CROSS-REFERENCE TO RELATED APPLICATIONS

FIELD OF THE INVENTION

[0001]The present invention relates generally to multiphase oscillators and more specifically to active devices that maintain the oscillation of such oscillators.

DESCRIPTION OF THE RELATED ART

[0002]Most electronic systems encounter noise from several sources, such as shot noise, thermal noise, and flicker noise, which usually arise from currents flowing in the electronic system.

[0003]Most electronic systems include one or more oscillators, which provide the clocks that pace or operate the system. Because oscillators are linear, time-varying (LTV) systems, electronic noise has unusual effects on oscillators. One way to characterize these effects is to determine the LTV unit impulse response Κ _φ for the oscillator. This function h _< p includes a key function called the impulse sensitivity function (ISF) Γ(ω ₀ τ), which captures the essential information about the oscillator and allows one to compute the oscillator's response to a phase impulse. The ISF for an oscillator shows that the various noise sources mentioned above give rise to phase 0(t) modulation cos[a> ₀ t -I- <j)(t)] of the oscillator frequency ω ₀ , which appears as a spectral "skirt" about the oscillator frequency ω ₀ . This spectrum, actually a power spectral density £- _ν1ια8β { ίώ), is usually separated into three regions. The region nearest the oscillator frequency is the 1/f ³ region. At some point ω ^ (the 1/f noise corner), this region breaks into a 1/f ² region, which continues until it reaches the noise floor of the system, which is a flat region at frequencies far distant from the oscillator frequency.

density of the noise, r _rms is rms value of the ISF, and Δω is the offset frequency from the oscillator nominal frequency. The PSD for an oscillator in the 1/f ³ region is L _phase ^) = power of the sidebands about the oscillator frequency indicate that there are both up and down frequency translations of circuit noise currents i _n into noise near the oscillator frequency. The Iff ² region results from down-converted white noise at integer multiples of the oscillator frequency and the 1/f ³ region results from up-converted 1/f circuit noise. The 1/f noise is particularly troublesome for some applications, but the presence of the c ₀ term (the dc term) in the PSD function suggests that to minimize this noise, one should make the term as small as possible. One way of doing this is to arrange the circuitry so that disturbances during the rising edge in the oscillation are the same as those during the falling edge, thereby making the net disturbances over the cycle almost cancel out.

[0005]Not only does the oscillator itself have a time-varying characteristic, but the oscillator can be subject to time- varying noise sources that vary in a periodic fashion. In fact, in a full analysis, another function called a noise modulating function (NMF) α(ω _α €) modifies the ISF to create an effective ISF (ISF _e ) that accounts for periodic time-varying noise. As with the other noise sources, one should attempt to find circuit topologies that minimize the effect of time-varying noise. One way to minimize the noise is to arrange circuitry so that time- varying noise sources have their maximum power at the minimum sensitivity point in the oscillator waveform.

[0006]Rotary wave oscillators are a particular type of oscillator, operating by means of a wave that travels in a closed loop that reverses the phase of the wave on every transit of the loop. FIG. 1 illustrates this type of oscillator. The direction of the traveling wave around the closed loop can be either clockwise or counter clockwise unless some technique sets the direction of rotation. Such oscillators have the advantage that multiple phases are available by physically tapping the oscillator. However, for the taps to be useful, the one must know the direction of the wave along the taps. It is thus desirable to have control over the direction of the traveling wave.

[0007]In summary, it is desirable to have an oscillator with low phase noise to make it useful in a great variety of applications and it is desirable, in the case of a rotary wave oscillator, to control the direction of travel of the wave.

BRIEF SUMMARY OF THE INVENTION

[0008]An embodiment of the present invention is an amplifier for generating and maintaining a traveling wave on a RTWO. The amplifier is connected to the phases of the RTWO such that it turns off at the critical time when noise would otherwise be injected into the oscillator.

[0009]One of the advantages of the present invention is that it eliminates crowbar current. In particular, the dwell time of the amplifier eliminates crowbar current, which flows from supply to ground, thereby contributing noise power but not energy to the oscillator.

[0010]Another advantage is that it enforces the wave direction of oscillator of a rotary traveling wave oscillator.

[0011]Yet another advantage is that, with proper location of the amplifier phase taps, it reduces the sensitivity of an RTWO due to process variations.

