ADAU1787 DMIC input noise floor increase high

Hello kermit.yu,

I do not have a great explanation except that all PDM modulators/demodulators will do that when they get within 6 or7dB of full scale. I will try to get a better explanation.

Dave T

Remember that PDM (pulse density modulation) represents a waveform as a series of high and low pulses. When the waveform spends an equal amount of time at logic 0 and logic 1, the average output is a zero signal. More time at 1 represents a positive signal, and more time at 0 represents a negative signal. Also remember that the width of pulses is quantized to the PDM clock rate so no pulse can be narrower than a half clock cycle in double-edge clocking as used in microphones.

At moderate amplitudes, the time quantization is not a huge problem since a slightly too long pulse can be offset by a subsequent slightly too short pulse. This is a key role of the noise shaper. However, as the signal approaches positive or negative full scale, the waveform is spending nearly all the time at either 1 or 0, respectively. The noise shaper has less "room" to do this balancing at high output levels.

The variation in pulse width is finite, and this quantization produces an increase in noise near the rails. The higher the clock rate for the same bandwidth, the lower this effect at the expense of radiated EMI, power, etc. However, in practice, the distortion from the mechanical elements of the microphone dominate over the modulator noise at high levels. Note that this is not only true of PDM. There is a very similar effect with PWM, and neither can achieve 100% modulation depth.

Ken