We have all shared the experience of driving down the road and listening to the radio when suddenly there is interference. Sometimes the interference is weak, and we can hear the station but have to strain a bit; other times the station is totally swamped by noise, and nothing can be heard above it. The same thing can happen to GPR signals because they are in the same frequency band as FM radio stations, CB radios, cellphones, walkie-talkies, 2-way radios, and other radio devices.
To GPR operators, these external radio sources are external noise.
It doesn’t matter if it’s a car radio or a GPR system, the definition of noise is the same: in-band, unwanted signals. If your depth (or time window) setting is long enough, the background radio noise can be seen towards the bottom of every GPR line, (Figure 1, left) as random or hash -like signal when time gain is applied.
An average trace amplitude (ATA) plot is an excellent way to compare the level of background noise and GPR signal strength, (Figure 1, right). In this case, the average background noise is around 0.9 mV while the peak GPR signal is around 50 mV.
An ATA plot graphically shows the GPR time that the GPR signal decays to the background noise level; in other words, the time the GPR signal can no longer be distinguished from the noise. This is the maximum depth of GPR penetration.
The noise level in Figure 1 does not prohibit use of GPR because it is much weaker than the GPR signals. The worst-case noise occurs when the noise signals are similar to or greater in strength than the GPR signals. When this type of noise is encountered (at a depth of 7 m in the example), the background noise floor is elevated to a very high level and, as an ATA plot shows, GPR exploration depth is severely comprised; in some cases, the GPR data can be unusable.
Figure 2 shows data from an extremely noisy site, less than 100 meters from a radio transmitting tower. The external radio transmitter produces a background noise level of 20 mV, in the bandwidth of the GPR receiver, and completely overwhelms the GPR signals, resulting in virtually no GPR signal penetration and making the data completely useless.
The background noise level is one of the factors that limits GPR signal penetration. Compare the depths of penetration for Figures 1 (7.5 meters) and 2 (1.0 meters). If noise and attenuation levels in Figure 1 were the same as the data in Figure 2, the data were from the same site, and therefore had the same GPR signal attenuation, the extreme noise has reduced the depth of penetration by 6.5 meters!
One way to suppress the type of random noise shown in Figures 1 and 2 is by increasing the number of stacks (Figure 3) . This is the basis behind the Ultra Receiver technology that Sensors & Software employs in the NOGGIN® Ultra 100 and the pulseEKKO® Ultra Receiver. Ultra-Receivers are a thousand times faster than standard receivers, increasing the maximum number of stacks from 2048 to 65,536. For more information on how the Ultra Receiver reduces noise to increase the depth of GPR penetration, see https://www.sensoft.ca/blog/ultra-receiver-revolutionizing-low-frequency-data/.
Figures 1 to 3 all show random noise, but background noise can take on various patterns depending upon the nature of the radio transmitting device. Sometimes it can appear over a narrow area of spatial location (Figure 4a) , fading in and out over time (Figure 4b) , or with a periodic pattern (Figure 4c) . The periodic noise patterns can be caused by the source starting and stopping transmission or changing position. A rotating directional transmitting antenna such as those found at airports might direct signals towards a GPR and then away.
Other variables for how the GPR system receives the external noise are the antenna bandwidth, the distance to the noise source, and orientation of the GPR antennas with respect to the noise source.
Figure 4d shows an interesting sloping noise pattern that is almost coherent. This means that the external noise source was transmitting at a repetition rate similar to the GPR system.
The best way to confirm that a suspicious signal in the GPR line is external noise and not somehow generated by the GPR system or a real GPR reflection, is to look at the data before first break (or zero time); all Sensors & Software systems collect approximately 10% of the data before the GPR transmitter fires (Figures 4a, b and c). If the signal deep in the section is the same as the signal before the first break, the source of the signal is definitely external.
Noise is not an issue for the vast majority of GPR surveys, but operators should recognize it when it occurs. When it does adversely affect the GPR survey, consider running the survey at a different time when radio transmitters are turned off or weaker; many radio stations reduce their transmitter power at night. Also, run some test lines to see if the noise is reduced when survey lines are collected in a certain direction, often orthogonal to the noise source.
