A Unified and Open LTSPICE Memristor Model Library

Round 1

Reviewer 1 Report

1. In this paper, a unified and open LTSPICE memristor library is described. The memristor models included in the considered library are related to the main applied materials- titanium dioxide, hafnium dioxide and tantalum oxide. The proposed LTSpice library contains the basic and frequently used existing standard and modified memristor models. The main contributions are the modification of several window functions. The modified models can simplify the existing models and improve the operation speed by increasing nonlinearity of window function. The research results of this paper are very useful for memristor circuit simulation.  
2. The proposed LTSPICE memristor library would be used in memristor circuit simulation. So the author needs to verify the accuracy of the library. It is suggested to find several memristor devices with different structure, test the performance, compare with simulation results and add the research results to the article.

Author Response

Dear Reviewer, thank you very much for your labour to review my paper proposal, for your valuable comments and useful remarks! According to your valuable notes for improving the paper's quality, new analyses and simulations were conducted and the simulation times and accuracy of the considered memristor models were addressed. The performance of the considered models is also analyzed and discussed. An example for the parameters estimation procedure is also added in the paper proposal. The English grammar is checked and corrected by a professional.

With best regards, Prof. DSc. Valeri Mladenov

Reviewer 2 Report

This manuscript focuses primarily on summarizing various mathematical models of memristors.  The idea is that these mathematical models can be used in SPICE simulation (particularly LTspice).  While the title and abstract focus on the LTspice models, the manuscript primarily focuses on the mathematical descriptions.  I would suggest expanding the content related to the actual implementation of the LTspice models.  I have several comments on this below. 

Commenting as someone who might be interested in using the LTspice models, I was left with many questions regarding how the mathematical models are actually implemented in LTspice.  One thing that could be done to help would be to provide at least one example of the implementation in LTspice.  While Figure 5 shows a simple subcircuit that is being used for the models, it is unclear how you are implementing the controlled sources.  What was the “code” used for implementing these (i.e. what is the LTspice syntax used for the behavioral model).  This information would be highly educational and help the user understand how the model is working before having to download the models from github first. 

Along these same lines, it was unclear how you are implementing the windowing function within the LTspice models.  Could you further elaborate on how you are translating the windowing functions to the LTspice models?  Also, how are you actually using these windowing functions?  An example would really help.  Of note, the Conclusions state that a major contribution of this work is the windowing function, and it really isn’t clear how they are working and how they are being implementing within the simulator. 

It is unclear how well these mathematical models are representing the real physical devices.  A comparison with real data from the devices would provide much more credibility to these models.  If real measured data are not available, is there a way to compare the mathematical/LTspice models to the actual physical devices?  What you have presented only shows that the models are implementing various equations, but it is not clear how well they match the actual workings of physical memristor devices. 

It is unclear if Figures 2-4 are from LTspice simulations, numerical analysis (such as Matlab), etc.  What level of abstraction are these models?  Could you please clarify? 

Are the examples of Section 5 included in the downloadable library?  If not, it would be very helpful for an end-user to have existing examples to run immediately to gain an understanding of how to use the models.  It was not clear from the manuscript if these example netlists are included in the library.   

Section 5.1 states that it is a simulation for a memristor crossbar application, but it seems that the results (Fig. 8) are primarily just a setup to generate a current-voltage relationship.  I’m missing how this is a crossbar application, or the related text is not particularly clear here. 

There are also a number of other items related to grammar, formatting, clarity, etc. that need to be fixed. 

Virtually all of the figures are hard to read.  Many are blurry.  Several are far too small.  Even after zooming in very far on the PDF, I had a hard time reading/deciphering them.  On a and on a printed-out copy, I could not read many things.  Please make sure that all Figures are clear and appropriately sized. 

Axes labels and trace labels are quite hard to read on several of the plots.  For example, I am unable to read the trace labels for Figures 2-4.  (Actually, I did not initially see that there were any labels since they were so small).   

In multiple places, you stated that something would be described in the next paragraph (“discussed in the next paragraph”).  However, the next paragraph had completely different subject matter.  Please rephrase and provide more clarity on where the appropriate information is located. 

At the end of the various subsections in Section 3, you include a summary figure (Figs. 2-4) for the related models.  I would strongly suggest separating these summary figures into their own subsection for clarity of presentation.  They are currently part of a subsection on a specific model – sometimes without even a paragraph break to set them apart.  These should be somewhat separated so that the reader understands that these are summaries, rather than part of a specific model. 

This manuscript needs another significant round of grammar checking.  Many parts of the manuscript were difficult to decipher because of issues with grammar.  Please use commas at appropriate locations.  The lack of appropriate comma usage made this manuscript harder to read than it should have been.  There are also a number of other typos that need to be fixed, scattered throughout the manuscript. 

A minor comment.  I would suggest looking into when to use the word “can” versus “could.”  Most of your usages of the word “could” in this manuscript should actually be “can.”   

The URL for LTspice is an old URL (line 49), with an auto redirect.  LTspice is now owned by Analog Devices, and I would expect they would appreciate using the new URL. 

Author Response

Dear Reviewer, thank you very much for your efforts and labour to review my paper proposal, thank you for your valuable remarks and useful comments for improving the paper's quality! All the changes according to your useful remarks and notes are marked by a red colour font.

1. According to your useful notes, I expanded the content related to the actual description of the LTSPICE incorporation of the considered memristor models and I added a detailed description of the basic equivalent schematic and a simple example of a LTSPICE netlist with detailed comments and explanations.
2. All your valuable comments and questions about the implementation of the memristor models in LTSPICE were addressed and a detailed description of their realization is included in the revised paper proposal.
3. An example for generation and implementation of memristor model K1 (Strukov-Williams) is presented with detailed information. The basic equivalent schematic for the models is described in details. All the specific details are also commented.
4. A detailed description for the dependent current sources denoted by G and their functioning is included. 
5. An example presenting a simple code of a memristor model in LTSPICE is included and explained in details, hoping that this will be a useful educational information for the interested readers.
6. The incorporation of the applied window functions responsible for limitation of the memristor state variable and representation the boundary effects for hard-switching mode in the code related to the state differential equations is also explained in details. A simple example for the Strukov-Williams model is presented for additional information. In several modified models the incorporated window functions are dependent on the applied memristor voltage which partially represents the nonlinearity of the memristor models.
7. Additional simulation for the models parameters estimation according to experimental current-voltage relationships is conducted and the derived results are discussed. The cost function used for this procedure is the root mean square error and its minimization is the purpose of this simulation.
8. The quality of all the figures is improved and the author tried all the paper's elements to be easily visible.
9. The models realized in LTSPICE were analyzed for different amplitudes and frequencies and several current-voltage and state-flux relationships are included in the paper proposal for expressing the basic fingerprints and properties of the considered memristors.
10. An example for simulation of a memristor memory crossbar is included in the considered LTSPICE memristor library for the interested readers to be able to make analysis and compare the models.
11. Additional analyses were conducted for a memristor memory crossbar and a new detailed schematic is included in the paper proposal.
12. According to your useful advice the English grammar of the revised paper is checked by a professional and the corrections are updated.
13. Having in mind your helpful advice, the link for the actual LTSPICE website and software is updated.
14. The author tried to address all your valuable remarks for improving the paper proposal and the changes are made by a red-coloured font. Several figures are corrected and several derived by new simulations are added.

     Thank you again for your remarks and valuable advices!

      With best regards, Prof. DSc. Valeri Mladenov