Method for reducing the out-of-band emission in am transmitters for digital transmission

Method of reducing out-of-band radiation in AM transmitters for digital transmission

description

The invention relates to the field of AM broadcasting (AM - amplitude modulation), which are to be converted in the course of digitization of analog broadcasting to digital transmission.

The usual types of transmitters are non-linear AM transmitters with RF input (radio frequency) and audio input, which should continue to be used. This has the following reasons:

- The AM transmitters work internally in switching operation and therefore have up to a factor of 3 better efficiencies than linear transmitters, which are usually used for digital transmission z. B. DAB (Digital Audio Broadcasting) and DNB (Digital Video Broadcasting) are used. This results in a saving in operating costs.

- The broadcasters are easier to convince to migrate from analog to digital, if no major investments in advance.

The use of a non-linear AM transmitter for digital modulation requires a special mode of operation of the transmitter. In the analog AM case, only the envelope of the RF (High Frequency) oscillation is affected according to the message signal. If a digital signal were fed into the modulator instead of an audio signal, the result would be "on-off-keying" (OOK) or something equivalent to "amplitude-shift keying" (ASK) as the digital modulation. In the vector diagram, OOK moves or ASK only on the positive real axis.

In digital modulations, however, it is common and, because of the better signal-to-noise ratio, it is also necessary for the entire complex level to be traversed by the digital signal. If one considers the respective sampling points of a digital signal in the vector diagram, one obtains the associated phase star. Since its nominal points are distributed on 4 quadrants, and not as in OOK or ASK are on a straight line, the minimum distances of the setpoints (at the same energy expenditure for the transmitted digital signal) are greater.

The generation of the modulated digital signal takes place by means of two mutually orthogonal partial signals (I and Q). The I-signal ("in phase") is modulated to a cosine oscillation with the frequency Ft (carrier frequency). The Q signal ("quadrature") is modulated to a sine wave of the same frequency Ft. The sum of both modulated oscillations gives the complex modulated data signal (cosine 0-180 degrees, sine -90

- +90 degrees). The modulated I / Q signal is shaped by filters to have exactly the prescribed waveform with the desired bandwidth.

However, the modulated I / Q signal must be converted so that the two signals amplitude signal (A signal) and phase-modulated carrier signal (RF-P) arise therefrom, which are suitable to drive the AM transmitter correctly (see Fig. 2) , At the output of the AM transmitter, the modulated I / Q signal with higher power results again afterwards.

The modulated I / Q signal corresponds to a Cartesian representation. This is converted into a polar representation with amplitude and phase. The amplitude signal (A signal) is thereby obtained for driving the AM transmitter at the audio input. From the initially generated phase signal (P signal), a phase-modulated RF (RF-P signal) is generated. Advantageously, the RF-P signal can be obtained directly without the intermediate step via the P signal. In this way, the necessary signals for driving the AM transmitter arise:

- Amplitude signal (A signal) to control the audio input

- Phase-modulated RF signal (RF-P signal) for controlling the RF input

The A signal is input to the modulator input (audio input) of the AM transmitter, and the RF P signal is used to drive the transmitter in the RF manner. In the transmitter output stage, the two signals A & RF-P are multiplicatively combined and form the high-frequency digital output signal. This is theoretically identical to the correspondingly amplified complex modulated I / Q signal before coordinate conversion.

This is state of the art, described z. In EP 0 708 545 or DE 197 17 169.

In practice, the digital output signal of a transmitter is not exactly identical to an amplified modulated I / Q signal. Rather, the output signal differs due to the non-linear distortion resulting from the modulation process and the resulting unwanted spurious emissions.

