Noise Floor Reduction in Frequency Delta-Sigma Modulation Microphone Sensors

Round 1

Reviewer 1 Report

The paper presented a design to lower the noise floor of a microphone using FDSM for data acquisition. The design has its merit but the related theory and design needs clarification. The size of the microphone is huge compared to the current microphone sensors on the market. Thus the integration of data acquisition with a microphone of such size is less attractive. 

Major concerns:

1. Please clarify fs. Is it the Nyquist frequency or the ADC sample rate? What is the relationship between the frequency of VCO and fs?
2. In figure 2, please explain the function of the register and XOR gate combination. A time sequence of the associated signals (output from the quantizer, fc and the output of the XOR gate) is expected.
3. It is unusual to use a plot (figure 5) to represent the phase noise model. What is the f0 in the chart? Why is the noise spectrum not continuous?
4. The derivation of equation 3 is required. Is f the same as f(t)?
5. The derivation of equation 5 is required. What is fc?
6. The mechanism how the sound is detected is not clear. Is the assumption that the displacement of the membrane cause the frequency shift in the VCO? The key issue is the quantification of the sound pressure. Is the ADC output (digital value) proportional to the sound pressure. This is missing in the paper. The ultimate goal is to integrate the ADC with the sensor. Therefore it has to be addressed. 

Minor Concerns:

1. Line 40. Should "few" be "fewer"?
2. Line 97. The previous sentence says that there is no feedback. Why does the next sentence say the accuracy of the feedback is extremely high?
3. Figure 3: Is "microstlip" "microstrip", It should be cylindrical, not cyrindlical.
4. Figure 10: The frequency scale is not correct.
5. Figure 12: Is it 32-bit or 6-bit after Counter in the digital filter. Use hyphen between the number and the word "bit".
6. Figure 16: What is the difference between a and b? There is not sound in both cases. Why is there a spike at 2kHz in c and spikes at 2kHz and 3kHz?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The manuscript reports continuing optimization of previously reported frequency modulated microphone. Significant improvement is obtained and the microphone performance can now be evaluated.

The manuscript requires some polishing in presenting the results:

1. Figures 10 and 11 have incorrect unit in x-axis (GHz => Hz)
2. Table 1 presents power but no units are given (dBm, dBc or something else)
3. The proposed application is microphone but no performance figures are given. The authors should calculate the lowest measurable sound level for comparison with other microphones. The measurements for this are readily available in the manuscript but this crucial step of converting measurements to microphone performance is not carried out.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors addressed most of the concerns in the previous review. There are still remaining concerns.

Major concern:

Equation 2 showed the nonlinear relationship between the L and fr. It is assumed that L changes to the sound pressure. Further discussion is needed to reveal that the non-linearity is trivial and fr is proportional to the sound pressure.

Minor concerns:

1. Line 87: Missing "," after XOR
2. Figure 3: Add fs next to Clock for clarification
3. Line 145: The authors used the discrete frequency spectrum to represent the noise spectrum. This can easily mislead readers that the noise is cyclic with the base frequency f0. It would be better to use the continuous spectrum and integration within a small frequency range df to represent noise power density.
4. After equation 6. The sentence "This indicates that the effect of ..." is not accurate. It would not be true if Δfs is much larger than Δfc. The sentence should be removed.
5. Figure 13:  Is the edge detector before the sampler? If not, how does the edge edge detector work with the 32-bit data?
6. Figure 16a: The right figure did not show the phase noise (blue)

Author Response

Please see the attachment.

Author Response File: Author Response.pdf