Cognitive Frequency Exclusion in EN 50561-1:2012
CONTENTS
22.2 Standardisation of ‘Cognitive Frequency Exclusions’
22.3 Overview of Shortwave Radio Broadcasting
22.4 Concept of ‘Cognitive Frequency Exclusion’
22.4.1 Radio Signal Spectrum on Power Lines
22.4.1.1 Fading Effects of Radio Stations
22.4.1.2 Setup to Record a Signal Spectrum
22.4.2 Sensitivity of SW Radio Receivers
22.4.3 Threshold to Detect Radio Services
22.4.4 Requirements of a Notch
22.4.4.1 Bottom Level of the Notch
22.4.4.2 Width or Slopes of the Notch
22.4.5 Adaptive OFDM, Channel and Noise Estimation
22.5 Implementation in a Demonstrator System
22.5.2 Notch Filter Environment
22.5.3 Detection of the Presence of Radio Services
22.5.4.1 Influence of Windowing on Spectrum and Notch Shape
22.5.4.2 Adaptive Band-Stop Filters
22.6 Verification of ‘Cognitive Frequency Exclusions’ in Buildings
22.7 Outlook for Future EMC Coordination
This chapter describes the ‘cognitive frequency exclusion’ concept in detail, specified in European Norm (EN) 50561–1. After an overview of the standardisation history, the high-frequency (HF) radio transmissions, the influence on the power line communication (PLC) spectrum, are shown. Thereafter, an excursion into the sensitivity of shortwave (SW) radio receivers follows. When the sensitivity of these devices is known, the threshold can be defined for receivable radio broadcast signals. If the frequencies to be excluded are known, requirements for a notch are defined. The process of adaptive modulation is described before calculating the impact of notching on the throughput of PLC. This allows the realisation of the notching concept with a minimal loss of throughput and implementation effort. The concept is verified by a hardware demonstrator system after discussing the theoretical aspects.
The EN 50561-1 [1] was approved in autumn 2012 by European national members. As described in Chapter 6, it includes new requirements for conducted disturbances and communications signals at PLC ports, which were never included in a harmonised electromagnetic compatibility (EMC) standard before.
In the past, EMC standards specified fixed and permanent limits which usually were described using a straight line for the given frequency range in a linear or logarithm graph (Figure 6.1). Manufacturers implementing such a norm guaranteed compliance by shielding their device at the time of production. The straight line must not be crossed at any time.
The new requirements in EN 50561-1 are adaptive, flexible and cognitive and have to function only when there is a risk of interference. This type of intelligent concept is also called ‘dynamic spectrum management’, ‘cognitive PLC’ or ‘smart notching’ and runs autonomous on the PLC modems without any user interaction. Compliance is guaranteed at the time and location, when and where, and the PLC modem is operated by the behaviour of the communication system. The principal philosophy is ‘a service which is not present needs not to be protected’. Therefore, EN 50561-1 requests flexible and/or permanent frequency exclusions in the power spectral density (PSD) mask of PLC modems. Compared to state-of-the-art PLC modem design, enhanced protection of radio services is provided in buildings where PLC is operated.
The general requirements of EN 50561-1 request the exclusion of sensitive frequency areas which are allocated by ITU-R Radio Regulations [2] to aeronautical, amateur radio and broadcasting services. The frequency exclusions for aeronautical and amateur spectra are permanent, where the radio broadcast frequencies could be notched permanently or dynamically. Permanent notching results in a significant loss of PLC communication resources versus dynamic notches. They only have to be applied at the few broadcast frequencies when the service is receivable at the location of operation.
22.2 Standardisation of ‘Cognitive Frequency Exclusions’
Reports about interference from PLC to HF radio broadcasts were published quite early by Stott and Salter [3]. The new high-bitrate PLC modems using multiple communication carriers at various frequencies with adaptive constellations enable the realisation of the cognitive frequency exclusion concept. The idea was discussed lively at ETSI PLT [4]. In order to prove that the new technology works reliably, prototype implementations demonstrated their functions under the umbrella of the ETSI special task force (STF 332) performing a plugtest with PLC modems and SW radio receivers. During the ETSI plugtest in October and November 2007, various tests and measurements were performed to stress the concept. A description of the demonstrator system and the plugtest highlights are given in the succeeding text.
After the plugtest and publishing the results, the European Broadcasting Union (EBU) issued a public statement that they are content that any system fully meeting the specification will offer adequate protection to HF broadcast transmissions. This led to the unanimous approval of ETSI TS 102 578 [5] specifying the ‘smart notching’ concept.
The Special International Committee on Radio Interference (CISPR) announced that they will work on adaptive dynamic notching in any future committee draft (CD) or committee draft for voting (CDV) related to PLC equipment in CISPR/I/257/CD [6]. Finally, they adopted the adaptive electromagnetic interference (EMI) mitigation techniques as a normative annex to their latest CIS/I/301/CD [7] and CIS/I/302/DC [8] documents. Accepting adaptivities is almost ‘revolutionary’ in the world of EMI standardisation.
The IEEE included the concept as ‘stand-alone dynamic notching’ to the IEEE 1901 standard [9].
22.3 Overview of Shortwave Radio Broadcasting
The frequency range of conventional PLC modems (2–30 MHz) overlaps with HF radio broadcast frequencies defined by ITU-R [2]. Power line wires in private homes are not shielded, and due to branches, distribution boxes, etc., the power line network is structured with a certain amount of asymmetry. As discussed in Chapters 1 and 5, the asymmetries of the power line network convert the differentially fed signals into common-mode signals (see Figure 1.3), which tend to interfere with radio devices. If an SW radio receiver (amplitude modulation [AM] or digital radio mondiale [DRM] [10]) is operated indoors where a PLC is active, the radio reception quality might suffer. When the radio device is connected to the mains power supply and the radio has an insufficient decoupling at its mains port, the conducted path is dominant in terms of interference. By design, SW radio receivers have insufficient decoupling at the mains port, because they use the mains grid – in the same way that PLC does – for something which it was not designed for: they use it as an antenna. Usually an SW radio receiver is equipped with a whip or monopole antenna. When the receiver is connected to the mains, the counterpoise of the mains grid allows it to generate a dipole antenna improving the reception quality of radio services.
