AD2S83

REV. E

–16–

SOURCES OF ERRORS

Integrator Offset

Additional inaccuracies in the conversion of the resolver signals

will result from an offset at the input to the integrator. This

offset will be treated as an error signal. The resulting angular

error will typically be 1 arc minute over the operating tempera-

ture range.

A description of how to adjust the zero offset is given in the

Component Selection section; the circuit required is shown in

Figure 1.

Differential Phase Shift

Phase shift between the sine and cosine signals from the resolver

is known as differential phase shift and can cause static error.

Some differential phase shift will be present on all resolvers as a

result of coupling. A small resolver residual voltage (quadrature

voltage) indicates a small differential phase shift. Additional

phase shift can be introduced if the sine channel wires and the

cosine channel wires are treated differently. For instance, differ-

ent cable lengths or different loads could cause differential phase

shift.

The additional error caused by differential phase shift on the

input signals approximates to

Error = 0.53 a

× b arc minutes

where a = differential phase shift (degrees).

b = signal to reference phase shift (degrees).

This error can be minimized by choosing a resolver with a small

residual voltage, ensuring that the sine and cosine signals are

handled identically and removing the reference phase shift (see

the Connecting the Resolver section). By taking these precau-

tions the extra error can be made insignificant.

Most resolvers exhibit a phase shift between the signal and the

reference. This phase shift will, however, give rise under

dynamic conditions to an additional error defined by:

Shaft Speed (rps)

× Phase Shift (Degrees )

Reference Frequency

= Error Degrees

Under static operating conditions phase shift between the refer-

ence and the signal lines alone will not theoretically affect the

converter’s static accuracy.

For example, for a phase shift of 20 degrees, a shaft rotation of

22 rps and a reference frequency of 5 kHz, the converter will

exhibit an additional error of:

22

× 20

5000

= 0.088 Degrees

This effect can be eliminated by placing a phase shift in the

reference to the converter equivalent to the phase shift in the

resolver (see the Connecting the Resolver section).

Note: Capacitive and inductive crosstalk in the signal and reference

leads and wiring can cause similar problems.

VELOCITY ERRORS

Some “ripple” or noise will always be present in the velocity

signal. Velocity signal ripple is caused by, or related to, the

following parameters. The resulting effects are generally addi-

tive. This means diagnosis needs to be an iterative process in

order to define the source of the error.

1.0 Reference Frequency

A ripple content at the reference frequency is superimposed

on the velocity signal output. The amplitude depends on

the loop bandwidth. This error is a function of a dc offset at

the input to Phase Sensitive Demodulator (PSD).

2.0 Resolver Inaccuracies

Impedance mismatch occur in the sine and cosine windings

of the resolver. These give rise to differential phase shift

between the sine and cosine inputs to the RDC and varia-

tions in the resolver output amplitudes.

2.1 Sine and Cosine Amplitude Mismatch

This is normally identified by the presence of asymmetrical

ripple voltages.

2.2 Differential Phase Shift between the Sine and Cosine Inputs

The frequency of this ripple is usually twice the input veloc-

ity, and the amplitude is proportional to the magnitude of

the velocity signal. The phase shift is normally induced

through the connections from the resolver to the converter.

Maintaining equal lengths of screened twisted pair cable

from the resolver to the AD2S83 will reduce the effects of

resistive imbalance, and therefore, reduce differential phase

shift.

3.0 LSB Update Ripple

LSB update noise occurs as the resolver rotates and the

digital outputs of the RDC are updated. For a correctly

scaled loop, this ripple component has a magnitude of

approximately 2 mV peak at 16-bit resolution.

3.1 Ripple due to the LSB rate given by:

LSB rate = N

× Reference Frequency

The PSD generates sums and differences of all its compo-

nent input frequencies, so when the LSB update rate is an

multiple of the reference frequency, a beat frequency is

generated. The magnitude of this ripple is a function of the

LSB weighting, i.e., ripple is less at 16 bits.

4.0 Torque Ripple

Torque ripple is a phenomenon associated with motors. An

ac motor naturally exhibits a sinusoidal back emf. In an

ideal system the current fed to the motor should, in order

to cancel, also be sinusoidal. In practice the current is often

trapezoidal. Consequently, the output torque from the motor

will not be smooth and torque ripple is created. If the load-

ing on a motor is constant, the velocity of the motor shaft

will vary as a result of the cyclic variation of motor torque.

The variation in velocity then appears on the velocity

output as ripple. This is not an error but a true velocity

variation in the system.