Proper compensation of reactive power with magnetic ballasts

Luminaires operated with magnetic ballasts cause a lot of inductive reactive power, much more than the share of active power normally is. The power factor (for a lamp together with its ballast under normal operating conditions) is always indicated on a ballast (Fig.4.1). In fact a luminaire with a lamp rated 58 W and a magnetic ballast has an overall active power intake between 64 W and 70 W, so with the 0.67 A current rating the apparent power is around 160 VA and the reactive component some 144 Var. So in the commercial and industrial sectors compensation becomes a must – which is old common practice and neither difficult nor expensive to realise.

Fig.4.1: The power factor is always indicated on a ballast
Fig.4.1: The power factor is always indicated on a ballast

General issues

Fig.4.2: An 11 W fluorescent lamp with magnetic ballast without compensation (top) and with parallel compensation (centre and bottom)
Fig.4.2: An 11 W fluorescent lamp with magnetic ballast without compensation (top) and with parallel compensation (centre and bottom)

The argument commonly forwarded for compensating is cost reduction, while in fact, as a rule, only prices are considered, the price the utility charges for reactive energy metered at the point of common coupling, not the cost the reactive current causes on its way from the device consuming (active) power to the PCC. Not (yet) so with lighting. As an exception, it is really common practice with ballasts to compensate the reactive power right in the place of origin, where this is most effectively done, say within the luminaire. This may happen in the usual way by paralleling the (approximately) ohmic-inductive load by a capacitance. However, the disadvantages or risks are as with any other static VAR compensator today:

  • Sound frequency signals in the mains, used for control of street lighting, night storage heating etc. may get lost.
  • Capacitive reactance drops proportionally as frequency rises, so capacitors may be overloaded since there are a lot of harmonics and other frequencies higher than the mains frequency rating superimposed upon the line voltage. At the top of Fig.5.1 the power intake of a small fluorescent lamp was recorded in an office environment without any compensation. The fundamental reactive power is really very high, with cosφ = 0.5 – while it nearly equals the load factor LF, which means that the current is approximately sinusoidal, as becomes obvious also from the graph. So compensation becomes a must, but a parallel capacitor adds a tremendous lot of distortion, say higher frequency constituents, to the overall current (centre of Fig.5.1). Although the capacitance is properly dimensioned, the reactive current cannot be brought to zero. When nothing in the wiring is changed but just the inverter driven elevator in the building starts to operate, the distortion and thereby the reading of reactive power once again increases substantially (bottom of Fig. 4.1). This provides evidence that indeed the additional current must consist of higher frequencies flowing through the capacitor.
Fig.4.3: Lead-lag compensation
Fig.4.3: Lead-lag compensation

Now, in static VAR compensators the usual approach to cope with these phenomena is detuning the capacitors, say connecting them in series with a reactance that at mains frequency compensates (takes away) only a few percent of the capacitor's reactive power rating. But why bother about an additional reactor with fluorescent lamps where a reactor is already there? Since current and phase angle with fluorescent lamps are practically invariable, there is another option, namely to use the ballast simultaneously for detuning a serial compensation capacitor (the so-called lead-lag connection, Fig.5.3). This means that every second lamp-and-ballast unit is (over-)compensated with a serial capacitor dimensioned – in theory – precisely in such a way as to make the current magnitude equal to that in an uncompensated lamp. The phase angle will then also be of the same absolute magnitude but with opposite sign.

Fig.4.4: Much better resilience to voltage variances with serial compensation
Fig.4.4: Much better resilience to voltage variances with serial compensation

So all the disadvantages of parallel compensation are avoided. Also the stroboscope effect is minimised through the phase shift between the leading and the lagging circuits usually installed within one luminaire. This is the reason why most luminaires come with 2 lamps. As a side effect, the compensated share of the lamps are much less sensitive to voltage variances and flicker (Fig.4.1) and entirely insensitive to possible direct voltages superimposed upon the feeding voltage, which otherwise, even if minimal in magnitude, may heavily affect inductive components.

Fig.4.5: Shares of the European ballast market
Fig.4.5: Shares of the European ballast market

The only disadvantage of this compensation principle is the risk to dimension the capacitor wrong. A bit of over- or under-compensation does not matter much in parallel, but in serial it means more than that! It means wrong lamp current, possibly lamp, capacitor and ballast overload or at least either higher loss level than necessary and premature failure or reduced light output. Therefore the tolerance rating of these capacitors is rather narrow, just 2%.

Care has to be taken with the selection of replacement, which should not be a problem, since the correct capacitance for serial compensation always used to be indicated on a magnetic ballast (Fig.4.1), but yet sometimes errors occur. Now that German lighting industry has decided to abandon serial compensation (instead of adapting the capacitance ratings to adequate values, which would be feasible without any risk, as both measurements and magnetic ballast experts confirm), the capacitance ratings on the rating plate (still to be found on the ballasts in Fig. 4.2 and Fig. 7.8) are now omitted.

