EP3308465A1

EP3308465A1 - Method of synthesis of analogue noise, noise synthesizer and coding chain using such a synthesiser

Info

Publication number: EP3308465A1
Application number: EP16725535.5A
Authority: EP
Inventors: Bruno LELONG; Patrice BARRET
Original assignee: Thales SA
Current assignee: Thales SA
Priority date: 2015-06-10
Filing date: 2016-05-27
Publication date: 2018-04-18
Also published as: FR3037456A1; FR3037456B1; WO2016198269A1; US10097194B2; US20180152195A1

Abstract

The method comprises at least the following steps:- Generating (1) pseudo-random noise in the digital domain coded on a number N of bits, sampled at a given frequency FH/N; - Multiplexing (2) in the digital domain the binary signals produced by each of the N bits at; sampling frequency FH so as to obtain noise coded on one bit at said frequency FH a - Transferring (3) the noise thus coded into the analogue domain via a low-voltage differential transmission interface; - Filtering (4) the analogue signal in a passband which may be centred on half the sampling frequency of an analogue/digital converter (23).

Description

METHOD FOR SYNTHESIZING AN ANALOGUE NOISE,

The present invention relates to a method of synthesizing an analog noise. It also relates to a high output level noise synthesizer implementing this method. Finally, it relates to a coding chain with very high linearity and very high dynamics using such a synthesizer. It applies in particular for the correction of non-linearity defects present in the analog-to-digital conversion chains.

In the field of analog-to-digital converters (ADCs), it is necessary to limit the non-linearity defects of these converters. These defects cause a degradation of the signal performance before sampling, which has the effect of introducing harmonic distortion. For radar applications in particular, this degradation greatly limits the sensitivity of the radar whose detection threshold must then be raised to not generate false detections on the harmonic ranks of the main signal.

To limit this phenomenon, a suitable solution consists in coupling to the signal a noise, which makes it possible statistically to limit the periodic passage on the points of non-linearity. However, for this to be effective, it is necessary that the power of this noise is large enough to address all the converter codes but it is also essential that this noise can be eliminated after digital acquisition. In the context of a radar application, the useful band of the signal to be processed being known, the choice is to inject a power of noise into a band not covering the useful band of the signal so as to easily eliminate the noise by filtering. high or low depending on the frequency positioning of the noise.

The generation of this noise and more particularly the implementation of this synthesis can pose a congestion problem, particularly in compact applications. Indeed, for compact applications it is essential to limit the surface of this functionality while the Useful bands to process are important, the frequencies involved are very high.

A technical problem to be solved is thus to synthesize an analogue noise in a high but limited frequency band, for example around 350 MHz to +/- 30 MHz, with a relatively high overall power, and this in a volume reduced.

Solutions are known to generate analog noise. In a first solution, the noise generation is performed by means of an amplified noise diode. This principle requires a relatively large volume and brings high constraints. Indeed, the noise power is extended over a very wide frequency domain and requires significant filtering to keep only the desired band. In addition, after filtering the noise power is very low, which requires several stages of amplification with risks of instability related to the high gain of the chain.

In another solution, the noise is generated from a digital-to-analog converter (DAC). The principle is then to synthesize noise in the digital domain and to convert it to analogue via a DAC. With this principle, it is possible to limit the noise band generated but like any sampled system, it is necessary to filter in analog to limit all the image bands. A major disadvantage of this solution with respect to the intended application is that the frequencies used require the use of a very high sampling rate DAC, which requires the generation of an additional external clock and a additional power consumed, and also does not remove the analog amplifier stage.

An object of the invention is in particular to overcome the aforementioned drawbacks, by allowing noise generation centered around an intermediate or microwave frequency with a frequency band sufficient to effectively linearize the encoders and that with very few components.

For this purpose, the subject of the invention is a method for synthesizing an analog noise, said method comprising at least the following steps: Generating a pseudo-random noise in the digital domain coded on an N number of bits sampled at a given frequency FH / N;

Multiplexing in the digital domain the binary signals produced by each of the N bits at a sampling frequency FH to obtain a noise coded on a bit at said frequency FH;

- Transfer the noise thus coded in the analog domain via a differential transmission interface.