[0012]The present invention thus reduces phase noise and provides a means for controlling the direction of the traveling wave in the oscillator.

BRIEF DESCRIPTION OF THE DRAWINGS

[0013]These and other features, aspects and advantages of the present invention will become better understood with regard to the following description, appended claims, and accompanying drawings where:

FIG. 1 shows a prior art rotary oscillator;

FIG. 2A shows an amplifier for use in an embodiment of the present invention in which the amplifier is the negative resistance device;

FIG. 2B shows an alternate version of the amplifier for use in an embodiment of the present invention;

FIG. 3 shows a portion of the transmission line on which a wave travels from left to right in the figure;

FIG. 4A shows a timing diagram of the various waveforms involved in the present invention;

FIG. 4B, the inset in FIG. 4A, illustrates the negative-going wave front and two times Tl and T2 at which the amplifier output impacts the wave front;

FIG. 5A shows an equivalent circuit for the amplifier of FIG. 2A when a wave travels in the preferred direction; and

FIG. 5B shows an equivalent circuit for the amplifier of FIG. 2A when a wave travels opposite to the preferred direction. DETAILED DESCRIPTION OF THE INVENTION

[0014]FIG. 1 shows a prior art rotary oscillator 10. The oscillator 10 includes a pair of conductors 14a,b connected to form a closed loop 12, an odd number of cross-overs 18 that reverse the phase of a wave traveling on the conductors 14a,b, and a plurality of regeneration devices 16 (of which 20a,b is one of many implementations) that create and maintain the traveling wave. Each of the regeneration devices 16 has negative resistance and operates to add energy to the traveling wave as it passes a device. For example, if the wave is traveling clockwise with a positive voltage on conductor 14b and has not reached 20a,b, then the regeneration device 20a,b has a polarity opposite to the approaching wave. As the wave passes the regeneration device 20a,b, the wave changes the state of the device 20a,b so that it has the same polarity as the wave, thereby adding energy to and reinforcing the wave.

[0015]Any negative resistance device that operates to add energy to the traveling wave will meet the requirements of a regeneration device for the rotary oscillator.

[0016]FIG. 2A shows an amplifier 40 for use in an embodiment of the present invention in which the amplifier is the negative resistance device. In the version shown in FIG. 2A, the amplifier 40 includes first 42 and second 44 p-channel transistors and first 46 and second 48 n- channel transistors. The p-channel transistors 42, 44 have their channels connected in series and between a first supply voltage VI 50 and the output out 52. The n-channel transistors 46,48 have their channels connected in series between a second supply voltage V254 and the output out 52. The gates of the second p-channel transistor 44 and the first n-channel transistor 46 are connected to a first input, inl 56. The gates of the first p-channel transistor 42 and the second n- channel transistor 48 are connected to a second input, in2 58.

[0017]FIG. 2B shows an alternate version 60, 62 of the amplifier for use in an embodiment of the present invention. In the alternate version, the amplifier 62 at one position on the loop comprises two p-channel transistors 64, 66 whose channels connected in series between the first supply voltage VI 50 and the output 52. The gate of one transistor is connected to the first input inl 56 and the gate of the other transistor is connected to the second input in2 58. At an adjacent position on the loop, up or downstream, the amplifier comprises two n-channel transistors 68, 70 whose channels are connected in series between the output 52' and the second supply voltage V2 54. The gate of one transistor is connected to the first input inl 56'and the gate of the other transistor is connected to the second input in2 58'. These two configurations thus alternate at each position on the loop, so that the traveling wave first encounters, say a p- channel amplifier 60, then an n-channel amplifier 62, and then a p-channel amplifier 60. When the wave encounters a p-channel amplifier 60 it receives a boost that increases the wave's higher potential. When the wave encounters an n-channel amplifier 62, it receives a boost that increases the waves' lower potential.