Of course, some might argue that HF radio broadcast is becoming less important. Frequency modulation (FM) radio provides a significantly better signal quality, but it is not a worldwide service. Satellite radio or the upcoming web-radio services compete with today’s AM transmissions. This might be true for the industrial countries or in the developed countries but the demand for information exists globally. HF frequencies have the unique property that a transmission can pass halfway around the globe. However, if the transmission goes halfway around the globe, so does the interference.
HF radio broadcast is also relatively cheap: the costs of constructing an HF radio transmission station, as well as the annual operational costs, are unrivalled by modern technologies. If a single HF transmission station is located in Thailand, the broadcast could reach 60% of the world’s population. Anybody who would like to receive this broadcast needs a receiver costing around 10€. HF radio broadcast is especially important in developing countries where there is little or no infrastructure in rural areas and HF radio is often the only option. Today, HF radio broadcast is used for the following reasons:
• In newly industrialising and developing countries where the transmission distances are very large and the installation of FM transmission infrastructure is too expensive.
• Tourists who like to receive their home services in a habitual manner. However, satellite or web services are available in most hotels of industrialised countries.
• Amateur radio listeners or hams. The PLC interference situation is different for them. Usually they do not use a kitchen radio equipped with a whip antenna. Furthermore, they have the knowledge of how to protect their equipment from interference by using additional filters, for example. However, due to their desire to receive extremely weak signals and the sensitivity of their equipment, interferences from PLC are relevant.
• Military services permanently use the HF frequency range but they also use professional antennas. Their operational area is often far abroad where no alternative to HF transmissions exists. Of course, militaries train the operation in their home countries. This should be done as closely as possible to real conditions. PLC is an interference source to an airplane flying over Europe. Tanks or marines usually do not operate in the vicinity of PLC modems.
• New upcoming digital services such as DRM [10]. Sales of DRM receivers are not as high as initially expected. However, it will be interesting to see how these services develop. If the frequency resources are polluted, once, a future installation is no longer possible.
• It is a fundamental freedom in democracies to receive information from everyone and everywhere. Broadcasters such as Deutsche Welle, BBC World Service, Voice of America and others transmit information in multiple languages using HF bands. The most frequently listened SW radio service today in Europe is Radio China. Unlike the Internet, it is difficult to censor HF radio broadcast; only jamming is possible. But more importantly, radio listeners cannot be monitored by government authorities.
• In the event of a crisis, disaster or earthquake, satellite dishes may no longer be aligned. In this case, HF radio broadcast is the most robust and proven technology and expected to be the first source of information to be reconstructed.
Countries without a well-developed wired telecommunication infrastructure expect to support people with Internet services over PLC. Such countries are also the main target areas of HF radio transmissions. Therefore, the coexistence between PLC and HF radio reception is very important.
EN 50561-1 specifies the solution to solve interference problems from PLC to SW radio receivers.
A permanent notching of all HF radio bands would result in the loss of 21% of the communication spectrum of a conventional PLC modem using the frequency range from 2 to 30 MHz.
Whether or not the HF frequencies can be used by radio transmissions depends on weather conditions and the reflection quality of the ionosphere. The structure of the ionisation layers in the ionosphere vary according to the time of day and seasonal changes. An 11-year sunspot cycle also affects radio reception (Annex A of ETSI TR 102 616 [11]). Radio broadcasters permanently operate monitoring stations in their target area to measure reception quality and schedule their services accordingly.
A service description channel is included in the specification for the new digital radio service DRM [12] for HF bands, which informs the receiver on which frequency the transmission will continue before a change in the transmission schedule is performed.
Usually, an HF transmission band is either fully allocated with radio services or relatively empty. This is why a permanent default notching of all HF bands is unnecessary and would result in too much throughput loss for PLC modems.
The cognitive frequency exclusion specified in EN 50561-1 provides optimum reduction of interference between PLC and HF radio broadcast and minimum impact on data throughput and quality of service (QoS) requirements of PLC.
22.4 Concept of ‘Cognitive Frequency Exclusion’
‘Cognitive frequency exclusion’ is an adaptive process which automatically excludes all frequencies – from PLC – being used by receivable radio services, without any user or network operator interaction.
The presence of broadcasting signals can be detected by PLC modems by sensing the ‘noise’ (including radio broadcasts picked up on the mains cabling) in an electrical socket. Frequencies where HF radio broadcasting signals are identified can then be omitted from the transmitted signal by inserting notches into the transmitting PLC spectrum.
Radio broadcast signals transmitted with a high power from the antenna of a radio station will be received by any wire acting as an antenna, for example, an electrical power grid (see Figure 22.1). The ingress of a broadcast signal can be detected within the reception range of the radio broadcast signals. The power line wires are passive and therefore reciprocity is valid. The transfer function, or the antenna gain, is identical for radiation as well as for signal reception. At frequencies with a high potential for interference, signal ingress is excellent and there is likely no ingress at non-radiating frequencies.
PLC modems are connected to the mains and are equipped with a very sensitive analogue front-end. To achieve adaptive modulations (see Figure 22.10), PLC modems monitor the noise signal on the mains. This noise information could be reinterpreted by the modems in order to identify a receivable radio station. If done, the frequencies of all receivable radio stations could be excluded from PLC transmission.
22.4.1 Radio Signal Spectrum on Power Lines
Of initial interest is a comparison of the signals measured when connected to the mains and the field measured with an antenna in the air. Comparing the signal strength of an HF radio broadcast station at two locations at once is very difficult. Due to the strong dynamic fading effects of radio transmissions in the time domain, a received HF signal never has the same level it had a moment ago. This time-variant effect was further investigated by monitoring the level of HF radio stations over a period of time.
FIGURE 22.1
Radio signals ingressing a house’s mains wiring.
22.4.1.1 Fading Effects of Radio Stations
Figure 22.2 shows the levels of two radio stations (6918 and 7106 kHz) which were monitored for a long time. The horizontal axis represents the time from 0 to 1000 s. The vertical axis is a relative level in dB. Both transmissions show strong fading effects. The transmission at 7106 kHz (bottom, dotted curve) is particularly interesting. 700 s after the recording started, the signal level dropped by 40 dB. Obviously, there was a change in the transmission schedule of the broadcaster. Changes such as these can be monitored quite frequently in HF broadcasting. An overview of frequency scheduling in HF broadcasting is given in ETSI TR 102 616 [11], Annex A.
