RU171560U1 - DEVICE FOR TRANSFORMING TIME INTERVALS TO DIGITAL CODE WITH AUTOCALIBRATION - Google Patents

Abstract

The utility model relates to the field of radio engineering and can be used in devices for converting time intervals into a digital code with automatic calibration. The device comprises a generator, a counter and a serial register connected in series with it, forming a “coarse” conversion channel, as well as a delay line with taps connected in series with the generator, a position register and a position code to binary converter forming an “exact” conversion channel. The outputs of the serial register and the positional code binary converter are connected to the first and second signal inputs of the code calculator. The clock input of the position register is connected to the output of the variable delay line, the input of which is connected to the signal input of the device. The clock inputs of the serial register and code calculator are also connected to the signal input of the device. The control inputs of the variable delay line and the position code to binary converter are connected to the outputs of the control unit. A filter unit, a dispersion calculation unit, and a minimum search unit, the output of which is connected to the third signal input of the code calculator, are connected in series with the output of the code calculator. The achieved result is the development of a converter of time intervals into a digital code with automatic calibration, in which calibration is carried out without interrupting the process of converting a useful signal, i.e. in the background. 2 ill.

Description

The utility model relates to the field of radio engineering and can be used in devices for converting time intervals into a digital code, used, in particular, for solving problems of phase detection in phase locked loop systems and tracking problems for the delay of pulse signals in temporary discrimination systems.

Also, a useful model can be used, for example, in solving problems related to measuring distances using a laser, controlling the parameters of high-speed data transmission systems, testing microcircuits, and constructing frequency synthesizers.

For all these, as well as a number of other applications, the task of measuring time intervals with an error of the order of units of picoseconds is urgent, which is achieved through the use of devices that convert the duration of the time interval specified by two events into a digital code, see, for example, works: [ 1] - Christiansen J. / Picosecond stopwatches. The evolution of time-to-digital converters // IEEE Solid State Magazine, 2012, pp. 55-59; [2] - Henzler S. / Time-to-Digital Converters // Dordrecht Heidelberg London New York Springer, 2010, p. 124.

In such converters, to increase the dynamic range (i.e., the bit depth of the conversion), “coarse” and “exact” channels (cascades) of conversion are used, where the data of the “exact” channel are used to refine the data received in the “coarse” channel. For example, in the “rough” channel, data is generated about the time interval, the duration of which is equal to an integer number of steps of “rough” quantization, and in the “exact” channel the error of the “rough” conversion is determined. The subsequent combination of the results of the “rough” and “accurate” transformations in the corresponding computing unit allows us to obtain the required accuracy of the conversion of the time interval.

The technical problem existing in such converters is the problem of matching the transmission coefficients of the “coarse” and “accurate” channels, i.e. determining the ratio of units of their least significant digits. This problem is solved by mutual calibration of the “coarse” and “accurate” channels, which can be carried out in automatic mode (auto-calibration).

A known method of calibrating time interval converters to digital code, based on the use of code density test, described, in particular, in the works: [3] - Samarah A., Carusone AS / A Digital Phase-Locked Loop with Calibrated Coarse and Stochastic Fine TDC // IEEE Journal of Solid-State Circuits, 2013 / V 48, p. 1829-1841; [4] - Ito S. et all / Stochastic TDC architecture with self-calibration // IEEE Asia Pacific Conference on Circuits and Systems, 2010, p. 1027-1030. This method consists in sequentially converting a large number of time intervals of random duration and building a histogram of the distribution of code combinations of the “exact” converter channel. The width of the histogram makes it possible to estimate the width of the quantization step of the “coarse” channel of the converter in units of the least significant bit of the “exact” channel. The data obtained are used to correct the conversion results. The disadvantage of this solution is the need to store a large amount of transformation results for building a histogram, as well as the complexity of data processing associated with the need to approximate the distribution of code combinations by an analytical function and finding characteristic points.

A known calibration method in which the ratio of the units of the least significant bits of the "coarse" and "exact" conversion channels is determined during the conversion of a time interval of a predetermined duration. This method is used in a device for converting time intervals into a digital code, presented in patent [5] - US 8362932 (B2), Н03М 1/48, 01/29/2013, which shows an example of its use in a phase-locked loop. This device, which is a converter of time intervals into a digital code with automatic calibration, is adopted as a prototype.

