WO2019223876A1

WO2019223876A1 - Delay line calibration

Info

Publication number: WO2019223876A1
Application number: PCT/EP2018/063748
Authority: WO
Inventors: Andrei Popescu
Original assignee: Huawei Technologies Co., Ltd.
Priority date: 2018-05-25
Filing date: 2018-05-25
Publication date: 2019-11-28

Abstract

A device for measuring the delay of a delay line in a phase-locked loop comprising an oscillator, the device being configured to: determine a difference between a first oscillator phase at a first time and a second oscillator phase at a second time after the first time, the phase- locked loop being phase-locked at the first and second times; and estimate a delay line delay in dependence on the difference between the oscillator phases at the first and second times.

Description

DELAY LINE CALIBRATION

This invention relates to measuring the delay of a programmable delay line, which may be used as part of a digital Phase-Locked Loop (PLL).

A delay line, also known as a‘Digital to Time Converter’ (DTC) is used in some PLL designs, especially in so-called‘All-Digital PLLs’ (ADPLLs). The function of the DTC is to delay an analogue signal, which is typically a clock signal in a PLL, by a programmable time. The purpose of the DTC in PLLs includes operating the PLL so that the oscillator frequency is not an exact multiple of the reference clock frequency and to apply phase modulation to the oscillator. Usually, the DTC has a digital control port and the delay is a function of the number applied at this port.

Inaccurate knowledge of the DTC delay characteristic typically leads to degraded PLL performance, such as degraded oscillator purity. For example, this can occur if the DTC delay is assumed to be proportional to the delay control, but the function is in reality non-linear.

Therefore, there is a need to measure the delay characteristic in both a characterization setting, where the aim is to evaluate the quality of the analogue design and identify possible improvements, and in an on-chip calibration setting, where the aim is to inform the on-chip delay line control algorithm of the actual delay characteristic, in order to minimize performance degradation due to inaccurate knowledge of this characteristic.

The delay characteristic can be measured directly in the time domain by bringing the delay line output signal out of the device and comparing it with a reference. However, this approach is not suitable for on-chip measurement and it presents difficulties in its implementation of laboratory measurement. For example, preserving the integrity of the DTC output signal edges, to femtosecond accuracy, is difficult. This approach also requires means to route the delay line output out of the device and to buffer it and often requires the implementation of a dedicated ‘measurement’ PCB.

Current solutions for on-chip calibration or compensation of a delay line approximate the delay line delay characteristic (i.e. the delay introduced by the delay line as a function of its delay control) to be linear or piecewise linear. The parameters of the approximated characteristic are estimated using the least mean squares (LMS) algorithm.

SUMMARY

A first disadvantage of prior art methods is that that they estimate an approximation of the delay characteristic, rather than directly measuring the characteristic itself. Therefore they are inherently inaccurate. A second disadvantage is that these methods rely on LMS training of the parameters of the estimated characteristic, whose convergence can be slow and depends on the training sequence used. These methods also have a high implementation cost to achieve a high-resolution approximation of the delay line characteristic.

It is desirable to be able to directly and accurately measure the characteristics of a delay line, in particular by performing measurements on-chip and using such measurements to apply compensation for non-ideality of the characteristic.

According to a first aspect, there is provided a device for measuring the delay of a delay line in a phase-locked loop comprising an oscillator, the device being configured to: determine a difference between a first oscillator phase at a first time and a second oscillator phase at a second time after the first time, the phase-locked loop being phase-locked at the first and second times; and estimate a delay line delay in dependence on the difference between the oscillator phases at the first and second times. This approach allows direct and accurate measurement of the delay line characteristic and is easy and low-cost to implement.

The device may be configured to, at the first time, apply a change to the delay line. This may induce a change to the oscillator phase in a phase-locked loop.

The device may be configured to hold the delay constant until the second time. The second time may be the next time at which the phase-locked loop is phase-locked after the first time. This may minimize the amount of time needed to measure the delay.

The difference between the oscillator phases at the first and second times may be calculated by measuring the phase of the oscillator at the first and second times respectively and calculating the difference between the respective measured values. This may allow the delay to be measured by an external measurement instrument in a characterization setting to evaluate the quality of the analogue design and identify possible improvements.

The phase of the oscillator at each of the first and second times may be measured by calculating the average phase over a time interval including a respective one of the first and second times. This may help to improve the accuracy of the phase difference measurements.

