We investigated the dependence of the signal amplitude and frequency deviation on IM. Fig. 6 depicts the schematic of the experimental setup. The two signal generators (SG) generated different frequency signals in the 420 MHz band. The SG1 and SG2 signals were input into the patient monitor with an input impedance of 50 ohm through a hybrid coupler. The simplified spectrum analysis function was performed on the patient monitor. We shall carefully explain this function in a subsequent section. In this experiment, we measured the IM-affected amplitude of the IM channel. Table 1 presents the scheme of the experiment. In condition No. 1, SG1 and SG2 generated 420.325 and 424.825 MHz, respectively, and the frequency deviation between them was 4.5 MHz. Hence, the IM channel was 429.325 MHz. In the actual setting, the IM effect will emerge at the lower frequency side of SG1 (415.825 MHz), although this frequency is outside of that allocated to WMTS. In this experiment, we adjusted the amplitudes of the SG 1 and SG 2 levels.
Table 1 Experimental conditions
|
SG1
|
SG2
|
IM channel
|
Δf
|
No.1
|
420.325 MHz
(ch1023, Zone 1)
|
424.825 MHz
(ch2028, Zone 5)
|
429.325 MHz
(ch3007, Zone 1)
|
4.5 MHz
|
No.2
|
420.2875 MHz
(ch1020, Zone 9)
|
420.5625 MHz
(ch1042, Zone 1)
|
420.8375 MHz
(ch106, Zone 1)
|
0.275 MHz
|
No.3
|
424.525 MHz
(ch2002, Zone 2)
|
425.2125 MHz
(ch2059, Zone 9)
|
425.925 MHz
(ch2116, Zone 2)
|
0.7125 MHz
|
Fig. 7(a) depicts the experimental results when the two input signal amplitudes were the same. The IM effect appeared strongly when the frequency deviation was narrow. Figs. 7(b) and (c) depict the experimental results when the two input signal amplitudes differed. Fig. 7(b) shows when SG2 (ch2028) was set to certain values, whereas SG1 (ch1023) was adjusted from low to high amplitude levels. Fig. 7(c) depicts when SG1 (ch1023) was set to certain values, whereas SG2 (ch2028) was adjusted. When the amplitude of SG1 was 1 dB higher, the IM amplitude increased by 1 dB. However, the IM amplitude increased by 2 dB when the amplitude of SG2 was 1 dB higher. Therefore, the amplitude of IM depends on the input signal near its frequency.
Intermodulation effect on the sensitivity of the WMTSs receiver
We disclosed the sensitivity of the WMTS receiver in our previous study. The required carrier-to-noise ratio (CNR) was ~15 dB when the electromagnetic noise, which can adopt a Gaussian approximation (GA) regardless of the amplitude of the noise power, was added [5]. However, CNR was 3–4 dB worse regarding the electromagnetic noise, which exhibited impulsive characteristics; hence, it could not adopt GA [9]. The IM effect on the sensitivity of the WMTS receiver remains unclear. Fig. 8 shows a schematic of the experimental setup. The type-A WMTS transmitter with a frequency of 429.325 MHz sends normal ECG signals. The transmitter and ECG simulator that generates normal ECG signals with 60 waves per minute were placed in a transverse electromagnetic (TEM) cell to extract the signal. Conversely, the 420.325 and 424.825 MHz continuous waves were generated by SG1 and SG2, respectively, and were input into a coupler, 1. The WMTS signal and coupled 420 MHz band continuous wave, including an IM frequency of 429.325 MHz, were input into the WMTS receiver through another oupler, 2. We evaluated the degradation of the wireless reception due to electromagnetic noise by visually examining the ECG waveforms on the display of the receiver. The critical WMTS signal level for normal reception was determined by decreasing the signal level with an attenuator (ATT). The decision criterion for defining a normal reception was if no abnormality was observed in the ECG waves during 100 wave periods. We varied the amplitude of IM by adjusting the SG power from high to low signal levels. The IM noise power, N, and WMTS signal power, C, were measured at the input port of the receiver at an RBW of 10 kHz using a SA.
Fig. 9 depicts the experimental results. The vertical and horizontal axes correspond to the WMTS signal power, and average IM noise power, respectively. The required CNR, which was obtained by subtracting the IM noise power, N, from the required signal power, C, was 17.7 ± 0.6 dB. The IM effect degrades the receiver sensitivity by several dB worse than the Gaussian Noise.
