Most engineers who design a magnetic position sensor already have a simulation tool available. It is usually a general-purpose finite element package: Ansys Maxwell, Altair Flux, COMSOL or something similar. Those tools are excellent, and they can solve almost any electromagnetic problem you can describe to them. That generality is exactly the problem.
A general-purpose solver does not know what a diametrically magnetized ring is, what a sensing IC reads, what an angle error curve looks like, or why the airgap matters more than almost anything else. You have to explain all of it, every time, in the language of the solver: geometry, materials, mesh, boundary conditions, post-processing scripts. The physics is handled for you. The engineering is not.
The second problem is time. A single accurate run of a three-dimensional magnetic assembly is not something you launch casually: you set it up, you start it and you come back later. That cost is bearable once, but sensor design is not one simulation. It is the magnet diameter against the airgap, then the same sweep with a different grade of material, then the whole thing again with the chip displaced by half a millimetre. When each variant costs hours, you stop exploring and start guessing, and you find out whether the guess was right on the bench.
MAGMagic starts from the opposite end. It is a cloud platform built for one job, the design and validation of magnetic position and current sensors, and everything in it is shaped around that job: it knows what you are designing, and it is fast enough that you can actually explore.
The most important difference is not visible in the interface, but it explains most of what follows.
MAGMagic does not use the finite element method. It is built on integral formulations, developed in house, and the practical consequence is simple: there is no mesh in the air. Only the sources are discretized, that is the magnets and the conducting parts. The air, which in a finite element model is the overwhelming majority of the elements, is not meshed at all.
This is what makes the platform fast. Simulations run in seconds rather than in hours, and how large the gain is depends on the problem. But the point of that speed is not the single run: nobody buys a tool to wait less for one answer. The point is that a design loop of thirty variants becomes something you do in an afternoon instead of something you plan for the following week, and that changes what you are willing to try. Exploring the design space stops being a project and becomes the normal way of working.
In a finite element model you have to enclose your sensor in a box, decide how large the box should be and impose a condition on its surface. If the box is too small the result is wrong, if it is too large the model is slow, and in either case the choice has to be justified. With an integral formulation the field decays naturally to infinity, because that is how the underlying kernel behaves. There is nothing to size and nothing to justify.
An infinitely long conductor is not an approximation in MAGMagic, it is a modelling primitive. This matters for current sensing and for stray field studies, where the wire carrying the disturbance is far longer than any box you would be willing to mesh.
Conducting regions are accounted for integrally rather than discretized. This helps twice: it removes the elements that would have been needed to represent them, and it removes the discretization error associated with them. Accuracy and speed improve together, which is unusual.
When you open a new simulation in MAGMagic you do not start from an empty domain. You start from a template of the sensing architecture you are working on: end-of-shaft, side-shaft, off-axis, linear, current sensing.
The magnets in those templates are real magnets, not idealized ones. Diametral, axial, multipole and more complex magnetization patterns are available directly, with the material library and the temperature coefficients attached to them. Shields, back-iron and shafts are components you place, not shapes you have to describe from scratch.
This is not only a matter of convenience. A large share of the mistakes in magnetic sensor simulation happens while translating a physical assembly into a solver model: a magnetization direction reversed, a symmetry assumed that does not hold, a material property left at its default. A template that speaks the language of the application removes an entire class of errors before they occur.
Angle reconstruction, deviation from the reference line and the resulting error over the mechanical sweep are produced automatically from the run. No export, no external post-processing, no spreadsheet to maintain.
Every sensing IC works properly only within a range of field amplitude. MAGMagic compares the computed field against that window, including the effect of temperature through the coefficients of both the magnet and the IC. A design that is fine at 25 degrees and out of range at 105 is caught in the browser, not on the bench.
The position and orientation of the sensing IC are not fixed inputs but variables you can sweep: translations and rotations, alone or combined. What comes back is not a single number, it is the behaviour of the sensor across the tolerance window of your assembly, which is what your production line will actually deliver.
If you need the field rather than the chip reading, it is there: free sampling anywhere in the domain, over rectangular or polar regions, in one, two or three dimensions, as vectors or as a colour map. You are simply no longer forced to start from the field and work your way up.
Here is the difference that engineers notice first.
A general-purpose solver gives you the magnetic field. What you want is the angle your sensor will report, and how wrong it will be. Getting from one to the other means exporting data, reconstructing the angle in Excel or Python, subtracting a reference line and repeating that for every variant. Everyone who has designed a magnetic sensor has that spreadsheet somewhere.
In MAGMagic you select the sensing IC and the platform closes that loop for you. You get the sensor output and the resulting angle error over the full revolution, computed automatically from the field at the sensing element. Not the field at a point, but the answer to the question you actually asked.
MAGMagic is entirely proprietary. The solver, the formulations, the templates and the platform were all developed at EMC Gems, on the back of more than twenty years of research in computational electromagnetics.
That has a consequence that matters more than it may seem when choosing a tool. When a customer of a general-purpose package asks for a feature specific to magnetic position sensors, that request enters a roadmap alongside thousands of others coming from motor designers, transformer designers and antenna engineers. It is nobody's fault, it is what a general-purpose product has to do.
Where it runsWe work in one domain only, so a request from a sensor designer is a request about the only thing we do. Several of the capabilities described above exist because a customer asked for them. If there is an operation you repeat by hand every time you design a sensor, that is precisely what we want to hear about, because automating it is our roadmap rather than a distraction from it.
MAGMagic is cloud-native. It runs in an ordinary web browser, on any machine, with nothing to install, nothing to maintain and no local licence server. Simulations run on our infrastructure, not on your workstation, so the time to a result does not depend on the hardware under your desk. Results are accessible to the team from anywhere, and the platform is verified with independent penetration tests to protect IP and data.
© 2026 EMC Gems |
P.IVA/VAT IT03026280309 |
REA UD – 359523 |
Codice SDI USAL8PV |
Cap. Soc. € 10.000,00 i.v.