The frequency range over which noise or interference can affect the AFM image extends from about 500 times the horizontal scan rate to about 500 times below it. The performance of electronic circuitry in this range is extremely good, and most of the sources of noise are external. If we assume the instrument has no vibration isolation or acoustic shielding, good results will be obtained with samples that have features at the micron level. Below this, seismic vibrations produced by foot traffic in the building, elevators, doors being closed, or street traffic outside are noticeable. This will depend largely on the actual site, with an isolated site on the ground floor being quite good compared to a large laboratory in an urban location. This type of interference is impulsive in nature, and shows up as horizontal streaks in the image.

       At a still lower frequency, dark bands in the image can be caused by random thermal expansions of the PZT and metal components of the head. This is more noticeable at lowest scan speeds, since there is more time for temperature changes to occur. This is usually caused by air currents.

       In the range above the scan rate, mechanical and acoustic noise can be caused by the fan in the EIU and the computer. These units should be placed on the floor, and as far from the stage as possible. The Fourier transform is useful in identifying the source of this noise. The rotation speed of the fans is in the range of 30 to 60 Hz, and the noise from the fan blades is 6 times higher. Higher harmonics can also be observed. Residual power supply ripple can also be observed at 60 Hz and higher harmonics, and may be indistinguishable from acoustic noise generated by line operated electrical equipment such as transformers or fluorescent light blasts. The cantilever responds as a microphone to conversations, or a ringing telephone.

       A potential source of interference is the light emitted by fluorescent lights. It will have a strong modulation at 120 Hz, and could be picked up by the laser beam detector. We have never identified this as a problem.

       Almost all of these noise sources are effectively eliminated by the acoustic cover/vibration isolator or AVIC. The 1 Hz cut-off frequency is below that obtained by most air tables, and by being designed specifically for the Quesant instrument, the cover is compact, adjustment free, and light weight. With the stepper ministage and video camera, the sample can be examined in a dust free environment without opening the cover.

       The signal from the laser detector in the head is extremely sensitive, and as noise sources are removed, the system amplifies this signal as much as necessary to provide an image with full scale deflection in Z. To obtain the least possible noise, it will be best to scan quite slowly. This allows the DSP to average more readings per pixel. In the configuration screen there is a number shown as samples/DP. This is the number of A to D conversions added together to give one pixel of data. In the broadband modes, the error signal has a bandwidth of 24KHz, and lower noise will be obtained using the standard modes.

       Noise acts as a dithering signal and removes any quantizing errors due to the 16 bit resolution of the A to D converters, and the noise is in turn removed by averaging many conversions. In any electronic system, the amount of noise in the signal will be dependent on the bandwidth transmitted through the system. The narrower the bandwidth can be made, the quieter the signal will be. In many systems, the frequency of the signal cannot be changed, but in the AFM, the signal can be lowered in frequency just by scanning at a slower rate.

       Before the conversion to digital, the signal passes through the PID loop, which is the principal means of limiting the bandwidth in the Z channel. By reducing the gains, the bandwidth is reduced. Of course, this slows the response of the system, necessitating slow scanning for good image resolution.

       There is some residual ringing of the scan tube at a frequency of about 660 Hz in spite of the efforts to suppress it using the two step reversal and the eight pole Bessel filter. Due to friction on the surface, there may be a slight twisting of the probe in contact mode, and this causes the very slight ringing to produce a slight modulation in Z. This can sometimes be seen in the Fourier transform as two diffuse vertical bars at 660 Hz. The amplitude of this ringing goes to zero as the scan speed is reduced. It is generally not seen in WaveMode due to the much greater cantilever stiffness and the reduction in surface friction.
The bandwidth can be reduced from below the scan frequency by operating in the error only mode. The PID gain settings now determine the low frequency cutoff, and spatial filtering can be used to limit the upper frequency cutoff. Images of mica atoms have been obtained using the error only mode at a scan frequency of 15Hz. The low frequencies are also eliminated by using tilt removal such as Line by Line H or Histogram/H.

       In conclusion, the quietest images will be obtained in the AVIC using a slow scan speed such as 1 Hz. The lowest gain settings should be used that provide adequate detail in the image. Periodic artifacts can be removed by the Fourier transform, and spatial filtering of the image can be performed. There are numerous modes of rendering the image, and careful adjustment of these can enhance the detail in the image.