Mar-2013

The impact of signal-to-noise ratio on guided wave radar performance

Guided Wave Radar (GWR), although a new technology relative to the industrial level market, has been used successfully in industrial applications for over a decade due to its superior performance, application flexibility, and immunity to changing process conditions. However, even with these benefits, and as is well known to experienced users of GWR technology, all applications are not created equal.

Bob Botwinski
Magnetrol International

Article Summary

Although several factors can determine the likelihood of success when applying GWR to a level measurement application, most notable is the dielectric constant of the process medium. Process dielectrics range from that of water (very conductive) having a nominal dielectric constant of 80, to very light hydrocarbons (nonconductive) having dielectric constants in the range of 1.4 to 1.7. The effective dielectric constant can be even lower when certain process conditions like boiling, flashing or foaming occur, or when bulk solids are considered. The corresponding “GWR Reflection Coefficients” of these media, which are critical for reliably detecting the process level, range from about 80% for water to less than 5% for hydrocarbons like propane and butane.

GWR is a “contacting technology” because the GWR probe is in direct contact with the medium being measured. Since the signal is “focused” in or around the probe, the primary advantage of having the probe contacting the process medium is that very little energy is lost as the signal travels down the probe. Disadvantages include mechanical issues related to any contacting sensor (such as buildup) and the complications created in the matter of probe selection.

As shown in Figure 1, the most “efficient” GWR probe design (one that has the best impedance control and highest sensitivity) is the coaxial probe.

Unfortunately, many applications will not support this probe. This is due to the inherently higher cost structure, and the possibility of the coaxial probe clogging or exhibiting build up, which often drives the user toward other less ideal probe designs. One such design is the single conductor (sometimes called single rod) probe. As shown in Figure 2, the single conductor probe is popular because of its simpler design, lower cost and greater resistance to buildup/clogging; however, it is the most difficult probe to work with due to its inefficient propagation and performance dependencies related to the application and installation.

For example, when inserted directly into tanks or in non-ideal side mounted chambers, the single rod probe has an unavoidable large impedance mismatch at the top of the probe that can interfere with level detection. Single rod probes are also much more affected by extraneous objects in close proximity (i.e. nozzles) and due to its dispersive nature, has much lower sensitivity than other probe types. This combination of unwanted reflections and low sensitivity can make the application of the single rod probe very problematic.

“Signal” and “Noise”
The entire principle of utilising GWR for level measurement is centred on the ability to detect and act upon the signal reflection (impedance mismatch) at the surface of the process medium. In an ideal world, this reflection from the process medium would be the only reflection present along the entire length of the probe. In this ideal case, the very small amplitude GWR signal reflected from a very low dielectric medium would be easy to detect and interpret, resulting in reliable level measurement and one probe could work for all applications in the world. There would be no need to choose between the different probe types.

In practice, however, this is simply not the case. Many other sources in a typical GWR application can produce unwanted reflections. The amplitudes of these “unwanted” reflections can be large, which makes it difficult to distinguish them from the actual level reflection and compromises reliable level measurement.

For this discussion, we will refer to these two cases as “signal” (desired signal from a level surface) and “noise” (unwanted reflections from anything other than the desired level signal, including electrical noise and other disturbances).

Transmit Pulse Amplitude and Signal-to-Noise Ratio
In recent years, much has been said in the industry about the importance of the amplitude (size) of the GWR transmit pulse. While the size of the transmitted radar pulse is certainly important, it is a fact that pulse amplitude alone will not always yield reliable, accurate level measurement under all process conditions. A far more important parameter in reliable level measurement in difficult applications is the signal-to-noise ratio (SNR), which essentially describes the difference between the desired signal and the unwanted noise.

As in our ideal case mentioned earlier, if the signal is much larger than any noise present, reliable level detection is a relatively simple matter. However, if the amplitude of the noise approaches that of the level signal, loss of accuracy or linearity is the first observed effect due to distortion of the level signal as it passes through and interacts with the noise. Worse yet, if SNR is bad enough, the adverse signal interaction can actually result in a loss of the level signal.

The rest of this paper discusses SNR in more detail, explains the effect of SNR on level measurement, describes how SNR is measured and uses actual test data to show how the Eclipse® Model 706 addresses this critical design issue. The following image provides an example of excellent SNR:

In this image of a coaxial probe measuring water, the water signal is large and there are no noticeable ripples, peaks, bumps or other noise in the baseline of the measurement region. Under these conditions, the signal is approximately 840 millivolts in amplitude. With no visible noise in the baseline, the ratio of the signal to the noise (signal ÷ noise) is very large:

SNR = 840mVsignal / 20mV noise = 42

The result: Detecting this level signal is a very simple matter.
On the other hand, things look considerably different in the scope trace below:
In this image, a single rod probe is measuring oil with a dielectric constant of 2. Although the desired level signal (oil surface) is evident, there are many other irregularities in the waveform baseline. Several factors can create these unwanted signals, and, if they are large enough compared to the signal amplitude, accuracy issues and/or loss of the level signal can result. With a signal of 88mV and 20 mV of noise, the SNR is:

SNR = 88mV signal /20mV noise = 4.40

Although much lower than the previous water example, the scope trace below shows that this SNR is still almost 3 times better than competitive GWR devices under similar conditions.