Lighting Tutorial [E-BOOK]

8.5 How to make magnetic ballasts more efficient than electronic ones

It is not so sure whether the undoubted, measurable rest of the efficiency improvement beyond the 4% difference of light output with electronic ballasts really bases on the high frequency – or perhaps rather on the current waveshape fed into the lamp? It was tried to find this out by means of another measurement at a special independent lighting institute. The idea behind this was another found statement that the efficiency of a fluorescent lamp is not optimal at rated current but better at lower current, as is the case with a lot of electrical equipment, incandescent lamps exempted. If this is valid for the TRMS or arithmetic mean value of same current, then it also goes for each and every instantaneous value along the curve. So, with sine current, efficiency drops within the range around the peak, since most of the light is generated during this time span. If the output current of an electronic ballast were rectangular, then there would be no efficiency drop at any point of the curve – and energy efficiency would be better, because this constant value would be considerably lower than the peak value of a sine wave. Indeed the output current looks more like a rectangle than like a sinus (Fig. 8.8).

Fig. 8.8: Output current curve of an electronic ballast (H.-G. Hergesell, Paderborn Airport), recorded with 3 different power analyzers

Fig. 8.9: Test samples for the measurements documented in Fig. 8.10 and Table 8.7

If this is so, then it should be possible to achieve the same efficiency improvement by lowering the overall current. The generalizing conclusion may be justified that higher power intensity is bad for efficiency.

Fig. 8.10: Efficiency of various ballasts with the same lamp at varying voltages according to Table 8.7

The values in the Directive refer only to rated power, but what happens at reduced power, e. g. when a lamp with magnetic ballast is fed only with the power rated for operation with an electronic ballast (Table 8.1) or even substantially less? To find out, 5 different ballasts for a 58 W lamp were taken under test (Fig. 8.9):

  • One stone-old ballast from an installation that had already been knocked down in 1987, still being rated 220 V and of course not efficiency classified and thereby falling into class D according to Table 8.1.
  • One new »superslim« magnetic ballast, inevitably falling into class C, since in electrical engineering restrictions of space nearly always come at the price of restricted efficiencies.
  • One new magnetic ballast efficiency class B2.
  • One new magnetic ballast efficiency class B1.
  • One mint condition electronic ballast rated efficiency class A3.

Now on each of these 5 samples all required parameters were measured, always using the same lamp: Active and reactive power across the whole system, active power (loss) across the ballast, and of course the light output of the lamp. All of the results have been compiled in Table 8.7) but, the graphic evaluation of this verbose table included in the download version of this text. Among others the 4% difference of luminous density in favour of magnetic ballasts (»5000 lm at rated voltage versus »4700 lm with electronic ballast) finds its confirmation here, but beyond this, the graphic evaluation of this verbose table provides much more ease of interpretation (Fig. 8.10). Unfortunately, on account of the high output frequency at the terminals of the electronic ballast, it was not possible to measure its output power. This is not a tragedy, though, since the most important data, system input power and light output, could be measured. The following can be concluded from the results:


Table 8.7: Measurements on 5 different ballasts at different line voltages with the same lamp

  • On the electronic ballast neither system input power nor light output vary with varying voltage. So the device under test fully compensates variances of the supply voltage within the tested range, which is usually seen as an advantage – and one commonly expected from electronic ballasts. A deliberate variation of power input and thereby of light output via the feeding voltage, however, is therefore not feasible.
  • Of course the energy efficiency comparison turns out best for the electronic ballast at 230 V, but at 200 V the A3 electronic one is only more about the same as the class B1 and even the class B2 magnetic ballasts, and at 190 V the electronic one performs poorer! So at 190 V supply voltage the B1 and even the B2 turns out as an A2, and the B1 should even be classified as A2, since the efficiency of the A3 model has not altered, but while those of both the B1 and the B2 and B1 models have exceeded it!
  • The information of the light output with electronic ballasts being about 4% reduced against that of efficient magnetic ballasts at rated input voltage (not necessarily rated input power – see next bullet point) finds its confirmation.
  • The rated lamp power is not always reached precisely at rated voltage. Other than the old ballast, the later magnetic ballast models of all classes reach their rated power only considerably above the rated system voltage. At 230 V, however, the electric lamp input power still falls considerably below the 58 W rating. After all that has been said so far, such design, e. g. deliberate utilisation of the permitted minus tolerance, must be seen as a reasonable approach.
  • Still, this does not yet mean that the electric values are now totally comparable to those of an electronic ballast! With classes C, B2 and B1, the light output is around 5000 lm, while the electronic ballast tested here provided only 4720 lm.
  • So the improved magnetic ballast models under test only feed about 53.5 W into the lamp instead of the rated 58 W, and still, the lamp shines 4% brighter than with the electronic ballast! Hence, for reasons of objectivity, in order not to compare apples with pares, the electronic ballasts’ light output at 230 V would rather need to be compared to those values metered on the improved magnetic models at about 220 V actual voltage.
  • At this point of operation the actual lamp inputs were only more around 50 W – matching the rating given for an electronic ballast. This makes the deviating lamp ratings for operation with magnetic versus electronic ballast operation appear relative and raises doubts about the improvement of efficiency at high frequencies. The confinement to this statement is the lack of measured electric output power at the electronic ballast. However, the systems’ power intakes with electronic A3 and magnetic B1 ballasts at the points of equal light outputs deviated from each other only more by exactly 2.1 W, interpolating between the two measured points at 220 V (4662 lm) and 230 V (4952 lm) to the 4720 lm the lamp performs with electronic ballast.
  • By switching from a poor class C magnetic ballast to a class B1 model the efficiency at rated lamp power is improved by 10% from 70.3 lm/W to 77,4 lm/W, since the share of ballast losses among the total input power drops from 22.9% to 15.0%. The price premium for the more efficient magnetic ballast therefore pays off in nearly all applications, short payback periods guaranteed.
  • Contrary to this, the persistent use of very old poor efficiency ballasts – especially if still designed for 220 V line voltage rating – leads to a significant lamp overload with highly over-proportional increase of losses and reduced lamp life but only little increase of light output.
  • By reducing the operating voltage from 230 V to 190 V, the efficiency e. g. of a lamp with a class C ballast is improved from 73.0 lm/W to 84.1 lm/W, that is by well over 15%. When a class B1 ballast is used, the light efficiency still rises from 80.6 lm/W to 89.1 lm/W and hence still by about 10.6%. So the reduction of the feeding voltage also pays off, especially in cases where poor magnetic ballasts are not replaced with better ones. However, this shall not be an excuse for further operating »old scrap« any longer, for also with high-efficiency magnetic ballasts the fairly simple and usually rather inexpensive voltage reduction technique provides pretty short payback periods. The upgrade from anything to a B1 ballast really is the bargain, and some greater or smaller voltage reduction may come on top of it as a perfection.