Fig. 4.6: Correct dimensioning of serial compensation capacitance
Fig. 4.6: Correct dimensioning of serial compensation capacitance

Another disadvantage – not of the principle but in common practice – is that the currents with and without serial compensation are not really equal. The ratings differ depending on whether inductive or capacitive coupling is applied (Fig. 4.2). At the rated current of a 58 W lamp the inductance of a 230 V 50 Hz ballast turns out to be 878 mH. This requires a capacitance of 5.7 µF to end up with a resonance frequency of 70.7 Hz, at which theoretically the lamp current magnitude at 50 Hz would be equal with and without the serial capacitor. Yet, for some reason, possibly the extreme distortion of the voltage across the lamp (Fig. 2.17) or nonlinearity of the ballast, currents turn out unequal. As a standard, 5.3 µF or 5.2 µF are used (Fig. 4.1) but this still by far does not offset the difference.

 

 

Fig. 4.7: Serial compensation capacitance dimensioned 20% wrong: Lamp, ballast and capacitor current 45% too high!
Fig. 4.7: Serial compensation capacitance dimensioned 20% wrong: Lamp, ballast and capacitor current 45% too high!

A measurement (Fig. 4.1) shows that 4.6 µF would be the correct value but it is argued this could not be used in order to avoid starting problems with the lamps, especially in cases of undervoltage and extremely low temperatures. It has nothing to do with the principle as such, once the lamp has been fired successfully, and the firing problems could very well be overcome by the use of electronic starters, which are the better choice anyway (section 2.1). Moreover, the question is whether there is any reason to worry at all. Rather, a further test revealed that absolutely no starting difficulties are to be expected: 3 electronic starters as well as 2 very old worn-out glow starters were tested together with 2 different types of 58 W lamps, both from the same manufacturer but of different light colour, with a modern efficient magnetic 230 V ballast. Both the reduced 4.6 µF serial capacitance and reduced voltage were applied, and all combinations started without any problems at first attempt with only 180 V, with just two exceptions where successful firing occurred »;only« at 190 V.

So it seems a revision of capacitance ratings is due here but industry rather seems to be hoping to replace all magnetic ballasts with electronic ones in the long run and therefore appears not too ambitious to adapt any old standards to new technologies as long as either of these refer to magnetic ballasts. However, even if the impression roused among experts may cause a different feeling, approximately 70% of the market is still being held by magnetics (Fig. 4.5).

In some countries the ratio is even a lot more extreme (Spain 91% magnetic ones). At least in terms of sold pieces this is so. In terms of turnover figures the share is only more around 50%, due to the much higher added value. Or should we rather speak of higher added price in this case? Howsoever, it is understandable that the lamp and luminaire industry is much keener on the promotion of electronic ballasts.

For reasons of fairness, however, it also needs mentioning that electronic ballasts more often than magnetic ones provide the option of operating two lamps on one ballast.

To top

Special aspects when compensating small lamps

Fig. 4.8: One and the same ballast is designed for 4 different lamp types as well as for 3 tandem connections (only one of them listed here for reasons of space); the power factor increases substantially with the lamp power rating connected
Fig. 4.8: One and the same ballast is designed for 4 different lamp types as well as for 3 tandem connections (only one of them listed here for reasons of space); the power factor increases substantially with the lamp power rating connected

The operating voltage drop across smaller, i. e. shorter fluorescent lamps of the same type family is lower than with the longer types of the same series. Thereby a larger part of the voltage drops across the ballast, and this voltage drop is greatly – in the ideal case would be wholly – inductive. So on the one hand the smaller lamp has a lower active power intake, but on the other hand it has a higher reactive power dissipation. Commonly, these two effects lead to a substantially lower power factor for the lower lamp power rating. So the compensation investment increases inappropriately. This can be observed very clearly on TC-S lamps with 5 W, 7 W, 9 W and 11 W power rating, since these 4 models are all operated on the same ballast (Fig. 4.8).

Fig. 4.9: Power factor as the ratio of active power (grey benches) plotted against reactive power (blue benches)
Fig. 4.9: Power factor as the ratio of active power (grey benches) plotted against reactive power (blue benches)

However, the operating voltage drop across the TC-S lamps rated 5 W, 7 W and 9 W is so low that the common mains voltage of 230 V allows two of these lamps to be operated in series on one ballast. In effect, this doubles the operating voltage drop again, of course. Since the same ballast is used for this so-called tandem connection as for the single operation, the actual current when operated in tandem lies slightly below the lamp current rating – though not very much, since the inductive voltage drop still prevails. One of the advantages of this operating mode is that two lamps together use less reactive power than one of them already does in single mode (Fig. 4.9). But the tandem configuration may very well claim even more advantages than this (see section 7.3).