Possible disturbances with electronic ballasts

The former of these two styles of drawing electric power, the direct rectification of the incoming AC, generates extreme periodic current peaks somewhere in the proximity of the voltage maximum, while during the rest of each semi wave no current flows at all (Fig. 5.1). This current waveform includes a high harmonic content, especially of the third and its multiples, which add up on the neutral instead of cancelling out and cause a bunch of problems that have recently been analysed and described in detail in various sources: Neutral overload, transformer overheating, substantial distortion of voltage waveforms if network impedances are high, and in TN-C resp. TN-C-S systems these permanent operating currents on the neutral also intrude into all earthed metalwork, including the screens of data lines. There they can cause an additional bunch of problems such as magnetic stray fields, corrosion of pipework and earthing electrodes and especially malfunction and damage of IT equipment.

Fig. 6.1: Comparison of CFL without PFC (left) to electronic ballast with PFC (right)
Fig. 6.1: Comparison of CFL without PFC (left) to electronic ballast with PFC (right)

Emission of disturbances from electronic ballasts

Fig. 6.2: 3 electronic ballasts of old design operated on 3 phases
Fig. 6.2: 3 electronic ballasts of old design operated on 3 phases

While these harmonic currents in modern office buildings originate from the multitude of PCs, their screens and peripherals, electronic ballasts below 25 W including CFLs, because of their limited use, contribute only a smaller fraction to this problem. However, operating all fluorescent lighting following this simple principle would be virtually impossible, for which reason the upgraded electronic ballast technique with electronic power factor correction (PFC, Fig. 5.2) was developed. One source says about 30% to 50% of the price for an electronic ballast is spent on avoiding disturbances. Most of this obviously goes into PFC – quite successfully, as a comparison shows (Fig. 6.1): The input current of a CFL without PFC, rated only 11 W, has approximately the same crest value as that of a ballast rated 58 W with PFC. The total harmonic distortion of the currents is 80% in the former case, but barely 19% in the latter. Although less than 12% were measured with a magnetic ballast, this value is low enough not to encounter any harmonics related problems.

Fig. 6.3: Resulting neutral conductor current of phase loads as in Fig. 6.2
Fig. 6.3: Resulting neutral conductor current of phase loads as in Fig. 6.2

This, however, gives rise to another type of disturbances. As the pulse width modulation on the input side »;chops« the incoming current into many »;thin slices«, this is equivalent with releasing a high frequency current into the network, which is largely attenuated, but not completely extirpated by a capacitive filter on the input side of each electronic ballast (Fig. 6.4). So the possibility of conducted as well as transmitted disturbances remains. It has happened, for instance, that the frequency was 77 kHz, equal to that of the Frankfurt long wave transmitter which broadcasts the time signal of the Braunschweig atomic clock. The interference caused radio-controlled clocks to malfunction inside buildings equipped with these ballasts. Typically these disturbances occur at two different frequencies, for obviously the HF transformer for generating the lamp current and the electronic power factor correction work at different clock frequencies (Fig. 7.5): The former is responsible for the radiated and the latter for the conducted disturbances. Moreover, this high frequency, since it is not sinusoidal, for itself consists of a theoretically infinite spectrum of harmonics, so that the highest frequencies occurring nearly reach right up into the megahertz range. In the meantime standards have been released to restrict the maximum permissible levels of such disturbances. Unfortunately the tests according to these standards are carried out individually in a lab on one sample of the ballast in question, while in the field some hundreds or even a few thousands of those are operated on one site, so the disturbance levels to some extend add up. Adding to this, there is a frequency gap in the standards, leaving a certain range of frequencies without any limitations. Witty engineers now design their appliances in a way as to displace all disturbances into this blank, just as if only standards did matter and disturbances did not. Lots of interferences have so far been reported informally but on account of the special market structure they never ever appear in print.

