digital control words denoted in
The product of R times C or the time constant RC must be substantially constant with temperature; otherwise, the frequency stability at the temperature-null phase is affected by temperature dependence of the RC time constant. Stabilizing the time constant RC versus temperature can be done using different techniques. A first example technique is obtaining R and C such that each of them is substantially stable versus temperature. A second example technique is to obtain R whose temperature dependence is substantially opposite to that of C and hence obtain a substantially stable product RC time constant.
As will be recognized by those skilled in the art, the innovative concepts described in the present application can be modified and varied over a wide range of applications. Accordingly, the scope of patents subject matter should not be limited to any of the specific exemplary teachings discussed, but is instead defined by the following claims.
Number | Name | Date | Kind |
---|---|---|---|
4918405 | Herleikson | Apr 1990 | A |
6606007 | Washburn | Aug 2003 | B1 |
20100045394 | Hanafi et al. | Feb 2010 | A1 |
Number | Date | Country | |
---|---|---|---|
20130314168 A1 | Nov 2013 | US |
Number | Date | Country | |
---|---|---|---|
61521669 | Aug 2011 | US |