MIMO control comes to Panther.
Multi-exciter random and sine control is coming to the Panther platform, built on the control law proven in Jaguar, extended with degree-of-freedom control and an AI setup agent that turns rig geometry into a working test configuration.
| DOF | Panther vs measured field spectra |
|---|---|
| X | within 0.1 dB |
| Y | within 0.1 dB |
| Yaw | within 0.1 dB |
MATCH · Jaguar's measured field files from a customer installation, decoded, and Panther run against a plant built from the same rig. Two independently constructed transforms gave the same answer.
One shaker is often the wrong answer
For decades the hard problem in vibration testing has been getting one shaker to reproduce a specified environment faithfully. The hard problem now is that real articles don't experience vibration along a single axis, and the standards that qualify them increasingly say so. Multi-exciter testing is no longer the exotic end of the market.
Spectral Dynamics has been controlling multi-exciter tests for years. Jaguar, our high-channel-count multi-axis controller, runs true real-time MIMO control on some of the most demanding rigs in service. What customers have asked for repeatedly is that capability at Panther's scale: the same control mathematics, on the platform their lab already runs, sized for the four-, six- and eight-shaker rigs that make up the bulk of real multi-axis work. That is what Panther MIMO is.
The control law Jaguar made the standard
Panther MIMO is not a second, parallel approach to multi-exciter control. It implements the adaptive matrix control that Jaguar established as the industry standard for real-time multi-exciter testing.
- Matrix power control. The response error is projected into drive space through the pseudo-inverse of the measured plant, H+·E·(H+)H, and applied to the Cholesky factor of the drive cross-spectral matrix, so the commanded drive spectrum is physically realizable by construction rather than by repair.
- An optimal step at every frequency line. Rather than a fixed correction gain, a cubic is solved per line, per iteration, for the step size that minimizes the predicted next error.
- An averaging gain matched to the statistics of the estimate, including the correction for overlapped frames that keeps the effective degrees of freedom honest.
What is new is everything underneath
Earlier generations of multi-exciter control were shaped by the signal-processing hardware available when they were designed. Panther MIMO is not, and it spends the headroom on refinements that matter in the field.
- 64-bit floating point end to end.
- Modern multicore processors with vectorized linear algebra.
- 24-bit acquisition with better than 110 dB of dynamic range and phase matching to within one degree between units.
- Per-line regularization scaled to local conditioning, coherence weighting, and a per-line throttle for frequencies that resist correction.
The control mathematics that made multi-exciter testing dependable, on a platform built three decades later, with an interface designed for the engineers using it now.
Four capabilities that change what a multi-axis test can prove
DOF control, done on both sides
A DOF test controls rigid-body motion of the table rather than individual accelerometers. Panther applies the geometric transform on both the response side and the drive side. Controlling DOF responses against physical drives leaves the plant coupled and silently under-tests the weak axis, a failure mode that looks like success on every plot. The drive-side transform diagonalizes it.
Multiple coherence, selected automatically
On a symmetric rig each drive explains only a fraction of any one response's power, so ordinary coherence reads low, around 0.33 for three drives, even when the measurement is excellent. A controller that throttles its correction gain on that number freezes. Panther detects the condition and asks the right question instead: how much of this response is explained by all the drives together.
Per-line limiting with attribution
In random MIMO the notch is not a number, it is a curve. Every frequency line can be derated independently, every frame, and a different limit channel can hold control at each line. Panther reports the deepest notch, the line it occurred at, how many lines were derated, and which channel governed each one.
Spatial view
Control error, coherence, safety state and channel roles rendered on the actual fixture layout rather than buried in a trace list. An accelerometer sitting on a structural node cannot be controlled, and that is far more obvious as a marker on a plate than as a dip in a spectrum.
An AI agent that sets the test up
MIMO's real barrier has never been the mathematics. It is that configuring a multi-exciter test demands expertise most labs have in one or two people.