Fig. 6.4: Input current of an electronic ballast at different time resolutions
Fig. 6.4: Input current of an electronic ballast at different time resolutions

Inspectors and official surveyors repeatedly report about an oscillation of the voltage amplitude in installations where there is a large coverage of electronic ballasts. At the feeding point the same can be observed with the current but with opposite phase, so this current variance must be the cause for the voltage variance. The inspectors speak of frequencies up to 3 Hz but usually only 0.3 Hz or often even a lot less than that, one period per 30 seconds is typical. They see a coherence with the usually capacitive power factors they find in these installations, while this cannot really be the cause. Truly electronic ballasts usually have a slightly capacitive power factor (Fig. 6.1), and truly installations are usually not metered or monitored, so nobody realizes the power factor correction is no longer a correction but the opposite of that and should be switched off or stepped down, but an oscillation at such a low frequency would require tremendous lots of both capacitance and inductance. Rather, the automatic output power control of the ballasts may be the cause: When there is a voltage sag for some reason, the input current into the ballast must be increased to keep the output power stable, and if the share of the total power that goes into such lighting equipment is high enough, this will increase the sag palpably. The voltage will continue to drop, and the overall current will go on rising until the input current increase of the ballasts is offset by a decrease of the input currents to some loads where it decreases as input voltage decreases. This is even indicated on the rating plates of electronic ballasts (Fig. 6.6). Now the process is inversed, and a voltage swell starts. The surveyors say the problem is usually solved by replacing failing electronic ballasts with magnetic ones without adding any compensation capacitance. When the share of magnetics reaches about 1/3 not only the electronic ballast failures stop but also the voltage oscillation ceases. So they think the shift of the power factor slightly into the inductive range was the solution, while the true explanation is probably that the behaviour of lamps with magnetic ballasts is inverse to that of electronic ones: Input current, both the active and the reactive share, drop over-proportionally as input voltage decreases. A linear drop might not suffice as an offset to stop the oscillation.

A high frequency expert reported he had tested some electronic ballasts and found out that their HF emission frequency also varies. It periodically hops to and fro between at least 2 frequency bands, obviously deliberately, by design. The background is probably that the relevant standard allows a certain amount of radiated energy at a certain frequency band, integrated over a defined period of time. So this standard is dodged by dispersing the disturbance across a wider range of frequencies. Unfortunately the expert was not able to say which standard it is that defines these values and procedures.

In another case the surge diverters in a brand new supermarket kept on failing. The whole market was equipped with electronic ballasts and the feeding lines with a properly designed overvoltage protection, comprising coarse, medium and fine protection downstream. However, the protective devices at the last stage, the fine protection, continuously failed, looking charred after failure, without any tripping of the coarse and medium stages. So the protection would have to be built up the other way round, coarse indoors and the fine stage upstream, since the disturbance came from inside the installation in this case.

Fig. 6.5: Frequency spectrum of the same electronic ballast
Fig. 6.5: Frequency spectrum of the same electronic ballast

A high frequency expert reported he had tested some electronic ballasts and found out that their HF emission frequency also varies. It periodically hops to and fro between at least 2 frequency bands, obviously deliberately, by design. The background is probably that the relevant standard allows a certain amount of radiated energy at a certain frequency band, integrated over a defined period of time. So this standard is dodged by dispersing the disturbance across a wider range of frequencies. Unfortunately the expert was not able to say which standard it is that defines these values and procedures.

Fig. 6.6: An advantage of electronic ballasts: Offset of voltage variances. Disadvantage of this: Current intake increases as voltage sags
Fig. 6.6: An advantage of electronic ballasts: Offset of voltage variances. Disadvantage of this: Current intake increases as voltage sags

In another case the surge diverters in a brand new supermarket kept on failing. The whole market was equipped with electronic ballasts and the feeding lines with a properly designed overvoltage protection, comprising coarse, medium and fine protection downstream. However, the protective devices at the last stage, the fine protection, continuously failed, looking charred after failure, without any tripping of the coarse and medium stages. So the protection would have to be built up the other way round, coarse indoors and the fine stage upstream, since the disturbance came from inside the installation in this case.

Fig. 6.7: Electronic ballast failures at Swiss Federal Institute of Technology Zurich within one year
Fig. 6.7: Electronic ballast failures at Swiss Federal Institute of Technology Zurich within one year

Susceptibility of electronic ballasts to disturbances

Figure 6.8:	Always the same type of failure: The filtering capacitor was overloaded …
Figure 6.8: Always the same type of failure: The filtering capacitor was overloaded …

The same goes for the vulnerability of electronic ballasts. It is frequently reported that under certain conditions they keep on failing (Fig. 6.7), while nobody is able to identify exactly which these conditions are. And again, there is an implicit vow of silence spelt over the affair. In one case, for instance, a major electrical contractor received a complaint from a customer where among a large number of newly installed electronic ballasts a substantial share malfunctioned right from the instance of installation. The contractor replaced the failed devices and passed the complaint on to the supplier, one of the European market leaders in lighting equipment. He got a letter back saying, in polite wording, an initial failure rate of 17% was absolutely normal for electronic ballasts. The electrician told this to his customer, who requested a copy of that letter but which was declined.