In a particular embodiment, the transfer step in the analog domain is followed by an analog filtering step by a bandpass filter.

Said analog noise being able to be combined with a useful signal at the input of an analog-digital converter, the bandwidth of said analog filter is for example centered on half of the sampling frequency of said converter, said bandwidth does not exceed not overlapping the frequency band of said useful signal. The filtering step may be followed by a noise amplification step.

Said analog noise being able to be combined with a useful signal at the input of an analog-digital converter, the noise coded on a bit at the output of the multiplexing step is for example centered on a central frequency Fc equal to the half of the sampling frequency of said converter.

The number N being for example equal to 4, said pseudo-random noise is sampled at a frequency equal to 2/3 of said central frequency Fc.

The subject of the invention is also an analog noise synthesizer comprising at least the following modules:

A module for generating a pseudo-random noise in the digital domain coded on an N number of bits, sampled at a given frequency FH / N;

A multiplexer performing in the digital domain the multiplexing of the binary signals produced by each of the N bits at a frequency FH sampling to obtain a noise coded on a bit at said frequency FH;

A differential transmission interface for transferring the noise thus encoded in the analog domain. The differential transmission interface is for example of the LVDS type.

Said synthesizer comprises, for example, an analog band pass filter at the output of the differential transmission interface. Said analog noise being able to be combined with a signal useful at the input of an analog-digital converter the bandwidth of said filter is for example centered on half of the sampling frequency of said converter, said bandwidth not overlapping the frequency band of said useful signal.

Said analog noise being able to be combined with a useful signal at the input of an analog-digital converter, the noise coded on a bit at the output of the multiplexer is for example centered on a central frequency Fc equal to half the frequency sampling of said converter.

In a particular embodiment, the number N being equal to 4, said pseudo-random noise is sampled at a frequency equal to 2/3 of said central frequency Fc. Advantageously, the pseudo-random noise generation module and the multiplexer are for example made in an FPGA, the differential transmission interface being a differential output of said FPGA.

The invention also relates to an analog-digital coding string, said string being able to encode a useful signal. It comprises at least: a digital-to-analog converter;

a synthesizer as described above;

a combiner combining said useful signal and the noise generated by said synthesizer; the output of said combiner being connected to the input of said converter, so that the combined signal is digitally converted. The digitized noise is for example filtered at the output of the converter by a digital filter.

The said coding chain is particularly suitable for use in a radar reception chain, the said useful signal being a radar reception signal.

Other characteristics and advantages of the invention will become apparent with the aid of the description which follows, given with regard to appended drawings which represent:

FIG. 1, a representation of an exemplary functional architecture of a noise synthesizer implementing the method according to the invention;

FIG. 2, an exemplary embodiment of a coding chain according to the invention.

FIG. 1 illustrates the functional architecture of a noise synthesizer implementing the method according to the invention.

In the method according to the invention the noise is synthesized on a bit shaped in mixed technology, digital and analog, around a high frequency. Initially, the digital noise signal is generated in one bit and transferred to the analog domain via a differential interface, and then optimized in the desired frequency band via a filter stage.

The functional architecture of Figure 1 shows the operating principle thus described in outline.

The generation of noise 1 is therefore first performed in the digital domain, for example in an FPGA, on the principle of a digital synthesis advantageously coded on a bit, the noise being presented as a succession of binary values sampled at a frequency clock FH.

More particularly, in a first step the noise generation 1 is carried out by parallelising N low frequency channels, each channel generating a noise clocked at the frequency FH / N. This is equivalent to Generate a 1-bit random sequence from N 1-bit subsampled channels.

In a second step, the noise is conformed by digital multiplexing 2 of the N channels giving a noise signal coded on a bit, sampled at the clock frequency FH. In practice, N is for example equal to 4. It is necessary to understand by consistent noise a noise which is not a white noise but a noise having a given bandwidth, in particular narrow. The central frequency of the noise is for example equal to F _eC h / 2, F _eC h being the sampling frequency of the encoder to be linearized.

