There are a number of necessary aspects of the design of the NPL Reference Flickermeter that are not defined in IEC 61000-4-15: 2003 [1]. These factors and other test parameters unspecified in [1] can have a significant effect on the Pst readings. This document illustrates some of the problems arising from these ambiguities and explains how they are minimised in the NPL Flickermeter design and calibration test set-up.

A full description of the design of the NPL Flickermeter can be found in Design of NPL Flickermeter.

Unspecified Design Parameters
Filter Charge Time

As explained in Design of NPL Flickermeter, IEC 61000-4-15: 2003 describes filters, which are used to simulate the human visual response to light flicker from an incandescent lamp. For flicker measurements, a voltage signal is applied to the filters and the resulting readings are used to calculate the Pst value. When a constant signal is applied the output of the filters will start at a high value before decreasing to a steady state level. The time allowed for the filters to reach this state, before a flicker test is begun, is not specified in IEC 61000-4-15: 2003.

The filters take around 20 seconds to reach a steady state value within 0.01 % for most waveforms. This is chosen as the pre-test time and any flicker perceptibility readings taken up to this time are not used in the final Pst evaluation.

Although the filters reach steady state within around 20 seconds, the Scaling Reference Voltage, described below, takes longer for certain waveforms, as explained in Effect of Pre-Test 1-Minute Averaging on Pst Readings, below. However, a longer pre-test time is perhaps impractical for commercial flickermeters, which tend to use a shorter pre-test, and so a time of 20 seconds is chosen.

Calculation of Scaling Reference Voltage

IEC 61000-4-15: 2003 requires that a one-minute sliding average RMS voltage must be used to scale the flicker readings, which is updated every half-cycle of the mains voltage waveform. It is not specified, however, how to calculate this voltage during any pre-test and the first minute of a flicker test.

For the NPL Flickermeter, up to the first minute, the average of all the half-cycle RMS values is used as the Scaling Reference Voltage and this is updated every half-cycle. After the first minute the average of all the half-cycle RMS values over the previous minute is used. The data from the pre-test is included in the calculation of the Reference Voltage used at the beginning of the flicker test.

Flickermeter manufacturers may choose to calculate the initial reference value differently and reset the value after the pre-test period. This can have an effect on the Pst readings, since calculating the reference value in this manner can effectively smooth the modulation waveform, if a change occurs near the beginning of a test. The behaviour of an appliance at the beginning of a flicker test can be important and the former approach (of continuing the averaging from the pre-test) is therefore chosen for the NPL implementation.

It can however be argued that the averaging should only take place with actual test data and as this is unspecified in IEC 61000-4-15: 2003, the NPL flicker tests are configured to minimise any difference arising from different implementations of the pre-test. For calibrations with complex waveforms (see Flicker Waveform Library) this is achieved by starting the modulation 1 minute after the beginning of the test.

For square wave modulation at the test points defined in Table 5 of IEC 61000-4-15: 2003 it is not possible to leave a one minute gap before starting the modulation, without deviating from the required Pst reading of 1. It is impractical to consider all implementations of Flickermeters and so the Pst reading of 1 ± 0.05 is used as a target Pst value for these tests. The readings given by the NPL Flickermeter are used only as a method of verifying that the settings of the applied waveform are correct, rather than for a direct comparison.

The Effect of Unspecified Test Parameters on Pst Readings

Table 5 of IEC 61000-4-15: 2003 the modulation depths and rates for a square wave modulated 50 Hz sine wave that should result in a Pst of 1, when applied to a Flickermeter. The table does not specify the phase of the modulation with respect to the carrier voltage or test start point.

Simulated signal

Simulated 50 Hz signals with rectangular and sinusoidal modulation are used for the results in the following sections.

For rectangular modulation the simulated signal can be given by the following equation.

(1)   

where f_c is the carrier frequency = 50 Hz, is the relative voltage change, f_F is the modulation frequency, f is the modulation phase and v(t) is the signal level at time t. sign(x) indicates multiplying by a value of +1 if the x is greater than zero and –1 if x is less than zero. If the x is zero and decreasing the resulting value is -1 and +1 otherwise.

Variation of Pst with Modulation Phase w.r.t. Test Start Time

A simulated mains voltage with a frequency of 50 Hz was amplitude modulated at the modulation depths and rates given in Table 5 of IEC 61000-4-15: 2003. The waveform is described by equation (1), above. The filters were charged with an unmodulated 50 Hz sinewave for 20 seconds before starting the tests.

The start time of the modulation was varied and the Pst readings for each modulation start time are shown in Figure 1. The modulation phase in Figure 1 indicates the point in the modulation cycle at which the test is begun after the pre-test. I.e. for 1 CPM a phase of 90 degrees indicates that the waveform starts in a high state and changes to a low state after 30 seconds (a quarter of one complete modulation cycle), or t in equation (1) is set to -30 s. The phase of the modulator with respect to the carrier voltage (f in equation (1)) was kept constant for all the points in Figure 1 and Figure 2.