The engineer supplies what they already know
- Which accelerometers sit near which shakers.
- Sensor and shaker positions and lines of action.
- Fixture type.
The agent runs a system identification, then
- Assigns channel roles and weights.
- Checks that the actuator layout can resolve the degrees of freedom being asked for.
- Builds the transform, and selects the coherence metric and reference type.
- Calculates loop gain, initial level, ramp step and iteration limits from the measured condition number and controllability, not from stored defaults.
It also catches what a person cannot. A rotational transform row is a matrix of signed coefficients, and a reviewer reads the value they expect to see; the agent checks that row arithmetically from geometry before a single frame is acquired. In one recorded case a single wrong sign meant half the intended yaw sensitivity and a translation leak. Every recommendation carries its reasoning, and a companion predict mode runs the whole analysis without changing anything.
Three translations and three rotations
Panther MIMO controls the full rigid-body set. The engineering lesson matters more than the number: six-DOF capability is a question of where actuators sit, not how many there are. Three vertical shakers on a circle at 120° spacing span vertical, roll and pitch between them, while a single central vertical shaker spans only vertical motion however much force it has.
Panther MIMO checks that arithmetic from your geometry and tells you what your rig can resolve before you build the test.
| Profile | flat 30–1000 Hz, and shaped to 1600 Hz |
|---|---|
| All six DOF | within 0.03 dB of reference |
| Frequency lines | every line converged |
| 12 physical sensors | span 0.17 dB |
The degrees of freedom are not being satisfied at the sensors' expense.
AI automation across design, development and test
Panther MIMO was developed against a full multi-input, multi-output plant simulator: per-path frequency response functions with resonances and anti-resonances, cross-coupling, noise floor and Q spread, validated against measured data from installed hardware and reproducing a customer's own rig to within 0.1 dB.
AI automation is embedded in that process rather than bolted onto it. Architecture and scoping documents are AI-authored against the source and reviewed by engineers, and every design decision is captured in knowledge bases that carry forward between sessions. The point is not speed: a control-law change made on Tuesday is measured against thirty-one plants by Wednesday morning. More on the simulated test system and nightly regression.
Panther MIMO Random and Panther MIMO Sine
Both run on the same channel table, the same transforms, the same spatial view and the same reporting.
Panther MIMO Random
Controls the full cross-spectral density matrix: auto-spectra plus the coherence and phase relationships between control points, with per-line limiting, abort channel protection and degree-of-freedom control.
Panther MIMO Sine
Controls magnitude and inter-channel phase across the sweep, on an adaptive impedance loop that needs no matrix inversion, with broadband identification ahead of the sweep.
Frequently asked
What is Panther MIMO?
Multi-exciter random and sine control on the Panther platform: Panther MIMO Random and Panther MIMO Sine. Both share one channel table, the same transforms, the same spatial view and the same reporting.
How does Panther MIMO relate to Jaguar?
Panther MIMO implements the adaptive matrix control that Jaguar established as the industry standard for real-time multi-exciter testing, on a modern 64-bit platform sized for four-, six- and eight-shaker rigs.
How was it validated?
Jaguar's measured field files from a customer installation were decoded and Panther was run against a plant built from the same rig. On a three-DOF test (X, Y and yaw) Panther's settled control spectra matched the measured field spectra to within 0.1 dB on all three degrees of freedom.
Can Panther MIMO control six degrees of freedom?
Yes: three translations and three rotations. Whether a given rig can resolve them depends on where its actuators sit, and Panther MIMO checks that from your geometry before you build the test.
What does the AI setup agent do?
From the geometry you supply and a system identification it runs, the agent assigns channel roles and weights, builds the transform, selects the coherence metric and reference type, and calculates the test parameters. Every recommendation carries its reasoning.
Bring multi-exciter control to your Panther lab.
Contact your Spectral Dynamics representative for configuration and scheduling on your rig.