This signal synthesized on a bit is transmitted on a buffer 3, this buffer being a low voltage differential transmission interface, also called LVDS (Low Voting Differential Signal), particularly well suited to very high frequencies, typically several hundred megahertz. This interface advantageously makes it possible to go from the digital domain to the analog domain without the use of a digital-to-analog converter.

This interface may advantageously be a differential output of the FPGA in which the digital synthesis is already programmed, which makes it possible not to use an additional component.

In a next step, the analog signal at the output of the buffer 3 is then filtered 4 by a band pass filter, the bandwidth of this filter being outside the band of the useful signal with which the noise is intended to be combined. At the output of the filtering 4, the noise can be amplified if necessary.

FIG. 2 shows an example of an analog-digital coding string according to the invention using a noise synthesizer of the type of that of FIG. 1, applied for example in a radar receiver. The useful signal 21, for example a radar reception signal, is combined with the noise signal by means of a combiner 22 before digital conversion by the ADC 23. The noise is then filtered downstream of the ADC.

As described above, the noise generation, also called "dither" thereafter, is first synthesized digitally. In the example of application of FIG. 2, the number N of channels is equal to 4.

A module 1 carries out the generation of a white noise on 4 bits sampled at 2/3 of the central frequency Fc of the desired dither, for example Fc equal to F _e ch / 2. A clock at the frequency FH = 2/3 Fc is used by the module 1 to synthesize noise on 4 bits. The module conventionally generates a pseudo-random code by means of a shift register offset looped back on itself and whose initialization value can be parameterized in the mask of the module. A brewing function can be added to increase the randomness of the generated code. It is possible to modulate the random signal, that is to say to decrease or increase its amplitude, by adjusting an amplitude parameter. This dither function is instantiated twice within the module with different initialization values to generate a noise sampled at 2xFH, ie at 4/3 of the center frequency Fc.

A function of formatting and concatenation of the different noises is performed. The multiplexing of two independent noises and their complemented values makes it possible to maximize the power of the noise around the central frequency Fc, which corresponds to a zone of the spectrum which will be maximized by the analog filtering 4. More precisely, as regards the maximization of the noise in the desired frequency band, it is assumed that this noise must be placed out of band useful, that is to say out-of-band operated by the CAN 23, and it conforms this noise so as to make sure that its contribution in the useful band after sampling remains negligible compared to the noise of the CAN and especially vis-à-vis the noise of the reception chain, in case of radar application for example.

In general, ideally, the noise spectrum is the "spectral complement" of the useful band that one wishes to exploit. Considering that the digital filtering at the output of the CAN 23 is capable of rejecting the conformed noise, the noise injected at the input of the CAN is eliminated at the output.

In practice, to achieve a simple one-bit digital conformation, the natural decay in sin (X) / X can be exploited for sustained sampling and the spectrum is shifted around the frequency where it is desired to maximize the power in performing interpolation by 2 by inserting complemented samples. The ratio between the sampling frequency F _eC h of the ADC and the sampling frequency F _eC hb of the noise makes it possible to place the spectrum spectrally in optimum manner with respect to this objective.

For example, Ni being equal to 0 or 1, for a series of samples (N1, N2, N3, N4, ... Nn) clocked at F _eC h / 2, after interpolation a result of the following form: (N1, -N1, N2, -N2, N3, -N3, N4, -N4 ... Nn, -Nn) sampled at F _eC h.

More generally, this conformation can be much more sophisticated and allow more complex spectra to be numerically conformed using digital Sigma-Delta shaping techniques. The increase in frequency and the miniaturization of the FPGA technologies can make it possible to minimize or even eliminate the analog filtering 4 which will be more particularly described later. Multiplexer 2 multiplexes the 4-bit coded noise to generate a 1-bit sampled noise at 4xFH, which is 8/3 Fc.