FIGURE 22.2
Fading of HF radio broadcast services in time domain.
The service at 7106 kHz (top line) shows relatively less fading compared to the 6918 kHz station with typical characteristics. These are flat hills and deep canyons in the shape of the curve. Fading effects are caused by multipath propagation between the transmitter and the receiver. The waves from individual paths may overlap constructively or destructively, depending on the phase difference when they arrive at the receiver. The phase of a path may vary with the dynamics in the ionosphere. Depending on weather conditions, an ionisation layer can move with more than 100 km/h. This is why Doppler effects impair HF reception and an automatic frequency control (AFC) in SW radio receivers is beneficial, even if the transmitter and receiver do not move. The signal level of a station changes by more than 30 dB within a few seconds. The AGC of an SW radio receiver must be very dynamic in order to follow such level fluctuations. If HF radio stations are to be detected on the mains, PLC modems have to consider such dynamics of the signal levels.
22.4.1.2 Setup to Record a Signal Spectrum
Figure 22.3 shows the equipment required to compare the levels of the electromagnetic field in the air and the ingress signals on the mains. Such measurements were performed in private flats, hotels and office buildings. The antenna was located in the centre of the room with a vertical alignment of the biconal probe. Photographs of these measurements can be found in ETSI TR 102 616 [11]. The Schwarzbeck electrical field probe EFS 9218 [13] allows calibrated field measurements. It provides a constant antenna factor for the frequencies of interest. A spectrum analyser is connected, using a probe, to the mains or alternatively using the antenna. For the measurements in Figures 22.4 and 22.5, the maximum signal level is recorded using the max-hold function of the spectrum analyser for about 1 min.
FIGURE 22.3
Detection of radio services. Setup for measurements in buildings.
FIGURE 22.4
Indoor electrical field snapshot of the 49 m band, any location.
FIGURE 22.5
Signal ingress in the 49 m band, measured connected at a power outlet.
Figure 22.4 shows a snapshot of the electrical field in one building of the 49 m band. The HF range is allocated by ITU-R [2] into multiple bands. Each individual band has its pros and cons in transmission properties. Some of them have better daytime reception (e.g. 19, 16 m), others perform better at night (41, 31 m). Some bands are used in summer, others in tropical regions. The x-axis of Figure 22.4 represents the frequency range from 5.9 to 6.3 MHz. There are 40 kHz per division. The y-axis in Figure 22.4 is interpreted by the spectrum analyser in dB μV, but considering the antenna factor of 18 dB(/m) [13], the y-axis is converted to the E-field from -2 dB (μV/m) to 98 dB (μV/m) using a scale of 10 dB per division. Some AM services can easily be identified: at 5954, 6035, 6155, 6120 kHz and more. At 5950 kHz, for instance, a DRM transmission is visible.
All sweeps in Figures 22.4 and 22.5 were recorded with max hold (top line) and min hold (bottom line) to get an impression of the fading of the individual services. In Figure 22.4, the AM service at 5920 kHz was switched off during the recording period. It is neither visible at the min-hold line nor in Figure 22.5. The service at 5954 kHz shows an average fading (around 20 dB), and the service at 6005 kHz shows almost no fading behaviour. The black line represents the latest frequency scan before the snapshot was recorded.
Taking the fading into account and that some stations might have been switched on or off during the measurements, Figures 22.4 and 22.5 look virtually identical.
Figure 22.5 recorded the conducted signals on the mains. The vertical axis represents the voltage in dB μV.
22.4.1.3.1 Characteristics of a DRM Spectrum
DRM transmissions use orthogonal frequency division multiplex (OFDM) – a modulation scheme with 88–228 carriers in a 10 kHz spectrum. The number of carriers used by DRM depends on the transmission mode corresponding to typical propagation conditions. An HF-DRM spectrum appears as a 10 kHz wide rectangle spectrum. The DRM specification [12] also allows bandwidth allocations of 5 and 20 kHz. In the HF band, 10 kHz spectra are usually used. Some DRM transmissions may also have a very high single carrier in the centre of their spectrum. This is necessary to ensure that legacy transmitting amplifiers keep their linearity. DRM was especially designed in order to use portions of older AM Tx facilities, such as antennas and amplifiers, avoiding major new investment. Simulcast transmissions use an AM-modulated centre carrier with the surrounding DRM carriers (e.g. at 5954 kHz in Figure 22.4). DRM has a more variable peak-to-average ratio (PAR) compared to AM services. This used to cause problems at some old transmitting stations. It explains why the carrier in the centre of the spectrum (known from an AM 30% modulation depth) has been kept. In this case, some OFDM carriers in the middle of such a DRM channel are not used and the central carrier has to be filtered away by the receiver.
22.4.1.3.2 DRM and AM
If a PLC modem has to detect these signals, it does not differentiate between AM or DRM. Today’s PLC modems have an OFDM carrier spacing fCS of around 20 kHz. The carrier spacing is provided by the system’s Nyquist frequency fNyquist (half of ADC/DAC sampling clock frequency) divided by the FFT_size implemented by the PLC modem. The parameters of the system described in [14] are as follows:
(22.1) |
The carrier spacing fCS of a transmitting OFDM system is identical to the resolution bandwidth of the receiving system using the same FFT_size and sample frequency. The resolution bandwidth of the PLC system is also relevant for noise measurements. Reference [14] presents methods to enhance the resolution bandwidth of noise measurements. If the bandwidth of the signal to be detected (AM or DRM) is smaller than the resolution bandwidth of the measurement system, its shape does not matter. Due to the fact that DRM is designed to reuse the transmission facilities (amplifiers, antennas) of AM equipment, it has the same signal power as an AM carrier.
The stations marked with a dotted ring in Figure 22.5 were receivable using the automatic frequency scan function of the Sony ICF-SW77 SW receiver. In total, at a single time instant during the day and at this location, 22 radio stations were receivable with the Sony ICF-SW77 with field strengths between 29 and 68 dB (μV/m). Test results in the same order of magnitude were measured during the ETSI plugtest and published in ETSI TR 102 616 [11].
J. Stott’s BBC R&D White Paper 114 [15] posts similar measurement results.
All these measurements help to answer the question: Which air-based HF carriers can be detected on the mains network? All radio signals in the air which are stronger than 20 dB (μV/m) (measured with a RBW of 9 kHz) significantly increase the noise floor on a quiet power line.
