dSpace vs. National Instruments

[Tim Erickson]

Hi All,

I have recently been exposed to a project that used a dSpace controller, Simulink, ControlDesk, and some other dSpace toolkits. Do any of you have any experience using these products as well as LabVIEW and NI hardware? What are the benefits/drawbacks of each?

Just trying to educate myself!

Thanks!

[Chip Grandits]

Tim,

I have experience with a complex engineering project involving the integration of work from two teams, one using dSpace & Simulink, the other using NI PXI with the RTOS and LabIVEW. Unfortunately, due to proprietary concerns, I cannot discuss the details of my experience.

Our team used N.I. technology. It seems that both the NI hardware and software cost less. Simulink was bought out by The MathWorks, the folks who brought us MATLAB, some time ago and is often integrated with MATLAB code. Simulink seems a bit of a higher level language that basically is a "solver" using extremely well tested code. Those with a control systems background and used to defining a simulation by setting up algebraic loops will probably find themselves more at home with Simulink. There is much more heritage for Simulink in the world of Aerospace.

LabVIEW of course, is endeavoring to become a higher level language. At the time (early 2006) we did not use express VIs, which basically presents LabVIEW at a higher conceptual level. LabVIEW does have a simulation toolkit that allows for algebraic loops (or data flow loops to use the language of LabVIEW). However in 2006 it was clear to us that LabVIEW simulation was not ready for prime time in Aerospace and not ready to take Simulink head-on.

Conversely we found NIs integration between sofware and hardware, specifically targeting simulation to real time targets, to be superior. Althought Simulink supports something called the Real Time Workshop to create real time code; we found using it quite painful in 2006. Both NI and The Mathworks are going after each other's audience and making strides in improving their respective weaknesses. Also we did not use true algebraic loops in LabVIEW, but implemented our simulation in pure data-flow paradigm LabVIEW (i.e. for control systems engineers: causal system in z-domain). For me this actually makes more sense when it comes time to separate out the controller from the plant and target the controller to an embedded processor - which obviously has to exist in a causal world. I.e., the pure data flow LabVIEW "translates" to real time code more naturally than a simulation implemented as an algebraic loop. However, we did find implementing a multi-rate simulation somewhat easier with Simulink.

One other NI advantage is that once we developed the SIM we could create an executable that runs requiring only the freely available run time engine. To the best of my knowledge, one always needs a Simulink development seat to run a Simulink simulation.

Again my experience is from early 2006, so I would defer to anyone who has more recent experience.