SystemC-AMS vs RNM

[stefaniecg]

I have to model analog parts of my IC.

I've heard the fastest simulation time is if I model them using the RNM approach.

My question is how does SystemC-AMS compares to RNM regarding simulation performance and simulation accuracy?

[karsten]

RNM uses the digital event driven model of computation (MoC) to allow abstract analog signalflow based behaviour. For such an purpose event driven simulation is not optimal - the dataflow MoC available additional in SystemC AMS is therefore more efficient and introduces less artefacts like unneccesarry activations.

Additional, modelling of dynamic behavior (transfer function, filter, ...) with RNM is difficult. You can transform your equations into a digital filter (e.g. using bi-linear transformation) or you have to use Verilog-A(MS) and thus the analog simulator which will slow down the simulation performance.

SystemC AMS provides dedicated language construct for describing linear differential equations - the ltf_nd,ltf_zp and ss objects.

Additional you will have the conservative ELN domain, which allows to describe electrical linear networks (via equivalent relation other physical domains can be described also). The resulting equation system will be solved by a dedicated light weight solver - so the simulation performance is usually orders faster than using a traditional analog simulator.

For more information you can also read the user's guide http://www.accellera.org/images/downloads/standards/systemc/OSCI_SystemC_AMS_extensions_1v0_Standard.zip

[Elvis Shera]

When it comes to modelling, you open the door to so many things. There are more options nowadays and It really depends on more things like:

• what effects/features you want your model to show?
• Is it a model for verification?
• is it for IP design?
• Is it a model you want to deliver to your customer? 

I don't believe there is one tool for everything. As usual, you need the right tool for the right task. Finally, it depends also on one modelling skills. While is common to talk about design and verification skills. It is also common to forget that modelling requires special skill-sets and understanding. 

RNM has been for me enough good on many occasions. However, depending on what you want to model it might not be anymore the right tool out there. Can you give an example of an IP you are trying to model and what kind of effects you want to model?

[Martin Barnasconi]

RNM is a simple approach to represent analog signals by a real-value (amplitude) on a  (discrete) event driven time axis. Depending on the type of signal, you need to generate a lot of samples (events) to follow the shape of the waveform (i.e. Nyquist rule). The event scheduling in a digital solver results in some simulation overhead, because the event list is dynamically scheduled and executed. The more events, the slower the simulation. Especially when you start modeling RF systems in RNM, your system simulation will get slow. In addition, the more input and outputs, the more events at these inputs and outputs, which need to be added to the sensitivity-list of the discrete-event solver. So the bigger the RNM system, the slow its gets.

Thanks to the Dataflow based simulation concept in SystemC-AMS, we do not have these issues. In SystemC-AMS the dataflow graph is based on the interconnected TDF modules, and computed before simulation starts. For each time step, this graph is executed only once, including signal input/output updates for all associated TDF modules. As such, the size of the TDF system does not matter much. Especially for bigger systems you will clearly see a difference between RNM in Verilog-AMS or SystemVerilog versus TDF modeling in Systemc-AMS, the latter being much faster.