Due to the strong variations in the amplitude of HF radio broadcast signals over time, PLC modems should periodically sense the ingress level of the radio signals. The level of these signals also depends on the modem’s location and the structure of the wiring in the electricity grid.
22.4.1.3.3 Noise on Power Lines
As described and measured in Chapter 5, many other noise signals may appear on the mains spectrum. Further noise sources such as switching power supplies frequently enhance the ingress level of radio broadcast services. Later in this chapter, both an absolute and a relative threshold are defined, when such signal ingress is receivable by a radio device.
22.4.1.3.4 Additional PLC Transmission
For further studies of the interference potential of PLC, the measurement setup from Figure 22.3 was used to check the level of radiation due to PLC in a building. Figure 22.6 – where a PLC transmission was also set up in parallel – shows an identical spectrum to the one measured in Figure 22.4. The radiated noise level covers most of the radio services.
22.4.2 Sensitivity of SW Radio Receivers
Reference [14] evaluated the sensitivity of HF radio receivers. A couple of AM receivers were stressed in an anechoic isolation chamber to derive the signal-to-noise ratio (SNR) in the demodulated audio signal. Additional tests were performed by checking at which level the automatic station scan stops. Furthermore, theoretical studies were performed based on [16,17] and the DRM radio planning parameters. All evaluations showed the minimum sensitivity of HF radio receivers to be at 22 dB (μV/m).
FIGURE 22.6
Radiation from PLC.
22.4.3 Threshold to Detect Radio Services
Propagation characteristics of HF radio transmissions are not stable. As already seen in Figure 22.2, the fading in the time domain generates heavy variations in the signal level at the receiver side. ITU-R specifies the probability of a transmission being receivable by the target device in [18]. If PLC modems monitor the ingress of HF radio broadcast signals using a max-hold detector, the short-term signal variations might be considered. Figure 22.7 presents a sketch of these calculations.
The x-axis of Figure 22.7 represents the time in arbitrary units, the y-axis the electrical field in the air. The bottom line ‘<’ is the intrinsic noise level that might be expected in a high-quality receiver. The 7 dB (μV/m) line ‘<’ takes the added man-made noise into consideration, and the Emin = 22 dB (μV/m) line ‘o’ includes the theoretical DRM [16] or measured AM receiver’s minimum sensitivity [14]. The fading line ‘+’ shows the expected signal at the receiver’s location with the statistics from [18]. Such signals are inevitably subject to a larger or smaller degree of fading. The maxima of the top fades are DuSh = 5 dB higher than the minimum receiver’s sensitivity [18]. The interleaver of DRM is designed to make receivers immune to such fading. If PLC modems need a margin Mto_detect_threshold of 1 dB for detection, the fading statistics specify the threshold to be exceeded with a probability of 30% in any interval longer than 10 s. PLC modems have to detect the fading line ‘+’ raising the 26 dB (μV/m) line ‘x’. In order to detect the top level of fading, the detection threshold of field strength Efield_to_detect in the air is calculated as follows:
FIGURE 22.7
Threshold to detect HF radio broadcast ingress:
‘x’ 26 dB(μV/m), max hold with 1 dB headroom;
‘+’ 14−27 dB(μV/m), fading of HF broadcast;
‘o’ 22 dB(μV/m), minimum receiver’s sensitivity;
‘<’ 7 dB(μV/m), intrinsic noise plus man-made noise;
‘<’ bottom line at 4 dB(μV/m), receiver’s intrinsic noise level (identical to man-made noise).
(22.2) |
The threshold of 26 dB (μV/m) is given by the top line ‘x’ in Figure 22.7. If PLC modems use an average detector instead of the max-hold one, the threshold has to be lowered by 5 dB. This is implementation dependent for the PLC modem manufacturer.
The reception factor (defined in the plugtest report ETSI TR 102 616 [11]) describes the relationship between the electrical field strength of a radio broadcast station in the air and the received power to be measured at outlets. The setup to measure the reception factor is visualised in Figure 22.8. An SW radio receiver is used to scan the spectrum. If a station is receivable, its E-field and its signal ingress level at the mains are verified.
Measurements deliver a cumulative statistical probability of the reception factor shown in Figure 22.9.
The lower the value in Figure 22.9, the better the antenna gain from the mains wiring. The median value of the reception factor (ReFa) is found to be 114 dB (μV/m) − dBm. The 80% worst case value is ReFa80% = 121 dB (μV/m) − dBm. A reception factor covering 80% of the cases, with an 80% confidence level, can be derived from the distribution function shown in Figure 22.9. With this value, the threshold of the signal level connected to the mains Pdetect_on_mains can be derived with
(22.3) |
This level can be verified with a spectrum analyser using a resolution bandwidth of 9 kHz and an average detector.
Besides the threshold level, a criterion has to be developed for the separation of SW radio stations from disturbance sources operated at mains. As shown in Figure 22.5, the ingress of a broadcast radio station appears as a needle in the noise measurement of a PLC modem. Today’s PLC modems are not able to demodulate the AM or DRM signals. However, a needle and even a very stable one in the frequency domain is something unique within all noise sources at power lines (see Chapter 5) and can be detected by modems. Two criteria must be fulfilled for the level of the needle to be identified as a receivable service and worth being protected by PLC modems. The first criterion is a relative threshold because the usable signal must have a minimum SNR. The second is an absolute threshold when the signal passes the minimum sensitivity level of radio receivers.
FIGURE 22.8
Definition of reception factor, measurement setup in a flat.
FIGURE 22.9
Cumulative probability of reception factor.
The PLC modem is not able to apply weighting windows on the demodulated AM signal and to measure SNR as specified in [19]. A simpler approach is required. A PLC modem might measure the noise as well as the peak level of signals. As described in ETSI TR 102 616 [11], the minimum SNR of an HF service is 14.6 dB for the most robust DRM transmissions. This is the distance between the noise and the peak signal level.
Taking the typical noise sources into consideration, there is not only one noise level. The noise is time and frequency dependent. To check the SNR of a broadcast service, the noise next to the service shall be used. As long as the transmission conditions are good, an HF transmission band is densely allocated with radio stations. Measuring the noise inside the band would result in a value accumulating all radio services, not in the surrounding noise. This is why EN 50561-1 specifies the noise floor to be measured at adjacent frequencies lower and higher than the HF radio bands. The adjacent frequency blocks must be completely monitored by the PLC modems without any gaps in order to avoid cherry picking of noise values by a PLC modem. Finally, the noise floor is the median value of all measured values. The median value is not affected by individual peaks of, for example, a strong ‘out of band’ radio station.
