The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl Modeling >> Transmission Lines and Other Distributed Devices >> Transmission Line Characterization & Simulation https://designers-guide.org/forum/YaBB.pl?num=1190983968 Message started by mkaragou on Sep 28th, 2007, 5:52am |
Title: Transmission Line Characterization & Simulation Post by mkaragou on Sep 28th, 2007, 5:52am I have a non standard cable that I would like to characterize by making measurements with a network analyzer and extracting the s-parameters. Now I am wondering about how I have to include the extracted s-parameters into the spectre simulator to be able to do some transient and ac simulations to see how the cable influences the system. I have noticed that there is an nport component available in spectre which I can use together with the extracted s parameters. But I have also read about the so called Eisenstadt method where the extracted s-parameters are converted into a RLGC parameter set which I could use for example together with spectre's mtline component. Now what is the difference between this two methods? Do I get realsitic results if I just use the nport component together with the extracted s-parameters or do I have to make the conversion into the RLGC parameter set? In addition I am thinking about if it is a problem to characterize an unshielded twisted pair cable with a two-port network analyzer or if it is necessary to make use of a four-port device to characterize the differential behaviour of the cable. |
Title: Re: Transmission Line Characterization & Simulatio Post by sheldon on Feb 19th, 2008, 6:24pm Mkaragou, Since no one has answered this, I will take a shot. The n-port is best approach for modeling cables. The mtline is more appropriate for package and board applications. When extracting parameters for the cable, you will need data points at low frequency to assure a good extrapolation to dc. For the twisted pair using mixed-mode s-parameters is better. Best Regards, Sheldon |
Title: Re: Transmission Line Characterization & Simulatio Post by Tawna Wilsey on Mar 3rd, 2008, 10:38am If you are simulating transmission lines, you should use the mtline from analogLib whenever possible. msline and tline (from analogLib), mline and tline3 (from rfExamples) are older components that are still in the libraries for legacy purposes. They are not nearly as accurate as mtline and should not be used. The nport family (nport, n1port, n2port, etc. from analogLib) can be used to simulate transmission lines provided that you have s-parameter data. However, mtline is more accurate and is the preferred use model. The nport was not designed/optimized to handle transmission lines with significant delay. The only time you use an nport to simulate a transmission line is when you have s-parameter data AND the s-parameter data also includes interconnect elements. (i.e. the s-parameter data models a transmission line that is connected to other components. It is not a pure transmission line.) When using the mtline from analogLib, you will need to decide which use model is best for your transmission line. For example: - Do you have RLGC data? - Do you have s-parameter data? - Do you want to use the tline use-model? - Do you have physical dimensions that you can enter into a 2-D field solver? Using the 2-D field solver and inputting frequency dependent RLGC data are the most accurate. best regards, Tawna |
Title: Re: Transmission Line Characterization & Simulatio Post by Tawna Wilsey on Mar 3rd, 2008, 10:44am One more piece of information that may (or may not!) be useful..... To model coax and cat5 cables, you have a couple of options available. The simplest approach to model a single cat5/coax cable is by specifying a constant characteristic impedance Zo and time delay td parameter in mtline(tline). For shorter cables or lower frequencies this may be a reasonable approximation, but for longer lossy lines, you may want to add skin and dielectric loss, fc/alphac and fd/alphad. This approach overall may have limited accuracy. We suggest that you make comparisons with measurements if possible. Another approach would be the use mtline with r, l, g and c data values along with rskin and gdloss to model the frequency effects. This data may be obtained from the manufacturer or by time-domain and frequency-domain measurements. See "EDA Development for 10 Gigabit Ethernet Measurement Based Hspice Model Library", Mayrand & DiMinico, DesignCon 2003. Coax lines have well known equations to obtain R, L, G and C which can be used with mtline. These expressions take into consideration the geometry and electrical characteristics of the cable. But, for increased accuracy you may need to use a field solver in order to obtain the frequency dependent R(f) L(f) G(f) and C(f) parameters. The above approaches rely on the mtline element. It should be noted that mtline assumes that a quasi-tem approximation is valid. This assumption breaks down for long lines and above some cut-off frequency where higher-order modes begin to propagate. We strongly recommend that you compare mtline results with measurements if possible. An alternative approach would be the use of NPORT with measured S-parameter data. Since the lines are likely to be long, you must provide a small (fmin) sampling frequency in order to ensure adequate impulse response modeling up to tmax, where tmax is the impulse response time including delay, i.e., fmin < (1 / tmax). If the cable is unusually long, then care must be taken to ensure an appropriate value of fmin sampling since the delay may be quite large. Again... for these types of lines, measurement results would be a useful resource to validate any of the approaches described above. Best regards, Tawna |
The Designer's Guide Community Forum » Powered by YaBB 2.2.2! YaBB © 2000-2008. All Rights Reserved. |