Determining accuracy in EMI measurement systems

During the last few years, EMI (electromagnetic interference) regulations have been introduced in most developed countries.

Compliance with regulations to meet C-tick, FCC or CE marking requirements or to satisfy military or automotive contractual requirements, generally requires type testing for emissions and immunity.

This article discusses the factors that contribute to measurement uncertainty in EMI measurement systems, with a view to helping users of EMI test reports from accredited laboratories and those performing in-house precompliance testing.

The accuracy of EMI measurement systems is of interest to a wide range of people, including anyone:

• Involved in operation of test facilities, both precompliance and full compliance;
• Interpreting test reports that include a statement of measurement uncertainty;
• Questioning test reports that do not include such a statement;
• Involved in the improvement of the EMI performance of a product (the measurement system needs to be able to resolve the difference in performance between versions of the product).

Definitions

Before embarking on a discussion of the particular measurement problems associated with EMI, some commonly used measurement definitions are repeated for reference:

Measurand - a physical quantity or condition that can be measured (eg, voltage, electric field strength, frequency).

Measurement - the process of associating a number with a quantity by comparison of the quantity with a standard.

Measurement System - a set of components that are interconnected to detect, transmit and process the physical quantity being measured.

Accuracy - the comparison of the measurement result to an accepted or true value.

Random Error - inaccuracy in the measurement result due to noise, drift, observational and other factors. This type of error has a Gaussian distribution about the mean.

Systematic Error - a fixed error between the mean measurement result and the true value of a measurement.

Transducer - a device that receives energy from a measurand and converts the energy to a form useable by a measurement system. (The output of all transducers used for modern EMI measurements is electric, but this definition also encompasses the simple mercury thermometer, which converts temperature to displacement.)

Measurement systems provide an estimate of the true value of a measurand.

Typical EMI Measurement Systems

There are two types of commonly applied EMI measurement:

1. Conducted disturbances, where a transducer (LISN, voltage probe or current clamp) is applied to a conductor or cable bundle to monitor the amplitude of low frequency (typically 9 kHz - 30 MHz) energy.

   The transducer is usually connected to a limited-bandwidth receiver by a network of one or more cables.

2. Radiated disturbances, where the energy released by a device into free space is measured, usually by placing a transducer (antenna) at a fixed distance from the device.

   The transducer is connected by a network of amplifiers and cables to a limited-bandwidth EMI receiver.

Conducted Measurements

The aim with most EMI measurements is to determine compliance with a standard that has specified some limit for the maximum amount of radiation that a device can emit.

In the case of conducted measurements, the limits are specified in absolute values of voltage or current, depending on the measurement system named by the standard. Most of the time, the limits are specified in units that are appropriate for the transducer named by the standard.

The standards require a measurement of the RF voltage at the output port of the EUT. A measurement is a comparison of a quantity with a known standard or reference.

To achieve this, we need to know the response of the transducer. This can only be done by providing the transducer with a known input energy and measuring its output energy.

Performing this calibration at many frequencies allows us to estimate the RF energy at the transducer's input (which we cannot directly measure in most cases) based on the RF energy present at the output.

Similarly, we need to know the loss of the coaxial cables that are present between the transducer and the receiver.

Again, applying a known RF input and measuring the output of the cable at a range of frequencies characterises the cable's response to a range of stimuli.

The preceding paragraphs give rise to correction factors. We have calibrated the response of two elements of our measurement system to take account of systematic errors.

To disregard cable loss or transducer response is to invite disaster by producing measurements with an in-built offset.

Radiated Measurements

Correction factors for this type of measurement system are typically cascaded, except that no correction factor is applied for the test site.

Test site calibration confirms that the site attenuation is within acceptable limits, but no correction can be made for a non-ideal site. The only way to deal with a test site that does not fulfil the site requirements is to revise the site's construction.

The receiver is calibrated to confirm that it is within specifications. The specification is incorporated in the uncertainty budget as a rectangular distribution.

Measurement System Uncertainty

Our estimate of the RF energy present at the output of the EUT is derived by the raw signal obtained at the measurement system output with correction factors for cables, transducers and other system elements.

The result of this process is an estimate of the true value of the amplitude of an emission. Like any measurement, there is an uncertainty associated with it.

This is a range either side of the measurement result that describes the extreme values within which the true value of the measured can be reasonably expected to lie.

When determining the overall measurement uncertainty, the uncertainty associated with the characterisation of each component in the measurement system is combined by the root-sum-of-squares (RSS) method. It works out that most of the time the resultant distribution is about Gaussian.

For a Gaussian distribution, about 66% of the values lie within one standard deviation of the mean, 95% lie within two standard deviations, and 99% lie within three standard deviations of the mean.

Conclusion: Real-World Implications

Typical measurement uncertainties for conducted measurement systems are around ±2 to ±3 dB. Radiated measurement uncertainties are more like ±4 to ±6 dB.

So, if a device under test has an emission that is very close (eg, within 0.5 dB) to the limit for compliance with a standard, the odds are only slightly better than even that the true value of the emission level is under the limit.

Regulatory interpretation of measurement uncertainty varies across the world. Most countries, Australia included, disregard measurement uncertainty: if the emission is under the applicable limit by 0.1 dB, the equipment complies with the standard.

However, precompliance testing that is undertaken under non-ideal conditions (eg, at a non-compliant test site) can have uncertainties much higher than the figures quoted here.

A marginal pass at precompliance testing can lead to a nasty surprise during final conformance testing, due to measurement uncertainty alone.

Thus, it is wise to perform precompliance testing on a compliant site; if possible, the same site that will be used for final testing. If this is impractical, work with very large safety margins during precompliance testing and product development.