In EN 50561-1, two criteria or thresholds are given, when a broadcast radio station is defined to be receivable:
• Criterion (1): 14 dB above the noise floor. It is 3 dB lower than the desired SNR of an AM receiver to understand voice and around 11 dB lower than that required by a normal DRM transmission (out of [16]).
• Criterion (2): the absolute threshold of -95 dBm which is derived from Equation 22.3.
For activation, a processing time of 15 s is conceded to PLC modems. If the two criteria were met once, the notch will be kept for at least 3 min. This timing hysteresis is a trade-off between consumer acceptance listening to a fading broadcast service and the processing capabilities of PLC modems creating a notch. The timings were found by the STF332 during plugtests where the concept was verified in ETSI TR 102 616 [11].
22.4.4 Requirements of a Notch
When signal ingress is identified as a receivable broadcast service, its frequency should be excluded from the PLC communication. The process of frequency exclusion in OFDM communication systems is called notching. EN 50561-1 characterises a notch by the bottom level and its width or slopes. In the case of a MIMO PLC modem, the notches have to be applied to both transmission channels.
22.4.4.1 Bottom Level of the Notch
The bottom level of the notch was derived from two different approaches: first, checking the levels of EMC standards [20] and secondly, feeding signals with this level into power outlets and checking whether it affects radio reception.
22.4.4.1.1 Checking Approach
CISPR 22 [20] specifies the mains class B (5–30 MHz) level to be UAMN = 50 dB μV (RBW: 9 kHz, average detector).
An artificial mains network (AMN) (specified in CISPR 16 [21]) is used to verify mains port limits. It measures half of the differentially fed voltage at a measurement output. It follows that at the outlet Uoutlet twice the differential voltage is allowed where the PLC modem is connected:
(22.4) |
When dB μV is converted to dBm using a characteristic impedance Z = 100 Ω, the power allowed at a given outlet is
(22.5) |
The PSDoutlet of PLC modem at the bottom level of the notch using Equation 22.1 is as follows:
(22.6) |
22.4.4.1.2 Subjective Evaluation
The noise is fed to the mains in the vicinity of the outlet where the radio receiver is connected. The level of noise is varied to check when interference is noticeable on the SW receiver. Human ears monitor reception to verify if the additional noise influences SW radio reception. The signal level is recorded, when the reception quality is deemed to be impaired (assessing the quality of signal, interference, noise, propagation and overall (SINPO) [22]). This is performed at many outlets in several buildings with the radio receiver tuned to various frequencies. ETSI TR 102 616 [11] documents some of these assessments. Noise levels up to those in Equation 22.4 (Uoutlet = 56 dB μV) were not noticeable by the receiver. Noise was hardly noticeable when signals were fed into the mains at exactly this level. If the noise was increased by another 10 dB, human ears were able to audibly detect the interference. However, if the radio was disconnected from the mains – and therefore battery powered – the interference was gone. Often, connecting the radio receiver to another outlet also solved the interference problem.
The value from Equation 22.4 was found to be a good choice for the bottom level of the notch, where EN 50561-1 specifies a verification setup for PLC modems to confirm this level. A resolution bandwidth of 300 Hz is selected to make the bottom level visible with the spectrum analyser performing a sweep. Care must be taken when comparing the absolute values of the ingress signal level, the values given in CISPR 22 and the bottom level of the notch. Individual resolution bandwidths are used. PLC modems with a bottom level of a notch as specified in EN 50561-1 no longer cause interference to an SW radio receiver.
22.4.4.2 Width or Slopes of the Notch
To avoid interfering with the bandwidth of an identified radio broadcast service, the minimum width of a notch should be at least 10 kHz (±5 kHz around the carrier frequency of the radio broadcast). Usually, the channels of radio broadcast services are allocated with a minimum spacing of 5 kHz. The centre frequency is a multiple of 5 kHz. If several neighbouring radio broadcast services are identified by the PLC system, the width of one notch may be scaled to integer multiples of 5 kHz.
In radio broadcasting, intercarrier interference (ICI) from other radio stations allocating adjacent channels is a serious problem. Signal amplitudes of adjacent carriers often differ by more than 30 dB. This is why slopes of potential ICI are precisely specified. References can be found in [10,16,23,24]. EN 50561-1 defines a notch where the slopes are approximated to the requirements of SW receivers’ (AM, DRM) protection ratios.
The resolution bandwidth of PLC modems’ noise measurements is usually identical to the width of an OFDM carrier. To protect one broadcast station with a single carrier notch, a PLC modem has to enhance the resolution bandwidth of the receiving fast Fourier transformation (FFT), in order to precisely locate the frequency position of the HF carrier.
22.4.5 Adaptive OFDM, Channel and Noise Estimation
Adaptive communications with feedback information was first presented in 1968 in [25]. Today’s PLC modems provide good starting conditions for implementing notching with minimal effort. Carrier adaptive OFDM [26] is used in wired and wireless communications to match the bit loading of a carrier (quadrature amplitude modulation [QAM] constellation) to the conditions on the channel attenuation and noise. The process of channel adaptation is dynamic. When the channel changes (e.g. a light switch is shifted), the adaptation process has to be retriggered. Some communication systems can do this within milliseconds. Adaptive modulation systems require the knowledge of the receiver’s (Rx) SNR at the transmitter (Tx). Its transfer function is measured at the Rx and fed back to the Tx. Adaptive modulation systems improve the rate of transmission and bit error rates by exploiting channel information that is present at the Tx. They exhibit great performance enhancements compared to other systems, especially for fading channels. As channel and noise are already estimated to realise an adaptive OFDM, noise information might be reinterpreted to identify receivable radio broadcast stations. In an adaptive OFDM system, every carrier loads an individual amount of information. It is no additional burden for the system if individual carriers do not carry information because they are notched. The concept is adaptive. The method of using a carrier, notch it later and reuse it again, does not cause any additional work.