The prototype device contains, see FIG. 1, generator 1, counter 2, serial (shift) register 3, code calculator 4, delay line 5 with taps, position register 6, position code to binary converter 7, variable delay line 8, control unit 9, and switch 10.

The output of the generator 1 is connected to the input of the counter 2 and the first input of the switch 10, the second input of which forms the signal input of the device, intended for receiving pulses, the time interval between which must be measured by conversion to a digital code. The output of the counter 2 is connected to the data input of the register 3, the output of which is connected to the first signal input of the calculator 4 of the code forming the output signal, which is the result of the conversion of the measured time intervals. The clock input of the code calculator 4 and the clock input of the register 3 are connected to the second input of the switch 10, i.e. with the signal input of the device. The first output of the switch 10 is connected to the input of the delay line 5 with taps, and its second output is connected to the input of the variable delay line 8. The taps of the delay line 5 are connected to the corresponding inputs of the bits of the register 6. The clock input of the register 6 is connected to the output of the variable delay line 8. The outputs of the bits of the register 6 are connected to the inputs of the bits of the Converter 7 position code in binary, the output of which is connected to the second signal input of the calculator 4 code. The control input of the position code converter 7 to the binary, the control input of the variable delay line 8 and the control input of the switch 10 are connected to the corresponding outputs of the control unit 9.

Generator 1 is used to generate clock pulses. Counter 2 and register 3 form a “coarse” conversion channel intended for “coarse” conversion into a digital code of the working time interval between pulses arriving at the signal input of the device. The delay line 5 with taps, register 6, the position-to-binary code converter 7 and the variable delay line 8 form an “exact” conversion channel designed to determine the error that occurs during the “rough” conversion and convert this error to a digital code. The calculator 4 code is used to combine the results of the "rough" and "accurate" transformations. The switch 10 provides receiving at its outputs signals corresponding to two operating modes, namely the operating mode and the auto calibration mode. The control unit 9 is used to control the operation of the device.

The operation of the prototype device in the operating mode, in which the conversion is made into a digital code of the time intervals between pulses arriving at the signal input of the device, as follows.

Generator 1 generates clock pulses that are fed to the input of counter 2 and to the first input of switch 10, from the first output of which clock pulses are fed to the input of delay line 5 with taps.

To the signal input of the device, i.e. to the second input of the switch 10, input pulses are received, the time interval between which (the working time interval) is converted into a digital code. The same pulses are supplied to the clock inputs of register 3 and code calculator 4.

Counter 2 counts the clock pulses coming from the output of the generator 1. At the moment of arrival of the k- th input pulse to the signal input of the device (and, accordingly, to the clock input of register 3), the contents of N _{k of} counter 2 are recorded in register 3. The fixing results, which are the results of a “rough” conversion, are output from register 3 to the first signal input of the calculator 4 code.

The same k- th input pulse, passing through the switch 10 to its second output, is fed to the input of the variable delay line 8.

The closest clock pulse generated by the generator 1, following the kth input pulse, enters through the switch 10 from its first output to the input of the delay line 5 with taps.

The time interval between the input pulse with number k and the nearest next clock pulse of the generator 1 is an error signal, which is further converted into an “exact” channel as follows.

The clock pulse of the generator 1, propagating along the delay line 5, sequentially appears on its taps with a delay, which determines the resolution of the “accurate” conversion channel.

When a delayed input pulse appears at the output of line 8 of the delayed input pulse, register 6 captures the logical state of the signals at the taps of the delay line 5. Further, the converter 7 positional code in binary provides a representation of the positional data of the register 6 in the required binary format, hereinafter referred to as n _k . These data from the output of the Converter 7 are transmitted to the second signal input of the calculator 4 code.

The next ( k +1) -th input pulse initiates a similar sequence of changes in the status of registers 3 and 6 with the receipt of data and N _{k + 1} and n _{k + 1} and their transmission to the computer 4 code.

The desired duration of the time interval τ _k between two consecutive input pulses ( k- th and k + 1- th) is determined in the calculator 4 code as:

where T ₀ is the frequency period of the generator 1,

τ _e is the resolution of the “exact” conversion channel.

The result of calculating the duration of the time interval τ _k is encoded and stored in the memory of the calculator 4 code. As necessary, the digital code of the measured time interval is removed from the output of the calculator 4 code.

To increase the accuracy of calculations, the pulse delay in the variable delay line 8 is changed according to a certain law established by the control unit 9. This allows the code calculator 4 to de-correlate the quantization error caused by the finite resolution of the “exact” conversion channel by averaging several estimates corresponding to different delay values in the variable delay line 8.

