Purpose of this document
This third part of the tutorial describes data collected on a crableg resonator at Carnegie Mellon. The design and simulation of the resonators are described in Part 1. Collecting data using the Microvision system is described in Part 2.
Corner of Chip showing Crableg Resonators (click on image for larger view)
SEM (scanning electron microscope) images are used to show the actual measurements of the device. There are several points in the manufacturing and release process that can cause the actual device to differ from the layout. For example, exposure of the photoresist with the mask may not be ideal. Etchants in the release may attack the metal slightly. Although the SEM image shows a scale, it may be more accurate to reference a more precise scale measured from the device itself. For example, the pitch of the comb finger teeth will be a constant. All of the SEM images are contained in this directory. The following image highlights the areas on the crableg resonator by their filename.
Crableg SEM showing Details (click on image for larger view)
Microvision Analysis Data
Microvision analysis data for the crableg measurements taken at Carnegie Mellon can be found in the directory here. The naming convention is as follows...
crab_(die #)_(resonator #)_(date of measurement)_(AC voltage applied).anl
The AC voltage may include a 'b' indicating a backward frequency scan.
A summary of the data in this directory can be found here. The data shows a maximum on chip frequency variation of 1.3%. Resonant frequencies vary based on the different voltages (2.5V or 5.0V) by roughly 100 Hz.
The following picture shows a plot of the data for die 5.
Plot of Crableg Data (Die 5) (click on image for larger view)
This tutorial was designed to provide an overview of steps from the design of a crableg resonator through the testing of the same bonded die. The design included hand analyis, simulation, and layout. The testing included taking measurements on the Microvision system as well as SEM images. Some of the data was also presented.
This work presents a number of questions for follow up. Can more of this process be automated? Can the simulation be more accurate by using physical dimensions from the layout? How does the data compare to measurements taken on different days?...or at different sites?
Back to Part 1
Back to Part 2