As previously described, the signal is then sent to buffer 3 to transfer digitized noise, on a bit, into the analog domain. This transfer is advantageously performed by a differential transmission interface, for example according to the LVDS standard. This interface may be a differential interface of the FPGA used elsewhere for the dither function or noise generation. This has the advantage of using an internal FPGA resource and does not require a specific conversion circuit. For example, when using an LVDS differential interface, the 350 mV output level across a 100 ohm resistor, which is unique to this standard, provides an output power of -8 dBm once the line impedance has been reduced to 50 ohms. This adaptation to 50 ohms is also very simple because the differential interface LVDS works in differential on 100 ohms.

The digital noise synthesis is optimized to maximize the power around the central frequency selected, for example F _eC h / 2, it results at the output of the LVDS interface, that is to say the buffer 3, a power of the order of -10 dBm.

Since the noise generated has a conformation, defined by its bandwidth, performed by digital synthesis, the following analog filtering 4 may be less constrained.

Analogue filtering 4 must be sized to ensure that there is no noise residue in the wanted band of the signal to be converted by ADC 23. The difference between the wanted band and the noise band must be so be enough. Here again, the choice of the central frequency of the noise is optimized and takes advantage of the principles of the sampling which is finally carried out by the CAN.

Indeed, the useful band of the signal can not be positioned astride a multiple of the point V2 F _eC h, F _eC h being as indicated above the sampling frequency of the CAN.

In an advantageous embodiment, it is precisely at this frequency V2 F _eC h that the noise is centered. Thus for a given band, the noise occupies after conversion by the ADC only half of its initial band, the sampling principle having folded the input signals to ^1/2 F _eC h, while having the efficiency of the total noise band.

The bandpass analog filtering is thus added at the output of the differential interface 3 in order to de-correlate the useful band of the noise band. An amplifier 5 may be added depending on the full scale of the CAN or the band of the noise is relatively narrow.

The filtered and possibly amplified noise is combined with the useful signal by the microwave combiner 22, the combined signal being converted into a digital signal by the ADC. The noise combined with the digitized signal is then filtered by a digital filter 24. The advantage of the consistent noise, that is to say with a narrow and controlled bandwidth, is that it can be easily placed out of the useful band signal and therefore simply filtered. Thus, after coding the elimination of the noise can be achieved by a simple digital filter 24 which is generally already present at the output of the ADC.

The invention thus makes it possible to guarantee a complete elimination of the noise in the end contrary to the conventional methods of noise synthesis where the noise occupies the entire sampling band. In these conventional methods the digital image is subtracted after conversion to digital and the effectiveness of this removal is dependent on the quality of the digital representation of the coded noise which is very difficult to obtain. The positioning of the central frequency of the noise at F _eC h / 2 thus makes it possible to obtain maximum efficiency in the correction of non-linearity defects of the ADC, and of the encoders in general, while pushing back the frequency band of the noise. as far as possible from the useful band. In addition to the guarantee of complete suppression of noise, the invention makes it possible to produce a noise generator in a reduced volume, or even zero or almost zero compared to existing equipment. In many applications, including radar, an FPGA is already present because used to interface with the CAN. This available FPGA can contain the one-bit digital noise generation function and the LVDS differential interface to switch to the analog domain.

It is not necessary to use an external clock because all synthesis operations can use the clock resources of other functions, especially in radar reception chains. Optionally, the analog bandpass filter 4 must be added if this resource is not available.

The invention advantageously makes it possible to dispense with a digital / analog conversion component of the noise. Since the noise signal is coded on a bit, it is advantageously possible to use a differential interface, for example of the LVDS type as has been described. Advantageously, without changing the synthesis principle of the invention, it is possible to increase the noise frequency as the frequency evolution of this type of interface.

The invention also has another advantage in that it does not require specific processing acquisition output other than the analog filter 4 which limits the noise signal to the useful band to eliminate the noise at the output of the encoder 23. .

In the case of an application to a radar reception system, the noise generated according to the invention may be in a location of the spectrum distant from the frequency band of the received signals, and this noise is naturally filtered by the radar reception chain , without the need for additional components or functions.

