Stats

  • Locked Locked
  • Replies 4
  • Subscribers 63
  • Views 13186
  • Members are here 0
This discussion has been locked.
You can no longer post new replies to this discussion. If you have a question you can start a new discussion

S-parameter bonding wire issue

Binhngo1210

Hi everybody, 

I am designing LNA and run into the issue of bonding wire S- parameter. The test bench for input LNA is described as below capture. 

My bonding wire inductor is about L=0.44 nH/ Q= 25 @ 2.5 GHz, to test whether performance is improved if using multiple bonding wires, I connect 10 wires (just for simulation, not realistic bonding case) in parallel (as seen in the captured) 

I hope that its equivalent L_bonding will be L/10 and in fact, with the simulation result ( use differential testbench 2 ports to measure equivalent L bonding) my prediction is correct (figure 1) - about 44p with Q almost similar 

However, it is quite troublesome when I do simulation with that S-parameter file, the Gain is dropped to 0 as well as input matching disappears, but the Gain gets normally recovered (Figure 2) once I replace that L bonding S-param file by an inductor having the same specs( Q=25, L = 44p @ 2.5 G).  It is quite weird and subtle to me even the passivity of the bonding wire S-param file was carefully checked. 

My question: Is that correct if I equivalate those 10 bonding wires by arraying 10 files S-parameters  2 ports in parallel? and How can i deal with it ? 

Very much appreciated receiving your help! 

Thanks

Binh Ngo


(capture 1)

figure 1

Figure 2

  • Dear Binhngo210,

    Binhngo1210 said:
    My question: Is that correct if I equivalate those 10 bonding wires by arraying 10 files S-parameters  2 ports in parallel? and How can i deal with it ? 

    No - you need to create a single S-parameter file with all 10 bonding wires if I understand your question correctly. I have direct experience with such a design using multiple wire bonds (in this case as part of a VCO), and if you simply try to model them as 10 distinct S-parameter files in parallel you are totally overlooking the mutual indutances between the wire bonds. The mutual inductance terms will be quite significant as their magnetic fields will overlap.

    Did I understand your question correctly Binhngo1210?

    Shawn

  • in reply to ShawnLogan
    Unknown said:
    Did I understand your question correctly Binhngo1210?

    Yes it is totally correct. 
    as your recommendation, I ran the EM simulation for 10 wire bonds, the overall L is  about 0.25 nH and Q = 50 @ 2.5 GHz. This value is a way much smaller then LG off chip (25n, Q = 31), however after post-layout simulation the gain also dropped to lower than 0 dB and S11 move to lower matching point (the link below shows its performance, please take a look)..

    there is a weird thing that when i switch the position of L bonding and L off chip (can see as the above capture post), the gain came back . I am quite confused for what  the simulation is going on@@. Have y ever met this bug? 

    https://drive.google.com/file/d/1x4pjbnWAwggzOZlCMHLnjRZuG72Y1Xk0/view?usp=sharing

    bonding performance : 

    https://drive.google.com/file/d/1kHDoicHdJsedspkNMmvbXeR_GPXVbkaq/view?usp=sharing

    log file: 

    https://drive.google.com/file/d/10MxdScg91t-fkgcYkSYYalRkw-Qht6Xo/view?usp=sharing

    Thanks for your time.

    Binh Ngo

  • in reply to Binhngo1210

    Dear Binhngo1210,

    Binhngo1210 said:

    as your recommendation, I ran the EM simulation for 10 wire bonds, the overall L is  about 0.25 nH and Q = 50 @ 2.5 GHz. This value is a way much smaller then LG off chip (25n, Q = 31), however after post-layout simulation the gain also dropped to lower than 0 dB and S11 move to lower matching point (the link below shows its performance, please take a look)..

    there is a weird thing that when i switch the position of L bonding and L off chip (can see as the above capture post), the gain came back . I am quite confused for what  the simulation is going on@@. Have y ever met this bug? 

    To be very honest, I am not clear at all about what you are indicating  with respect to "LG", "off-chip", and "gain" as  I do not understand the test bench diagram in your initial post. It looks as if you have several S-parameter file based ports connected in parallel - maybe only one is used in each simulation? Further, I don't see the amplifier nor any matching network and have no idea where you are measuring the gain.

