Using SMART

A quick guide to using the SMART (Scanning Microscopy Analysis and Resolution Testing) Macro in

conjunction with SCION Image or NIH Image to measure SEM resolution, imaging performance and

drift.

Problems, questions, suggestions for improvements ? - please let me know ……... djoy@utk.edu

Version 6.0 gamma

What does SMART do?

SMART currently provides three options (1) Manual and Automatic Fourier analysis

resolution testing (2) An alternative Autocorrelation mode to

measure image resolution (3) A superposition diffractogram mode for

precision resolution measurement including time dependent drift and instabilities

An implementation of the Frank and Al Ali algorithm to measure signal to noise ratios

and please note...

Although the MACRO language of IMAGE is good there are some things I could not make it do. If you can help improve the interface and usability of SMART please let me know how.

This is a work in progress so check back on our website for periodic improvements

The code is copyright 2001 to the author. Feel free to use it and modify it as you wish. If you use it in a publication a citation or an acknowledgement would be welcome

Go forth and magnify…….

Setting up SCION or NIH Image

Before using SMART you must set up SCION or NIH Image

Go to the Options menu and click on Preferences

Edit the “Undo &Clipboard Buffer Size” to at least 2000k, ideally 4000k, then click OK. Unless this is done this macro may not work properly and may even fail completely

Go to File and Exit the Image program. This saves the preference setting to disc

Then restart Image

Loading SMART Go to the Special menu and

click on “Load Macros” Navigate your way to the

folder where you stored SMART and click Open in the dialog box

HINT - it is a good idea to store SMART in the MACRO folder of Image

HINT -all output data appears in the INFO box. It is also a good idea to try and position this box some where it is always visible.

Ready to go

If you click on the Special menu item again then the drop down menu will now show the new options that are offered by the SMART macro

Clicking on any of these new entries in the menu will start that function

Automeasure resolution

Step1 - Load the micro-graph you wish to analyze (using the File menu). Note NIH and SCION image can use almost any standard image file format. The number of pixels in the image is not important.

Go to the SPECIAL menu and select an ROI (128 up to 512). Position the ROI by clicking and dragging.

Automeasure Resolution

Go to SPECIAL menu and click the first option - AutoMeasure Resolution.

A dialog box appears and asks for the width of image - this is is width of the field of view of the entire image in microns. Calculate or measure this value from the micron marker

Enter the data and click OK

Total image width (microns)

Calculate width from the scale marker

The FFT display

After some processing the screen now shows the FFT

Hint - if the dialog box covers the FFT drag it out of the way, but don’t click OK until you are ready

The program has attempted to separate the signal from the noise. In many cases the guess that is made is good but you can adjust the result to your own needs...

The threshold toolbar

Click on the Threshold tool

Then put the cursor in the LUT window and drag the cursor up and down to change the threshold level

Too low a value of the threshold will reveal the noise in the background, too high a value may cut off some of the signal and give a pessimistic result

When satisfied click OKThreshold tool

LUT window

Threshold too low

Resolution result

The info box now shows the computed resolution (in nm) determined from the average of the length of the major and minor axes of the ellipse fitted to the FFT.

The eccentricity of the FFT is a measure of the stigmatic error in the beam. Ideally the value should be zero - anything less than about 0.1 is evidence of good astigmatism.

Manual Resolution Mode

A manual mode is also provided.

The initial steps in this are identical with the automatic mode

Instead of relying on fitting an ellipse to the signal region calibrated rings are superimposed on the power transform so the user can visually determine resolution

In cases where the signal transform is not circular this is more accurate

0.35nm resolution

Mode 2 - Autocorrelation

This mode does an auto-correlation of the FFT of the image to determine the information cut-off. No user intervention is needed

Load the image then go to the Special menu and select your ROI

When the ROI is set then go to the Special menu again and select the AutoCorrelation mode

Autocorrelation - step 2

After you answer the question about the image width the program will display the 2-D Auto-Correlation plot as shown

The macro measures the width of the correlation peak at 20% of maximum

This width gives the spatial resolution of the image at the information cut-off limit.

Line profile across correlation

Mode 3 - Two Image Analysis

This part of SMART uses two images to give an accurate measure of resolution and drift

The two images should be recorded one directly after the other. Do not adjust the stage, focus, gun etc..

Load the two images then go to the SPECIAL menu select and position the ROI on the top image. The ROI must be smaller than the image size

Image 1

Image 2

Mode 3 - step 2

After you calibrate the image size as usual the Macro copies the image in the ROI from image 1, puts the same ROI on image 2 but shifted horizontally by 16 pixels, copies that image area, then forms a composite image from these two fragments. The 2-D power spectrum is then calculated and displayed

2-D power spectrum from composite image

Resolution analysis Only detail present on a

pixel by pixel basis in both images appears modulated by the fringes in the FFT. Noise is unmodulated and so is readily distinguished from image detail

Calibrated rings are superimposed over the fringed FFT so that the resolution can be visually determined

Here fringes can be seen extending to below 2.5 nm

Measuring Drift The fringes should be

vertical because the images are offset horizontally by 16 pixels

If drift or instability occurs the fringes will rotate through some angle . The actual drift direction is then in the indicated direction.

fringes

Drift vector

Note that…….

The resolution result given by this mode averages over the time interval required to record both images and so is a good test of real world resolution.

This routine has failed to work on some computers, apparently because of screen driver problems with the SWITCH command. If the macro fails to work for you, load both images as usual, position the ROI on image 1, then manually put the ROI in position on image 2 as well, then run the Two Image routine.

On small monitor screens, or when using big images, the INFO box may sometimes be hidden. Try and move it to a convenient corner of the screen before starting an analysis.

S/N ratio measurement To use this mode, load the

image of choice then select S/N mode. The Frank and Al Ali algorithm is used on alternate line pairs of data

The computation takes a few seconds so please be patient

The result appears in the Info box

This works best when the detail in an image is large compared to the pixel size (i.e. images with empty magnification)