Circuit Simulation packages?

[Shocker]

Hi,

Are there any differences in circuit simulation packages and what they're used for?


I have a circuit, where the signal has step raise/fall times of approximately 1ns. I want to see how the signal propagates through the circuit and i also want to see how parasitics of the circuit affect signal integrity, as i want the steps in voltage to settle within a nanosecond or two. With a rise time of about 1ns, i'm looking at harmonics at up to about 300-400MHz.

The circuit takes a digital input, amplifies it, where it turns it in to an analog signal and continues to be an analog signal until it reaches the output.

I've tried simulated traces as lossy transmission lines, and components with their parasitic elements in Multisim but it doesn't seem to like it because after so many picoseconds the simulation halts and it tells me that the step size is too small. I only have the step sizes at a moderate 0.1ns, which is only 10 times greater than my signal. I can't increase my step size because then i will completely loose the detail that i'm interested in.

The circuit simulated before i turned the traces in to transmission lines. I thought about simulating the transmission lines as RLC circuits, but this will give inaccurate results.

I've thought about going to other simulation packages such as LTSpice, Hyperlynx, ADS but i don't know if they'll be advantageous? Or whether i'll be just wasting my time and should stick with Multisim?

My PCB is designed using Mentor Graphics, and so Hyperlynx is a potential option because it comes as part of the package. I'm also led to believe it has the ability to simulate each trace with how it's laid out on the PCB. Not sure if this is correct or not.


I've tried doing a bit of Googling but i haven't found anything that gives me the type of answers that i've looking for.

Thanks

[TimFox]

Have you tried artificial time dilation, where you make all transmission-line dimensions larger so that longer time steps will work?

[Marco]

IME simetrix is most forgiving convergence wise. With highly complex components you might quickly run into the node limit of the free version though.

You'd need Hyperlynx Analog to use it as a simulator and if you'd wanted to use it with Hyperlynx parasitic extractions you'd need Eldo as well (seems a rather artificial limitation) :

Collaborates with Simulate Signal Integrity to extract PCB trace parasitic models from the physical layout and incorporate them into the functional simulation when using either the Eldo or Questa ADMS simulation kernel upgrades

[John_ITIC]

This is exactly what Hyperlynx was designed for. Since your PCB is designed with a Mentor EDA software you should be able to directly export a HYP file that Hyperlynx directly can load. You then assign models for all the drivers/inputs of the net you are interested in simulating. Then you can choose to check the step response or eye diagram (depending on what you want to analyze). For higher bandwidth simulations, you may want to look into IBIS models, which are intended to give a much faster result than normal SPICE-type simulators. Parasitics can also be simulated. Hyperlynx is however very pricey so if you don't have the budget, just buy a couple of books to learn how to calculate the reflections. Granted, you have to put in a lot of work when analyzing a net manually via calculations so any "real" project should use a simulator.

Given the 1 ns rise-time, it seems your signal does not have high clock frequency? I strongly suspect any reflections will have died out by the time the next clock edge comes around? It is impossible to give accurate advise without knowledge of the specific trace topology.