Low jitter direct digital synthesizer

This invention relates generally to signal generation and more specifically to generation of periodic signals with high purity.

BACKGROUND OF INVENTION

Direct Digital Synthesis ("DDS") is a technique used to generate periodic signals where control over one or more signal properties is desired. Analog signals can be generated with the period and the wave shape controlled through DDS.

<figref idref="f0001">Fig. 1</figref> shows a traditional DDS architecture with (DDS) 100 being used to generate a sine wave that is then converted to a two-valued clock. The DDS 100 receives an accumulator clock CLK_ACC and a digital input signal Φ_Inc indicating a phase increment. The DDS outputs an analog signal F_OUT. The frequency of F_OUT can be set by varying the frequency of CLK_ACC and/or varying the phase increment Φ_Inc.

In operation, accumulator 110 produces a new output value Φ_Acc on each cycle of clock CLK_ACC. To produce each new value, the accumulator 110 adds Φ_Inc to its current contents. <figref idref="f0001">FIG. 2</figref> shows a block diagram of an accumulator, such as is known in the art.

The value in accumulator 110 serves as a control input to Sine Unit 112. Sine Unit 112 converts each phase value Φ_Acc into a corresponding amplitude value. In the illustration, the DDS signal generator is producing a sine wave. Accordingly, the amplitudes are related to the value Φ_Acc by the function sin (Φ_Acc). Sine Unit 112 might generate the required outputs using a math engine - a circuit configured to produce an output signal having a specific mathematical relationship to an input signal. Alternatively, it is possible to implement a sine unit by pre-computing the required output value for each value of the control input. These pre-computed output values are then stored in a memory at locations addressed by the control input. In operation, the control inputs are applied as addresses to the memory, resulting in the required output value of sin (Φ_Acc) being read from the memory for each of Φ_Acc applied as an input. Such a look-up table is shown in <figref idref="f0001">FIG. 3</figref>.

The output of sine unit 112 is periodic. Periodicity is achieved because of overflow in accumulator 110. The value stored in accumulator 110 increases (or decrease if negative values of Φ_Inc are used) for every cycle of CLK_ACC. Eventually, the value in accumulator 110 overflows (or underflows - if negative values of Φ_Inc are used). The full scale value of the accumulator is selected to correspond to a phase of 2pi radians. If the addition of Φ_Inc would cause the value of Φ_Acc to exceed 2pi radians by an amount <i>x</i>, after the overflow, the accumulator stores only the value <i>x</i>. As a result, an overflow of the accumulator has the same effect as starting a new cycle of the periodic waveform, with the appropriate phase relationship maintained between the end of one cycle and the beginning of the next cycle.

The duration of one cycle of the waveform F_out can be controlled by altering the time it takes for accumulator 110 to overflow. This time can be controlled by changing the frequency of clock CLK_ACC. This time can also be controlled by changing the value of Φ_Inc.

Digital values representing sin (Φ_Acc) are then fed to a digital to analog converter, such as DAC 114 which converts them to a quantized analog signal. Usually, a filter is attached to the output of the digital to analog converter to smooth out the quantized signal. Where a sine wave is required, the filter is likely a bandpass filter because a bandpass filter that includes the desired frequency of the sine wave in its pass band will increase the "spectral purity" of the signal.

If a digital signal, such as a clock, is desired, the analog signal may be fed to a comparator 118 to square off the signal. Thus, the DDS signal generator provides a convenient mechanism to generate a clock of controlled frequency. Where the DDS signal generator is being used to generate a clock, spectral purity is also important. Lack of spectral purity in the signal F_out appears as "jitter" in the digital clock. For precise measurement applications requiring a clock, low jitter is important. Therefore, it would be highly desirable to provide a DDS signal generator with high spectral purity.