The invention advantageously makes it possible to obtain a coding chain with very high linearity and very high dynamics. The noise frequency can reach frequencies of the order of 1 GHz.

Claims

1. Process for synthesizing an analog noise, characterized in that it comprises at least the following steps: generating (1) a pseudo-random noise in the digital domain coded on an N number of bits, sampled at a given frequency FH / N;

- Multiplexing (2) in the digital domain the binary signals produced by each of the N bits at a sampling frequency FH to obtain a noise coded on a bit at said frequency FH;

- Transfer (3) the noise thus encoded in the analog domain via a differential transmission interface.

2. Method according to claim 1, characterized in that the transfer step in the analog domain is followed by an analog filtering step (4) by a band pass filter.

3. Method according to claim 2, characterized in that said analog noise being adapted to be combined with a useful signal (21) at the input of an analog-digital converter (23), the bandwidth of said analog filter (4). ) is centered on half of the sampling frequency of said converter, said bandwidth not overlapping the frequency band of said wanted signal.

4. Method according to any one of the preceding claims, characterized in that it comprises a step of amplifying (5) the noise in the analog domain.

5. Method according to any one of the preceding claims, characterized in that said analog noise being able to be combined with a useful signal (21) at the input of an analog-digital converter (23), the noise coded on a bit at the output of the multiplexing step (2) is centered on a central frequency Fc equal to half the sampling frequency of said converter (23).

6. Method according to any one of the preceding claims, characterized in that the number N being equal to 4, said pseudo-random noise is sampled at a frequency equal to 2/3 of said central frequency Fc.

Analogous noise synthesizer, characterized in that it comprises at least the following modules:

A module (1) for generating a pseudo-random noise in the digital domain coded on an N number of bits sampled at a given frequency FH / N;

A multiplexer (2) performing in the digital domain the multiplexing of the binary signals produced by each of the N bits at a sampling frequency FH to obtain a noise coded on a bit at said frequency FH;

A differential transmission interface (3) for transferring the noise thus encoded into the analog domain

8. Synthesizer according to claim 7, characterized in that the differential transmission interface (3) is of the LVDS type.

9. Synthesizer according to any one of claims 7 or 8, characterized in that it comprises an analog bandpass filter (4) at the output of the differential transmission interface.

10. Synthesizer according to claim 9, characterized in that said analog noise being adapted to be combined with a useful signal (21) at the input of an analog-digital converter (23), the bandwidth of said filter is centered on half of the sampling frequency of said converter, said bandwidth not overlapping the frequency band of said useful signal.

11. Synthesizer according to any one of claims 7 to 10, characterized in that it comprises an amplification stage (5) at the output of said analog filter.

12. Synthesizer according to any one of claims 7 to 11, characterized in that said analog noise being able to be combined with a useful signal (21) at the input of an analog-to-digital converter (23), the noise coded on a bit at the output of the multiplexer (2) is centered on a central frequency Fc equal to half of the sampling frequency said converter (23).

13. Synthesizer according to any one of claims 7 to 12, characterized in that the number N being equal to 4, said pseudo-random noise is sampled at a frequency equal to 2/3 of said center frequency Fc.

14. Synthesizer according to any one of claims 7 to 13, characterized in that the module (1) for generating pseudo-random noise and the multiplexer (2) are made in an FPGA, the differential transmission interface being an output. differential of said FPGA.

15. Analog-digital coding chain, said string being able to encode a useful signal (21), characterized in that it comprises at least: an analog-digital converter (23);

a synthesizer according to any one of claims 7 to 14;

a combiner (22) combining said useful signal and the noise generated by said synthesizer; the output of said combiner being connected to the input of said converter (23), so that the combined signal is digitally converted.

16. Coding chain according to claim 15, characterized in that the digitized noise is filtered at the output of the converter by a digital filter (24).

17. encoder chain according to any one of claims 15 to 16, characterized in that it is suitable for use in a radar reception chain, said useful signal (21) being a radar reception signal.