The transformation Cartesian - »polar required for the digital operation of AM transmitters is very strongly nonlinear. This has the consequence that both the A signal and the RF-P signal have very large bandwidths and thus considerably wider than the bandwidth of I and Q signal corresponds. This results in the following problems:

1. The transmitter must have a much larger bandwidth (factor 5 or more) in its A branch than is necessary for analog AM. Second The transmitter must also have a larger bandwidth (factor 5 or more) in the RF branch than is necessary for analog AM. Third The signal propagation times of the two branches must be identical (except for fractions of a microsecond) so that the A and RF-P signals can be combined simultaneously in the transmitter output stage. 4th The over the channel boundaries (i. H. Bandwidth of the I / Q signal)

Spectral components of A-signal and RF-P signal must compensate each other in the transmitter output stage. Otherwise, unwanted spurious emissions (Out of Band Emissions - OOB & Spurious Emissions - SE) are generated. Since a transmitter output stage is an analogue structure, the necessary compensation of the spectral components projecting beyond the channel boundaries is only incompletely possible. Practice shows that these unwanted spectral components with reasonable effort only by approx. 30 dB (factor 1000 in the power) to approx. 36 Suppress dB (factor 4000). These are the distances between the size of the useful spectrum and the interference spectrum of the spurious emissions. These distances are called "shoulder distances". Shoulder distances of 40 dB (factor 10 000) to 55 dB (factor 320 000) are required so that the permissible spectrum mask is adhered to according to the specifications of the ITU (International Telecommunications Union, Geneva).

The ITU Spectrum Mask (ITU-R SM.328-9, Spectra and Bandwidth of Emissions: 3.6.1.3 Out of Band Spectrum) determines how large the unwanted emissions of an AM transmitter may be in the highest case. The international coordination of the AM stations sets the absolute maximum values ​​of the unwanted transmissions using the ITU spectrum mask.

In the immediate vicinity of the useful spectrum, the mask specifies a tolerance field with the aid of which it is possible to comply with the conditions of the ITU mask. After this, for a given shoulder distance, the out-of-band radiation must only decrease sufficiently quickly so that the ITU mask is not exceeded. FIG. 1 shows this for some examples, assuming here that the digitally generated sideband power is exactly the same as the nominal analog sideband power (-12 dB below carrier power) for the ITU mask.

From this representation it can be seen that the slope (resp. the waste) of the out-of-band radiation must be greater, the lower the shoulder distance achieved in the transmitter. If for technical reasons in the transmitter output stage z. B. Therefore, only a shoulder spacing of 35 dB is achievable, therefore, the digital modulation signal must be controlled to provide at least an OOB slope of 10 dB / 7.2 KHz at 10 KHz channel bandwidth, generally 10 dB / (0.72 x bandwidth) comes. Influencing the digital modulation signal in the transmitter and its effect on the OOB slope to reduce unwanted spurious emissions is the content of this invention.

The digital modulation signal, when no particular precautions have been taken, moves in the representation of the vector diagram from any allowable complex point to any other allowable complex point on curved paths that connect all allowable points without kinking. This has the consequence that in the vector diagram very many zero crossings or almost zero crossings come about.

In the transformation Cartesian -> polar thereby arise in the A-signal sharp peaks with zero touch or almost zero touch and jumps to Pi in the phase. Both properties mean that both the A signal and the RF-P signal receive a very large bandwidth. This is undesirable because of the compensation required in the final stage and must be avoided.

If the peaks in the A signal and thus at the same time the phase jumps are avoided, the direction reversal in the A signal is retained, but is not so fast. Likewise a phase transition of Pi is preserved (instead of a jump), but also not as fast as in a jump. As a result, the bandwidths of A-signal and RF-P signal are much narrower and the slope of the spectra in question becomes larger. Overall, this makes the compensation process in the transmitter output stage less critical and the unwanted emissions decrease.

A closer analysis of the processes shows that the spectrum is not very close to

Changes frequencies that are close to the useful signal (channel boundary), which means that the shoulder distance can not be significantly increased by this measure. Its size is also dependent on the degree of compensation available. On the other hand, the slope in the spectrum of the out-of-band radiation increases, so that the desired value of 10 B / (0.72 x bandwidth) can be achieved if 35 dB of shoulder distance result from the compensation.

The avoidance of "zero contacts" of the A signal and thus of the phase jumps is achieved by virtually "drilling" the vector diagram at the 0/0 point. This means that a modulation must be chosen that avoids the point 0/0 in the vector diagram.