Only at Paderborn-Lippstadt airport, a small but rapidly growing regional airport in Germany, two cases could be documented:

  • Out of ≈80 electronic ballasts no less than 30 had failed within 4 months in one part of the installation. The same luminaires with the same type of ballasts, same manufacturer and even same batch, work without a single problem in an adjacent part of the installation being fed from a different subdistribution but from the same transformer. No indication of the reasons for these failures have been found so far, except that from the branch with the faults exclusively this lighting arrangement was fed, while the other one also fed some other loads. This would mean that the ballasts kill each other, unless other loads absorb their litter, and provides further scope for speculation about the causes, but still no evidence.
  • About half a year later the same problem occurred in another location of said airport, but with different ballasts from a different manufacturer.

 

Since the failing ballasts are now being replaced by high-quality magnetic ones, the failures have come to a halt. This provides further scope for speculation about the causes, but still no evidence.

At Kaufbeuren hospital about 480 luminaires were integrated into the ceiling, each fitted with 2 fluorescent lamps, rated 2*13 W, with 1 common electronic ballast. By end of 2004, some 800 lamps had to be replaced. The filaments had blown. After long vain efforts to find out about the causes, the electrician in charge found a coherence with the relatively long lines in the installation: On account of some very fast voltage fluctuation the electronic ballasts switched over to pre-heat mode. In the lab it was possible to reproduce this effect with a 50 m long line and a drilling machine, whereas it did not have to be a drilling machine but any other electronic device with a filtering capacitor at the input side did the »;job«. It need not even be set into operation, just connecting it was enough to produce an extremely short (few microseconds) but very steep current rise time edge with an according voltage dip. The ballast misinterpreted this dip as an instance of switch-off and switch-on again and started to heat the filaments, waiting for the lamp current to rise as a signal of successful start, to shut off the heating current. But the lamp current did not rise because the lamp was already in operation, so the pre-heat current remained on and overloaded the filaments.

Another case occurred so to say right in place with a fluorescent lamp manufacturer at the final test of the production line for T5 lamps rated 80 W. The lamps are tested individually, so the test rack tests 1 piece every 6 seconds. Now the electronic ballasts installed in the test equipment did not bear this frequent switching and kept on failing, this making production stall each and every time it happened, along with all the cost impacts this brings about. But unfortunately T5 lamps cannot be operated with magnetic ballasts. Why can they not? With the 80 W lamp it does not work because the required lamp operating voltage is too high. At least as long as the applied voltage equals 230 V it is not possible but in commercial areas there is always a second supply level of 400 V available. At present a 400 V magnetic ballast is being developed with one of the ballast manufacturers. A prototype was exhibited at the 2004 Frankfurt Light & Building fair and is now being used in the shipment test procedure of said lamp manufacturer. Note that this implicitly means this manufacturer specifies its T5 lamps as fit for 50 Hz operation, since final test is carried out exclusively in this manner! The required 400 V electronic starter has already been made available [7] and is now being used in the test line – under the tough conditions of permanent response requirement, but without failure!

From another site it was reported the cause for permanent electronic ballast failures in a large hall had been searched for approximately two years until it was found out that they were due to mechanical oscillations. Fork lifters ran into and out of the hall all day long, and each time an automatic swinging door caused an air pressure wave that made the ceiling swing. Certain electronic components on the PCBs in the ballasts could not bear this and came loose.

Strangely enough, none of such failures have been reported so far about CFLs, although they employ the same working principle except for the electronic PFC. This may be because they are not used in such large quantities within a constrained area. It is more likely, however, that the PFC electronics is the main source of failures in electronic ballasts, since it needs to be located right at the input side of the inverter, where it is exposed to all surges and other disturbances coming in from the network.