The device may be further configured to integrate the frequency applied to the oscillator between the first and second times, to determine the difference between the oscillator phases. This may allow the measurement of the delay to be made on-chip and used to inform the on- chip delay line control algorithm of the actual delay characteristic, in order to minimize performance degradation due to inaccurate knowledge of the delay.

The device may be further configured to perform the following steps in order one or more times and then estimate the delay in dependence on the average value of the calculated phase differences, the said steps being to: allow the phase-locked loop to reach a first phase-locked state; allow the phase-locked loop to reach a subsequent phase-locked state; determine the difference between the phase of the oscillator at the time at which the phase-locked loop was first phase-locked and the time at which the phase-locked loop was subsequently phase-locked. A change to the delay may be applied each time the phase-locked loop reaches a first phase- locked state. The delay may be held until each time the phase-locked loop reaches a subsequent phase-locked state. Repeating the phase measurements and using the average phase difference to calculate the delay may improve the accuracy of the delay measurement.

The phase-locked loop may comprise a phase detector for sensing a phase difference between a delayed signal received from the delay line and an output signal of the oscillator, and a loop filter, wherein the oscillator is configured such that its frequency is dependent on the output of the loop filter. The delay line may be configured to receive a reference clock signal and delay it to form the delayed signal. This is a convenient configuration.

The device may be configured to receive as an input a signal from a phase-locked loop, wherein the signal is acquired when the phase-locked loop is in a phase-locked state. According to a second aspect there is provided a method for measuring the delay of a delay line of a phase-locked loop comprising an oscillator, the method comprising: determining the difference between a first oscillator phase at a first time and a second oscillator phase at a second time, after the first time, the phase-locked loop being phase-locked; and estimating a delay of the delay line in dependence on the difference between the oscillator phases at the first and second times. This approach allows direct and accurate measurement of the delay line characteristic and is easy and low-cost to implement.

The method may further comprise, at the first time, applying a change to the delay line. This may induce a change to the oscillator phase in a phase-locked loop.

The method may further comprise holding the delay constant until the second time. The second time may be the next time at which the phase-locked loop is phase-locked after the first time. This may minimize the amount of time needed to measure the delay.

The difference between the oscillator phases at the first and second times may be calculated by measuring the phase of the oscillator at the first and second times respectively and calculating the difference between these measured values. This may allow the delay to be measured by an external measurement instrument in a characterization setting to evaluate the quality of the analogue design and identify possible improvements.

The step of calculating the difference between the oscillator phases may comprise integrating the frequency applied to the oscillator between the first and second times. This may allow the measurement of the delay to be made on-chip and used to inform the on-chip delay line control algorithm of the actual delay characteristic, in order to minimize performance degradation due to inaccurate knowledge of the delay.

The method may further comprise performing the following steps in order one or more times and then estimating the delay in dependence on the average value of the calculated phase differences, the said steps being: allowing the phase-locked loop to reach a first phase-locked state; allowing the phase-locked loop to reach a subsequent phase-locked state; calculating the difference between the phase of the oscillator at the time at which the phase-locked loop was first phase-locked and the time at which the phase-locked loop was subsequently phase-locked. A change to the delay may be applied each time the phase-locked loop reaches a first phase- locked state. The delay may be held until each time the phase-locked loop reaches a subsequent phase-locked state. Repeating the phase measurements and using the average phase difference to calculate the delay may improve the accuracy of the delay measurement.

The method may comprise receiving as an input a signal from a phase-locked loop, wherein the signal is acquired when the phase-locked loop is in a phase-locked state.

BRIEF DESCRIPTION OF THE FIGURES

Figure 1 illustrates a phase-locked loop incorporating a delay line.

Figure 2 illustrates apparatus for measuring the delay line delay characteristic.

Figure 3 illustrates apparatus for applying compensation according to the measured delay line characteristic.

Figure 4 illustrates a method for measuring the delay of a delay line.

Figure 5 illustrates the delay versus time when the delay is changed by a relatively large amount (120 ps). Figure 5(a) shows the delay estimated based on integrating the frequency control values applied to the oscillator and Figure 5(b) shows the delay estimated using an external measurement instrument to measure the oscillator phase at times first and second times, when the phase-locked loop is phase-locked.

Figure 6 illustrates a further method for measuring the delay of a delay line.

Figure 7 illustrates delay versus time when the delay is repeatedly changed back-and-forth between two values differing by a relatively small amount (3 ps). Figure 7(a) shows the delay estimated based on integrating the frequency control values applied to the oscillator and Figure 7(b) shows the delay estimated using an external measurement instrument to measure the oscillator phase at times when the phase-locked loop is first and subsequently phase-locked.

