A research report  by Francis MASSEN ,  Jean MOOTZ  and Claude BAUMANN 

francis.massen@education.lu

jean.mootz@education.lu

claude.baumann@education.lu

 version 1.03  09 Sep 2009

Abstract:

Using a CFL instead of a traditional incandescent bulb makes some heavy changes to the usual harmonic flow of current and to the magnitudes of apparent and reactive power. We show this in a comparison of 3 lamps: a 25 W incandescent and two CFL's rated 21W and 25W.

1.  Mean, apparent and reactive power.

In AC circuit, current and voltage are not always in sync. In the case of an electric AC motor having many inductive copper windings, there is a time lag between the flow of the current through the motor and that of the voltage measured on the connections of the device. In the case of an inductive load L, current is late in respect to voltage; if the load is capacitive,  voltage is late on current. Both parameters still vary as sinus functions of time; the time-lag  dt can be expressed in time units (ms for instance) or as an angle phi, with phi = (2*pi/T)*dt in radians. Commonly one speaks of current and voltage being out of phase.

fig. 1: The voltage signal crosses the vertical zero axis in advance to the current: the  current is late on voltage.
The phase-difference here is 1 rad; as the period of the signal is 20ms (corresponding to 2*pi radians), the time-lag between voltage and current 
is 1/(2*pi)*20 ~ 3.2 ms

The effective electrical power used by a device is defined as P = U*I*cos(phi), with P measured in Watt [W], U = effective voltage measured in Volt [V] and I = effective current measured in Ampere [A].

The effective voltage or current is that measured by an usual cheap multimeter or a traditional mechanical instrument; actually U = Û/sqrt(2) and I = Î/sqrt(2) when voltage and current are strictly sinusoidal. The "peak" voltage of a 230 VAC lamp is Û =230*sqrt(2) = 325 V. 

We may represent voltage and current by vectors (in a so-called Fresnel diagram); for the power this gives the following 2D picture:

fig.2: The different power types in a AC circuit (English and German expressions). Commonly abbreviations used:
          S = apparent power, Q = reactive power,  P = effective power

A client driving an electrical motor consumes (and pays) the amount of effective power used ; the power station producing the electrical power has to produce the larger amount of the apparent power. The "oscillating" reactive power Q (in German: Pendelleistung) represents an excessive power that has to be produced and put into the grid. The supplementary costs correspond to the transportation losses ( approx. 8% [5]) of the reactive power. As a consequence: a home user who pays only his effective power but generates a big phi (as will be seen later) is a nuisance for the power station operator. Actually, large industrial clients must pay steeper tariffs when they are unable to bring back their phi to decent small values. "Normal" clients do not bother (for the moment), as the operators assume that for an usual household phi will be very small, and do not monitor it (traditionally one uses the cos(phi) as a convenient expression for the phase difference).

If voltage and current are not sinusoidal anymore, everything changes! Run of the mill multimeters give faulty readings, and the 2D power diagram becomes a much more complicated 3D picture. We will analyze this in the following chapters, using lamps bought from the shelf.

2. Experimental setup

The following picture shows the experimental setup:

fig.3: Experimental setup

Power is delivered through a power supply having a security transformer (output has no direct connection to the grid). The Hameg HM407-2 oscilloscope is used to visualize both current and voltage (current is represented as the voltage drop on a 1 Ohm resistor); it has a digital storage function, very convenient to freeze the screen before taking a picture with a digital camera. The spectrum analyzer is a Pasco [1] Scientific Workshop 750 interface with FFT software running on the laptop; the small black box is the current sensor feeding the interface. The watt & cos(phi)-meter is a Siemens B4305 instrument.

We use 3 lamps of similar electrical power ratings, so that the comparison of electrical properties makes sense. The luminous flow of the 25W ACEC bulb is evidently much lower than that of the CFL's.

4. The incandescent lamp

 The lamp used is a rather old ACEC 25W bulb. This resistive load should not create a noticeable phase difference, as voltage and current swing synchronously: 

fig.4: The ACEC 25W incandescent lamp

fig.5: Incandescent bulb: current and voltage are in sync.

Here are the measurement results:

The cos(phi) shown could be due to an instrument problem. The oscilloscope cearly points to a phi = 0 ( i.e. cos(phi) = 1)
As current flow remains sinusoidal, the ordinary and the RMS ampere-meters show the same result !

The Fourier spectrum has one single peak at 50 Hz (the frequency of the grid); the amplitude of the peak is 0.112 A,  the same reading as those shown by the other ampere-meters.

fig.6: FFT spectrum of the current through the Incandescent bulb. Amplitude in [A]

5. The Osram Dulux Superstar CFL lamp

This lamp is rated at 21W, 170mA and 220-240 VAC on the packaging. The fluorescent tubes are of the commonly found inverted U-form type.

fig.7: Osram Dulux Superstar 21W CFL

The flow of current is extremely different from a smooth sinus pattern: i(t) is still a periodic function, but shows very steep peaks with a characteristic bumb in the falling part:

fig.8: Osram CFL: current and voltage are not in sync. Current flow is extremely spiky! Voltage is late on current!