Operation of Figure 2A [0018]FIG. 3 shows a portion of the transmission line 14a, 14b (in FIG. 1), on which a wave travels from left to right in the figure. Amplifier 40 has input 1 and input 2 connected to different positions Ph _j 84, Ph _k 82 on conductor 14b and its output connected to a position

Ph ₀ 86 on conductor 14a. The amplifier embodiment 40 in FIG. 2A operates as follows, adopting the convention that positive j and k values lag the wave front, while negative values lead it. If the wave front of a wave traveling in the direction from j to k is at Ph ₀ , then the wave front arrives at Ph _j before it arrives at Ph _k . Before the wave arrives at Ph _j or Ph _k on the RTWO, input 1 and input 2 are relative low voltages compared to their complementary voltages on the other conductor of the RTWO. This means that both p-channel transistors are on, sourcing current from the first supply voltage VI to the output 86, which is at a voltage that is relatively high.

[0019]When a positive-going wave front V(Ph _j ) arrives the Ph _j point 82 on the RTWO, input 2 of the amplifier 40 goes positive first. This causes one of the p-channel transistors (42 in FIG. 2A) in the amplifier 40 to turn OFF, and one of the n-channel transistors (48 in FIG. 2A) to turn ON. Because now there is no pair of transistors in amplifier 40 connected between a supply voltage, VI or V2, and the output 52, the output 52 is in a high-impedance state, neither sourcing nor sinking current to the RTWO at position 86.

[0020] When the positive-going wave front V(Ph _k ) arrives at the Ph _k point 84 on the RTWO, now input 1 of the amplifier goes positive. This has the effect of turning ON both n- channel transistors (46, 48 in FIG. 2A) and turning OFF both p-channel transistors (42, 44 in FIG. 2A). The output of the amplifier V(Ph ₀ ) now sinks current to the second supply voltage V2, which reinforces the negative-going wave front at the point 86 (Ph ₀ ) on the RTWO to which the output of the amplifier 40 is connected.

[0021]Conversely, if we assume that the output of the amplifier is at a relative low voltage, then both n-channel transistors (46, 48 in FIG. 2A) are ON, sinking current from output 52 to the second supply voltage V2. Under this condition, both inl and in2 are relatively positive. When a negative-going wave front arrives at the Ph _j point 82 on the RTWO, input 2 of the amplifier 40 goes relatively low. This causes one of the n-channel transistors (48 in FIG. 2A) to turn OFF and one of the p-channel transistors (42 in FIG. 2A) to turn ON. Because there is no pair of transistors connected between a supply voltage and the output, the output 52 is in a high- impedance state, neither sourcing nor sinking current to the RTWO at position 86. When the negative-going wave front arrives at the Ph _k point 84 on the RTWO, input 1 goes relatively low, causing both n-channel transistors (46, 48 in FIG. 2A) to turn OFF and both p-channel transistors (42, 44 in FIG. 2A) to turn ON, thereby sourcing current from the first supply voltage VI to the output 52. Again, this reinforces the positive-going wave front at the point 86 on the RTWO to which the output of the amplifier 52 is connected.

[0022]FIG. 4A shows a timing diagram of the various waveforms involved. V(Phj) is the waveform of a positive-going wave at Phj, V(Phk) is a positive going wave a Phk, and V(Ph0) is the output of the amplifier 40 at position 86 in FIG. 3. 1(out p) is the output current through the p-channel transistors (42, 44 in FIG. 2A) and I(out n) is the output current through the n-channel transistors (46, 48 in FIG. 2A). I(out n+p) is the resultant output current from the amplifier 40. ISF ² is the impulse sensitivity function, α(ωί) is the cyclostationary noise function, and Teff is the effective impulse sensitivity function. As is clear from the timing diagram, the output of the amplifier I(out n+p) has a dwell time X _pc ^ 92, which separates the time when the amplifier 40 is sourcing current from the time when the amplifier 40 is siriking current. It is the time when the amplifier 40 is in the high impedance state. It is also clear that the time when the amplifier40 is in the high impedance state aligns with the ISF ² for the oscillator, which is the time when the oscillator is most sensitive to phase noise. Therefore, arranging the inputs of the amplifier 40 to create a dwell time 92 reduces the phase noise of the oscillator, because it reduces the energy input to the oscillator during the time when the oscillator is most sensitive. The size of the cyclostationary noise source is also minimized, resulting in a lower effective ISF. [0023]The size of the delay or dwell time I _pC (Ph _j - Ph _k ) is adjustable, as shown in FIG. 3. The tap for Ph _j can be adjusted from Ph ₄ , to Ph ₃ , to Ph ₂ while the tap for Ph _k can be adjusted to to Ph ₀ , to Ph^ To be most effective in reducing noise, the time difference T _pc between Ph _j and Ph _k should be on the order of the output transition time or rise and fall time of the wave front on the oscillator. This allows precise alignment of the amplifier's switching with the transition of the wave front at Ph ₀ . In fact, the goal here is to arrange the input connections so that the amplifier switches at a desirable point on the transition of the wave front at Ph ₀ . FIG. 4B, the inset in FIG. 4A, illustrates the negative-going wave front and two instants in time Tl and T2 at which the amplifier output can be designed to switch. Time Tl is at a time above a detection threshold of the wave front and occurs when the amplifier switches after the wave front has passed Ph ₀ . Time T2 is a time below the detection threshold and occurs when the amplifier switches before or at the time the wave front passes Ph ₀ . It is preferable to impact the negative- going wave front at a time Tl, which means that it is preferable to connect the inl input of the amplifier to a point Ph _k after Ph _Q . This permits the oscillator to run at its natural oscillation frequency, which is that determined only by the transmission line. If the inl input is connected to a point before Ph ₀ so that the switching of the amplifier occurs at T2, i.e., before Ph ₀ , the oscillator is forced off its natural oscillation frequency, which consumes more power, and raises the noise floor.