 
Figure 1 – Variation of Pst reading with modulation start time. Charging for 20 seconds with unmodulated voltage

Figure 2 shows the results of repeating the above test, this time starting the modulation at the beginning of the pre-test. This significantly reduces the variation in Pst readings.

Figure 2 – Variation of Pst reading with modulation start time. Charging for 20 seconds with modulated voltage

As can be seen in Figure 1, the time at which the modulation is started can have a significant effect on the Pst readings. It is therefore important that the test conditions are carefully specified. Even if this is the case an additional uncertainty contribution results from the difficulty of synchronising the start time of the Flickermeter under test. This uncertainty is low for low modulation frequencies (e.g. 1 and 2 CPM), where the test can easily be started in the middle of a high state. The uncertainty contribution is also small for high modulation frequencies, where the variation is smaller. However, for 7 CPM the waveform is only constant for short time periods and the uncertainty contribution is therefore significant (~0.5 %).

This is illustrated in Figure 3, which shows how an error in the test start time can cause an error in the Pst reading when compared to the expected reading. The desired modulation start time is when the waveform is halfway through a high state.

Figure 3 – Potential error in Pst reading if test start time cannot be precisely determined

For positive errors in test start time the modulation has already begun when the test is started and therefore the variation is small. However, for negative errors the modulation does not start until some time later into the test. As a result fewer changes occur during the test and the Pst reading reduces for larger delays.

The 4000 CPM point is a special case in that the Pst reading is lower if the modulation is started early. The beginning of the modulation causes a spike in the flicker reading, which becomes less significant as the filters charge and reach a steady state. The spike is ignored if the start of the modulation occurs during the pre-test and so the Pst reading is lower.

To reduce uncertainties, it is therefore desirable to start the modulation some way into a test, or before the test is started. The Pst readings for some of the points tend to tail off when the modulation is started late. Therefore, at NPL, for flickermeter calibrations with square wave modulations, the modulation is started before the test is begun. The smaller variation seen in Figure 2, where the modulation is present throughout the test and pre-test, further illustrates this point.

Effect of Pre-Test 1-Minute Averaging on Pst Readings

Although the variation in Pst reading with start time is reduced by starting the square wave modulation before starting a test, there is still some variation for certain modulation frequencies. At 2 CPM, a maximum variation of ~0.23 % is seen when a pre-test time of 20 seconds is used, as seen in Figure 2. This is caused by a variation in the Scaling Reference Voltage, which is averaged over 1-minute. Using a pre-test time of greater than 60 seconds minimises this and reduces the variation in Pst reading to less than 0.002 % (and <0.003 % for 1 CPM).

A pre-test time of 60 seconds would therefore appear to be a sensible choice. However, this is perhaps a little long to use in practise and a time of 20 seconds was chosen to bring the implementation of the NPL Flickermeter closer to that of commercial flickermeters, which tend to have shorter pre-test times.

Variation of Pst with Modulation Phase w.r.t. Carrier Voltage

The phase of the modulator with respect to the carrier voltage is important for faster rates of modulation. The phase does not affect the Pst reading per se, but it does have an effect on the mean RMS voltage, used as a reference voltage to scale the Pst readings. This change in reference voltage has a proportional effect on the Pst reading. The Pst readings at a modulation frequency of 4000 CPM and dV/V of 2.4 % for various modulation phases (f in equation (1), above) are shown in Figure 4. For all other points in Table 5 of IEC 61000‑4‑15, changing the phase of the modulator with respect to the carrier voltage, f, has a similar effect to changing the start time, and the results are similar to those given in Figure 2 largely due to the period of the modulation waveform being significantly greater than that of the carrier.

The simulated modulated voltage waveform is applied to the filters of the Flickermeter for 20 seconds before starting each test. Modulation phase in Figure 4 refers to the phase of the modulator with respect to the carrier voltage.

Figure 4 – Variation of Pst reading with modulator phase, f, at 4000 CPM

Influence on NPL Flickermeter Calibrations

In summary, although not detailed in IEC 61000-4-15: 2003, it is clearly important to specify the phase of the modulator with respect to both the carrier and the measurement start time as these can both have a significant effect on Pst readings for all the required measurement points. To reduce the measurement uncertainty for square wave modulation, the modulation is begun before starting the test and configured so that the signal is in a high state at the beginning of a test for modulation frequencies of 1 and 2 CPM. A target Pst reading of 1 ± 0.05 can then be used to assess the flickermeter.

For complex waveforms, the target Pst reading must be obtained from a measurement using the NPL Reference Flickermeter. This target reading is based on assumptions about the pre-test methods that should be implemented in a flickermeter and the effect of these assumptions must be minimised. This is achieved by configuring the waveforms so that any modulation is begun 60 seconds after the beginning of a flicker test.

References

[1]    IEC 61000–4–15, Electromagnetic Compatibility (EMC) – Testing and measurement techniques – Flickermeter – Functional and design specifications, Published by The International Electrotechnical Commission, Amendment 1, 2003.