The current flow starts with a very rapid rise to the peak amplitude of about 0.38 A (vertical grid scale  is 100 m A), then falls down showing a slowing bumb during the last part of the impulse. The peak of i(t) precedes that of u(t) by about dt = 1.7ms what represents a phi = 1.7/20* 2*pi = 0.53 rad, and cos(phi) = 0.86. As u(t) lags on i(t), this CFL behaves like an capacitive load, due to the capacitors of the electronic driver circuit!
The time of peak of the sine-voltage has been found by using the middle of the 2 points where the voltage curve reaches its half-height amplitude.

Here are the readings of the different instruments:

The non RMS readings are a complete nonsense, whereas the RMS ampere-meter gives a result relatively close to the 170mA rating. 
The cos(phi) shown by the instrument is about one quarter of that computed from the oscilloscope. Thus the simple definitions valid for sinus-type functions can not be used anymore in this case. Actually cos(phi) as a measure of effective power must be replaced by a power factor PF computed in a different way.

The FFT of this lamp shows a large number of harmonics: only uneven ones are present (50Hz, 150Hz, 250Hz ...), in accordance to  the theory  of Fourier series.

fig.9: FFT spectrum of the Osram CFL with odd-numbered harmonics. 

The RMS intensity can be computed as I_RMS = sqrt(I₁² + I₂² + I₃² + ...) where the I_i are the amplitudes in Ampere of the peaks; the harmonics shown are approximately:

Sum of the I_i² = 0.0215; sum of the I_i² starting at I₂ = 0.0123

Square-root of these sums = 0.147 A and 0.111 A. Note that the first result is close to the I_RMS reading!

Total harmonic distortion THD

This parameter  is defined as THD = (1/I₁)*sqrt(I₂² + I₃² + I₄² + ...) the sum being extended up to the 40th harmonic at 2000Hz. 
An incandescent lamp without any higher harmonics has a THD = 0. 

Our measurements give for the 21W Osram Dulux a total harmonic distortion greater than THD = (1/0.096)* 0.111 = 1.15 or 115%. (as we used only 11 upper harmonics, the sum in the numerator is too small).

Power factor PF

The traditional cos(phi) must be replaced by a power-factor  PF defined as PF = (effective power P)/(apparent power S)  [3]

Effective power P

It can be shown [3] that the effective power is P = U*I₁* cos(phi) where cos(phi) is computed from the time-lag shown on the oscilloscope.  
Here P = 230*0.096* 0.86 = 19.0 W, close to the power given on the bulb and practically the same as the reading of the Siemens instrument. It should be noted that only the first harmonic contributes to the effective power!

Apparent power S

The apparent power is defined as S = U*sqrt(sum of the I_i²) = 230 *0.147 = 33.8 VA and the power factor PF replacing the cos(phi) becomes:

PF = P/S = 19/33.8 = 0.56 

Reactive power Q

The first type of reactive power  Q = U*I₁*sin(phi) = 230*0.096*0.51 = 11.3 VA (phi derived from oscilloscope); for non-linear devices it must be augmented by a second type of reactive power.

Current distortion reactive power D

The current distortion reactive power ("Strom-Verzerrungsblindleistung") D is defined as D = U*sqrt(I₂² + I₃² +I₄² + ...) 

Here D =  230*0.111 = 25.5 VA.

Both reactive powers Q and D travel far and forth through the grid, and the 8% losses must be accounted for. Total reactive power becomes sqrt(Q2+D2) = 28.5 VA

The 2D vector diagram valid for a sinusoidal current (and voltage) must be replaced by a 3D diagram when the current is non-sinusoidal:

fig.10: The different power types for non sinusoidal current. S² = P² + Q² + D². Total reactive power = sqrt(Q² + D²)

6. The  Brelight 25W twisted CFL

 Our second (cheaper!) CFL is of the twisted type, commonly found in the USA but nowadays also showing up in European shops. It is rated at 25W,  same as the ACEC bulb. Current is given as 190mA, very different from I = P/U = 109mA !

fig.11: Brelight 25W CFL, twisted type.

The oscilloscope gives a similar picture to that of the Osram CFL:

fig.12: Brelight CFL: current and voltage are not in sync. Current flow is visibly more distorted than that of the Osram CFL! Voltage is late on current.

The current flow has two discernable bumbs ; phase lag is capacitive. Phase lag phi = 1.8/20*2*pi = 0.57 and cos(phi) = 0.84 (to be compared to 0.86 of the Osram CFL)

Readings of the different instruments:

Again the non RMS measurements and the cos(phi) shown by the Siemens instrument are meaningless .

fig.13: The spectrum of the Brelight CFL is slightly different from that of the Osram CFL, mostly in the region of higher harmonics.

