Roadmap
MBE3Dstudio is early. The most useful thing this page can do is be specific about where the edges are, so you can judge whether it is worth your afternoon.
All of this is in the build you can open right now, and is covered by the test suite.
These are real constraints in the current build, not things we are being coy about. Some are deliberate; some are just not done yet.
There are three primitives. Complex or swept shapes have to be approximated by several parts. This is deliberate at this stage (envelopes are meant to be cheap enough to build before the CAD exists), but it does mean an awkwardly-shaped bracket is modelled coarsely.
Separation is computed by the separating-axis theorem over oriented boxes. For overlapping parts the penetration depth is exact; for disjoint parts the reported gap is a lower bound on the true distance. It can therefore report a clearance violation that is not quite real, but it will not report a pass that is not real, which is the direction a clearance check should err. A cylinder is bounded by its box for this test, so two nearly-touching cylinders read tighter than they are.
It reports dissipation, volumetric heat density and which hot parts are close to each other. It does not model conduction paths, airflow or steady-state temperatures. It is meant to tell you what deserves a real thermal analysis, not to replace one.
A STEP file imports: it is parsed, and the part and its geometry arrive as entities held for review rather than merged silently. There is no mesh import, and no geometry export at all beyond the viewport PNG, so this is a way into the model and not a round trip. Envelopes are still box, cylinder and sphere primitives, so what comes back out of MBE3Dstudio is the model, never your CAD.
Importing a specification pulls out the requirements it can identify, and on a long prose document that is a fraction of them. A structured requirements document of ours, all EARS patterns and tagged identifiers, yielded twelve requirements out of sixty-six, and the import reported no errors doing it. A clean result is not the same as a complete one, so treat the import as a first pass that saves you the bulk of the typing, then read what came through against the source.
A model saves to the server as well as to a file, so it is not trapped on one machine. What there is not is collaboration: no shared editing session, no locking and no merge. Two people working on the same model coordinate the way they would over any other file.
Eight kinds cover mass, power, clearance, CG, harness length, voltage drop, sensor coverage and heat density. There is no user-defined constraint expression, so a limit outside that list cannot currently be made executable.
Ordered by how often they have come up, not by commitment. Nothing here has a date, and some of it will not be built. What gets picked up depends heavily on what the people using it actually hit.
Replacing the lower-bound distance with a true closest-feature distance, so a tight-but-legal gap stops reading as a violation.
Capsules and extruded profiles would cover most of what currently has to be faked with several boxes.
An expression over the quantities the analysis already computes, so a programme-specific limit can be executable rather than remembered.
Two revisions side by side, with the deltas in mass, power, coverage and findings: the question actually asked at every review.
Importing envelopes from existing CAD, and handing envelopes forward once the design is worth detailing.