One application where a variable frequency, but low jitter, clock is desirable is in automatic test equipment. <figref idref="f0004">FIG. 8</figref> shows in greatly simplified form a block diagram of an automatic test system 800 of the type that might be used to test semiconductor chips. An example of such a system is the Tiger^™ test system sold by Teradyne, Inc. of Boston, MA, USA.

The test system includes a work station 810 that controls the test system 800. Work station 810 runs test programs that set up the hardware within tester body 812 and reads back results of tests. The work station also provides an interface to a human operator so that the operator can provide commands or data for testing a particular type of semiconductor device. For example a program running on work station 810 might change the frequency of a clock within the test system by changing the value of a register inside tester body 812 holding the value of Φ_Inc.

To fully test many types of devices, both analog and digital test signals must be generated and measured. Inside tester body 812 are digital "pins" 820 and analog instruments 818. Both are connected to the device under test 850. Digital pins are circuits that generate or measure digital signals or DC voltages and currents. In contrast, analog instruments generate and measure analog signals.

Pattern generator 816 provides control inputs to the digital pins 820 and the analog instruments 818. These control inputs define both the values and the time at which test signals should be generated or measured. To ensure an accurate test, the actions of the digital pins and the analog instruments often must be synchronized. Timing generator 814 provides timing signals that synchronize operation of the various components within tester body 812.

Automatic test equipment is made programmable so that it can test many different types of devices. It is often desirable to be able within automatic test equipment to generate a digital clock of a programmable frequency that has very low jitter. An example of such an application is called an arbitrary waveform generator (AWG). AWG 822 creates a waveform that can be programmed into an almost arbitrary shape using a clock of controllable frequency. In the prior art, a DDS signal generator 100 has been used as the clock for the AWG. The generated waveform would be more accurate if the clock provided to AWG had less jitter.

Automatic test equipment also sometimes contains an analog instrument called a digitizer. Digitizer 823 also relies on a clock, which should preferably be programmable. Digitizer 823 could also be made more accurate if it were provided with a clock with lower jitter.

More generally, there are many applications where sine waves of high spectral purity are desirable for testing devices such as semiconductor chips. It would therefore be desirable to provide automatic test equipment with an improved synthesizer circuit to generate signals with improved spectral purity.

We have recognized that the DAC strongly influences the overall spectral purity of signals produced by a DDS. We have also recognized that spectral purity strongly depends on the sampling rate at which the DAC operates. In particular, impurities arising from signal-to-noise ratio (SNR) and spurious-free dynamic range (SFDR) decrease approximately linearly with DAC sampling rate. We have recognized the benefits of operating the DAC of a DDS at its highest possible sampling rate. However, as DACs with faster sampling rates become available, there are limitations on the availability of circuitry that can generate the data stream into the DAC. The circuitry needed to operate a DDS signal generator for high spectral purity is generally not available or undesirably expensive, consumes too much power, takes up too much space or is otherwise undesirable.

Disclosures <patcit id="pcit0001" dnum="US4454498A"><text>US-A-4 454 498</text></patcit>, <patcit id="pcit0002" dnum="US4958310A"><text>US-A-4 958 310</text></patcit> and <patcit id="pcit0003" dnum="US5467294A"><text>US-A-5 467 294</text></patcit> illustrate a method of generating a periodic signal, the method comprising the steps of providing a plurality of streams of digital values, each value having a predetermined number of bits, and each stream representing samples of the periodic signal to be generated, of interleaving values derived from the plurality of streams to produce a final stream of digital values and of converting the final stream to an analog signal forming the periodic signal in a digital to analog converter having an input receiving said digital values. The disclosure <nplcit id="ncit0001" npl-type="s"><text>B. J. Dinteman: "Phase Coherent, Event Synchronized Test System Architecture", European Test Conference, 1993. Proceedings of ETC 93, Third Rotterdam, Netherlands 19-22 April 1993, Los Alamitos, CA, USA, IEEE Comput. Soc, US, Pages 312-319</text></nplcit>, XP010031761 show an automatic test system with digital and analog testing components wherein the analog component comprises a digital synthesizer.

