Not Applicable.
Not Applicable
1. Field of the Invention
This invention relates generally to automatic test equipment for electronics, and, more particularly, to techniques for generating periodic signals for testing electronic devices.
2. Description of the Related Art
Electronics manufacturers commonly use automatic test equipment (ATE) for testing semiconductor components and electronic assemblies. ATE reduces costs to manufacturers by allowing products to be tested early in the manufacturing process. Early testing allows defective units to be identified and discarded before substantial additional costs are incurred. In addition, ATE allows manufacturers to grade different units according to their tested levels of performance. Better performing units can generally be sold at higher prices.
One of the basic functions of ATE is to generate signals of predetermined frequency. These signals may include, for example, digital clocks, analog waveforms, and RF waveforms. Often, particular testing scenarios require a test system to produce multiple signals of different frequency. Commonly, frequency and phase differences between different signals must be precisely controlled. Phase-locked loops are commonly used in ATE systems to produce signals with precisely controlled frequency and phase.
The conventional PLL 100 is a closed loop feedback system that operates essentially as follows. The phase detector 110 compares the input signal FIN to the feedback signal FFB to generate an error signal, which varies in relation to the difference in phase between FIN and FFB. The loop filter 112 smoothes the error signal and generally helps to stabilize the feedback loop. The VCO 114 converts the filter's output signal into an oscillatory signal, FVCO, which has a frequency that varies in relation to the filter's output signal. The feedback divider 116 (generally a counter) divides the frequency of FVCO by an integer, M, to produce the feedback signal, FFB. Outside the loop, the output divider 118 divides the frequency of FVCO by an integer, N, to produce FOUT. As the feedback tends to drive the difference between FIN and FFB to zero, it consequently drives the frequency of FVCO to a value equal to the frequency of FIN*M, and therefore tends to drive the frequency of the output signal FOUT to a value equal to the frequency of FIN*M/N.
The conventional PLL 100 provides many benefits. For example, output frequency FOUT can be varied, through appropriate selection of N and M, over a wide range of values. In addition, phase noise in the PLL can generally be reduced by setting the bandwidth of the loop filter 112 to arbitrarily low values.
Nevertheless, we have recognized certain shortcomings in the PLL 100, which limits its usefulness in many ATE applications. High frequency applications, such as RF signal generation, require high frequency VCOs. The speed of the VCOs in these applications often greatly exceeds the speed of the phase detectors. This problem is conventionally addressed by making the value of M in the feedback divider 116 very large.
Making the value of M large involves certain drawbacks, however. For instance, the larger the value of M, the greater the reduction in the open-loop gain of the PLL 100. As is known, reducing open-loop gain increases loop tracking errors. It also impairs the ability of the loop to reject noise. To illustrate this effect, consider that the feedback divider 116 not only divides the frequency of FVCO by M, but it also divides any variations (i.e., phase noise or, equivalently, timing jitter) by the same value of M. Sensitivity is therefore reduced.
The frequency divider 116 also adds noise directly. Frequency dividers are commonly implemented as counters, which are known to create spurious noise at their outputs. Although this noise can be attenuated by the loop filter 112, attenuation cannot generally be achieved without setting the bandwidth of the loop filter to a much lower frequency than the offending noise components of the divider 116. Reducing bandwidth to this degree, however, has the effect of reducing programming speed of the PLL 100, which can negatively impact ATE system performance and throughput.
What is desired is a phase-locking circuit that can produce high frequency signals with low phase noise, without sacrificing programming speed.
In accordance with the present invention, a phase-locking circuit employs a sampler for producing aliased feedback signals, upon which a circuit is caused to lock.
The ensuing description will be better understood by reference to the accompanying drawings, in which—
The circuit 200 also includes a circuit path 220, coupled from the output of the VCO 214 to the input of the sampler 202, for providing the feedback signal, FFB. Bandpass filters 230a-230n are preferably provided in the circuit path 220. These bandpass filters are preferably individually selectable via switches 240a-240n. Each filter preferably has a different center frequency.
During operation, the sampler 202 is made to sample the feedback signal, FFB, at a sampling rate FS. The phase detector 210 receives the sampled feedback signal, SFFB, and outputs an error signal, Φ-Err. The error signal varies in response to the difference between SFFB and FIN. The loop filter 212 filters the error signal and helps to stabilize the loop. The VCO 214 converts the filtered error signal into an oscillatory waveform, FVCO. The frequency of FVCO varies in response to the level of the filtered error signal.
One of the bandpass filters 230a-230n is selected for filtering noise from FVCO. The selected filter is preferably the one having the center frequency that is closest to the expected frequency of the FVCO. The desired filter is selected by closing its associated switch (one of 240a-240n) and opening the remaining switches.
The circuit 200 behaves in an essentially normal manner when the frequency of FVCO is less than the Nyquist rate (FS/2) of the sampler. However, significant differences arise when the frequency of FVCO is greater than the Nyquist rate.
As is known, a phenomenon called “aliasing” arises in discrete-time systems when a signal being sampled at a rate FS contains frequency components greater than FS/2. Aliasing causes out-of-band frequencies, e.g., those above the Nyquist rate, to appear as images within the system's bandwidth. These images are normally regarded as errors. However, we have recognized that these aliased images can be used to improve performance.
