Commercial automotive AM/ FM radios must perform dependably under a variety of operating conditions. Because designers of these radios employ digital-signal-processing (DSP) algorithms to overcome the effects of reflections, signal multipath, and fading, they often spend weeks in the field analyzing the effects of different signal conditions. A more practical and less time-consuming solution is the use of actual recorded radio signals to simulate the conditions faced by an automotive AM/FM radio design in the field.

A particularly challenging test of AM/FM radio performance is an evaluation of adjacent- and/or alternate-channel performance in which the radio must receive a moderately weak RF signal in the presence of a strong adjacent RF signal. To evaluate the subjective reception quality between radios, OEM car radio manufacturers often apply test plans based on specific in-field routes at different locations. When driving in cities, for example, it is common to have an RF environment called "urban canyon" where large buildings will create a complex pattern of multipath and shadowing specific for FM stations. Channel simulators are often used to recreate typical multipath models for a specific channel. The approach is no longer practical, however, when multiple simulators are needed to simulate adjacent interferers one of the reasons that the automotive industry still performs radio optimization using field test drives.

The test drive approach presents major repeatability problems, however. Since propagation conditions change due to weather conditions and the proximity of large vehicles, the results are never the same from one test drive to the next. Also, since test drives are performed over a wide range of locations, test repeatability is critical for evaluating an AM/FM radio design for use in different areas. Fortunately, due to advances in RF signal recording and storage, recording of actual FM broadcast signals offers a viable alternative to traditional test drives for evaluating commercial AM/ FM radio designs.

For example, the RF Record & Playback System from Averna can capture the full FM band (20-MHz bandwidth) with 14-b resolution. Based on a PXI hardware architecture from National Instruments, the digital recorder can also make parallel recordings of the GPS location, the radio's audio, and a video of the drive test from an onboard camera. The system has an 80-dB spurious-free dynamic range (SFDR) that may seem large, but FM receivers can handle a much wider range of signal levels from -2 to over +110 dBuV (-109 to +3 dBm).

The Ann Arbor area of Detroit represents a "hard-to-reproduce," extreme-dynamic-range test case, which can be used to evaluate a radio's capability to mitigate the effects of very strong adjacent interference. In the test drive, a local 3-kW transmitter at 107.1 MHz interferes with a signal from Detroit at 106.7 MHz. During the test drive, the interferer will range from 65 to 95 dB V while the desired channel will vary from 25 to 50 dB V. Besides multipath fading, there is a shadowing effect from a tall building that at times blocks line-ofsight reception. This type of fading is not well represented by models found in RF signal generators and is better suited for record-and-playback signal simulation.

Figure 1 shows this test drive superimposed on a map, with the 3-kW transmitter located on top of a tall building. While the antenna is in direct line of sight, the interferer is at its peak level and the reception of the desired frequency at 106.7 MHz is seriously degraded. Figure 2 shows a typical frequency spectrum of the signal strength while Fig. 3 shows the strength of both signals over the duration of the test drive.

Although the dynamic-range specification of the RF Record & Playback System is 80 dB, practical recording has shown the usable dynamic range to be around 60 to 65 dB. This reduction can be explained by digitizer saturation and the peak-to-average ratio (PAR) of multicarrier signals. To avoid possible saturation of the digitizer (signal clipping), recorded data must be at least 5 dB below saturation. For multicarrier signals, such as FMband or COFDM signals in an urban environment, the recorded signal will commonly have a 10 to 15 dB PAR due to vector addition of multiple RF signals present within the 20-MHz passband of the RF signal chain. The combination of the two factors leaves 60-to-65 dB of usable dynamic range for good-quality recording of multiple carriers.

Despite this limitation, the RF Record & Playback system has proven to be very effective in capturing multipath and weak signals if strong interferers (greater than 40 dB) are not present in the band of interest. The system employs a low-noise amplifier with better than 2-dB noise figure to capture weak signals in rural areas with no noticeable degradation, taking into account impedance mismatches. The RF Record & Playback system is designed for a 50-Ohm impedance while an automotive FM antenna and/or a radio input are traditionally matched to 75 Ohms for FM and to a high-impedance (above 1.5 kOhms) for AM.

Figure 4 presents a block diagram of the proposed solution that uses the Universal Receiver Tester (URT) to replicate a test condition such as the radio environment found in Ann Arbor; a weak signal is received in the presence of a strong interferer and combined with an impediment such as multipath effects. The first generator (Gen 1) provides the strong interferer. Since the noise floor of this source is high at the frequency of the weak signal, a notch filter is required to suppress the noise before combining it with the weak signal from a second generator (Gen 2) to simulate the desired signal. The notch filter, a three-pole cavity filter that can be tuned across a 10-MHz range, has at least 50-dB rejection (Fig. 5).

Gen 2 is configured with a Dynamic Range Extender (DRX), which is a programmable attenuator. By providing attenuation of the desired signal, the noise floor is also attenuated. Also, the effect of multi-path fading can be imposed on the weak signal since the DRX response time is sufficient to react to the fading profile (nominally, 40 times per second). For the purpose of the test, the radio is mounted within a shielded enclosure to suppress potential interference from external signals.

With the hardware configuration and the performance as described, an accurate representation of the RF signal condition must be provided from the output of the generators. The strong interferer is provided through Gen 1 from an on-site RF recording of the FM band. The Universal Receiver Tester (URT) RF recorder was used to capture the signal on a test drive during typical weather conditions. The weak signal is generated through Gen 2 by using the FM Fading Simulator, a .wav file format music track, and playback of the power profile derived during a field recording of the weak signal.

Multipath fading is a common type of signal impairment on mobile signal reception. In the case of FM signals, it will cause deep dropouts of the signal, to several times per second, depending on the speed of the vehicle. The FM Fading Simulator applies a flat fading response to the test signal compared to the more comprehensive effects of a standard channel simulator. However, this approximation of multipath is normally sufficient for testing FM tuner performance.

