Nearly all stereo decoders use a 38 kHz square wave to demodulate the L−R signal, which occupies 23 to 53 kHz in the stereo composite. An unintended consequence is that the waveform's fifth harmonic demodulates power near 190 kHz. HD Radio digital sidebands, which appear at 129 to 198 kHz for hybrid service mode MP1, can cause an annoying background noise when the fifth harmonic demodulates them. Extended hybrid signals, whose spectrum begins near 116 kHz for service mode MP3, cause additional noise due to the third harmonic at 114 kHz. A lowpass filter between the FM detector and stereo decoder can eliminate this HD Radio self-noise.
These images from an HP 141T/8553B/8552B spectrum analyzer show the detected spectrum in several tuners for various IF filters. I sampled the signals at the detector output prior to any postdetection filter. The horizontal span is 0 to 200 kHz, the vertical scale is 10 dB/div, and the analysis-filter bandwidth is 3 kHz.
This is the demodulated spectrum of KUSC in Los Angeles in a Yamaha T-1020 tuner. The tuner had two Murata SFE10.7MA5 280 kHz ceramic filters installed. The digital sidebands occupy spectrum about 129 to 198 kHz from the center of the RF channel. Analog modulation smears the demodulated digital spectrum by moving the carrier with respect to the digital sidebands. The smearing is visible in this image and in the ones that follow.
The 190 kHz fifth harmonic of the stereo decoder's 38 kHz oscillator can demodulate the sideband power between 175 and 198 kHz to baseband. This is the area of interest. (The digital sidebands may get smeared to within demodulation range of the 114 kHz third harmonic, but the resulting noise is likely to be audibly masked by the modulation that caused the smearing.)
This compares the spectra for two SFE10.7MXs, which came with the tuner, and a pair of SFE10.7MP3s. Both are specified as 250 kHz wide at the −3-dB point. The lower trace for the MP3s is nearly identical to that of the 280s in the previous image.
This is the spectrum for two SFE10.7MS2 230-kHz filters. Noise near 190 kHz is 13 dB lower than with the MXs, and about 10 dB lower than with the MP3s.
This overlaps the spectra of the 230s and the MXs. The 230 spectrum is slightly corrupted by an adjacent-channel carrier near 200 kHz, but the difference between the filters is clear.
This overlaps spectra for pairs of 280, 180, and 150 kHz filters in a Kenwood KT-45 for KCRW in Santa Monica. The 180s were stock. This is the only single-bandwidth tuner I've come across that came with narrow filters installed.
This image overlaps wide and narrow spectra for KYSR in Los Angeles in a Technics ST-G5 tuner. The wide filter is a 230 followed by a 280, while narrow adds a pair of 150s. The narrow filter drops the response more than 30 dB at 190 kHz. Because the ST-G5 has no postdetection filter, the audible difference in HD Radio self-noise between the two bandwidth settings is pronounced. The 19 kHz pilot has been mostly cancelled at this test point.
This shows wide and narrow spectra for KWVE in San Clemente in a Technics ST-9030 tuner. The station was transmitting in mono. The wide filter is a phase-linear LC block with 23 dB of alternate-channel selectivity. Narrow is a pair of 110 kHz SFE10.7MHYs.
This image compares four 230s vs four 250s (MP3s) in a Sony ST-J75 for KKJZ in Long Beach. (The traces are noisier than in previous images because the spectrum analyzer was set for less video filtering.) Using the pilot as a reference, cascading four filters instead of two drops the digital sidebands an additional 20 dB near 190 kHz. The 230s drop the sidebands another 18 dB. The ST-J75 has a very effective LC postdetection filter and HD Radio self-noise isn't a problem for either set of IF filters.
88–108 MHz