    Your output log file, was useful and there are two specific comments I can make that will impact your findings...

    1. You are using a very old version of the Cadence tools (Cadence (R) Virtuoso (R) Spectre (R) Circuit Simulator Version 14.1.0.848.isr14 32bit -- 6 Oct 2015) and there have been many, many improvements as to how it handles S-parameter files. I highly recommend you consider getting access to a much more recent version! This alone, may explain some of your simulation results that you consider "weird".

    2. More importantly, however, your log file clearly indicates your S-parameter file "HHM1902B1.s3p" does NOT have a DC term!

    The specific log file contents are:

    Warning from spectre during DC solution estimation, during DC analysis, during SP analysis `sp'.

        WARNING (CMI-2134): Risky extrapolation to DC of data given in S-parameter file `/home1/mpw7911/03_sp_files/balun/HHM1902B1.s3p'.

        WARNING (CMI-2134): Risky extrapolation to DC of data given in S-parameter file `/home1/mpw7911/03_sp_files/balun/HHM1902B1.s3p'.

    I may have overlooked it, but perhaps some of your other S-parameter files are also missing DC terms. I recommend several items:

    a. View your S-parameter files in ViVA using a more recent version of the Cadence tools. ViVA can read an S-parameter file directly and plot it in rectandular or polar coordinates. Verify that a DC term exists and the S-parameters are consistent with what you expect for each port. Please view the methodology to view S-parameter files in ViVA directly at URL:

    support.cadence.com/.../ArticleAttachmentPortal

    b. If not, re-generate the S-parameter files of interest and verify that sufficient frequency resolution exists, that there is a DC term, and the resulting plots of its S-parameters are consistent with what you expect.

    c. Study the following document from the Cadence On-line support site entitled "Seven Habits of Highly Successful S-parameters". This excellent document by Mrs. Tawna Wilsey of Cadence outlines a number of potential S-parameter file issues as well as a number of troubleshooting tips when using an S-parameter file in a simulation. You should get familiar with this document if you are planning to run many S-parameter analyses. The document is at URL:

    https://support.cadence.com/apex/ArticleAttachmentPortal?id=a1O3w000009bgt4EAA&pageName=ArticleContent

    or in Webinar format:

    https://support.cadence.com/apex/ArticleAttachmentPortal?id=a1O3w000009y0apEAA&pageName=ArticleContent

    Shawn

  • in reply to ShawnLogan

    Dear  Shawn Logan, thanks for your very helpful suggestion. However, even resolve all problems regarding tools as well as obtain completeness of wire bonding-S param files. I cannot get the Gain of LNA performance as my expectation. I try all ways but did not know whether such fixings are enough.

    First of all, I replaced mmsim141 with mmsim161 and that is the highest version that I can upgrade (because this tool is managed by another center in my university, and they installed 2 versions on the server). 

    then, I replaced the balun S param files "HHM1902B1.s3p" as your mention in No.2  with an available ideal balun component: "balun_com"  in "rfLib" to avoid the DC issue.

    second, for bonding wire S param files, I regenerated the S param file such that it includes DC frequency point. 

    finally, to avoid a similar problem of passivity or DC point  with L, C S param files, I replaced it with components from analog Lib).. --> finally, it really did not work out 

    However, when I switch positions of L_bonding and L_G (L_bonding is on the left of LG) as illustrated in testbench (violet arrow) and run again the circuit with post-layout netlist of 03_LNA_RX_TB --> the gain is recovered (yellow line in figure 1) and the violet line in figure 1 represents for the Gain that L_bonding is on the right of LG.
    I am aware that the result is unreasonable because when we reduce the effect of bonding the gain should be as close as to the post-layout simulation performance and the L_bonding should be a part of L_G because its inductance is very tiny and Q is high enough in this simulation. 

    I will include the test bench below. Could you hint me another check? 

     test-bench 

    Thanks for your time, 
    Binh Ngo 

Community Guidelines

The Cadence Design Communities support Cadence users and technologists interacting to exchange ideas, news, technical information, and best practices to solve problems and get the most from Cadence technology. The community is open to everyone, and to provide the most value, we require participants to follow our Community Guidelines that facilitate a quality exchange of ideas and information. By accessing, contributing, using or downloading any materials from the site, you agree to be bound by the full Community Guidelines.