More precisely, the document <patcit id="pcit0004" dnum="US4454486A"><text>US-A-4 454 486</text></patcit> relates to a signal synthesizer providing high frequency synthesized waveforms for the user. By converting phase information into digital outputs in parallel and by selectively coupling these digital outputs, an ordered digital output is formed to provide the high frequency waveforms. This ordered digital output, which represents points on a sine function, is converted to an analog signal for the synthesizer output. Furthermore, frequency and phase modulations of the synthesized waveforms are easily implemented with this signal synthesizer; the modulation information is simply added to the digital outputs prior to selectively coupling. Thus when the ordered digital output is converted to an analog signal, the analog signal contains the modulation information.

SUMMARY OF INVENTION

One aspect of the present invention is a method of generating a low jitter digital clock as defined in independent claim 1. Another aspect of the invention is an automatic testing system incorporating an apparatus for generating a low jitter digital clock as defined in independent claim 11. Further embodiments of the invention are specified in the respective appended dependent claims.

The invention inconcerned with a DDS signal generator which generates a sine wave. This sine wave is used to create a low-jitter digital clock.

In another aspect, the DDS signal is used to generate a low-jitter digital clock of variable frequency. That clock is used within an automatic test system to clock an arbitrary waveform generator.

BRIEF DESCRIPTION OF DRAWINGS

The foregoing and other objects will be better appreciated by reference to the following detailed description and accompanying drawings in which: <ul id="ul0001" list-style="none" compact="compact"><li><figref idref="f0001">FIG. 1</figref> is a block diagram of a prior art DDS signal generator used to generate a digital clock;</li><li><figref idref="f0001">FIG. 2</figref> is a block diagram of a prior art accumulator;</li><li><figref idref="f0001">FIG. 3</figref> is a block diagram of a prior art lookup table;</li><li><figref idref="f0002">FIG. 4</figref> is a block diagram of a DDS signal generator used to generate a digital clock;</li><li><figref idref="f0002">FIG. 5</figref> is a block diagram of the fill-in unit used in the DDS signal generator of <figref idref="f0002">FIG. 4</figref>;</li><li><figref idref="f0003">FIG. 6</figref> is a block diagram of an alternative embodiment of a DDS signal generator used to generate a digital clock;</li><li><figref idref="f0003">FIG. 7</figref> is a block diagram of the accumulator bank used in the DDS signal generator of <figref idref="f0003">FIG. 6</figref>; and</li><li><figref idref="f0004">FIG. 8</figref> is a block diagram of a test system which can be improved through one of a DDS signal generator of <figref idref="f0002">FIG. 4</figref> or <figref idref="f0003">FIG. 6</figref>.</li></ul>

DETAILED DESCRIPTION

This invention is not limited in its application to the details of construction and the arrangement of components set forth in the following description or illustrated in the drawings. The invention is capable of other embodiments and of being practiced or of being carried out in various ways.

<figref idref="f0002">FIG. 4</figref> shows a synthesizer 400 used to generate a low jitter digital clock. Synthesizer 400 uses a modified form of direct digital synthesis.

As in the prior art, synthesizer 400 is clocked by a clock CLK_ACC. Register 402 stores a value of Φ_Inc. For each cycle of clock CLK_ACC, the value stored in accumulator 410 increases by the value of Φ_Inc stored in register 402.

The value in accumulator 410 is fed to accumulator fill-in unit 420. Accumulator fill-in unit is shown in more detail in <figref idref="f0002">FIG. 5</figref>. For each new value of Φ_Acc, accumulator fill-in unit 420 produces N new phase values, Φ₀ ... Φ_N-1. These N phase values represent values between the value in accumulator 410 and the value that will be in the accumulator 410 when the accumulator is incremented the next time accumulator 410 is clocked by CLK_ACC.