To minimize the errors in the calculations performed in the code calculator 4, it is critically important to know exactly the ratio T ₀ / τ _E. In the prototype device, for determining the ratio T ₀ / τ _E , a self-calibration mode is provided, in which, instead of the process of converting the time interval between input pulses (time interval with an indefinite duration) described above, a time interval of a predetermined duration is converted.

The calibration process is as follows. According to the corresponding signal from the control unit 9, the switch 10 is transferred to a state in which a clock pulse A generated by the generator 1 is issued to the delay line 5 with taps, and a pulse of the generator 1 separated from the pulse A is issued instead of the input pulse to the variable delay line 8 on j periods T ₀ , where j can take the values “0”, “1”, “2”, etc. As a result of the sequential implementation in the "exact" channel of two transformations of error signals with different values of j ₁ and j ₂ and the subsequent solution of the system of linear equations in the calculator 4 of the code, the desired ratio T ₀ / τ _{E is} determined:

where n ₁ is the result of the conversion of the error signal for j = j ₁ ;

n ₂ is the result of the conversion of the error signal for j = j ₂ .

Denoting the error of the j- th transformation caused by the quantization effect in the “exact” transformation channel, as δ _j , expression (2) can be represented as:

Substituting the resulting expression (3) into expression (1), we obtain:

where δ _k and δ _{k + 1} are the quantization errors that occur during the conversion of error signals generated by the pulses of the useful signal with numbers k and k +1.

The first term in expression (4) is quantized with a step τ _{Э the} duration of the working time interval, the second is an additional error caused by the error of the mutual calibration of the channels, and the third is the quantization error.

Random errors δ ₁ , δ ₂ , δ _k , δ _{k + 1} have the same nature, therefore their variances are equal. As a rule, the value ( Nk ₊₁ - Nk ) is a random variable, and ⎪ Nk ₊₁ - Nk ⎪ >> ⎪ j ₁ - j ₂ ⎪, which implies that the calibration error is the dominant source of the random transformation error.

раз. Типовая величина - ⎪j ₁-j ₂⎪ равна 1…2, а величина разности ⎪Nk ₊₁-Nk⎪ может иметь порядок 10…100. Таким образом, полагая шум квантования независимым и равномерным, требуется усреднять 10²…10⁴ калибровочных преобразований перед выполнением преобразования рабочего интервала, что существенно увеличивает длительность процесса калибровки.To reduce the contribution of calibration errors in the prototype device, it is proposed to carry out several calibration iterations, followed by averaging their results. From the expression (4) it follows that to ensure the error caused by the calibration error at least at the level of quantization error in the “exact” transformation channel, it is necessary to reduce the value (δ ₁ -δ ₂ ) in

time. The typical value - ⎪ j ₁ - j ₂ ⎪ is 1 ... 2, and the difference value ⎪ Nk ₊₁ - Nk ⎪ can be of the order of 10 ... 100. Thus, assuming that the quantization noise is independent and uniform, it is necessary to average 10 ² ... 10 ⁴ calibration transformations before performing the conversion of the working interval, which significantly increases the duration of the calibration process.

Due to changes in external factors affecting the device, as well as under the influence of internal noise of the active elements of the device, the ratio T ₀ / τ _E changes, which requires periodic calibration for applications in which high conversion accuracy is required.

Thus, the prototype device, due to the applied auto-calibration, provides a solution to the technical problem of determining the ratio of the units of the least significant bits of the channels of the “rough” and “accurate” transformations, i.e. determine the ratio of T ₀ / τ _E.

However, this solution requires the need for periodic (and for a significant period of time) interruption of the process of converting the working time interval for the calibration procedure. For a given accumulation time, this reduces the number of transformations of the working time interval, i.e. the sample size over which the results are averaged, which can lead to an increase in the quantization error. In addition, in some applications, for example, in phase-locked loops, switching to calibration mode with the suspension of converting the working time interval leads to opening the phase-locked loop for a time equal to the duration of the calibration procedure, which in some cases is unacceptable.

The technical result, which the utility model aims to achieve, is to create a converter of time intervals into a digital code with automatic calibration, which provides a solution to the technical problem of determining the ratio of the units of the least significant bits of the channels of the "coarse" and "exact" transformations (determining the ratio T ₀ / τ _E ), which calibration is carried out without interrupting the process of converting the useful signal, i.e. in the background, while converting the working time interval into a digital code.

