Updating IEEE 185-1975

IEEE 185-1975, Standard Methods of Testing Frequency Modulated Broadcast Receivers, has long been used to evaluate consumer FM products in the U.S. Although its IEEE status is withdrawn standard, it remains an informal reference because of its wide acceptance and familiarity. However, some sections of the standard specify tests that are inconvenient, unrealistic, or inherently inaccurate. Other tests promote confusion, inadequately probe worst-case signals, or impose difficulties with modern equipment. Finally, some useful tests are missing.

Meter Response

IEEE 185-1975 §3.9 specifies an average-responding audio level meter. This introduces an error of −1.05 dB when measuring Gaussian noise. The standard does not compensate for this error. Sensitivity and S/N figures for older receiver specifications and test reports that used such meters should be 1 dB high. To obtain actual power ratios, I use a true RMS meter for all audio level measurements.

Test Frequencies

§5.2 specifies 90, 98, and 106 MHz (or 90.1, 98.1, and 106.1) as standard test frequencies. However, local interference that leaks into the test setup may preclude use of a frequency, especially for S/N measurements. Since I don't have a screen room or screened enclosure to attenuate such signals, I use the clear frequency closest to that specified.

Stereo Modulation

§5.3 specifies L = −R as standard stereo modulation. This is a curious choice, one not explained in the standard. It puts all of the modulation in the stereo subband and none at baseband. It is not representative of typical broadcast signals. In fact, no uncontrived sounds exhibit L = −R. I use L-only, R = 0, which spreads the test signal over the composite spectrum like typical broadcast modulation.

Tuning

§5.4 says to fine-tune the receiver or signal generator frequency for minimum distortion using instrumentation. This can unrealistically minimize mistuning. Instead, I set the signal generator to channel center and tune the receiver as a user would. For an analog receiver this may mean using a center-tuning meter for strong signals and fine-tuning weak signals for cleanest sound.

Output Level

§5.5 tells how to set an adjustable output level for testing, but nowhere does the standard specify an output level measurement for a standard test signal. I define output level as the voltage generated across the standard audio load (100kΩ || 1000 pF) for a 100%-modulated, 1-kHz, 65-dBf monophonic signal. When adjustable, I report the maximum output level.

Output Impedance

The standard does not specify an output impedance measurement. A high impedance can cause extreme-treble loss when driving high-capacitance cable or circuitry. An impedance measurement can help distinguish this loss from that due to deemphasis-network error. If I can't determine the output impedance by schematic inspection, I report the load resistance that drops the audio output level 6 dB at 15 kHz.

Frequency Response

§6.9 says to report frequency response as 30–15,000 Hz ±X dB with respect to the response at 1 kHz. I separate the bass and treble components. I report as treble response the response from 1 to 15 kHz +X/−Y dB. I report as bass response the frequency below 1 kHz where the response is −1 dB.

Modulation Acceptance

§6.11.2 calls for distortion tests to 120% modulation, but this does not test a receiver's ability to handle signals accidentally or intentionally deviated much higher, which occasionally occur. I increase the deviation of a 65-dBf monophonic signal modulated with a 1-kHz tone until THD reaches 1%. I report the modulation percentage as modulation acceptance. If it is significantly worse at another modulation frequency, I report the value.

Capture Ratio

§6.13 gives a convoluted procedure for measuring capture ratio, one that involves an approximation. I make a direct measurement using the stated definition, which is simply how far below an unmodulated signal a 100%-modulated, 1-kHz interfering signal must be to suppress the modulation 30 dB. Sometimes I also report the figure for 50-dB suppression, a more realistic level for interference tolerance. All signals are monophonic.

Receivers are much more susceptible to co-channel interference and multipath distortion in stereo, but §6.13 addresses monophonic reception only. I define stereo capture ratio in the same way as the monophonic figure, but with both signals in stereo. Any FM signal generator can provide the unmodulated signal with stereo pilot.

I measure capture ratio with the desired signal at 65 dBf. When I have the patience, I follow the standard by remeasuring at 45 dBf and reporting the worst result.

Selectivity

§6.14 calls for the use of a 1-kHz bandpass filter when testing selectivity “to reduce errors from noise.” But the filter only increases measurement error. For conventional receivers, a dozen or more high-level harmonics accompany the 1-kHz fundamental of an adjacent-channel test signal that leaks into the tuned channel. The harmonics are distinctly audible and their power contribution should be included. For DSP receivers, reciprocal mixing of local-oscillator phase noise may limit selectivity. Neither the modulation fundamental nor its harmonics ever appear. Instead, the background noise rises. Using a bandpass filter in this case can yield a grossly inaccurate result.

RF Spurious Response and Intermodulation

§6.15 and §6.16 define RF spurious response and intermodulation tests that impose several difficulties, particularly for modern receivers. One problem is that monophonic usable sensitivity is the reference signal level for unwanted responses. This is the RF level at which audio output for a 100%-modulated, 1-kHz tone drops 30 dB when measured through a 1-kHz notch filter. Not only is an audio filter needed for this measurement, but its Q is unspecified. Notch bandwidth can affect the result for receivers that use adaptive noise reduction, which concentrates residual noise near a test tone. A more fundamental problem is that modulation-induced noise for narrow IF filters may cause 30-dB usable sensitivity to be several dB worse than 50-dB quieting sensitivity. This perplexing reversal confounds sensitivity and distortion. Moreover, a signal at usable sensitivity is too noisy for enjoyable listening but too quiet to degrade intelligibility. It is an inappropriate test level. A further problem is that the standard calls for unwanted-signal levels to be measured using residual unmodulated noise. Such two-signal measurements are highly susceptible to receiver and signal generator phase-noise sidebands, which can be strong enough even for nonsynthesized oscillators to invalidate a test. An intrinsic confusion is that the standard defines spurious response ratio as the worst result of several specified tests, one of which has the same name. It's never clear whether a quoted spec refers to the single test or to the ensemble. Another inconvenience is that all results are given as ratios with respect to usable sensitivity. Determining the absolute signal level at which RF impairment commences requires the addition of two numbers. Finally, since due to its limitations you may not ordinarily measure usable sensitivity, the tests may impose unnecessary additional work. To address these issues, I modified and renamed the tests.

