Three-Channel Fully Integrated Galvanic Isolation Interface in GaN Technology
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThis paper designs and experimentally demonstrates three galvanic isolation channels based on a TSMC GaN-on-Si technology. It seems to be a modified design of the authors' previously published papers [45] and [46].
1. The authors claim that this work improves the receiver robustness compared with the previous work due to circuit design innovations.
(1) It is suggested to discuss the drawbacks of the previous work first before introducing the new design, so that the motivation of the design modifications can be better understood.
(2) It is better to show the receiver robustness improvement by compare the PVT and MC simulation results of this work with the previous design.
2. Although it might has been discussed in previous publications, it is suggested to include some discussion about the coupling coils design for completeness.
(1) The coupling between the coils should be quite weak with the side-by-side configuration. Could the authors provide more information on the exact values of the voltage gain of the coupled coils and the various receiver stages of this specifc design?
(2) With three channels put together, are the crosstalks between channels serious?
(3) It seems that different coil designs are used for different channels according to Fig. 7. What are the corresponding design considerations?
3. The authors assumes that the molding compound could achieve a dielectric strength of 50V/um so that reinforced isolation can be achieved with 250um separation. Reference should be provided for this data of 50V/um.
4. Could the authors have some comment and comparison in terms of the power consumption?
5. There are quite a few errors in the writing of this paper.
(1) In the last paragraph of Section 1, the authors said "Section 4 describes measurements and Section 5 draws conclusions"， in fact Section 5 discusses the measurements and Section 6 draws the conclusions.
(2) The same labels are used for different devices in Fig. 3(a) and Fig. 3(c), it is difficult to understand which devices the text is refering to.
(3) Please check if it is correct that the No. of dice of this work is 3 in Table I.
Author Response
This paper designs and experimentally demonstrates three galvanic isolation channels based on a TSMC GaN-on-Si technology. It seems to be a modified design of the authors' previously published papers [45] and [46].
Many thanks for your comments.
(1a) The authors claim that this work improves the receiver robustness compared with the previous work due to circuit design innovations. It is suggested to discuss the drawbacks of the previous work first before introducing the new design, so that the motivation of the design modifications can be better understood.
(1b) We have followed someway your suggestion. After first describing the new RX solution, we highlighted the drawbacks of the previous implementation (lines 184-206) and then how the proposed solution overcomes these drawbacks (line 207-224).
(2a) It is better to show the receiver robustness improvement by compare the PVT and MC simulation results of this work with the previous design.
(2b) Compared with typical integrated analog circuits in GaN technology, our system is rather complex implementation with a lot of functionalities. It was a big effort for the authors to describe the contents in a typical paper length. We wouldn’t like to overcharge the paper with simulations. Moreover, all the limitations and improvements discussed can be easily understood by circuit inspection.
(3a) Although it has been discussed in previous publications, it is suggested to include some discussion about the coupling coils design for completeness.
(3b) Sorry, but this topic was faced by the authors in many previous papers in CMOS, BCD, and GaN technologies. We would like to avoid introducing descriptions in which there are no novelties.
(4a) The coupling between the coils should be quite weak with the side-by-side configuration. Could the authors provide more information on the exact values of the voltage gain of the coupled coils and the various receiver stages of this specific design?
(4b) The coupling between coils is very weak as you say. Indeed, the attenuation from the output of the TX to the input of the RX is about 38 dB. We have included some numbers in Section 3 (lines 151-153) from which this attenuation can be derived. The only gain stage in the receiver is the rectifier and is followed by the SC comparator.
(5a) With three channels put together, are the crosstalks between channels serious?
(5b) The crosstalk is in general a problem, but the crosstalk signal can be made low with respect to the desired signal by properly setting the distance between the antennas and selecting different RF carrier frequencies for the channels. This problem was extensively discussed in the previous paper ref. [46], although a brief mention is also given in Section 5 of this paper.
(6a) It seems that different coil designs are used for different channels according to Fig. 7. What are the corresponding design considerations?
(6b) We have used different coils to work with different carrier frequencies as well depicted in Fig. 13. This was done to reduce crosstalk as mentioned before, but also to increase gain from the TX to the RX transfer as in the driver channel.
(7a) The authors assumes that the molding compound could achieve a dielectric strength of 50V/um so that reinforced isolation can be achieved with 250um separation. Reference should be provided for this data of 50V/um.
(7b) A reference to our previous paper (ref. [43]) in CMOS technology has been given, which extensively discusses on dielectric materials for high galvanic isolation and gives various references that clarify this aspect.
(8a) Could the authors have some comment and comparison in terms of the power consumption?
(8b) The current consumptions of the three channels are reported in the beginning of Section 5. A comparison with other works is in general difficult since we don’t have similar implementations in GaN technology and a comparison with implementations in silicon technologies wouldn’t be fair.
(9a) There are quite a few errors in the writing of this paper. In the last paragraph of Section 1, the authors said "Section 4 describes measurements and Section 5 draws conclusions"， in fact Section 5 discusses the measurements and Section 6 draws the conclusions. The same labels are used for different devices in Fig. 3(a) and Fig. 3(c), it is difficult to understand which devices the text is refering to. Please check if it is correct that the No. of dice of this work is 3 in Table I.