The DRM system (Digital Radio Mondiale) for digital transmission in the AM areas, which uses an OFDM (Orthogonal Frequency Division Multiplex) method of transmission, is recommended by the ITU for standardization because of the noise-like nature of the Transmit signal not so readily possible to create a "hole" in the vector diagram. Therefore, the DRM system places very high linearity requirements on the transmitter, which therefore requires correspondingly large shoulder distances. This means that new, highly linear AM transmitters have yet to be developed for the DRM system in order to meet the conditions for unwanted emissions according to the ITU spectrum mask. This is expensive and tedious and may jeopardize the introduction of the DRM system as a whole.

By using a modulation method with a "hole" around the 0/0 point, it is possible to use existing AM transmitters with today's conventional shoulder distances smaller than 35 dB for digital transmission in the AM bands, as otherwise by DRM in the "Service Requirements" is required. The reusability of existing channels has the highest priority for broadcasters. In addition, only a digital modulator for the preparation of the A signal and the RF-P signal is required (see FIG. 2). This makes it possible to change from analogue to digital without major costs in relation to the transmitter. For the digital transmission in the previous AM ranges, modulation methods are proposed which have a "hole" around the point 0/0 in the vector diagram. Such modulation methods are known as offset modulations or coded modulations.

Such modulations are applied z. For example, in transmitters that are unable to reduce the amplitude of the radio frequency to zero. Examples can be found in satellite transponders with traveling wave tubes or in GSM mobile phones with transmitter amplifiers in C mode.

Particularly suitable is a modulation type derived from 16APSK (16-point amplitude-phase-shift keying). This can be explained by the fact that, especially for difficult shortwave transmissions, a smaller number of set points (eg 16) leads to a higher net data throughput than the use of a higher level modulation (eg 64 setpoints) because the lower level Modulation is inherently more robust and requires less error protection coding.

Example 1: Modification of an OFDM signal

OFDM signals have a fairly rectangular spectrum, but in the time domain noise-like character, both for the I component and for the Q component of the time signal. This is a consequence of the superimposition of many independent subchannels.

If a certain degradation, ie a slight increase in the bit error rate with given signal / noise ratio, is accepted in the case of OFDM signals, a "hole can be drilled" in the vector diagram.

To do this, the I (t) and Q (t) portions of the OFDM baseband signals must be modified. Only then can the I / Q-A / RF-P conversion and the transposition to the frequency of the transmitter take place. The I (t) and Q (t) baseband signals are AC voltages and therefore each have their own zero crossings. The critical case leading to an approximation or touching of the point 0/0 in the vector diagram arises when I (t) and Q (t) both have a zero crossing at the same time or only slightly differently in time. This can be seen if I (t) is considered as x and Q (t) as y in an x ​​/ y coordinate system (Cartesian). So if x = 0, then y = 0 must not be the same time, otherwise the origin of coordinates 0/0 is hit, which must be avoided, however.

The problem can be solved by setting thresholds + S₀, + S_U, -S ", -S_u with S₀> P_u are defined and the size of I (t) and Q (t) is constantly compared with these threshold values ​​(see Fig. 3). The arrows mark a location where correction is required to shift the I (t) signal in its zero crossing. The I (t) and Q (t) signals are rounded in practice and have no kinks.

Considering the case that Q (t) is the first of the two signals, coming from positive values, the lower threshold + S_u If it has fallen short of negative values, it would be the threshold - S_u, I (t) will immediately follow a zero crossing. For I (t) it is examined whether the signal, coming from positive values, is the threshold + S₀ or coming from negative values, the threshold - S₀ has fallen below. If this is the case, then it is to be expected that shortly afterwards also the corresponding S_uThreshold is exceeded and then the zero crossing comes. This meant that the zero crossings of the signals I (t) and Q (t) follow each other too closely, which should be avoided after the previous consideration.

Zero crossings of the I (t) and Q (t) signals can not be avoided completely. What is to be achieved here is that the zero crossings are pushed apart so far in time that not both signals at the same time have low amplitude values. For the example described, this means that the I (t) signal must be influenced in such a way that the zero crossing takes place sufficiently far in time from the zero crossing of the Q (t) signal. For this purpose, in the example, a rounded pulse is added to the I (t) signal whose sign depends on the sign of the S_uThreshold is selected. This briefly "bends" the I (t) signal so that it makes its zero crossing at a sufficient distance from the Q (t) signal.