Figure 22.10 shows an example of an SNR estimation of a PLC link. The horizontal axis represents the frequency where the vertical axis represents the SNR in dB estimated by the system. Frequencies with excellent SNR utilise 4096-QAM and are marked with ‘12 bit/carrier’. Frequencies with lower SNR are allocated with lower constellations from 1024-QAM down to binary phase shift keying (BPSK). Frequencies with less SNR than what the most robust implemented constellation requires can no longer be used for communication and can be notched or suppressed. This is the case for carriers around 10 MHz in Figure 22.10.
There are two cases for omitting a carrier from communication: (1) the desired SNR for the minimum constellation is not available at this frequency or (2) a radio service has to be protected at this frequency. OFDM only provides low side-lobe suppression (see Chapter 14, Section 14.3.3), where the shape of the notch might be improved by additional filters. These filters will only be applied if the notch was initiated to protect a radio service.
Both the channel transfer function and the noise have to be measured in order to estimate the SNR. This is done by re-encoding the Rx’s demodulated data and comparing them with the received signal. The noise measurement during communication even works in notched carriers. If the noise is measured by calculating the variance of the received OFDM symbols, it does not matter if the carrier is allocated or notched. Figure 22.11 shows a result of a channel and noise measurement performed with the use of four training symbols per data burst. Figure 22.11 is a snapshot from a PLC modem prototype implementation [14]. The horizontal axis shows the index of the OFDM carriers. It represents the frequency range from 0 Hz up to 40 MHz with a carrier spacing of 19.53 kHz. The vertical axis shows a relative level in dB. The carrier with the index of 211 is the first carrier used for communication; the one with index 1506 is the last. The topmost curve (solid black) is the channel transfer function derived by averaging four training symbols. It shows a relatively flat channel with only one fading of 30 dB. The three other curves are the noise measured by calculating the variance values of the four received training symbols. The tiny, dashed curve shows the noise signal last measured before the screenshot was taken. It shows a high variation (>10 dB) of the individual values. The fat line represents the median with the latest 20 noise shots and the third noise curve (dotted, black) is the max hold of these 20 shots. The average and max-hold values are better suited for detecting thresholds. The ‘needles’ of the ingress from HF radio broadcast are clearly visible.
FIGURE 22.10
Adaptive channel modulation. Simulation of a frequency scan using the channel data of a measured in-house PLC channel.
FIGURE 22.11
Channel and noise measurement of the PLC demonstrator performed during communication.
When multiple OFDM symbols are sequentially chained, there is the disadvantage that noise measurements performed during communication do not detect frequencies at integer multiples exactly matching the training symbol’s repetition frequency. Such noise does not enhance the variance between the received training symbols. To overcome this, variable guard intervals have to be inserted before preamble, frame control and data symbols. The length of each guard interval should be unique to avoid that a constant sine wave interfere with a frequency of a multiple of the training sequence repetition frequency influencing each training symbol identically. The HomePlug [27] or IEEE 1901 specifications support multiple guard intervals within a data burst. The IEEE 1901 (see Chapter 13, Figure 13.4 – PHY protocol data unit [PPDU]) data burst starts with the preamble (10 repetitions of a 5.12 μs sequence), followed by the frame control symbol with a 18.32 μs guard interval, followed by two data symbols with 7.56 μs of protection and further data symbols with a guard interval that can be selected out of 13 alternatives depending on channel characteristics.
If the noise is measured during communication, a further disadvantage is that it is recorded at the receiving modem. The transmitting modem is responsible for the detection of radio broadcast ingress and notching. However, in a time-division duplexing communication – which is usually the case with PLC – the return path of the noise measurement can be used. In order to capture the communication signal as well as the noise signal within the dynamic range of the Rx ADC, the implementation of dynamic power control as specified in this chapter is a prerequisite for measuring the noise during communication.
Alternatively, the noise could also be measured during any period of no PLC. The minimum length of such a quiet period is given by the basic OFDM system parameters such as carrier spacing and symbol duration. Equation 8 calculates the carrier spacing of an OFDM system. The OFDM symbol duration is reciprocal to the carrier spacing. For example, a carrier spacing of fCS = 24.41 kHz like in IEEE 1901 (see Chapter 13) results in a symbol duration TSymbol of
(22.7) |
The FFT of the system has to be filled once, in order to capture a noise shot. The time needed here is equal to the system’s symbol duration TSymbol. In a multi-node communication system, the MAC layer organises by device when the resources are allocated. For example, in a carrier sense multiple access (CSMA) medium access layer (MAC) layer, the contention-free interframe spacing is often longer than the symbol duration. All such gaps could be used to measure the noise.
To conclude, to activate a notch, the noise can be measured during communication or within a transmission break. For a PLC system to reuse the notched frequency, the noise has to be measured in a quiet period or within a notch during communication, but no signal may be transmitted on the notched frequency measuring the noise.
22.5 Implementation in a Demonstrator System
A feasibility study is implemented in order to prove the concept of ‘cognitive frequency exclusions’. The PLC system is described in detail in [14]. Its main focus is desired applications and rapid development. The system is a proprietary PLC technology and does not follow any PLC standards such as HomePlug [27], HD-PLC [28] or ITU-G. Hn [29]. As an application, the system transports a high-definition video stream from Tx to Rx and measures maximum payload data. Its maximum throughput on PHY layer is 212 Mbps.
The PLC demonstrator was implemented using a similar platform as described in Chapter 24 but without MIMO features. Here, statistical evaluations were implemented from various noise recording techniques: max hold, median and latest shot (as shown in Figure 22.11). The main unit distinguishing this feasibility study from a conventional PLC modem today is the additional notch filtering function.
22.5.2 Notch Filter Environment
The implementation of the notch filtering function requires a noise measuring unit, a function to detect the presence of radio signals and the notch filter.
22.5.3 Detection of the Presence of Radio Services
A trade-off has to be found in the number of noise shots to be recorded and the detection speed to activate a notch. Timings in EN 50561-1 are 15 s to activate a notch. As discussed in connection with Figure 22.7, it is recommended to implement a max-hold detector to limit the huge number of noise records and to ensure that the top 1 dB margin of the fading signals is captured. These values have to be compared if the threshold is exceeded. The notch has to be activated at frequencies where the threshold is passed.
