Research on Cosine-Sum Windows with Maximum Side-Lobe Decay for High Precision ADC Spectral Testing
Round 1
Reviewer 1 Report
The main contribution of this paper is to optimize a general method to estimate the parameters without requiring coherent sampling. But currently, there are some other optimization methods for this purpose, why don’t you compare them with the method you proposed, in terms of power consumption, speed, complexity, or other specifications, to prove the effectiveness of your method. What’s more, the ADC under test is a low-speed and middle-resolution sample(<100KHz, 18bit), so how to prove this proposed method is also applicable to other types of ADCs, especially for high-speed and high-resolution ADC(>1MHz, 24bit)?
Author Response
Response to Reviewer 1 Comments
Point 1: There are some other optimization methods for this purpose, why don’t you compare them with the method you proposed, in terms of power consumption, speed, complexity, or other specifications, to prove the effectiveness of your method.
Response 1: This paper is focused on optimizing the side lobe decay rate of the classical window, so only the classical window functions are given in the simulation, as shown in Figure 5. In order to compare with other state-of-the-art methods, the comparison of the computation time with different methods is presented in Table 4. It can be seen that the proposed window is more computationally efficient and accurate than other methods.
Point 2: The ADC under test is a low-speed and middle-resolution sample(<100KHz, 18bit), so how to prove this proposed method is also applicable to other types of ADCs, especially for high-speed and high-resolution ADC(>1MHz, 24bit)?
Response 2: Noncoherent sampling is a common problem in high-precision ADC testing because the lower noise floor is sensitive to spectral leakage. The high-speed ADC usually has less number of bits, so this method is suitable for high-speed ADC. For high-speed ADC, a more common problem is the clock jitter, which will be processed by another method. Our group also researches this problem at the same time.
Special thanks to you for your good comments.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The paper is very interesting and concrete contribution to the topic of non-coherence in ADC. The obtained results are impressive. I have just few minor remarks:
1) Eq. (3). Usually the division by M is applied for the IDFT (see Proakis and Manolakis "Digital Signal Processing"). Please, explain in the text why you are dividing by the number of data in the DFT.
2) Eq. (5). Please, replace "NoiseFloor" with "Signal to Quantization Noise Ratio. Moreover, it should be reported in the text that Eq. (5) implies that the input signal amplitude covers the full dynamic available for the conversion.
3) Fig. 6. Even if Table 3 reports aggregates data, it would be interesting to add in Fig. 6 the errors simulated in the non-coherent case of Table 3, so that to compare the data in Fig. 6 with a worst case.
Author Response
Response to Reviewer 2 Comments
Point 1: Eq. (3). Usually the division by M is applied for the IDFT (see Proakis and Manolakis "Digital Signal Processing"). Please, explain in the text why you are dividing by the number of data in the DFT.
Response 1: The spectrum obtained by dividing the DFT by M can directly reflect the amplitude of the input signal, so it is very common in spectrum testing. This formula is also cited extensively in classic books and papers, such as Equation (9.49) in ‘An Introduction to Mixed-Signal IC Test and Measurement.’ (Gordon Roberts, Friedrich Taenzler and Mark Burns. Oxford University Press) and Equation (4) in ‘FIRE: A Fundamental Identification and Replacement Method for Accurate Spectral Test Without Requiring Coherency.’ (S.Sudani.; D.Chen. IEEE Transactions in Instrument and Measurement).
Point 2: Eq. (5). Please, replace "NoiseFloor" with "Signal to Quantization Noise Ratio. Moreover, it should be reported in the text that Eq. (5) implies that the input signal amplitude covers the full dynamic available for the conversion.
Response 2: We replace "NoiseFloor" with "Signal to Quantization Noise Ratio" and add more description about Equation (5).
Point 3: Fig. 6. Even if Table 3 reports aggregates data, it would be interesting to add in Fig. 6 the errors simulated in the non-coherent case of Table 3, so that to compare the data in Fig. 6 with a worst case
Response 3: In order to compare the estimation error of the different window, a classic Blackman-Harris window (3-term) is used to suppress the leakage. The simulation is performed 100 times to verify the functionality of the proposed method. For each simulation, the estimation error of SINAD and SNR are plotted in Figure 7 for both Blackman-Harris window and 3-term MSLD window.
Special thanks to you for your good comments.
Author Response File: Author Response.pdf
Reviewer 3 Report
The paper entitled "Research on cosine-sum windows with maximum side-lobe decay for high precision ADC spectral testing" meets the unanimously agreed format for a scientific article: theoretical model, simulation and experimental validation. The conclusions are supported by the results from the simulation and the experiment. The content provides the requirements for these results to be reproduced by an informed reader of the journal Electronics. In this sense, the analyzed paper largely meets the qualities of a tutorial in the methods of evaluating high-resolution ADCs. The paper can be published in this form.
Author Response
Response to Reviewer 3 Comments
Thank you very much for your comment. I have revised our paper to make it more complete. These changes will not influence the content and framework of the paper.
Author Response File: Author Response.pdf