Figure 8 illustrates splitting the frequency control value-based delay estimate of Figure 5(a) into sections centered on the times of delay changes and averaging the sections to reduce estimation noise.

DETAILED DESCRIPTION

Figure 1 shows a phase-locked loop 100 incorporating a delay line 101 and an oscillator 102. The PLL may comprise a phase detector 103 for sensing a phase difference between a delayed signal 104 received from the delay line 101 and an output signal 105 of the oscillator 102, and a loop fdter 106, wherein the oscillator is configured such that its frequency is dependent on the output of the loop filter. The delay line may be configured to receive a reference clock signal 107 and delay it to form the delayed signal 104.

Figure 2 shows the PLL of Figure 1 with an additional circuit comprising a module 200 for measuring the delay line delay characteristic. A calibration sequence may be applied to the delay line control port, shown at 201. This sequence may include applying a step change to the delay line control port. The delay of the delay line is a function of the number applied at this port. For example, the delay may be proportional to the number applied at the delay line control port, or the delay may be a non-linear function of the number applied at the port.

The module 200 may be configured to receive as an input a signal from the PLL indicating the phase of the oscillator, wherein the signal is acquired when the PLL is in a phase-locked state.

For example, a change to the delay can be applied at a time Tl, when the PLL is phase-locked. When the PLL is phase-locked, it can be arranged (for example by using a type-II PLL) for the oscillator and reference clock rising edges to be aligned, such that the time difference between them is a constant, plus noise.

Applying a delay change to the delay line results in a change in the PLL oscillator output phase. Changing the delay line delay by an amount p results in a change in the position of rising edges of the oscillator by the same amount. If before the delay change, the rising edges of the oscillator are at times nTosc, after the delay change they are at nTosc+p (plus noise). In terms of oscillator phase, this equates to a phase change of 2^p/Tosc. Hence p can be measured directly by observing the oscillator phase change resulting from a delay change.

The delay is held until a second time T2 when the PLL achieves phase lock again. T2 may be the subsequent time after T 1 at which the PLL is phase-locked. The difference between the oscillator phases at times Tl and T2, denoted phase(Tl) and phase(T2) respectively, is then calculated.

A measurement instrument that is external to the circuit, such as a Vector Signal Analyzer (VSA), may be used to measure the oscillator phase at times Tl and T2. The difference between the oscillator phase at T 1 and T2 may then be determined at 203.

On-chip, the oscillator phase cannot be measured directly in the same way. However, the change in oscillator phase between two times Tl and T2 is equal to the integrated oscillator frequency of the oscillator in-between these times, plus noise. In turn, the oscillator frequency is determined by the frequency control values (FCWs) applied to the oscillator, which for a digitally-controlled oscillator are directly available for on-chip processing, shown at 202. Assuming that the FCWs are updated at discrete times, Tref, once per reference clock period, the oscillator phase change can be estimated on-chip as: å_fe 2nf(U)Tref (Equation 1)

where the sum is over all reference clock periods k occurring between T 1 and T2 and f(k) is the difference between the oscillator frequency corresponding to FCW(k) and the oscillator frequency corresponding to the mean FCW in locked state.

Hence, the delay p can be estimated as: p = ToscTref å_kf(k ) (Equation 2) If the delay line is used to implement fractional ratios between the oscillator and reference frequencies, then preferably for this measurement the PLL is locked in integer mode.

Measuring the delay p relies on knowledge of the mapping between frequency control values FCW(k) and oscillator frequency f(k). Methods for measuring this function are outside the scope of this invention. However, it is possible and easily feasible to measure the oscillator frequency against a reference clock frequency by means of frequency counters.

After the delay has been estimated or determined, for instance measured, at 203, the delay value may be stored in a lookup table, shown at 204 in Figure 2.

The determined delay may be used to calibrate the PLL. In a characterization setting, the aim is to evaluate the quality of the analogue design and identify possible improvements. In this case, external measurement equipment can be used as well as on-chip means.

The determined delay line characteristic can also be used to apply compensation to achieve an accurate desired delay in a PLL. Figure 3 shows a PLL with an additional circuit comprising a compensation module 300 for adjusting the applied delay in dependence on the measured delay line characteristics to accurately achieve a desired delay. The desired delay is input to the compensation module at 301. The module looks up the measured delay from a lookup table 302 and maps it to the delay control at 303 to adjust the applied delay to the desired value. Thus in an on-chip calibration setting, the aim is to inform the on-chip delay line control algorithm of the actual delay characteristic, in order to minimize performance degradation due to inaccurate knowledge of this characteristic.