There are various techniques for creating notches. One simple technique for creating notches is using wavelet transformation [30], where omitting a carrier from the communication is sufficient and the spectrum is notched with a depth depending on the side lobes of the applied wavelet. The waveform of the wavelet is responsible for the shape of the notch. The notching feasibility study described here uses an FFT transformation process. The output of the FFT is very sharp in the time domain but provides weak slopes in the frequency domain compared to wavelet transmissions. Windowing can be applied for spectral shaping of FFT systems. Chapter 14 discusses the influence of windowing on the shape of a notch. Another alternative would be to design notches with additional filter stages in the Tx spectrum of the data. This solution is presented in the following.
22.5.4.1 Influence of Windowing on Spectrum and Notch Shape
The technique of efficient notching was considered when drafting the HomePlug AV2 specification. It is described in Chapter 14 (Section 14.3.3).
22.5.4.2 Adaptive Band-Stop Filters
Another alternative method of creating notches is to filter the unintended frequencies using tunable band-stop filters. In the feasibility study described here, a cascaded structure of second-order infinite impulse response (IIR) filters is implemented. A filter block consists of five multipliers, four delay lines and an adder. Depending on the frequency, up to three filter blocks are needed to create a notch as specified in EN 50561-1. The algorithm to calculate the filter coefficients can be explained using the filters unity circle. The zeros are on the unity circle where the angle specifies the frequency. The poles are close to the zeros at identical angles, inside the unity circle to guarantee stability. The distance between zero and pole defines the attenuation of the notch. Reference [14] described the algorithm in detail.
22.6 Verification of ‘Cognitive Frequency Exclusions’ in Buildings
The prototype system should be tested in buildings under real and noisy conditions in order to verify the implementation of ‘cognitive frequency exclusions’.
As shown earlier, HF radio broadcast transmissions change their transmission frequency from time to time. Frequency hopping by an HF radio station requires a ‘cognitive frequency exclusion’ implementation in order to comply with the EN 50561-1 under all criteria. Such a scenario was very interesting to monitor under live broadcasting conditions in a private building. During the ETSI plugtest, a radio broadcast station in Skelton (United Kingdom) was available to schedule the transmission of any radio service according to the plugtest demands. To verify the dynamic behaviour of the PLC system, the Skelton transmission toggled from 7225 to 7320 kHz. Two SW radio receivers were used to monitor this event at the test location in Stuttgart (Germany). Each of the radios is tuned to one of the frequencies. There also was a PLC transmission running in parallel inside the building from the ‘cognitive frequency exclusion’ demonstrator system.
Figure 22.12 shows a timetable overview of the actions occurring at the test site. The horizontal axis represents the time in seconds. Before the station hop took place, the first radio receiver which was tuned to 7225 kHz received a good-quality AM signal. This frequency was notched by the PLC system. There was no interference from the PLC system to this radio station. The PLC signal was clearly noticeable on the second radio receiver, which was tuned to 7320 kHz.
The trigger for the time axis in Figure 22.12 was set to 0 s when the radio broadcast signal went silent on the first radio receiver. The Skelton transmitter had stopped its broadcast. Thirteen seconds later, the start of the transmission at 7320 kHz was noticed on the second radio receiver. A further second later, the AM started at 7320 kHz. The service could be heard on the second radio receiver, but it was still interfered with by the PLC transmission. The PLC device detected the presence of this radio station another 3 s later and inserted a notch to protect 7320 kHz. This was noticed when the interference stopped on the second radio device. Around 1 min later, the frequency at 7225 kHz was reused by the PLC system. The PLC signal was now noticed on the first radio receiver which was still tuned to this frequency. EN 50561-1 today specifies a time period of 3 min until a frequency might be reused by the PC system. In the early days when the draft specification of ETSI TS 102 578 [5] was still undergoing modifications, a time hysteresis of 1 min was sufficient.
FIGURE 22.12
Frequency hopping of a notch recorded at verification of the demonstrator.
To assess the quality of an AM radio station, the SINPO assumption was standardised in [22] with the properties of signal strength, interference, noise and propagation individually estimated. The signal strength and propagation can be measured using a spectrum analyser. Noise and interference levels are estimated by human ears. This way, it is difficult to identify the source of the noise. The listener’s impression is noted without taking any further action. PLC was often the dominant interferer during the ETSI plugtest, when the ‘cognitive frequency exclusions’ were not activated. Finally, the overall estimation is given as an average of individual properties. Radio signal quality assessment was conducted according to SINPO [22] 168 times during the plugtest. Each of them is noted in the plugtest report ETSI TR 102 616 [11] including the signal levels at the mains as well as the E-field. Figure 22.13 shows a histogram of the occurrence of the overall SINPO estimation. The horizontal axis represents the SINPO level:
• Level 1: Unusable. No listener will stay tuned to a service with such bad quality.
• Level 2: Poor. A human voice might be understood.
• Level 3: Fair. Music might be enjoyed, with limited quality.
• Level 4: Good. AM audio quality.
• Level 5: Excellent. Usually, an AM service will never reach level 5. Only DRM supports this level in the HF band.
The SINPO assumption was done three times for each radio station received at the site of the plugtest: initially with PLC switched off (right [bright] column at each SINPO level in Figure 22.13) and later with a PLC transmission running with the ‘cognitive frequency exclusion’ concept activated (middle column) and finally, when PLC transmission was running, but no notches were in place (left [black] column). To ensure that every station was captured, a station scan was executed in all HF transmission bands using the Sony ICF-SW77. The SINPO assumptions were performed at every station where the station scan stopped.
FIGURE 22.13
Histogram of subjective assessment of sound quality.
Figure 22.13 represents a histogram of how frequent an overall signal quality was monitored.
Figure 22.13 shows that there was no difference in the reception quality of HF radio services when the PLC system was transmitting data with ‘smart notching’ activated and when the PLC was off. The right and middle columns have an identical height. When the PLC system was transmitting data without inserting notches, many radio stations degraded to an unusable quality.
22.7 Outlook for Future EMC Coordination
Cognitive radio or dynamic spectrum allocations is a huge area in research today. There are many ideas at DVB, digital dividend, OFDM overlay systems, frequency management with mobile satellite services, dynamic spectrum management for DSL, etc., where this concept is relevant. Frequency resources are an extremely rare good. Motivation is high to allocate these resources as efficiently as possible. If intelligent devices are able to adapt to the local situation without causing interferences to other applications, the classical concept of EMI is overthrown. In the past, there were always constant limits for EMI emissions as well as immunity. Cognitive systems change this paradigm. Thanks to adaptive OFDM transmission with a high number of carriers, ‘cognitive frequency exclusions’ or ‘smart notching’ only causes a minor decrease in transmission bitrate as only low SNR carriers are lost. Continuous analysis allows the system to minimise interference and to optimise throughput depending on the current conditions.
