Frequency modulation detection is level-independent but thrice frequency-difference detection

Evidence

The data produced for justnoticeable frequency modulation (open circles in the right part in Fig. 7.4) show no dependence on level. The same is true for the just-noticeable frequency difference (dots) although the data are about a factor of 3 lower than for frequency modulation. The interquartile ranges indicate very clearly that the two sets of data do not overlap, i.e. just-noticeable frequency modulation and just-noticeable frequency difference not only lead to different results but may also be produced by different signal processing in our hearing system. — Professor Dr.-Ing. Hugo Fastl, Professor Dr.-Ing. Eberhard, Psychoacoustics- Facts and Models, 1034

The just-noticeable value of 2L1f depends on the frequency of modulation. This dependence is outlined in Fig. 7.10 and illustrates again that our hearing system is most sensitive to frequencies of modulation around 4 Hz. Figure 7.10 holds for a frequency of 1 kHz and a level of 60 dB. The minimum value of 2L1f i:::: about 6 Hz, corresponding to the data given in Fig. 7.8. For modulation frequencies between 10 and 50 Hz, 2L1f increases markedly with a slope that corresponds to the square root of modulation frequency. This increment ends sooner for low carrier frequencies than for higher carrier frequencies. For a carrier frequency of 8 kHz, the increment continues up to modulation frequencies of about 300 Hz, but for a carrier frequency of 1 kHz (Fig. 7.10) the end of the increment is reached at a modulation frequency of about 70 Hz. This effect is due to the frequency selectivity of our hearing system. Modulation in frequency produces sidebands similar to amplitude modulation. The narrow spacing of the many sidebands for a low frequency of modulation is much wider for high frequencies of modulation. The sidebands are spaced at integer multiples of fmod from the carrier frequency and become audible separately at high frequencies of modulation. In this case, the subject no longer listens for a change in frequency but for additional sidebands. At very low frequencies of modulation, the increment of 2L1f seems to be due to a limited memory. We do not remember the pitch of a tone very precisely after several seconds have elapsed. Consequently, the value 2L1f increases towards very low frequencies of modulation. — Professor Dr.-Ing. Hugo Fastl, Professor Dr.-Ing. Eberhard, Psychoacoustics- Facts and Models, 1065

Categorization

• XREF: Suggests separate neural mechanisms for detecting modulation vs. static differences — relevant to models distinguishing temporal coding from rate-place coding.
• SHARE: For a psychoacoustics colleague specifically: FM detection being level-invariant while AM strongly depends on level is a clean design-constraint fact for any hearing-model work.
• DEFINE: Explains what 2L1f means (twice the frequency-modulation depth at threshold) and why it has a U-shaped dependence on modulation rate.
• corroborated: 2 independent passages support this claim — Professor Dr.-Ing. Hugo Fastl, Professor Dr.-Ing. Eberhard, Psychoacoustics- Facts and Models, 1034; Professor Dr.-Ing. Hugo Fastl, Professor Dr.-Ing. Eberhard, Psychoacoustics- Facts and Models, 1065