The essence of the claimed utility model is as follows. The device for converting time intervals into a digital code with automatic calibration contains a series-connected generator, a counter and a serial register, the output of which is connected to the first signal input of the code calculator, the delay line with taps is connected in series, the position register and the position code to binary converter, the output of which is connected to the second the signal input of the code calculator, as well as the variable delay line, the output of which is connected to the clock input of the position register, and the control unit a line whose corresponding outputs are connected to the control input of the variable delay line and the control input of the position code to binary converter, while the clock input of the code calculator and the clock input of the serial register are connected to the signal input of the device. Unlike the prototype, a filtering unit is connected in series to the device, the input of which is connected to the output of the code calculator, the dispersion calculation unit and the minimum search unit, the output of which is connected to the third signal input of the code calculator, while the signal input of the variable delay line is connected to the signal input devices, and the input of the delay line with taps to the output of the generator.

The essence of the utility model and its feasibility are illustrated by the drawings shown in FIG. 1 and 2, where:

in FIG. 1 shows a structural diagram of a prototype device,

in FIG. 2 is a structural diagram of the inventive device.

The device for converting time intervals into a digital code with automatic calibration in the considered implementation example contains, see FIG. 2, the generator 1, the counter 2 and the serial (shift) register 3 are connected in series, the output of which is connected to the first signal input of the code calculator 4. The device also contains a series-connected delay line 5 with taps, a position register 6 and a position code to binary converter 7. The input of the delay line 5 is connected to the output of the generator 1. The taps of the delay line 5 are connected to the corresponding inputs of the bits of the register 6. The outputs of the bits of the register 6 are connected to the inputs of the bits of the position code converter 7 into binary, the output of which is connected to the second signal input of the code calculator 4. The clock input of the register 6 is connected to the output of the variable delay line 8. The control input of line 8 of the variable delay and the control input of the converter 7 of the positional code in binary are connected to the corresponding outputs of the control unit 9. The signal input of line 8 of the variable delay is connected to the signal input of the device, to which the clock input of register 3 and the clock input of the calculator 4 of the code are also connected. The output of the code calculator 4 is connected to the input of the filtering unit 11, the output of which is connected to the input of the dispersion calculation unit 12. The output of the dispersion calculation unit 12 is connected to the input of the minimum search unit 13, the output of which is connected to the third signal input of the code calculator 4.

The generator 1 serves, as in the prototype, for the formation of clock pulses. Counter 2 and register 3 form a “coarse” conversion channel intended for “coarse” conversion into a digital code of the working time interval between input pulses arriving at the signal input of the device. The delay line 5 with taps, register 6, the position code to binary converter 7 and the variable delay line 8 form an “exact” conversion channel designed to convert the error that occurs during the “rough” conversion into a digital code. The code calculator 4, the filtering unit 11, the dispersion calculating unit 12, and the minimum search unit 13 combine the results of the “coarse” and “accurate” transformations with estimating the variance of the output signal and searching for the true calibration ratio T ₀ / τ _E according to the criterion of minimizing the output noise variance devices.

In an advantageous embodiment, the device is implemented in the form and according to the manufacturing technology of an extra-large integrated circuit (VLSI), where all the functional elements and the conductors connecting them are located on one substrate, closed by a housing and equipped with leads for external connections.

The operation of the device is as follows.

Pulses are received at the signal input of the device, the time interval between which (the working time interval) is to be converted into a digital code. The same input pulses as clock signals arrive at the clock inputs of register 3 and code calculator 4.

The generator 1 generates clock pulses that are fed to the input of the counter 2 and to the input of the delay line 5 with taps.

Counter 2 counts the clock pulses coming from the output of the generator 1. At the moment the kth input pulse arrives at the signal input of the device (and, respectively, at the clock input of register 3), the contents of N _{k of} counter 2 are recorded in register 3. The fixing results from the output of register 3, which are the results of a “rough” conversion, are sent to the first signal input of the calculator 4 code.

The same k- th input pulse is fed to the input of line 8 of the variable delay.

The time interval between the kth input pulse and the next next clock pulse of the generator 1 is an error signal, which is then converted into an “exact” channel as follows.

The clock pulse of the generator 1, propagating along the delay line 5, sequentially appears on its taps with a time delay, which determines the resolution of the “accurate” conversion channel.