The additional pulse has a favorable cos²Form or a Gaussian bell curve, which must be chosen so that the bandwidth of the transmitted signal is not increased. Its amplitude is determined from the slope of the I (t) and Q (t) signals, respectively, where the amplitude is to be chosen in proportion to the slope.

The modified I (t) and Q (t) signals are converted to I / Q → A / RF-P and fed to the AM transmitter as described above. As a result of the modification, the signals A (t) and RF-P (t) in the transmitter have a lower bandwidth than in the modulation with the OFDM baseband.

In the transmitter output stage, where the signals A (t) and RF-P (t) combine to produce the lower bandwidth output, namely channel bandwidth, less compensation is required due to the modification in the spectrum, resulting in lower out-of-band radiation (OOB). is reached.

The reduction in out-of-band radiation is practically barely noticeable in the height of the shoulder distance, but increases the slope with which the OOB decreases on both sides. The pitch increases the larger the "hole" in the vector diagram is selected.

However, the "hole" in the vector diagram can not be made arbitrarily large because the OFDM signal is corrupted by the additional pulses. This is synonymous with the intentional addition of a fault. Therefore, it is necessary to choose a favorable compromise between OOB and bit error rate. It has an advantageous effect that the disruption also affects the pilot symbols transmitted in the OFDM signal to which the receiver refers, because the disturbance thereby has a less serious effect.

Example 2: Modification of a BPSK Test Sequence

For the measurement of the radio channel and also for the synchronization of the receiver, a BPSK test sequence (binary phase shift keying = biphase shift keying - pseudo random or cazac) can be used. The properties of the test sequence must be determined so that no inadmissibly high spurious emissions occur.

An unmodified BPSK test sequence has a frequent change of sign. In the transformation Cartesian -> polar thereby arise in the A-signal sharp peaks with zero touch or almost zero touch and jumps to Pi in the phase. Both properties mean that both the A signal and the RF-P signal receive a very large bandwidth. This is undesirable because of the compensation process required in the transmitter output stage and must be avoided.

The BPSK test signal is therefore modified to create a "hole" in the vector diagram. The modified test signal thus belongs to a class of modulations that have a "hole" around the point 0/0 in the vector diagram.

The unmodified BPSK test sequence looks in the time domain in sections as shown in Fig. 4-A. The pulses have the size 1 and a regular time interval (clock) to each other. The known algorithms for evaluating the test sequence should also be applicable to the modified test sequence.

The modification consists in that in a first step the clock rate is doubled and in each case between two pulses of the same sign a further pulse is arranged exactly between them, as shown in FIG. 4-B shown in dashed lines. In a second step, everywhere (and only there), where a change in the sign of the pulses takes place in the unmodified test sequence, a pulse is inserted exactly in between (see FIG. 4-C). The pulses are smaller than 1 in size (eg. 0.2) and in the direction such that a positive pulse is formed in a transition from + to - in the unmodified test sequence and a negative pulse in the case of a sign change from - to + (definition I).

The definition is arbitrary, so that the signs of the pulses in the modified test sequence can just as well be chosen the other way around, which should be referred to as Definition II.

The modified signal (Figure 4-C) produces a rotation in the modulated signal when viewed in the complex plane. The rotation has an asymmetry in the spectrum. To avoid the imbalance, the pulses are defined at each repetition of the sequence according to the definition II. This creates a constant alternation of definitions for the pulse sequence.

Before modulation, the signals of the modified test sequence are passed through rounding filters (eg root raised cosine with roll-off = 0.2).

The modulator for the modified and filtered test sequence has I / Q shape, wherein the signal according to FIG. 4-B is modulated as I-signal and the signal according to FIG. 4-C as Q-signal.

This ensures that the vector diagram of the modulated test signal receives a "hole" in its origin.

At the receiving end, no complex correlation is required. Rather, it is sufficient to correlate the reception-side I & Q signals separately with the signal formed in the receiver according to FIG. 4-B. Subsequently, a separation of the received I and Q shares is possible.