Measurement of the radio propagation of WMTS for evaluating the potential of the intermodulation effect
We measured the radio propagation of WMTS in the simulated environment to evaluate the actual potential of the IM-induced interference. Fig. 10 depicts the ground plan of the simulated environment. In this measurement, the whip antenna of WMTS, which was connected to SA, was placed at Point, A, of the second floor. The WMTS transmitter was moved from Point B to H per second to the fifth floor. Table 2 presents the measurement results of the WMTS signal level. When the transmitter was placed on the second floor, the received signal level was –40 dBm/10 kHz to −98 dBm/10 kHz. However, the signal levels coming from the fourth and fifth floors were approximately −100 dBm/10 kHz.
Table 2 Received signal level of WMTS
|
2F
|
3F
|
4F
|
5F
|
B
|
−68.6
|
−83
|
−99.9
|
−100
|
C
|
−84.8
|
−99.6
|
NM
|
NM
|
D
|
−97
|
−83
|
−99.7
|
−100.2
|
E
|
−88.8
|
−103
|
−99.2
|
−102
|
F
|
−77.2
|
−97.5
|
NM
|
NM
|
G
|
−98
|
−100.3
|
−100.3
|
−100.1
|
H
|
−40.4
|
NM
|
NM
|
NM
|
NM: not measured, unit: dBm / 10 kHz
Electromagnetic interference caused by electrical devices installed in switched-mode power supplies and CPUs
As already described, many recent electrical devices employ switched-mode power supply to save energy and achieve low power consumption. Additionally, CPUs are typical components that control communication signals, although they may generate high-frequency emissions. These devices may generate wideband emissions and cause poor reception for radio communications. For instance, LED lamps and security cameras are installed in many hospitals, and these devices generally utilize switched-mode power supplies and/or CPUs.
Fig. 11 depicts the assumed EMI scenario, including WMTS and LED lamps. Generally, LED lamps, including their power-feeding lines, and the receiving antenna of WMTS are installed in the ceiling of hospital wards. Therefore, the reception signals of WMTS and the electromagnetic noise generated from the switched-mode power supply installed in LED lamps are readily received and transmitted to the patient monitor in the same transmitting line.
Fig. 12 depicts the frequency spectra of the electromagnetic noise generated by an LED lamp and security camera. These spectra were measured on SA using the whip antenna of WMTS at a distance of 10 cm away from the devices in a semi-anechoic chamber. The spectra were measured at a frequency of 30 MHz–1 GHz and an RBW of 100 kHz. The green line indicates the background noise of the measurement system that was measured using the maximum hold function, and the orange line indicates its 100 times average function. The electromagnetic noises generated from the LED lamp and security camera are represented by yellow and blue lines, respectively. Here, by assuming the worst case, we used extremely strong noise sources. The electromagnetic compatibility standard, CISPR 15 “Limits and methods of the measurement of the radio disturbance characteristics of electrical lighting and similar equipment,” for LED lamps has been established for the protection of radio communications [10]. However, this standard only targets bulb-shaped LED lamps; therefore, tubular lamps are not targeted. Additionally, as CISPR 15 only considers a measured distance of 3 or 10 m, it is not ideal for WMTS as LED lamps and their power lines may cause near-field EMI at much smaller distances.
As described above, considerable caution is required to ensure that electromagnetic noise is not generated by the main body of electrical devices. Electromagnetic noise is delivered via the connected cables, including the power-feeding line and communication cables, and it radiates wideband emissions. Unfortunately, the distance between the receiving antenna of WMTS and the power-feeding line, which is an EMI source, may be much lower than a few centimeters. Generally, good reception is ensured for WMTS at a distance of 7 m under a line-of-sight propagation condition and in the absence of a noise source. However, the electromagnetic noise generated by electrical devices may shorten this distance and cause poor reception. Thus, hospitals must investigate the presence or absence of electromagnetic noises generated by electrical devices before installing them. Additionally, it would be desirable to maintain a separation distance between the receiving antenna of WMTS and electromagnetic noise sources.
Managing the safe operation of WMTS
Simplified spectrum analysis function installed in the WMTSs receiver
In the above sections, we introduced recent EMI issues in WMTS. Notably, EMI sources may exist in some situations before they are noticed. Therefore, it is crucial to manage the electromagnetic environment around the WMTS frequency band, including the intentional emission of other types of radio communication systems and electromagnetic noise generated from electrical devices. Monitoring the frequency spectrum is key to visualizing the electromagnetic environment around the 400 MHz frequency band. However, most hospitals cannot perform such management operations because of the lack of staff members and cost.