EN 50561-1 is the first EMC standard that embeds dynamic cognitive interference mitigation technologies.
Some videos illustrate the implementation of the PLC ‘smart notching’ system in a flat:
• PLC with Sony ICF SW 77 [31]
• PLC with Sangean ATS 909 [32]
• PLC with Roberts DRM receiver [33]
1. EN 50561-1, Power line communication apparatus used in low-voltage installations – Radio disturbance characteristics – Limits and methods of measurement – Part 1: Apparatus for inhome use.
2. ITU-R Radio Regulations, edition of 2004.
3. Stott, J. and Salter, J., BBC R&D White Paper WHP067, The effects of powerline telecommunications on broadcast reception: Brief trial in Crieff. http://downloads.bbc.co.uk/rd/pubs/whp/whp-pdf-files/WHP067.pdf, accessed May 2013.
4. European Telecommunication Standardization Institute, Technical Committee on Power Line Transmissions (ETSI TC PLT). http://www.etsi.org/WebSite/Technologies/Powerline.aspx, accessed May 2013.
5. ETSI TS 102 578 V1.2.1 (2008-08), PowerLine Telecommunications (PLT); Coexistence between PLT modems and short wave radio broadcasting services.
6. IEC, CISPR/I/257/CD, CISPR22 – Limits and method of measurement of broadband telecommunication equipment over power lines, February 2008.
7. IEC, CIS/I/301/CD, Amendment 1 to CISPR 22 Ed. 6.0: Addition of limits and methods of measurement for conformance testing of power line telecommunication ports intended for the connection to the mains, July 2009.
8. IEC, CIS/I/302/DC, Comparison of the RF disturbance potential between Type 1 and Type 2 PLT devices compliant with the provisions of CISPR/I/301/CD and EUTs compliant with the limits in CISPR 22 Ed. 6.0, July 2009.
9. IEEE Std 1901-2010, IEEE Standard for Broadband over Power Line Networks: Medium Access Control and Physical Layer Specifications. http://grouper.ieee.org/groups/1901/, accessed May 2013.
10. Minimum Receiver requirements for DRM, Draft version 1.5.
11. ETSI TR 102 616 V1.1.1 (2008-03) PowerLine Telecommunications (PLT); Report from Plugtests™ 2007 on coexistence between PLT and short wave radio broadcast; Test cases and results. http://www.etsi.org/plugtests/plt/plt1.htm, accessed May 2013.
12. ETSI ES 201 980 (V2.2.1). Digital Radio Mondiale (DRM); System Specification. http://www.drm.org/, accessed May 2013.
13. Schwarzbeck EFS 9218, Active electric field probe with biconical elements EFS 9218 and built-in amplifier. http://www.schwarzbeck.com/Datenblatt/m9218.pdf, accessed May 2013.
14. Schwager, A., Powerline communications: Significant technologies to become ready for integration. Doctoral thesis, University of Duisburg-Essen, Essen, Germany, 2010. http://duepublico.uni-duisburg-essen.de/servlets/DerivateServlet/Derivate-24381/Schwager_Andreas_Diss.pdf, accessed May 2013.
15. Stott, J.H., BBC R&D White Paper WHP114, Co-existence of PLT and radio services – A possibility? June 2005. http://downloads.bbc.co.uk/rd/pubs/whp/whp-pdf-files/WHP114.pdf, accessed May 2013.
16. ITU-R Rec. BS. 1615, Planning parameters for digital sound broadcasting at frequencies below 30 MHz.
17. ITU-R Rec. P.372-8, Radio Noise.
18. ITU-R Rec. P.842-2, Compotation of reliability and compatibility of HF radio systems.
19. EN 60315-3:2000, Methods of measurement on radio receivers for various classes of emission. Receivers for amplitude-modulated sound-broadcasting emissions.
20. CISPR 22:1997, Information technology equipment – Radio disturbance characteristics – Limits and methods of measurement.
21. CISPR 16-1-1, Specification for radio disturbance and immunity measuring apparatus and methods – Part 1-1: Radio disturbance and immunity measuring apparatus – Measuring apparatus.
22. ITU-R Rec. BS. 1284, General methods for the subjective assessment of sound quality. See http://stason.org/TULARC/radio/shortwave/08-What-is-SINPOSIO-Shortwave-radio.html, accessed October 2007.
23. ITU-R Rec. BS. 703, Characteristics of AM sound broadcasting reference receivers for planning purposes.
24. ITU-R Rec. 560-3 1, Radio-frequency protection ratios in LF, MF and HF broadcasting.
25. Hayes, J.F., Adaptive feedback communications, IEEE Transactions on Communication Technology, COM-16, 29–34, February 1968.
26. Lee, J.-J., Cha, J.-S., Shin, M.-C. and Kim, H.-M., Adaptive modulation based power line communication system, Advances in Intelligent Computing (Lecture Notes in Computer Science), Springer, Berlin, Germany, 2005, Vol. 3645, pp. 704–712.
27. Homeplug. http://www.homeplug.org/, accessed May 2013.
28. HD-PLC Alliance. http://www.hd-plc.org/, accessed May 2013.
29. ITU-T. 2011. G.9960, Unified high-speed wireline-based home networking transceivers – System architecture and physical layer specification.
30. Sandberg, S. D. and Tzannes, M. A., Overlapped discrete multitone modulation for high speed copper wire communications, Journal on Selected Areas in Communications, 13(9), 1571–1585, December 1995.
31. Video ‘Smart Notching’ demonstrator and AM receiver Sony ICF-SW77. http://plc.ets.uni-duisburg-essen.de/sony/SmartNotching_ICF-SW77.wmv, accessed May 2013.
32. Video ‘Smart Notching’ demonstrator and AM receiver Sangean ATS 909. http://plc.ets.uni-duisburg-essen.de/sony/SmartNotching_Sangean.wmv, accessed May 2013.
33. Video ‘Smart Notching’ demonstrator and DRM receiver Roberts MP-40. http://plc.ets.uni-duisburg-essen.de/sony/SmartNotching_DRM.wmv, accessed May 2013.