Operation of FIG. 2B

[0024]The amplifiers in FIG. 2B operate in a fashion similar to those of FIG. 2A, except that the wave front encounters the p-type amplifier 60 in FIG. 2B and the n-type amplifier 62 in FIG. 2B at different times. In particular, for the p-type amplifier 60 when a positive-going wave front approaches input 2, the amplifier 60 goes to a high impedance state. When the positive- going wave front reaches input 1, amplifier 60 is shut off, thereby preventing any boost to the negative-going wave on conductor 14a when it reaches position 86 in FIG. 3. The wave then continues to the next amplifier, which is now an n-type amplifier 62 in FIG. 2B. When the positive-going wave reaches input 2, amplifier 62 is enabled and when the wave reaches input 1, amplifier 62 is turned ON, thereby sinking current on the negative-going wave front traveling on conductor 14a. This means that there are times during the traveling of the wave when a positive- going wave front is not reinforced, while the negative-going wave front is a short time later and vice-versa. An advantage of the configuration of FIG. 2B is that the capacitive load of the RTWO is reduced, thereby reducing RTWO power consumption. The configuration also simplifies the layout wiring of the RTWO.

Direction Control

[0025]To encourage a wave front to travel in a preferred direction, an additional constraint must be imposed on the amplifiers' input connections. In particular, if the input connections are subjected to a constraint that input 2 is reached before input 1 and \Ph _j — Ph ₀ \ > \Ph _k — Ph ₀ \, then the wave front propagation from Ph _j to Ph _k in FIG. 3 is encouraged, while a wave in the reverse direction is discouraged. FIG. 5A shows an equivalent circuit for the amplifier of FIG. 2A when a wave travels in the preferred direction. Having the positive difference between Ph _j and Ph ₀ larger than that between Ph _k and Ph ₀ provides sufficient time for the p-channel and n-channel transistors 42, 48 shown in FIG. 5A connected to the first VI and second V2 supply voltages to reach very low resistance (resistance in the triode or linear region (R _P ON, R _II ON)) by the time the wave front reaches input 1, after first passing input 2. When the wave front reaches input 1, the amplifier 40 then switches with maximum gain to maximally boost the wave.

[0026]FIG. 5B shows an equivalent circuit for the amplifier of FIG. 2A when a wave travels opposite to the preferred direction. A wave front traveling in the reverse direction, upon reaching input 1 first, would turn ON the transistors in the amplifier 44, 46 connected to the output Ph _Q , as shown in FIG. 5B. However, these transistors 44, 46 have high resistance Rp, Rn due to their position in the amplifier. When the wave front reaches input 2, the amplifiers switch but with gain insufficient to maintain the wave. Therefore, waves traveling in the direction shown in FIG. 3 are encouraged and waves traveling in the opposite direction are discouraged.

[0027]Although the present invention has been described in considerable detail with reference to certain preferred versions thereof, other versions are possible. Therefore, the spirit and scope of the appended claims should not be limited to the description of the preferred versions contained herein.