The harmonics have following amplitudes:

Sum of the I_i² = 0.0473; sum of the I_i² starting at I₂ = 0.0345

Square-root of these sums = 0.217 A and 0.186A. The first result is close to the rated current of 190mA !

The total harmonic distortion THD is at least THD = (1/0.141)* 0.186 = 1.32 or 132%, higher than that of the Osram CFL (and in agreement with the visibly more distorted oscillogram).

Effective power P = U*I₁*cos(phi)  = 230*0.141* 0.84 = 27.2 W,  close to the power of 25W given on the bulb, but distinctly higher than the power reading of the Siemens instrument. 

Apparent power S = U*sqrt(sum of the I_i²) = 230 *0.217 = 49.9 VA 

Power factor PF = P/S = 27.2/49.9  = 0.55,  lower than that of the Osram lamp. 

Reactive power  Q = U*I₁*sin(phi) = 230*0.141*0.54 = 17.5 VA 

Current distortion reactive power D = U*sqrt(I₂² + I₃² +I₄² + ...)  = 230*0.186 = 42.8 VA

Total reactive power sqrt(Q2+D2) = 46.2 VA, close to double of the Osram lamp.

7. Summary and conclusion

Let us resume our findings in one table:

Compared to the Osram CFL, the cheaper Brelight CFL (~6 Euro) has a visibly more distorted current flow;  total reactive power is much higher but both power factors are about the same.

Our research shows that to find the power factor (the most important characteristic!) the following measurements have to be made:

- voltage U measured by a traditional or an RMS voltmeter

- phase lag phi measured from the oscillogram of i(t) and u(t) (dual channel oscilloscope)

- apparent power and the two types of reactive powers calculated from the FFT spectrum, U and sin(phi)

- effective power calculated from U and cos(phi)

Our measurement results are close to the ratings printed on the lamps. 

The high  reactive power produced by the CFL's is a supplementary burden to the grid, inexistent when all lamps are of the incandescent type. This reactive power causes excess losses in the grid that must be accounted for.

As a general rule, the grid becomes "dirtier" with the number of CFL's and other non-linear consumers [4]. One annoying consequence is that in a 3-phase grid, the neutral line will possibly carry large to extreme large currents (caused by the 3rd, 9th and 15th triplen harmonics which are in phase and whose currents add up  [4][7]), and may blow if it's section is too small (in a well balanced 3 phase grid, the neutral line carries no or very little current!).
 

As an example let us compute the current in the neutral line  for the Brelight lamp, which would be normally connected between a phase-wire and the neutral-wire of the 3-phase grid. [7] gives the following approximate formula:

I_neutral = sqrt(I₃² + I₉²)

which amounts using the harmonics found above to I_neutral = sqrt(0.113² + 0.048²) = 0.368 A . Thus the  current in the neutral wire of the 3-phase grid is 0.368/0.217 = 1.7 times higher than the phase current. With many lamps in use and driven by the 3 phase lines, this might overload the neutral wire.

There are reports of such a neutral meltdown when a large block of buildings switched from incandescent lamps to CFL's  (reference).

 

The rather old  1997 NUTEK field test [6] done in 1 house switching 5 bulbs to CFL's showed a relatively small increase of 25% in total neutral current (but a remarkable 100% increase in the 3rd harmonic (fig.14); when the test was done in 17 houses with a total of approx. 100 CFL's installed the increase was found to be negligible.

fig. 14:  Changes in neutral current in a house replacing 5 incandescent bulbs with CFL's [7]

The European standard  EN 61000-3-2

Starting January 2001 this standard  defines the limits of maximal allowed harmonic currents for devices having a power > 75 W or lightning devices > 25 W. For lightning devices (class C) the first 4 limits are given in the following table [4], together with a check for conformity of the two CFL's (as all lamps have a rated power < =25 W, this check should be considered as an academic exercice!).

This short check shows that neither CFL would conform to the European standard if it was applicable to lamps with less than 25 W power ratings! 

REFERENCES

[1]     http://www.pasco.com

[2]    Wikipedia: Blindleistung

[3]    Blum J.: Abschied von der Glühlampe bringt höhere Netzbelastung. Elektropraktiker, Berlin 63 (2009) 6

[4]    Höck G.: Dirty Power. Oberschwingungen durch nichtlineare Verbraucher.

[5]    Mackay D.:    Sustainable Energy - without the hot air. UIT, Cambridge, England, 2009. (PDF)

[6]    GOTHELF N.:    Power quality effects on CFL's - a field study, 1997. (PDF)

[7]    AINTABLIAN H.:    Harmonic currents generated by personal computers. Dissertation, 1994. (PDF)

[8]    ELLIOTT R.:    Should there be a ban on incandescent lamps? (LINK)

An excellent free software package to study power, harmonics and distortion is Alex McEACHERN's  Power Quality Teaching Toy!

Thanks to Guy Schintgen for providing reference [3]

History:

last modified: 09 Sep 2009

go back