Figure 4 summarizes a method for measuring the delay of a delay line of a phase-locked loop comprising an oscillator. The method comprises, at 401, determining the difference between a first oscillator phase at a first time and a second oscillator phase at a second time, after the first time, the phase-locked loop being phase-locked; and estimating a delay of the delay line in dependence on the difference between the oscillator phases at the first and second times at 402. As described above, a change may be applied to the delay line delay at the first time and held until the second time to induce a change in the oscillator phase which is measured, either by external equipment or by integrating the FCWs, and used to estimate the delay. Figure 5 shows the delay that is estimated by applying a change of 120 ps to the delay of the delay line using the two approaches described herein. Figure 5(a) shows the delay estimated based on integrating FCWs, which can be used for on-chip measurements, and Figure 5(b) shows the delay estimated using an external measurement instrument to measure the oscillator phase at times Tl and T2.

The measurement accuracy may be further improved by, instead of measuring a single oscillator phase value at two discrete times Tl and T2, measuring the average phase over two time intervals. The first time interval includes Tl (for example, Tl-Atl to Tl) and the second time interval includes T2 (for example, T2 to T2+At2). Alternatively, the two time intervals may be centered about T 1 and T2 respectively. The phase difference may then be calculated from the average oscillator phase values over these time periods.

Another way of improving the measurement accuracy is to repeat the phase measurement described above and average the phase differences across multiple individual measurements.

This is illustrated in Figure 6, which summarizes a method for measuring the delay of a delay line of a phase-locked loop comprising an oscillator. The method comprises, at 601, allowing the phase-locked loop to reach a first phase-locked state. At 602, the PLL is allowed to reach a subsequent phase-locked state. At 603, the difference between the phase of the oscillator at the time at which the phase-locked loop was first phase-locked and the time at which the phase- locked loop was subsequently phase-locked is calculated. Steps 601, 602 and 603 may be repeated a plurality of times. For example, 20, 50 or 100 phase differences calculations may be performed. At 604, the delay is then estimated in dependence on the average value of the calculated phase differences. As described above, a change may be applied to the delay line delay each time the PLL reaches a first phase-locked state. The delay may be held until the time the PLL reaches the subsequent phase-locked state to induce a change in the oscillator phase which is measured to estimate the delay.

The delay may be estimated in this way from the phase difference measurements obtained either by integrating FCWs, which can be used for on-chip measurements, or measured using an external measurement instrument to measure the oscillator phase at the required times. The on-chip delay estimate (for example as shown in Figure 5(a), calculated using Equation 2) is noisier than the estimate based on observing the oscillator phase directly using a measurement instrument. This is because the frequency control values applied to the oscillator by the PLL include compensation for‘open-loop’ low frequency oscillator noise, which is not present in the‘closed-loop’ oscillator signal.

To reduce inaccuracy in the on-chip delay estimation, for example due to oscillator noise, multiple phase difference measurements can be made and the results averaged, the average result being used to calculate the delay as described above. This can be implemented by changing the delay line delay back and forth multiple times, adding and removing the delay increment. For small delay steps, such as measuring a‘Least Significant Bit’ (LSB) increment in DTC delay, the loop settles quickly. Therefore it is possible to perform multiple measurements relatively quickly.

Figure 7 illustrates a measurement involving 50 cycles of varying the delay between 0 and 3 ps, each cycle taking 25 ps. The average phase difference measured across these 50 cycles can then be determined and used to estimate the delay. Figure 5(a) shows the delay estimated based on integrating FCWs, which can be used for on-chip measurements, and Figure 5(b) shows the delay estimated using an external measurement instrument to measure the oscillator phase at times Tl and T2.

Figure 8 illustrates separating the phase variation versus time of Figure 7 into 50 sections centred on the times when the delay line delay changes and averaging the 50 sections to reduce the delay estimate noise.

For delay measurement using the approach described herein, one LSB calibration may take 1.25 ms. To measure the 255 DTC steps of an 8-bit delay line would take 320 ms. It is possible to fine tune the compromise between measurement accuracy and measurement time, for example by shortening each cycle and by averaging fewer delay switching cycles for each delay estimate.

Previous solutions for on-chip calibration or compensation of a delay line approximate the delay line delay characteristic to be linear or piecewise linear. The parameters of the approximated characteristic are estimated using the least mean squares (LMS) algorithm. These methods estimate an approximation of the delay characteristic, rather than measuring the characteristic itself. Therefore they are inherently inaccurate and also rely on LMS training of the parameters of the estimated characteristic, whose convergence can be slow and depends on the training sequence used. These methods also have a high implementation cost to achieve a high-resolution approximation of the delay line characteristic.