As shown in <figref idref="f0002">FIG. 5</figref>, the values of Φ₀ ... Φ_N-1 are formed by a bank of adders, of which 510, 512, 514 and 516 are shown. One of the inputs to each of the adders is the value stored in accumulator 410.

The second input to each of the adders is based on the value of the phase increment, Φ_Inc, stored in register 402. A different fraction of the value of Φ_Inc is input to each of the adders. As shown in <figref idref="f0002">FIG. 5</figref>, the adders are ordered 0...(N-1). The fraction of Φ_Inc input to each adder is based on the position of the adder in this order. The first adder receives (0/N)* Φ_Inc. The next adder receives (1/N) * Φ_Inc. The pattern continues in this fashion, with the final adder in the order receiving an input of (N-1/N)* Φ_Inc.

Thus, for each cycle of CLK_ACC , accumulator fill-in unit outputs N linearly increasing phase values. The signals Φ₀ ... Φ_N-1 are provided to sine bank 412.

Sine bank 412 can be implemented as a bank of sine units 112. Each of the sine units in sine bank 412 receives one of the signals Φ₀ ... Φ_N-1 as a phase control input. Each of the sine units within sine bank 412 outputs a digital value representing a point on a sine wave. The specific point on the sine wave is controlled by the phase input applied to the specific sine unit.

The first control input Φ₀ represents the control input to a sine unit as used in a prior art DDS circuit. Accordingly, the output of the first sine unit is a value of a sine wave as in the prior art DDS circuit. Each of the successive inputs Φ₁ ... Φ_N-1 represents a phase that is shifted relative to Φ₀. Accordingly, the output of each successive sine unit within sine bank 412 represents the value of a sine wave shifted in time relative to the value produced by the preceding sine unit in the sine bank.

The outputs of each of the sine units within sine bank 412 is fed to one of the switchable inputs of selector 422. Selector 422, sometimes called a "multiplexer," connects one of the switchable inputs through to its output based on the value of a control input.

The control input of selector 422 is provided by counter 426. Preferably, counter 426 is a counter that counts at least from 0 to (N-1). Counter 426 either overflows to zero or is reset to zero for each cycle of CLK_ACC.

Counter 426 is clocked by frequency multiplier 424. Frequency multiplier receives an input from CLK_ACC. The output of frequency multiplier 424 is a higher frequency clock with a frequency that is N times the frequency of clock CLK_ACC. In this way, for each cycle of clock CLK_ACC, N successive values are clocked through selector 422.

Accumulator fill-in unit and sine bank 412 can be thought of as multiplying by N the number of samples of a sine wave available in each cycle of CLK_ACC. To allow the circuitry to operate at a relatively slow speed, these samples are generated in parallel. Frequency multiplier 424, counter 426 and selector 422 convert these parallel samples of a sine wave into a stream of digital values representing one sine wave. However, the sample rate of that stream of values is increased by a factor of N.

This stream of digital values is applied as the input to digital to analog (DAC) converter 414. DAC 414 is similar to DACs used in prior art DDS circuits. However, it needs to have a higher sampling rate to handle the higher data rate out of selector 422. By providing a data stream with a higher sampling rate to DAC, the output of DAC 414 has much greater spectral purity than signals generated by prior art DDS circuits.

The output of DAC 414 is shown in the embodiment of <figref idref="f0002">FIG. 4</figref> to be presented to a band-pass filter 416. Such a filter can further increase spectral purity of the signal generated by DAC 414. However, bandpass filter 416 might not be used in all implementations.

The output of the bandpass filter 414 is provided to comparator 418. As in prior art synthesizers, a sine wave with high purity can be used to generate a digital clock with low jitter using a comparator as shown in <figref idref="f0002">FIG. 4</figref>.

