ASAP is a micron resolution kilosample/sec position tracker based on PSD (Position sensitive device) sensing. ASAP was originally developed for characterizing hand tremor and for ground truth in Micron development (using inertial sensors), but when the inertial approach didn't work, we change to using ASAP inside the Micron feedback loop. See High-Speed Microscale Optical Tracking Using Digital Frequency-Domain Multiplexing and Micron: an actively stabilized handheld tool for microsurgery.

Since '09 when the ASAP paper was published I've reimplemented the sensor head so that it is much less bulky and uses a 60 degree verging angle instead of 90 degrees. I used pro-grade photographic quick releases and ball heads to create a stable and flexible mounting arrangement for the sensor head. This make deployment of ASAP in the operating room a more practical proposition. We now also track 6 LEDs so that we can measure the full pose of both tip and handle in Micron.

The PSD camera contains only four transimpedance amplifiers. The PSD itself is on the other side of the PC board. The PSD is very beneficial for high-rate, high-resolution tracking because it intrinsically measures the light centroid. This gives a much lower bandwidth and processing burden than would result from using a conventional imaging camera. Achieving similar performance would require a 10k x 10k pixel sensor with 1000 FPS output. Such a camera simply isn't available, even if you could process the data in real time.

I initially came onto the Micron project because the existing ASAP used an analog lock-in amplifier to demodulate the position signals, and it was malfunctioning (due to poor soldering technique.) Even with only one source, lock-in detection is common with PSDs because it rejects ambient light and 1/f noise. It was fairly obvious to me that the lock-in approach could be extended to multiple sources by using different modulation frequencies, but the hardware complexity of analog demodulation seemed unwarranted, given the kHz modulation frequencies. Instead, I implemented demodulation using Labview software. This requires oversampling the signal by 30x or so, but even then there were no significant performance issues using a Labview implementation.

ASAP implementation has also involved a considerable amount of MATLAB coding for performance characterization and calibration. The current calibration uses the optimization toolbox to solve for the camera calibration parameters, and also computes a sensor calibration lookup table.