Thus, we recommend a simplified measurement method using a WMTS receiver. Recently, WMTS receivers were installed with simplified spectrum analysis functions that measure the amplitude of received signals and/or electromagnetic noises in each WMTS frequency channel. This simple function can measure and display the received signal levels of every WMTS frequency channel. Fig. 13 depicts the screen of this function installed in a WMTS receiver. A channel 5021 WMTS signal is detected (Fig. 13(a)), and an increase in the noise floor owing to the electromagnetic noise generated from the LED lamp is shown (Fig. 13(b)) [11].
We investigated the accuracy of the signal amplitude of the receiver. Fig. 14 depicts the schematic of the experimental setup. The RF player that captured the WMTS signal replayed the recorded signal with a frequency of 429.25 MHz via the variable ATT and input it into the WMTS receiver and SA. We used two WMTS receivers with different input impedances of the input port: one was 50 ohm (Receiver A), and the other was 75 ohm (Receiver B). In this experiment, the simplified spectrum analysis function was run, and the received signal amplitude was measured.
Additionally, we measured the signal power at a bandwidth of 10 kHz using a real SA, and Fig. 15 depicts the evaluation results. The horizontal axis represents the received signal level of the real SA, the left side of the vertical axis represents the received signal level of Receiver A, which was displayed as RSSI [dB], and the right side corresponds to the signal received by Receiver B, which is displayed as voltage [dBµV]. Each result was displayed by 50 times the average with a standard deviation. The indicated values of the received signal levels of both simplified spectrum analysis functions correlated with that of a real one.
However, this function is highly limited—the measurable frequency band is only the Japanese WMTS band (400 MHz)—and cannot configure detailed parameters, such as RBW, SWT, and a detector. Nevertheless, this function can easily measure the received signal level in each WMTS frequency channel. The WMTS signal and electromagnetic noise generated by electrical devices or other radio communication signals can be confirmed in clinical settings. The simplified spectrum analysis function facilitates the management of the electromagnetic environment in hospitals, such as electromagnetic noise detection and the management of the WMTS frequency channel.
Measurement of the electromagnetic environment using software-defined radio
We propose another approach for spectrum management using SDR. SDR is a radio communication system in which the components that have been conventionally implemented in analog hardware (e.g. amplifiers, filters, mixers, modulators/demodulators, detectors, etc.) are implemented using software on a personal computer or embedded systems [12]. Similar to SA, SDR can be used as a radio communication receiver facilitated by signal processing, such as fast Fourier transform. The greatest advantage of SDR is its inexpensiveness. For instance, RTL–SDR (RTL-SDR.com) is an 8-bit SDR with a frequency range of 0.5–1766 MHz, and it costs approximately 30 US Dollars. ADALM-PLUTO (Analog Devices) is a 12-bit SDR with a frequency range of 325–3800 MHz, and it costs approximately 230 US Dollars. Moreover, the free software for SDR SA was recently released. We investigated the receiver sensitivities of RTL–SDR and ADALM-PLUTO. Fig. 16 depicts the schematics of the measurement setup. SG generated a continuous wave with a frequency of 429 MHz and was input into SDR and a real-time SA that were connected to a personal computer. The input signal level was adjusted between −10 and −130 dBm. We used two SDRs: RTL–SDR and ADALM-PLUTO, and Fig. 17 depicts the experimental results. The horizontal axis represents the output power of SG; the left side of the vertical axis represents the signal level that displays as [dB] received, measured by SDRs; and the right one represents the received signal that displays as the [dBm] level measured by SA. In this experiment, we adjusted the amplifier-gain built-in SDR. RTL–SDR and ADALM-PLUTO ranged from 0 to 20 and 0 to 50 dB, respectively. The signal level was correctly measured from −120 to 10 dB when no amplifier was used at both SDRs. When the amplifier gain was high, a lower-level signal of −140 dBm was measured, although the high-level signal was not measured correctly as the receiver circuit was saturated. Conversely, the real-time SA was measured at lower- to higher-level signals between −140 and 10 dB without an amplifier. Fortunately, the indicated values of the received signal levels of both SDRs correlated with those of real-time SA. Moreover, we can evaluate the electromagnetic environment by adjusting the amplifier gain.
We introduced an applicative trial using SDR with machine learning (ML). We developed a novel ML model to estimate the CNR of WMTS using the time-domain waveform data measured by RTL–SDR [13]. In this model, the lower to higher levels that ranged from 1 to 58 dB of CNR were correctly estimated, with a 99.5% R-square and 0.844 dB mean absolute error, using a gradient-boosting regression tree. RTL–SDR performed satisfactorily in estimating CNR despite using only an 8-bit resolution and inexpensive SDR. We implanted our model onto a single-board computer that is equipped with a graphic processing unit, such as the NVIDIA Jetson series. A novel inexpensive electromagnetic environment evaluation system that can be easily measured using SDR can be realized soon.