The approach described herein allows direct and accurate measurement of the delay line characteristic and is easy and low-cost to implement. The invention can be used for characterization of the delay line characteristic, using measurement equipment external to an integrated circuit, and/or on-chip measurement of the delay line characteristic, using the measurement result to apply compensation for non-ideality of the characteristic.

The applicant hereby discloses in isolation each individual feature described herein and any combination of two or more such features, to the extent that such features or combinations are capable of being carried out based on the present specification as a whole in the light of the common general knowledge of a person skilled in the art, irrespective of whether such features or combinations of features solve any problems disclosed herein, and without limitation to the scope of the claims. The applicant indicates that aspects of the present invention may consist of any such individual feature or combination of features. In view of the foregoing description it will be evident to a person skilled in the art that various modifications may be made within the scope of the invention.

Claims

1. A device for measuring the delay of a delay line in a phase-locked loop comprising an oscillator, the device being configured to:

determine a difference between a first oscillator phase at a first time and a second oscillator phase at a second time after the first time, the phase-locked loop being phase-locked at the first and second times; and

estimate a delay line delay in dependence on the difference between the oscillator phases at the first and second times.

2. The device of claim 1, wherein the device is configured to, at the first time, apply a change to the delay line.

3. The device of claim 2, wherein the device is configured to hold the delay constant until the second time.

4. The device of any preceding claim, wherein the second time is the next time at which the phase-locked loop is phase-locked after the first time.

5. The device of any preceding claim, wherein the difference between the oscillator phases at the first and second times is calculated by measuring the phase of the oscillator at the first and second times respectively and calculating the difference between the respective measured values.

6. The device of claim 5, wherein the phase of the oscillator at each of the first and second times is measured by calculating the average phase over a time interval including a respective one of the first and second times.

7. The device of any of claims 1-4, wherein the device is further configured to integrate the frequency applied to the oscillator between the first and second times, to determine the difference between the oscillator phases.

8. The device of any preceding claim, wherein the device is further configured to perform the following steps in order one or more times and then estimate the delay in dependence on the average value of the calculated phase differences, the said steps being to:

allow the phase-locked loop to reach a first phase-locked state;

allow the phase-locked loop to reach a subsequent phase-locked state;

determine the difference between the phase of the oscillator at the time at which the phase-locked loop was first phase-locked and the time at which the phase-locked loop was subsequently phase-locked.

9. The device of any preceding claim, wherein the phase-locked loop comprises a phase detector for sensing a phase difference between a delayed signal received from the delay line and an output signal of the oscillator, and a loop filter, wherein the oscillator is configured such that its frequency is dependent on the output of the loop filter.

10. The device of claim 9, wherein the delay line is configured to receive a reference clock signal and delay it to form the delayed signal.

11. A method for measuring the delay of a delay line of a phase-locked loop comprising an oscillator, the method comprising:

determining the difference between a first oscillator phase at a first time and a second oscillator phase at a second time, after the first time, the phase-locked loop being phase-locked at the first and second times; and

estimating a delay of the delay line in dependence on the difference between the oscillator phases at the first and second times.

12. The method of claim 11, the method further comprising, at the first time, applying a change to the delay line.

13. The method of claim 12, further comprising holding the delay constant until the second time.

14. The method of any of claims 11-13, wherein the second time is the next time at which the phase-locked loop is phase-locked after the first time.

15. The method of any of claims 11-14, wherein the difference between the oscillator phases at the first and second times is calculated by measuring the phase of the oscillator at the first and second times respectively and calculating the difference between these measured values.

16. The method of claim 15, wherein the phase of the oscillator at each of the first and second times is measured by calculating the average phase over a time interval including a respective one of the first and second times.

17. The method of any of claims 11-14, wherein the step of calculating the difference between the oscillator phases comprises integrating the frequency applied to the oscillator between the first and second times.

18. The method of any of claims 11-17, further comprising performing the following steps in order one or more times and then estimating the delay in dependence on the average value of the calculated phase differences, the said steps being:

allowing the phase-locked loop to reach a first phase-locked state;

allowing the phase-locked loop to reach a subsequent phase-locked state;

calculating the difference between the phase of the oscillator at the time at which the phase-locked loop was first phase-locked and the time at which the phase-locked loop was subsequently phase-locked.