Add support for the KPC101 K-Cube Piezo Strain Gauge controller

[MahsaHabibi]

Summary

Adds MIC support for the Thorlabs KPC101, which integrates the piezo driver and strain gauge reader into a single K-Cube with a single serial number. MIC predates this device: the existing mex wrappers only cover the split-device architecture (KPZ101/TPZ001 driver + KSG101/TSG001 gauge, PCC_/SG_ C APIs), so classes like mic.linearstage.KCubePiezo and mic.stage3D.NanoMaxPiezos cannot drive it. This PR adds wrappers for the KPC_ C API (Thorlabs.MotionControl.KCube.PiezoStrainGauge) and a new linear stage class, following the existing MIC patterns throughout.

New files

Mex wrappers (mex_source/MIC/Kinesis_KPC_*/mexFunction.cpp + compiled mex64/*.mexw64)

Wrapper	Functionality
Kinesis_KPC_Open	Opens the device, loads settings, identifies, starts 200 ms polling
Kinesis_KPC_Close	Stops polling and closes the device
Kinesis_KPC_SetPosition	Commands closed-loop position (uint16, 0-32767 = 0-100% of travel)
Kinesis_KPC_GetPosition	Reads position from the strain gauge feedback (same units)
Kinesis_KPC_SetPositionControlMode	Sets loop mode (1 = open, 2 = closed)
Kinesis_KPC_SetZero	Starts the strain gauge zeroing routine
Kinesis_KPC_GetStatusBits	Reads the status word (used to wait for zeroing to finish)
Kinesis_KPC_GetMaximumTravel	Best-effort read of actuator travel in 100 nm steps

Build script (mex_source/MIC/buildKPCMex.m)

Compiles all eight wrappers against the Kinesis import library into mex64/. Requires a configured C++ compiler (mex -setup C++) and Kinesis at C:\Program Files\Thorlabs\Kinesis.

Class (src/+mic/+linearstage/@KCubePiezoStrainGauge/KCubePiezoStrainGauge.m)

mic.linearstage.KCubePiezoStrainGauge extends mic.linearstage.abstract. On construction it opens the device, sets closed loop, runs the zeroing routine (~30 s, waits on status bits), and centers the stage. Positions are in microns; the class converts to/from the API's percentage-of-travel WORD internally. Because the device reports true position through its built-in strain gauge, no Slope/Offset calibration is needed (unlike KCubePiezo).

PZ = mic.linearstage.KCubePiezoStrainGauge('113251934','Z');   % 20 um default travel
PZ.setPosition(5);
PZ.getPosition()

MaxPosition is an optional third argument (default 20 um) and must match the Maximum Travel configured in Kinesis — KPC101 firmware was observed to return 0 from KPC_GetMaximumTravel even after zeroing, so the device query is best-effort only.

Runtime DLLs (mex64/)

Adds Thorlabs.MotionControl.KCube.PiezoStrainGauge.dll and updates all existing Thorlabs DLLs in mex64/ from 1.14.10/1.14.11 to 1.14.59. The device DLLs and DeviceManager.dll are mutually version-locked: mixing releases fails at load time with missing-module/missing-procedure errors, so they must all come from the same Kinesis release.

Testing

Verified on hardware (KPC101 S/N 113251934, NanoMax 300 Z axis, 20 um travel, MATLAB R2025b, MSVC 2022):

• Construction, zeroing, and centering complete cleanly
• Commanded 5 / 10 / 15 um; strain gauge read back 4.9995 / 10.0003 / 15.0011 um

Known limitation (shared with the other Kinesis mex classes): re-opening the device in the same MATLAB session can crash MATLAB; construct the object once per session.

Notes for reviewers

• DLL updates affect existing workflows: everyone using this repo's TCube/stepper/laser-diode mex files gets the 1.14.59 DLLs after merging. The Thorlabs device DLLs and DeviceManager.dll only load as a matched set, so a partial update is not possible. Existing mex binaries were not recompiled and load fine against the new DLLs, but other lab setups should be sanity-checked after pulling.
• Docs workflow fix included: the GenerateDocumentation workflow could never pass on pull requests (PR runs check out a detached HEAD, so its git push step always exits 128). This PR guards the commit-and-push step to run only on push events; on PRs the workflow now just validates that genDoc runs. Happy to split this into a separate PR if preferred.
• Zeroing on every construction (~30 s): the constructor always runs the strain gauge zeroing routine to guarantee positional accuracy after power-up. If this proves annoying in daily use, an optional skip-zeroing flag would be an easy follow-up.

🤖 Generated with Claude Code