I define RF intermod as the 50-dB quieting level for a third-order intermodulation product. I use three signal generators for this measurement. First, I modulate one generator 92% (69 kHz deviation) at 400 Hz. This yields the same audio level after 75-΅s deemphasis as would a 100%-modulated, 1-kHz tone. I set this generator 1600 kHz above the tuned frequency and sum it with an unmodulated generator 800 kHz above. Next, I set a third generator at the tuned frequency to the 50-dB quieting level for a 100%-modulated, 1-kHz tone. I sum all three generators and adjust the equal RF levels of the off-frequency generators until the levels of the 400-Hz and 1-kHz components in the audio output are equal. The tuned-frequency generator and the intermodulation product compete for capture, and no amount of phase noise will alter the relative audio-component levels. I repeat the measurement with the untuned generators below the tuned frequency and average the two results.

I define RF spur as the 50-dB quieting level for an untuned signal. I tune a signal generator from 88 to 108 MHz, increasing its level (to a 10-mW maximum of 130 dBf) until I find a spur at the tuned frequency. Because the spur may be the result of high-order intermodulation, I modulate it with 400 Hz using a deviation that yields the same audio output level as that from a tuned-frequency generator 100%-modulated with 1 kHz (sometimes less than 10 kHz suffices). I do this with both generators at a high enough signal level that the audio level is independent of RF level. Then I reset the tuned-frequency generator to the 50-dB quieting level and sum the signal from the spur generator, adjusting its RF level until the levels of the 400-Hz and 1-kHz components in the audio output are equal.

I define RF image as the 50-dB quieting level for a mixer image. I use the measurement procedure for RF spur but tune the spur generator to the image frequency (tuned frequency + 21.4 MHz). Because the IF may not be exactly 10.7 MHz, I adjust the generator frequency for maximum response. Subtracting the 50-dB quieting level yields a figure equivalent to image response ratio defined in §6.15.2, but the updated test avoids phase noise and provides an absolute level.

Whenever phase noise is low enough, you can use one fewer signal generators and a simplified procedure. Just set the unwanted response to the 50-dB quieting level. Normally this procedure is valid when measuring RF spur and RF image.

In all cases, the 50-dB quieting level is that defined in §6.4: the RF signal level at which the audio output drops 50 dB when 100% 1-kHz modulation is removed. It is not the unmodulated RF signal level that quiets the background noise 50 dB.

RF Mismatch Loss

§6.20 specifies RF input impedance measurements that yield SWR at the antenna port. I measure the equivalent return loss but convert it to RF mismatch loss in dB so that sensitivity degradation can be directly assessed. I give the range of values across the FM band. I use a signal level below the RF AGC threshold to avoid any change in input impedance.

RF AGC Threshold

Some receivers reduce mixer and IF intermodulation by applying AGC to the front-end. But when activated by a strong untuned signal, AGC may lower tuned-signal S/N by degrading the front-end noise figure. I define RF AGC threshold as the tuned RF signal level at which the front-end transfer function falls 1 dB below linear. Usually I estimate this point by determining when the AGC control voltage reaches the gain-reduction threshold. RF AGC threshold is defined for tuned signals, but it provides a lower bound for untuned signals attenuated by front-end selectivity.

Noise Figure

Noise figure is a useful RF measurement the standard does not address. Unlike quieting sensitivity, noise figure characterizes weak-signal performance independently of IF and audio bandwidth. It provides an easily interpreted absolute metric. Mainly it characterizes the front-end, which normally limits receiver noise figure. Above the FM threshold, audio noise varies linearly with RF signal level. I measure it with an AC voltmeter and calculate noise figure with the Y-factor method.

I connect a signal generator and a calibrated noise source to a 50Ω power combiner that drives the receiver through a 50:75Ω matching network. I set the RF signal level above the point where any soft-muting or variable high-cut filter engages, usually between 60 and 70 dB of quieting. Typically the receiver noise floor is 10–20 dB below. To improve accuracy, I subtract the noise floor from a measurement with the MC.EXE measurement corrector in this set of noise figure utilities. I measure the audio noise level with the noise source on and off. The difference between the two measurements, after correcting each for the noise floor, is the Y-factor. Then

NF = ENR - L - 10log(10Y/10 - 1)

where NF is noise figure, ENR is the excess noise ratio of the noise source, L is the sum of power combiner and matching network losses, and Y is the Y-factor, all in dB. Another utility in the set, Y.EXE, will do the calculation.

Stereo S/N

§7.3 specifies an RF signal level of 65 dBf to measure stereo S/N. Because this level may not fully quiet a receiver, manufacturers sometimes also report S/N at 85 dBf. This is a useful addition to the standard.

The standard specifies an unweighted measurement. Some manufacturers specify A-weighted S/N, while others use RF levels above 65 dBf without noting it. Stereo S/N figures above 80 dB are not likely in accordance with §7.3.

For all S/N and quieting sensitivity measurements, I use a 200–15,000 Hz bandpass filter that meets the specifications in §3.8.

Audio Latency

The processing delay of a DSP receiver may be large enough to cause an echo when played along with a conventional receiver, perhaps located in another room. I measure audio latency in milliseconds with a two-channel oscilloscope and pulsed sinewave modulation.
March 5, 201388–108 MHz