The high variance of efficiency even with moderate voltage reduction on a lamp circuit with whatever type of magnetic ballasts has three main reasons:

  • Copper loss and approximately also iron loss in the ballast rise by the square of the current. Therefore the power lost in the ballast drops overproportionally when current is reduced (see Table 8.7).
  • Lamp voltage increases when lamp current decreases (Fig. 1). Therefore lamp power decreases underproportionally with decreasing supply voltage, while lamp efficiency moderately increases and simultaneously ballast losses dramatically drop.
  • On account of this, current drops overproportionally to the voltage reduction and accelerates the former effects.

To offset the lower absolute light output, about 150 magnetic ballast luminaires operated at 190 V would have to be used to replace 100 electronic ballast luminaires. Now since the 150 magnetic ballast luminaires are simultaneously the more energy efficient solution, a cost premium would be acceptable in replacing the electronic with magnetic ballasts in order to save energy, inverting the usual approach. Still, this need not necessarily be any more expensive. Cases have been reported where the solution with 100 electronic ballasts has been bid higher. So the payback time may assume a negative value! Adding the cost for voltage reduction, it is still very short. In two example cases from Switzerland 50 open longitudinal 58 W luminaires were bid alternatively with electronic ballasts at 2575 SFr and 50 commensurate luminaires with magnetic ballasts, regardless of efficiency class, at 1700 CHF. So no premium was charged at all for a better efficiency class of the magnetic ballast, but it was very well possible to get 150 lamps equipped with these at a lower price than 100 luminaires with electronic ballasts. Whether the price premium in such cases really improves the electrical contractors’ businesses or whether the electronic ballast merely adds to the turnover but cuts revenues is yet another question to be critically scrutinized in each individual case.

In May 2000, being informed about this, the EU made an amendment to their document that any other measure judged appropriate to improve the inherent energy efficiency of ballasts and to encourage the use of energy-saving lighting control systems should be considered.

Indeed, in Germany there are at least three manufacturers of dedicated voltage reduction plant that is meant to operate fluorescent lighting at reduced voltages. Refurbishment in existing installations is easy as long as dedicated power lines for the lighting have been installed. Occasionally voltage reducers are also offered for the general supply but these have to be treated with care. Many power consuming devices have the inverse behaviour as fluorescent lamps with magnetic ballasts. Incandescent lamps, whenever living a lot longer, yield a dramatic loss of energy efficiency. Induction motors as well as practically all electronic devices, including decent electronic ballasts with constant light output regulation, have an increased instead of decreased current intake with reduced line voltage. Ohmic losses in the mains and especially inside the motor increase instead of decreasing. Also electronic ballasts of the type tested here, with constant regulated light output, react in this way and therefore cannot be influenced by varying the line voltage. With fluorescent lighting, however, the loss of luminous density can be offset by installing additional lamps – or simply taken for granted, which often is acceptable.

On the other hand, the undervoltage extends the lamp life by about 33% ... 50%, the voltage reduction plant manufacturers claim. The trade association of German lamp manufacturers points out that also the opposite can happen because the optimum filament temperature is not reached. So far it can only be concluded from the conflicting statements that this issue has not yet been experimentally investigated. Life time tests of longlife devices take a long time by definition.

Log in to post comments

Follow us