Of course there is no alternative to the use of electronic ballasts wherever one and the same lamp is to be used on various voltages and frequencies or on DC. On many railway vehicles, for instance, lighting can reasonably be fed on DC only, as the vehicle is fed on DC or 16.7 Hz. Since the DC feeding makes the active power factor correction in the ballast superfluous, no mass failures have been reported so far, which again confirms that the PFC is the weak point. The older German »;InterRegio« railway carriages may be counted as an exception, where quite obviously the ceiling lamps, which can be switched individually by travellers, are operated with magnetic ballasts and conventional starters, as can be concluded from the well-known flicker during start. This means that a dedicated power system is created inside the carriage, fed by an inverter converting either the 16.7 Hz power from the locomotive or the 24 V DC supply of the carriage into 50 Hz, since using the 16.7 Hz would end up not only with a ballast of triple volume and weight, which would be a serious issue on a vehicle, but also with a stroboscope light. It is reported that this was done because typical disturbances on a train, such as pantograph sparking, had caused failures of electronic ballasts, but obviously this problem has been overcome, and today's trains use electronic ballasts (but those without the dispensable electronic PFC) without causing any major trouble.

As for the voltage dependency or independency of the light output, one company in Germany carried out a test among various electronic ballasts and CFLs, an incandescent lamp (for comparison) and halogen lamps with electronic and conventional transformers. Surprisingly enough, just one type of electronic ballast from each of the three leading manufacturers performed a complete compensation of input voltage variance (constant light output). It may be speculated that these three were the top models of the three brands. Some of the CFLs at least managed to come from a square relationship between voltage and power, as for resistive loads, down to a linear behaviour.

Figure 6.9: … by a strong HF current due to a minor HF voltage superimposed upon the line voltage
Figure 6.9: … by a strong HF current due to a minor HF voltage superimposed upon the line voltage

At Kaufbeuren hospital about 480 luminaires were integrated into the ceiling, each fitted with 2 fluorescent lamps, rated 2*13 W, with 1 common electronic ballast. By end of 2004, some 800 lamps had to be replaced. The filaments had blown. After long vain efforts to find out about the causes, the electrician in charge found a coherence with the relatively long lines in the installation: On account of some very fast voltage fluctuation the electronic ballasts switched over to pre-heat mode. In the lab it was possible to reproduce this effect with a 50 m long line and a drilling machine, whereas it did not have to be a drilling machine but any other electronic device with a filtering capacitor at the input side did the »;job«. It need not even be set into operation, just connecting it was enough to produce an extremely short (few microseconds) but very steep current rise time edge with an according voltage dip. The ballast misinterpreted this dip as an instance of switch-off and switch-on again and started to heat the filaments, waiting for the lamp current to rise as a signal of successful start, to shut off the heating current. But the lamp current did not rise because the lamp was already in operation, so the pre-heat current remained on and overloaded the filaments.

Another case occurred so to say right in place with a fluorescent lamp manufacturer at the final test of the production line for T5 lamps rated 80 W. The lamps are tested individually, so the test rack tests 1 piece every 6 seconds. Now the electronic ballasts installed in the test equipment did not bear this frequent switching and kept on failing, this making production stall each and every time it happened, along with all the cost impacts this brings about. But unfortunately T5 lamps cannot be operated with magnetic ballasts. Why can they not? With the 80 W lamp it does not work because the required lamp operating voltage is too high. At least as long as the applied voltage equals 230 V it is not possible but in commercial areas there is always a second supply level of 400 V available. At present a 400 V magnetic ballast is being developed with one of the ballast manufacturers. A prototype was exhibited at the 2004 Frankfurt Light & Building fair and is now being used in the shipment test procedure of said lamp manufacturer. Note that this implicitly means this manufacturer specifies its T5 lamps as fit for 50 Hz operation, since final test is carried out exclusively in this manner! The required 400 V electronic starter has already been made available [7] and is now being used in the test line – under the tough conditions of permanent response requirement, but without failure!

From another site it was reported the cause for permanent electronic ballast failures in a large hall had been searched for approximately two years until it was found out that they were due to mechanical oscillations. Fork lifters ran into and out of the hall all day long, and each time an automatic swinging door caused an air pressure wave that made the ceiling swing. Certain electronic components on the PCBs in the ballasts could not bear this and came loose.

