Analog Angle Blog

Simulating Mixed-Signal ICs: Good or Bad News?

Terms such as simulator and simulation can have many different meanings, depending on what you are doing and the problem you have. A recent small item in NASA Tech Briefs made this very clear. The item, Victim Simulator for Victim Detection Radar, described the development of human body simulators which could be used to test “victim detection radars,” which locate people trapped under rubble. This is not a trivial exercise, and until the development of this simulator, the tests required human subjects who were willing to lie for hours in uncomfortable positions in rubble-like conditions — not an easy task. The new simulators have the dielectric properties of humans, motions corresponding to heartbeats and breathing, and a skin-like enclosure — pretty impressive.

Using simulators and simulations are standard steps in the engineer's toolkit and design-validation process. For IC designers, simulations increase the likelihood that a new chip design will work as intended, using a specific set of manufacturing processes.

Of course, IC users don't care about the simulation used by the designers of an IC, but they do care about the models of the IC which the vendor provides. For building-block IC such as op amps, these models use Spice and other simulations of the overall circuit design.

Large mixed-signal ICs have similar layers of complexity beneath their skin. They contain lots of subcircuits and functions, and have many internal signals in the circuit path which are invisible to the user. The simulation and models provided by the vendor to users must hide these. After all, there is no way to make them visible. As a larger-scale black box IC, all users can see in these devices are its pins and I/O — everything else is inside the black box.

But at the same time, these internal signals may be of interest to the circuit designer, because they reveal what the complex IC is doing to the analog signal in terms of filtering, rise and fall times, distortion, and other intentional (or unintentional) processing. Not being able to see them means users have to have hope and faith, and maybe some luck, too.

In contrast, simulations of circuits using the simpler building-block ICs don't have this restriction, since designers can see many of the points along the circuit path if they want to. While the virtue of larger mixed-signal ICs is that they give you more of what you need in a single package — and that's usually a good thing for many reasons, including design simplicity — it can also be a limitation.

Regardless of whether you're using smaller or larger mixed-signal devices, though, there's one factor which every user of simulations and simulators has to keep in mind: they are just approximations, with many explicit and implicit assumptions. That's their nature, and when you forget it, you can find yourself in big trouble.

Has a simulation or simulator ever been a source of headache for you? And have you found that simulations and models of larger mixed-signal ICs are easier to use and better than those for basic building-block ICs — or was it the other way around? Please tell us about your experience in the comments section.

Related posts:

5 comments on “Simulating Mixed-Signal ICs: Good or Bad News?

  1. eafpres
    September 24, 2013

    @Bill–most of my simulation exposure is in the purely analog domain of RF interconnect and antennas.  It is interesting that in the world of passive and active components, it is expected to provide SPICE models and other tools so circuit designers can include them in a higher level simulation.  On the other hand, for RF connectors this is not the case.  In general, RF connectors are assumed to be “perfect”, but they frequently are not.

  2. PeteG101
    September 25, 2013

    Like eafpres, most of my difficulties with using simulators have come from things neglected. This could be interconnect parasitics. Nothing is a perfect, lossless conductor – even in the low MHz region. Simple lumped parameter representations can be all that's needed to illuminate the behaviour you're interested in. Also, exported noise from other system elements sometimes gets forgotten. As has been said many times, 'it's only a model' should never be forgotten either!

  3. Brad_Albing
    September 30, 2013

    @eafpres – I did often wonder about connectors – like whether there were SPICE models of them. So, I guess not.

  4. SunitaT
    September 30, 2013

    though, there's one factor which every user of simulations and simulators has to keep in mind: they are just approximations, with many explicit and implicit assumptions.

    @Bill, thanks for the post. I totally agree with you that the models are just an approximation with many explicit and implici assumptions but these models have improved a lot over time. The existing models closely mimic the exact behaviour of the original models. Getting improvement over the existing models would be very difficult.

  5. SunitaT
    October 29, 2013

    Synopsys lectures the most serious issues in AMS confirmation. HSPICE is hired by leading semiconductor foundries to deliver designers with the golden standard for correctness in device modeling.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.