Hello Bob: I don't know if you remember me but I worked at National Semi a while back, both in quality assurance and for the automotive group as an apps engineer. I have a Spice story for you. Recently, while preparing for a job interview, I had my first serious encounter with Spice. I had a first-round phone interview with the hiring manager, and the one minor concern he had was that I had no experience with Spice.
I am a systems guy, and most of my board designs in the past really never required Spice. I usually used a white board, some quick transmission-line calculations, and some good common sense. However, with newer RAM technologies and much higher frequencies being utilized, using Spice to do some prelayout simulation is becoming more commonplace. Therefore, I decided (never having used this beast) to download a student version of PSpice from Cadence just to play with. The one nice thing about Spice is that you can make changes to things on-the-fly to see how certain parameters affect a circuit. With transmission lines, this is kind of hard to do in the lab. Varying line lengths and the parasitics on termination resistors is really hard to do in a lab. (You can always snip a transmission line to be shorter, but it's not so easy to relengthen it! /rap)
I have always wondered what the exact effects are of the distance of your source terminator from a driver and the effects of the termination resistor's parasitics. I had initial ideas, but I kind of wanted to confirm it.
Well, I thought to myself, here is a good thing to try that Spice is perfect for. I suspected that moving my source terminator too far from the driver should set a smaller, but nonetheless distinct, reflected waveform on my signal. The parasitics of the resistor should tend to filter out high-frequency sharp edges and whatnot. This all made sense because that's what the parasitics used to do on the gates of the power MOSFETs that I used to work on at National. My plan was set.
I entered the schematic into OrCAD and fired up the PSpice window, and I got the nice reflected waveform on top of my signal. It kind of even looked like stuff I had seen in the lab with poorly terminated lines. This thing really works! So I immediately switched back to OrCAD, shrunk the length of the line to the terminator down to some very short value, and reran the simulation. To my absolute shock and dismay, the reflected waveform was still there. This made no sense.
So I closed the PSpice program, went back to change all the parasitics to very small numbers, and then reran the simulation. And the reflected waveform was gone. How could this be? It made no sense. I repeated the experiment several times. But each time, the result was the same. (Here, you're talking about a case where you know what the answer ought to be—MUST BE—so you're very skeptical when it gives an absurd answer! It's important to know what will make sense! /rap)
I concluded, therefore, that maybe Spice was right. Maybe there was some kind of resonant circuit being set up by the RLC parasitics that was doing this. So I shut the machine off and pondered the subject over lunch. Finally, when I just could not accept it, I powered the machine back on and reran the simulation. This time, I got the results I expected with the shorter line. After fooling around, I figured out what had been the problem. Apparently OrCAD and PSpice are not really integrated but are two separate programs. For whatever reason, unless you kill the PSpice program, even though the computer resimulates, you always get your previous results. The two programs only seem to talk to each other when PSpice is started. So what happened was that when I only changed line length, I changed just one parameter. So I just switched programs and closed the waveform window. When I changed all the parasitics, which takes a lot longer, I sort of got frustrated and killed the whole PSpice program. You might want to add this one to your list of Spice gotchas. Always close Spice after simulating.
• Michael Rally (via e-mail)
• Pease: Wow, what a mess! I always close Spice—BEFORE simulating!
[email protected] —or:
Mail Stop D2597A, National Semiconductor
P.O. Box 58090, Santa Clara, CA 95052-8090