IRTF TCS Servo concepts I Peter Onaka Note: this is a compilation of information blatantly copied from multiple sources. 11/4/03.

IRTF TCS Servo concepts I

Peter Onaka

Note: this is a compilation of information blatantly copied from multiple sources.

11/4/03

TCS Servo primary functions

• Tracking = primarily velocity control– Most important = Only time when we are actually

acquiring science data.

• Offset = small position move – Finishes with track mode, observer waiting…

• Slew = large position move– Observer waiting….

Why is tracking most important?

• Poor tracking means– Smeared image/spectra– May render AO useless– Image quality invalidated– Bad or poor data– * because all or our instruments integrate on object

• Poor Offset means – Wasted time– Possible failure to acquire data (sequences)

• Poor Slew means– Wasted time

What PIDs do

• PIDs close a position loop

Ground based telescopes

• A PID alone will probably not be adequate. We also have a more complicated drive arrangement. We need to close a velocity AND position loop.

What does the velocity loop do?

Resonances

Position loop

Slew and Feedforward

What does each loop do?

• We shouldn’t expect a PID alone to work well for velocity control due to “disturbances and nonlinearities” affecting the dynamics of the servo.

What disturbances?

• It’s important to understand that the disturbances have a power spectrum (they have frequency terms).

Dynamic wind effects

Mauna Kea

Power at and above

telescope resonances

of 3.5Hz (RA) and 8.5Hz

(DEC)

Wind effect PSD (VLT model)

Torque magnitude

could still be significant

Power at and above

telescope resonances

of 3.5Hz (RA) and 8.5Hz

(DEC)

Frequency/tuning challenge

Decrease to avoid

resonances

Increase for disturbance correction

• But remember

“Increase bandwidth”

“stick/slip” and nonlinear disturbances

Other nonlinear disturbances

How to fix nonlinear disturbances

Position loop is low

bandwidth

High bandwidth

velocity loop is “ essential”

What a PID might do

• We could easily get this with a PID alone.

velocity

position

20Hz PID position updates

Desired velocity

time

High inertia case

Peaks excite resonances

The old servoPosition command pulse rate

Tach

Tach

Pwr amp

Rect/ lead/lag/

sum

3Hz HP filter

preload Bull gear

3.7Hz LP filter + sum

sum

Pwr amp

Rect/ lead/lag/

sum

3.7Hz HP filter

preload

Torque split

inc encoder

Position error counter

14 bit DAC

Integrator + feed-forward

Limiters + ramp

gen

Feedforward & offset

Position loop

LP velocity loop

HP velocity loop1

HP velocity loop2

LP velocity loops

• Rect = unipolar for anti-backlash• Lead/lag = PI controller plus R/C lead circuit• Sum = velocity command, tach HP and

preload

Tach

Bull gear

Pwr amp

Rect/ lead/lag/

sum

3Hz HP filter

preload

HP velocity loop1

Lead Lag circuit

Lead Lag = stiffness

• This stiffness increase compensates for torque disturbances (wind, cable wrap loading, stick-slip etc.)

Why we shouldn’t use the bull gear encoder feedback for HP velocity control

Tach

Pwr amp

Rect/ lead/lag/

sum

3Hz HP filter

preload Bull gear

3.7Hz LP filter + sum

inc encoder

HP velocity loop2

Ev: ” It was discovered early on in the development of the TCS, that there existed a significant amount of torsion between a given motor and the bull gear twisting of motor shaft). The simplest method of

reducing the affect of this torsion was to create two separate servo loops, one for each motor/tachometer combination in the frequency

range where this torsion affect dominates.”

Torsion difference