The creation of aliased images has significant consequences in the phase-locking circuit of
Output frequency ambiguity can arise if the VCO 214 operates over too large a frequency range. For instance, if the output range (maximum frequency minus minimum frequency) exceeds FS/2, then the phase-locking circuit may be able to satisfy its feedback conditions at two or more different VCO frequencies. Preferably, this condition is avoided by limiting the bandwidth of each of the bandpass filters 230a-230n to less than FS/2. Alternatively, it may be avoided by selecting a VCO 214 that has an output range less than FS/2.
Significant performance benefits arise from the use of aliased images in the phase-locking circuit 200. These are best understood with reference to
The use of aliased signals therefore allows the phase-locking circuit 200 to be operated at high gain (where FOUT is much greater than FIN) without the need for feedback dividers. It allows open loop gain and therefore precision to be kept high. Since feedback dividers are not required, the noise spurs normally introduced by these devices are avoided. Therefore, the need to slow down the loop filter and suffer the consequent reduction in programming speed is also avoided.
The harmonic generator 540 receives a filtered version of FVCO and generates one or more harmonics of that signal. These harmonics, or overtones, have frequencies that are integer multiples of the frequency of FVCO, i.e., the fundamental frequency.
A second bandpass bank 550 is optionally coupled to the output of the harmonic generator 540. The second bandpass bank 550 may be used to select one or more specific harmonics to be presented to the sampler 502. Selection of particular harmonics is not required, however.
The harmonic generator 540 effectively multiplies the width of noise bands fed back to the sampler 502. It therefore further increases open loop gain and sensitivity of the phase-locking circuit 500.
The elements of the phase-locking circuits 200/500 can be implemented in a wide variety of ways. The phase detector 210/510 can be either an analog phase detector or a digital phase detector. Similarly, the loop filter 212/512 can be either an analog loop filter or a digital loop filter. Analog and digital phase detectors and loop filters are well-known in the art.
If an analog phase detector is used, the sampler 202/502 is implemented as an analog sampling circuit, such as a sample-and-hold circuit or a track-and-hold circuit. These devices are well-known and readily available off the shelf. In this arrangement, the input signal FIN is preferably an analog signal, such as the output of a crystal oscillator.
If a digital phase detector is used, the sampler 202/502 preferably includes an analog sampling circuit (described above) coupled to an analog-to-digital converter (ADC). The analog sampling circuit and ADC are both clocked at FS. Preferably, a sampling ADC is used, i.e., one which includes both an analog sampling circuit and an ADC in a single device package. Digital values are thus provided to the phase detector at a rate FS. In this arrangement, FIN is preferably a digital signal.
The VCO 214/514 is preferably a conventional type. VCOs are well-known and are commercially available off the shelf.
The harmonic generator 540 is preferably implemented as a non-linear analog circuit, such as a clipping circuit or a commercially available RF comb generator. As is known, clipping circuits flatten the positive and negative peaks of a sinusoid, thus introducing harmonics of the sinusoid's fundamental frequency. Optionally, the harmonic generator 540 may be equipped with an amplifier for boosting low amplitude harmonics.
The digital loop filter 714 offers a particular advantage in the circuit 700. If any of the circuit elements, such as the ADC 712 or DAC 716, are found to repeatably generate noise at known frequencies, or if noise at certain known frequencies is injected into circuit from its environment, the digital loop filter 714 can be programmed to have low gain, or a “zero,” at each offending noise frequency. Designing the loop filter 714 in this fashion reduces noise in the output signal, FOUT, and contributes to the overall precision of the circuit.
Notably, the automatic test system 812 includes a plurality of phase-locking circuits 816a-g. These phase-locking circuits are of the same general type shown in any of
Having described certain embodiments of the invention hereof, numerous alternative embodiments or variations can be made. For example, although phase-locking circuits shown and described preferably include a bank of bandpass filters (230, 530, and 722) coupled to the output of the VCO, these filters are not strictly required. In addition, although the bandpass filters are preferably implemented as analog filters that precede the sampler (202, 502) or the sampling ADC (712), they can alternatively be implemented as digital filters provided at the output of the sampler or sampling ADC.
A particular advantage of the phase-locking circuits disclosed is that they provide closed loop frequency gain without requiring frequency dividers (such as counters) in their feedback paths. This should not be taken to mean, however, that feedback dividers are prohibited. Certain instances may arise wherein feedback dividers are deemed desirable in the context of the circuits disclosed. Aliasing will occur, even with feedback dividers, provided that the overall frequency gain (output frequency divided by input frequency) of the circuit path between the VCO and the sampler is greater than FS/2FMIN, where FMIN represents the lowest frequency provided by the VCO.
The sampling rate FS, with which the sampler (202, 502) or sampling ADC (712) is operated, is preferably fixed. However, this is not required. It may also be variable. According to one variant, FS may be derived from the output of the VCO.
As shown and described, the VCO is made to operate at frequencies higher than the Nyquist rate (FS/2); however, this is not required, either. Aliasing can occur with VCO frequencies below the Nyquist rate if a harmonic generator (540, 730) produces harmonics above the Nyquist rate.
Those skilled in the art will therefore understand that various changes in form and detail may be made to the embodiments disclosed herein without departing from the scope of the invention.
The following patent document is incorporated by reference herein in its entirety: U.S. patent application Ser. No. 10/817,780, entitled “High Performance Signal Generation,” filed Apr. 2, 2004.