Olydyne › Guides › Current FFT

Understanding

Motor current FFT: what the spectrum says about a turntable

A motor fed with a perfect sine wave should draw a sinusoidal current at the same frequency. It never quite does. The current keeps a trace of what goes on inside the motor and in the mechanics it drives, and the FFT is the tool that lets you read it. Isostase uses it under one constraint: there is no accelerometer on the turntable. Everything it knows comes from the current.

What the current tells you

The voltage sent to the motors is a sine wave built by calculation and referenced to a quartz clock: 50 Hz at 33 rpm. If the motor were perfect, its current would also be a 50 Hz sine wave. In practice, two kinds of imperfection are added to it.

The first are harmonics: components at 100 Hz, 150 Hz, multiples of 50. They reflect a distorted sine wave, caused for instance by magnetic saturation of the iron or by an asymmetry between the windings. The second are sidebands: small lines on either side of the 50 Hz. A mechanical disturbance that repeats once per rotor revolution, play in a bearing, an unbalance, modulates the current and makes these lines appear a few hertz away from the fundamental. Industry calls this motor current signature analysis.

An FFT, in practice

An FFT (fast Fourier transform) takes a signal that varies over time and breaks it down into frequencies: how much at 50 Hz, how much at 100 Hz, how much elsewhere. Isostase records 4096 samples of the current at 1000 per second, a little over four seconds of signal. That gives a resolution of about 0.24 Hz: two lines half a hertz apart remain distinct. The highest readable frequency is 500 Hz, half the sampling rate, which leaves plenty of room for the fundamental and its first harmonics.

The signal is first weighted by a Hann window. Without it, the 50 Hz peak would smear across its neighbours and drown the very area where the sidebands are sought.

What comes out of it

Three things.

• The fundamental. The peak is searched for close to the frequency actually commanded, not the nominal one: a pitch adjustment by the user moves it by a few hertz.
• THD. The energy of the 2nd and 3rd harmonics compared with that of the fundamental, as a percentage. The starting threshold is 12% at 33 rpm. It was chosen a priori: the board is not assembled yet, so no real measurement was used to set it.
• Sidebands. The energy of all the lines between 1 and 30 Hz on either side of the fundamental, relative to it. No precise fault frequency is sought: the geometry of the motor's bearing is internal, hence unknown, and the theoretical frequencies of its faults cannot be calculated. What is measured instead is the general spread of energy that a deteriorating bearing causes, whatever its exact frequency. This quantity has no absolute threshold.

The test is run on each winding of each motor, one after the other. The result kept is the less favourable of the two: a fault that shows up on only one winding must still be reported.

Always at 33 rpm, always at full amplitude

An FFT test is only worth something when compared with the previous ones, so it has to be run under the same conditions. During the test, Isostase imposes 33 rpm, whatever speed the user has selected, and gives it back afterwards. It also imposes maximum amplitude, regardless of the quiet-listening setting, and waits five seconds of steady running after the ramp before recording.

From a test to a trend

A single result says little. What matters is how it evolves. Isostase keeps the last 16 tests of each motor, using the motor's hours of use rather than the date as the horizontal axis, and computes a slope in THD points per 100 hours. A slope is given a status only once 100 hours of use are covered: three tests close together, with two tenths of a point of measurement noise, would give an enormous slope and a false "critical".

For the bearing, the logic is relative. After three valid tests, Isostase sets a reference specific to that turntable, then tracks a smoothed average of the sidebands and flags it when it reaches twice the reference. The threshold therefore does not depend on an absolute value that cannot yet be calibrated, but on how that particular motor behaved at the start.

What it does not tell you

It is an indicator of change, not a diagnosis. A drifting signature may come from the motor's bearing, from the belt, or from something else: it says that something changed, not what caused it. It is also the motor's current that is analysed, not the vibration of the platter. Finally, the starting values are reasonable but not calibrated on real hardware, and none of this guarantees what the ear will hear.

Related videos: Motors and Belt