Register 6 captures the logical state of the signals at the taps of the delay line 5 at the time of the appearance of the variable delay of the kth input pulse at the output of line 8. Further, the Converter 7 positional code in binary provides a representation of the positional data of the register 6 in the required binary format, which is denoted as n _k . These data from the output of the Converter 7 are transmitted to the second signal input of the calculator 4 code.

The next ( k +1) -th input pulse initiates a similar sequence of changes in the state of registers 3 and 6 to obtain data Nk ₊₁ and n _{k + 1} and transfer them to the computer 4 code.

The duration of the time interval τ _k between two consecutive input pulses with numbers k and k +1 is determined similarly to the prototype device in the code calculator 4:

where T ₀ / τ _E - calibration factor,

T ₀ - the period of the frequency of the generator 1,

τ _e is the resolution of the “exact” conversion channel.

From the output of the code calculator 4, the conversion results are transmitted through the filtering unit 11 to the dispersion calculation unit 12.

To improve the accuracy of the calculations, the delay of the input pulses in the variable delay line 8 is changed according to a certain law established by the control unit 9, which allows working on different sections of the conversion function even at a constant working time interval and providing quantization noise decorrelation.

The dispersion of the conversion results at the output of the code calculator 4 has several components: quantization noise, calibration error, and also random noise τ _w introduced by line 8 of the variable delay. Random noise τ _W is a nuisance in estimating the variance of quantization noise. Therefore, the control of the variable delay line 8 is such that the spectrum of the noise sequence (τ W ) lies outside the frequency band of the useful signal and can be suppressed to the required level in the filtering unit 11. In this case, the frequency parameters of the filtering unit 11 are selected so that the spectrum of the result of the conversion of the working time interval has minimal distortion.

In the block 12 for calculating the variance of a sample of a given volume, the variance of the conversion results is estimated. The sample size can be calculated based on a given error and reliability of the variance estimate.

The dispersion estimates obtained in block 12 are fed to the input of the minimum search block 13, in which, using one of the known algorithms for minimizing the objective function, a new value of the calibration ratio T ₀ / τ _{E is} calculated. A new value T ₀ / τ _{E is} supplied to the third signal input of the code calculator 4, in which for this new value the variance is estimated using a new sample of conversion results.

Thus, the device performs an iterative search for the true calibration coefficient T ₀ / τ _E , which corresponds to a more accurate result of converting the time interval into a digital code.

The above shows that the utility model is feasible and ensures the achievement of a technical result, which consists in creating a converter of time intervals into a digital code with automatic calibration, in which calibration (determination of the calibration coefficient T ₀ / τ _E ) is carried out without interrupting the process of converting the useful signal, i.e. in the background (simultaneously with the conversion of the working time interval into a digital code).

Information sources

1. Christiansen J. / Picosecond stopwatches. The evolution of time-to-digital converters // IEEE Solid State Magazine, 2012, pp. 55-59.

2. Henzler S. / Time-to-Digital Converters // Dordrecht Heidelberg London New York Springer, 2010, p. 124.

3. Samarah A., Carusone A.C. / A Digital Phase-Locked Loop with Calibrated Coarse and Stochastic Fine TDC // IEEE Journal of Solid-State Circuits, 2013 / V 48, p. 1829-1841.

4. Ito S. et all / Stochastic TDC architecture with self-calibration // IEEE Asia Pacific Conference on Circuits and Systems, 2010, p. 1027-1030.

5. US 8362932 (B2), H03M 1/48, publ. 01/29/2013.

Claims (1)

A device for converting time intervals into a digital code with automatic calibration, containing a series-connected generator, counter and serial register, the output of which is connected to the first signal input of the code calculator, series-connected delay line with taps, the position register and the position code to binary converter, the output of which is connected to the second signal input is a code calculator, as well as a variable delay line, the output of which is connected to the clock input of the position register, and the unit the circuit, the corresponding outputs of which are connected to the control input of the variable delay line and the control input of the position code to binary converter, while the clock input of the code calculator and the clock input of the serial register are connected to the signal input of the device, characterized in that the filter block is connected in series, the input of which is connected to the output of the code calculator, the dispersion calculation unit and the minimum search unit, the output of which is connected to the third signal input of the subtraction code splitter, while the signal input of the variable delay line is connected to the signal input of the device, and the input of the delay line with taps is connected to the output of the generator.

Images