<figref idref="f0003">FIG. 6</figref> shows an alternative embodiment in which a different circuit is used to generate phase values, Φ₀ ... Φ_N-1. Synthesizer 600 uses an accumulator bank 610 in place of accumulator 410 and accumulator fill-in unit 420.

As shown more fully in <figref idref="f0003">FIG. 7</figref>, accumulator bank 610 includes N accumulators, of which 710, 712, 714 and 716 are shown for simplicity. Each of the accumulators, such as 710, 712, 714 and 716 receives the value of Φ_Inc as an input. This value is added to the current value stored in the accumulator once for each cycle of CLK_ACC. To generate the signals Φ₀ ... Φ_N-1 representing values staggered in phase, each of the accumulators is initially set to values that are slightly different.

For example, accumulator 710 can initially be set to 0. Accumulator 712 can initially be set to (1/N)*Φ_Inc. Accumulator 714 can initially be set to (2/N)*Φ_Inc. This pattern continues in order, with accumulator 716 being initially set to (N-1)/N*Φ_Inc.

<figref idref="f0003">FIG. 7</figref> shows that each of the accumulators contains a reset line. Each time the accumulators are reset, the initial value should be loaded into the circuit. It should be appreciated that an accumulator having reset circuitry that computes and loads the appropriate initial value could be constructed. The circuitry could be part of each accumulator unit. Alternatively, in a system such as automatic test system 800, a computer controller, such as work station 810, might compute the required value and store it in the accumulator register before synthesizer 600 is enabled to operate. Setup of electronic circuits before enabling them and loading values in accumulator registers are known functions of electronic circuits.

Various known techniques could be used to construct a synthesizer, such as synthesizer 400 or 600. To achieve high spectral purity, it is desirable that DAC 414 have both a high sample rate and relatively large number of bits. In a preferred embodiment, DAC 414 will receive at least 2 Giga-values (2*10⁹) per second. More preferably, DAC 414 will receive 4 Giga-values per second. In a presently contemplated implementation, a 4.8 Giga-Hertz DAC is used.

Each digital value input to DAC 414 in a presently preferred embodiment has 10 bits. It is possible, as a way to reduce the effects of computational roundoff, that the circuits generating digital values applied to DAC 414 will generate values with more bits than DAC 414 uses in a conversion. For example, each value will preferably have at least 14 bits of resolution. In a presently contemplated implementation, the generated values have 18 bits of resolution.

Frequency multiplier 424, counter 426 and selector 422 have components that operate at the sampling rate of DAC 414. Accordingly, relatively high frequency components are needed to implement these components. Technologies to make circuit components operate at this speed are known. For example, circuits made using an ECL or SiGe processes can operate at the required speeds. In a preferred embodiment, these components will be implemented as a single ASIC or as part of an ASIC in automatic test 800 that is used for other functions. These features could, for example, be implemented as part of the same integrated circuit chip that holds DAC 414.

One advantage of the above described architecture is that the sine bank 412 and accumulator bank 610 or accumulator 410 and accumulator fill-in unit 420 do not need to operate at the same data rate as DAC 414. Thus, these components, and any other components clocked by CLK_ACC, can be implemented with relatively low speed circuitry. In a preferred embodiment, CLK_ACC will operate at a frequency less than 500 MHz. In a presently preferred embodiment, CLK_ACC will operate at a frequency less than 200 MHz. These clock rates correspond to a value of N that is at least 32. In a presently preferred embodiment, N is 64.

Many techniques are known to implement circuits operating at this range of frequencies. For example, CMOS is widely used to construct circuits operating in this range. A CMOS ASIC could be used for these components. However, because of the simplicity of the design, a relatively low-cost CMOS gate array could be used to implement these components. Moreover, the small size and relatively low power consumption of CMOS allows a synthesizer as described to be practically implemented in applications that are sensitive to size, power consumption and/or cost, such as automatic test equipment.