(9b) Thanks. We have corrected the errors.
Reviewer 2 Report
Comments and Suggestions for AuthorsThe reviewer would like to thank the author for the manuscript presented. Recommendations for the manuscript are as follows.
1-It may be good to place Figures after it is first stated.
2-Discussion section can be added to discuss the results given in Table 1.
3- Conclusion should be rewritten emphasizing major contributions of the paper.
4-Because the lack of analytical design equalities of the proposed system, the manuscript may better fit Applied Sciences Journal.
Author Response
The reviewer would like to thank the author for the manuscript presented. Recommendations for the manuscript are as follows.
Many thanks for your comments.
(1a)-It may be good to place Figures after it is first stated.
(1b)-Given the high number of figures, it wasn’t possible to place figure close to their description.
(2a)-Discussion section can be added to discuss the results given in Table 1.
(2b)-Unfortunately, Table 1 reports available implementations that are in silicon technologies with the only exclusion of the previous work of the authors. The comparison with implementations in silicon technologies is in general not fair. This is the reason of the lack of a comparative discussion.
(3a)-Conclusion should be rewritten emphasizing major contributions of the paper.
(3b)-The various improvements on the previous implementation [46] have been well described along the paper and in the abstract. The authors wanted to avoid repetition of contents.
4a-Because the lack of analytical design equalities of the proposed system, the manuscript may better fit Applied Sciences Journal.
Reviewer 3 Report
Comments and Suggestions for AuthorsThe review report of the paper can be found in the attached file.
Comments for author File: 
Comments.pdf
Author Response
Many thanks for your comments.
Sorry, but being the file of comments in pdf I am not able to report the individual comments. Therefore, I only give the answers numbering the answers
- Electrical resistances in the conduction mode are important for the power transistors that are used for the implementation of power switches. In this work, transistors are used for low-power control circuits. They usually work in saturation where they approximate a voltage-controlled current generator rather than a resistance.
- In line 91 is specified that the dielectric material is “a standard molding compound”. Moreover, a reference has been given.
- The antennas are near-field planar components. They exploit the magnetic coupling. The antennas can well be seen in the photograph in Fig. 7. They were discussed in the previous paper [46].
- The capacitances are of course real components. However, the inductance losses in the RF circuits are well dominant on the capacitance losses that can be neglected. Therefore, the quality factor of the LC networks in both TX and RX is dominated by the inductance quality factor.
- Perhaps the reviewer refers to Fig. 6b (not to Fig. 5b). They are typical signals in the time domain. The authors think that this aspect cannot be confused.
- Sorry, but formatting this paper was very difficult, and we prefer not to change it.
- Sorry, but I don’t understand this comment.
- Unfortunately, we were not able to achieve a better resolution for Fig. 7 with our equipment.
Round 2
Reviewer 1 Report
Comments and Suggestions for Authors1. The authors refer the cross talk to ref [46]. However, it is also not thoroughly discussed in [46]. When signal is transferred through one channel, what are the outputs of the other channels during measurement?
2. The authors refer the 50V/um dielectric strength of molding compound to ref [43]. However, there is no corresponding measurement result in [43].
3. Since the improved receiver robustness is one major claimed contribution of this work, besides theoretical analysis, it is still suggested to have some simulation or measurement data to support this claim.
Author Response
(1a) The authors refer the cross talk to ref [46]. However, it is also not thoroughly discussed in [46]. When signal is transferred through one channel, what are the outputs of the other channels during measurement?
(1b) In ref. [46], it is stated the crosstalk is mitigated with the distance between adjacent antennas and different carrier frequencies. The level of crosstalk between adjacent channel is lower than the hysteresis threshold.
(2a) The authors refer the 50V/um dielectric strength of molding compound to ref [43]. However, there is no corresponding measurement result in [43].
(2b) As I have replied in the previous revision, if you look at the paper of ref. [43] at the end of section 2, you can find two references [20][21], which give exhaustive measurements on the dielectric robustness of different types of molding compounds. The two references are:
Paye J.; Claudi V.; Stecher M. High voltage robustness of mold compounds under different environmental conditions. Proc. of IEEE Int. Reliability Physics Symp., Monterey, CA, April 2015, pp. P.5.1‑CP.5.6.
Paye J.; Claudi V.; and Stecher M. High voltage robustness of mold compounds after different treatments. Proc. of IEEE Electrical Insulation Conf. (EIC), Montreal, QC, August 2016, pp. 162-165.
(3a) Since the improved receiver robustness is one major claimed contribution of this work, besides theoretical analysis, it is still suggested to have some simulation or measurement data to support this claim.
(3b) The limitations of the previous receiver [46] were discussed in Section 3 with many details. In a GaN technology, the mismatch between transistors and the variation of the transistor threshold voltage are much higher than the CMOS counterpart. Therefore, it is easy by inspection to see the drawbacks of the previous receiver in worst case conditions and how the proposed new solution overcomes them.