Strangely enough, none of such failures have been reported so far about CFLs, although they employ the same working principle except for the electronic PFC. This may be because they are not used in such large quantities within a constrained area. It is more likely, however, that the PFC electronics is the main source of failures in electronic ballasts, since it needs to be located right at the input side of the inverter, where it is exposed to all surges and other disturbances coming in from the network.

Of course there is no alternative to the use of electronic ballasts wherever one and the same lamp is to be used on various voltages and frequencies or on DC. On many railway vehicles, for instance, lighting can reasonably be fed on DC only, as the vehicle is fed on DC or 16.7 Hz. Since the DC feeding makes the active power factor correction in the ballast superfluous, no mass failures have been reported so far, which again confirms that the PFC is the weak point. The older German »;InterRegio« railway carriages may be counted as an exception, where quite obviously the ceiling lamps, which can be switched individually by travellers, are operated with magnetic ballasts and conventional starters, as can be concluded from the well-known flicker during start. This means that a dedicated power system is created inside the carriage, fed by an inverter converting either the 16.7 Hz power from the locomotive or the 24 V DC supply of the carriage into 50 Hz, since using the 16.7 Hz would end up not only with a ballast of triple volume and weight, which would be a serious issue on a vehicle, but also with a stroboscope light. It is reported that this was done because typical disturbances on a train, such as pantograph sparking, had caused failures of electronic ballasts, but obviously this problem has been overcome, and today's trains use electronic ballasts (but those without the dispensable electronic PFC) without causing any major trouble.

As for the voltage dependency or independency of the light output, one company in Germany carried out a test among various electronic ballasts and CFLs, an incandescent lamp (for comparison) and halogen lamps with electronic and conventional transformers. Surprisingly enough, just one type of electronic ballast from each of the three leading manufacturers performed a complete compensation of input voltage variance (constant light output). It may be speculated that these three were the top models of the three brands. Some of the CFLs at least managed to come from a square relationship between voltage and power, as for resistive loads, down to a linear behaviour.

To top

Reliability of electronic ballasts

There is little quantitative evidence for the reliability of electronic ballasts. One statement speaks of a failure rate below 2% per 1000 hours of operation. That sounds quite nice, but for an average supermarket with 3000 h/a of operation this amounts to 6% dropouts per year. Under constant duty, like in a subway, it already means replacing more than 1 per every 6 ballasts annually. Considering this, it seems slightly strange to find this figure in a publication speaking very much in favour of electronic ballasts.

It may be rather unspectacular if the manufacturers of certain dedicated plant to be discussed further below do not state a single word in favour of electronic ballasts, since their products are applicable to magnetic ballasts only. But it is very well worth considering why official surveyors, inspectors and site electricians have serious qualms with the use of electronic ballasts. The use of electronic ballasts is, from today's viewpoint, inevitable if special high-end control functions including dimming are required, for as mentioned, dimming techniques for magnetic ballasts do no longer match today's ideas of functionality and comfort, such as in conference centres. Yet, for the common »;area lighting« in warehouses, supermarkets, ordinary offices, subways, schools, industry, especially in EMC sensitive environments or under extreme temperatures or vibrations, the best EEI class (see section 8 below) of magnetic ballasts will be the optimum choice. Their failure rates are next to zero in nearly all environments, as long as indicated maximum ambient temperatures are not substantially exceeded, while where a lot of electronics is integrated a lot can fail. Just like an instance of »;the irony of destiny«, a severe power quality problem occurred during a power quality conference in a large modern conference building in Brussels. Sophisticated electronic lighting control got out of control and turned off the lighting every other minute. The conference centre management felt quite embarrassed and compensated the loss of usability to their client with a 50% price reduction. This financial loss may equal the electricity consumption of 1000 conferences and the energy savings achievable with high-tech lighting, if working properly, of at least 4000 conferences. It is evident that energy saving is not the prevalent reason for installing such technique in a conference room. It is the opportunity to provide optimal lighting for virtually everything one might want to do in a conference room. Even so, the loss of reputation caused by such embarrassing occurrence is probably a lot worse than through providing a less sophisticated, less versatile, less impressive technique but which just functions.

An advantage at least of many electronic ballasts is that they function with any frequency including DC. This cannot be expected from a magnetic ballast. Just by coincidence, a European ballast rated 58 W, 230 V, 50 Hz would do its job just as fine in an American 277 V 60 Hz office environment, but that is sure pure fluke in this individual case.