As described above, a synthesizer generates a signal with high spectral purity by creating a stream of digital values at a high rate. These values represent samples of a sine wave and serve as an input to the DAC. This stream of values is constructed by interleaving, for each period of the low frequency clock, the outputs of multiple sine units. The outputs of the sine units represent samples of sine waves, each having the same frequency as the sine wave in the interleaved data stream. However, the sampling rates of these sine waves is lower. Other implementations of a circuit that generates streams of digital values representing samples of sine waves and interleaving them would also be possible.

Having thus described several aspects of at least one embodiment of this invention, it is to be appreciated various alterations, modifications, and improvements will readily occur to those skilled in the art. Such alterations, modifications, and improvements are intended to be part of this disclosure, and are intended to be within the scope of the invention as defined in the appended claims. Accordingly, the foregoing description and drawings are by way of example only.

For example, the invention is described in connection with a synthesizer that generates a sine wave that is then converted to a low jitter digital clock. The invention could be used to generate a sine wave that is used for other applications.

Also, the invention is illustrated as used in connection with automatic test equipment. However, the invention is applicable in any other area where sine waves of high spectral purity or digital clocks with low jitter are desired.

Furthermore, it should also be appreciated that DDS can be used to generate signals other than sine waves. Sine bank 412 could be replaced by a memory or math engine that implements a function other than a sine function.

<figref idref="f0002">FIG. 4</figref> shows that each stage of synthesizer 400 is clocked, allowing for the circuit to be pipelined. It is not necessary that the circuit be pipelined.

Also, it was described above that the DDS circuit generates the multiple streams of digital values, each representing a sine wave offset by a uniform phase from sine waves represented by the other streams of digital values. These streams of data were achieved by supplying sine bank 410 with multiple phase inputs, each offset from the other by a uniform amount. The same effect could be achieved by supplying one control input to each sine unit in a sine bank, where each sine unit of the sine bank responded with a sine wave of a slightly different phase. For example, rather than receive N inputs representing N different phases, sine bank 412 might contain N memories that each stored values for a sine wave with a different phase.

As another example, it should be noted that the adders, sine units within the sine bank, and the inputs to the selector are "ordered." This ordering is a "logical" ordering in the sense that the ordering is used, for example, in determining which phase input to connect to which sine unit or which sine unit to connect through to the selector output. Such a logical ordering does not mean that any specific physical ordering is required when a chip containing the unit is constructed. Any convenient construction technique might be used. Also, the ordering may be established by any convenient system. For example, it is not necessary that the control input 0 to selector 422 switch the first value in the order to the output. Any numbering system or convention that preserves the desired ordering of signal may be used.

Also, counter 426 is shown to be clocked by a clock that is generated in a frequency multiplier. Any convenient method of generating two clocks with a known frequency relationship might be used. For example, a higher frequency clock might be generated to clock counter 426, with CLK_ACC being generated in a frequency divider having the higher frequency clock as an input. Also, though preferable, it is not strictly necessary that the clocks be generated from the same clock.

Further, it was described that a sine unit could be implemented as either a look-up table or a math engine. It is also possible that a combination of look-up and computation could be used to generate the data streams that are provided by each sine unit. For example, a look-up table might have only 2⁹ locations, meaning that the memory is addressed by only 9 address lines. However, accumulator 410 might have more than 9 bits of resolution. In this case, the high order bits of the accumulator would be used to select a base value from one of the look-up tables. The low order bits of the accumulator could then be used to interpolate between the base value and the value at the next address in the memory. A simple linear interpolation might be used, though other more complicated forms of interpolation could be constructed.

Also, it was described that a DSS synthesizer is used to generate a sine wave of high spectral purity. The same circuitry could be used to generate a sine wave at a higher frequency than is possible with a traditional DDS circuit. In a traditional DDS circuit, the frequency of a generated sine wave is limited to ½ the frequency of CLK_ACC. In the described embodiment, the frequency of the generated sine wave is limited at Nx1/2 the frequency of CLK_ACC.