Quadrature Interpolation Glass Scale

Quadrature Interpolation / Multiplication

I will be retrofitting at motorized height gauge using LinuxCNC to drive the various measuring routines.

I have been thinking about the glass scale measurement system:

I have a basic understanding of how my glass scale quadrature encoder works.....
I have determined that the scale is a 20 micron grating and it say is has an accuracy of 5 microns.
I understand this to mean the sin cos output of the read head has been converted to a square wave (which divides the grating by 4 giving 5 microns).

Are these rising and falling 'ticks' positionally accurate and repeatable at slow speed?

If so.
Here is my idea for getting accurate height readings.

I will use a calibrated Mahr Federal LVDT gauge head as the touch sensor. I envision the touch routine as working like this:
The servo system drives the touch probe carriage down toward the measurement surface, at first contact the drive slows and feeds the touch sensor to near a preset measurement force value (readout preset value). The routine will stop the carriage at the nearest encoder tick. I can then add or subtract the dioistance the probe shows from the readout preset value.

For example let say I am reading a 20.0180mm gauge block. The carriage will drive the probe until the glass scale reads 20.0200mm and the LVDT will be showing a value of .0020mm 'before' the readout preset value.

If the gauge block is 20.026mm the carriage will drive to 20.0250mm but the LVDT will say the reading is .0010mm 'after' the readout preset value. I am hoping to work to 1/5 micron accuracy.

Does my logic seem sound?

▶ Show quoted text

my 2c on this subject

The gratings will have a tolerance on absolute position from the manufacturer - so they will not be perfect.
add to that expansion due to temperature.

but I would think that will be consistent to a large degree,
linuxcnc has a way to compensate for this by mapping a correction set of data to add/subtract from position. but that requires calibration.
( I assume you are using linuxcnc's motion controller for carriage movement.)
You could add temperature compensation using HAL if you required it.

You didn't say exactly how you intend to drive this machine. Does the drive have higher resolution then the scales?

Why would you not use a linear scale that has higher resolution?
( Maybe you wanting to use existing hardware? )

Chris M

Scale and LVDT'S are existing hardware. Will use linuxcnc motcontrol and the glass scene as feedback. I hope to also use the lvdt inside the pid loop.

Will probably eventually include temp feedback.

Will calibrate using length standards and will use the linuxcnc compensation.

The system will be calibrated by touching off the surface plate as absolute zero.