The FM Fading Simulator allows test tones to be generated as a standard AM/FM generator, or a given audio track in the form of a .wav file to be modulated. Additionally, the envelope of the generated signal can be modulated through playback of a tab-delimited text file. The power pattern envelop can be extracted through simple signal processing of any prerecorded FM signal by use of a simple utility within the playback toolkit.

Figure 6 illustrates the basic functional aspects of the test setup. The capability to play back any given audio signal or track, with the impairment of multipath flat fading on the desired signal, provides an additional degree of freedom not possible during field test. This is important considering that some DSP algorithms are designed to react differently, according to the type of audio program.

For the sake of an example analysis, assume that the desired weak signal is at a level of 30 dBV and a strong interferer exists at 110 dBV, 400 kHz away from the desired signal. This would be one of the most stressful tests that can be applied to an FM radio, and typical of what one would encounter in the worst part of the Ann Arbor test drive (Fig. 7). With these experimental assumptions, it is now possible to calculate the effective dynamic range of the system. With an in-field ADC recording dynamic range of about 65 dB and FM cavity notch filter with about 50 dB suppression, the test system dynamic range is about 115 dB at the frequency of interest.

Continue to page 2

Page Title

It must be determined if 115 dB is sufficient for testing a standard FM radio. The required dynamic range for the test is based on several assumptions: that the FM receiver has a 4-dB noise figure, that the wideband AGC threshold is about 80 dB V, that the noise figure increases 1 dB due to the interference signal strength beyond the wideband AGC threshold, and that the RF interference signal is only 110 dB V. Based on the last three assumptions, the receiver will desensitize by 80 to 110 dB or a 30-dB range. Therefore, the noise floor of the FM radio receiver in this case is -15 dB V + 4 dB (for the noise figure) + 30 dB (for desensitization), or 19 dB V. Therefore, the RF dynamic range needed for the test is 110 ?? 19 = 91 dB, allowing better than 20 dB of margin from the set-up for the testcase measurements.

Before using the system, the measurement setup was tested in order to confirm that it would not introduce any significant noise or distortion that would degrade the measurements. Power levels were calibrated by adjusting the relative insertion values. The specified levels were measured at the output of the combiner, the point of the radio antenna connection. Linearity was measured at above the required signal level. This was done by setting both generators to CW mode at the respective frequencies and at a total power level that is higher than that used for the test case. Referring to Fig. 8, Gen 1 is set to 0 dBm and Gen 2 to -20 dBm. There is no measurable third-order intermodulation (IM3). Figure 7 illustrates the signals set to near the maximum of test condition.

The spectrum analyzer plots in Figs. 8 and 9 prove that the power levels are accurate and that there is no IM3 introduced by combining the two signal generators. An additional test was performed to confirm that the notch filter does effectively suppress the noise from Gen 1 and that the subsequent measurements at the frequency of interest are limited by thermal noise, and not by that from the test system. Audio measurements were performed on the output of the test receiver using the URT Audio Analyzer over different offsets of the interferer signal center frequency. In this way, the effect of receiver desensitization would demonstrate that the noise floor of the setup does not impact the test-case measurements.

Both generators were set to FM mode: Gen 1 at 107.1, 107.3, 107.5, and 107.7 MHz with +3 dBm output power, 1-kHz modulating tone, stereo mode, and 75-kHz deviation; Gen 2 at 106.7 MHz, -62 dBm output power, 1-kHz modulating tone, mono mode, and 22.5-kHz deviation.

At the different Gen 1 frequency settings, the following was observed. If the test setup is the limiting factor, the audio signal-to-noise ratio (SNR) will not change as the interference RF signal moves further away. As the table illustrates, the desensitization of the receiver is caused by the strong interferer, and therefore proves that noise from the test setup is not the limiting factor at the test frequency of 106.7 MHz.

The signals used to stimulate the test receiver were derived from broadcasted RF signals captured during a test drive at Ann Arbor, MI. Audio listening tests were performed at Averna by expert listeners Hans Troemel from Nfuzion LLC and an engineer from a Japanese automotive OEM. (Nfuzion LLC is a full product design engineering consulting company that services the automotive OEM and consumer electronics markets.)

The audio tracks were taken from the output of the same test receiver during the on-site recording and in Averna??s laboratory (Montreal, Quebec, Canada). Also recorded in Ann Arbor was the RF signal played back through Gen 1 of the setup. Finally, Gen 2 was used with the URT FM Fading Simulator to generate the weak signal at 106.7 MHz. The .wav file used with the FM fading simulator was the same artist and track as the off-the-air audio recording made during the test drive.

The test plan can be summarized as follows:

The URT RF Player is started and the FM Fading Simulator with the desired RF signals for testing are passed to the test receiver, which is tuned to the weak signal at 106.7 MHz.

Next, the audio from the test receiver is recorded during the RF playback.

Then, listening is performed to the recorded audio collected from the site by trained, experienced listeners.

Following this, listening is performed to the recorded audio collected from the test setup in the lab.

The two audio recordings are then compared in terms of fidelity and quality in order to determine whether signal degradation has taken place due to environmental or signal fading effects.

After a comparison of the audio recordings, it was determined that the recordings were similar. Both audio recordings contained the same mitigation from the DSP within the receiver, such as switching from stereo to mono and audio high-frequency cutoff. Further listening tests were performed with and without multipath fading applied to the weak signal. In this way, the effect of receiver desensitization was isolated. This method is a valuable engineering test tool to isolate the impairment variables of the site in order to gain greater understanding and optimization of the audio quality. In addition, the approach can be applied to a wide range of radios.