Compact Fluorescent Lamp Review

[lima]

Compact Fluorescent Lamp A long time friend, K8AQC, asked about the AC current and power characteristics of the compact fluorescent lamps (CFL) he uses. Revision History Originally written June 2008 revised 27 Sept 2008 Introduction and Economics I've made measurements of the AC current and power and also looked at the broadband and line spectrum noise coupled back into the AC line by a Sylvania 100 watt (equivalent) 23 watt actual screw-in CFL, model CF23EL/BL/1. This lamp appears typical of those sold as of mid 2008. I purchased this one at a Sears Hardware store for about $7.00. The packaging says the expected life is 8000 hours with an output of 1600 lumens. The packaging also says a typical 100 watt incandescent lamp has an output of 1680 lumens. The electricity and replacement lamp savings are stated as $61. (The price on CFLs continues to drop and with a bit of shopping around, you can find 60 watt equivalent lamps for $2 to $3 and 100 watt equivalent lamps for $5.) Before looking at the electrical and noise characteristics of this lamp, are the cost savings figures approximately correct? The industrial supply house I deal with, McMaster-Carr, sells a 1600 lumen 100 watt incandescent lamp, with an estimated 750 hour life, at $0.98 for a pack of two lamps. For a reasonable quantity purchase, and including shipping, these lamps would run about $0.70 each, delivered. In 8000 hours, therefore, our lamp capital costs are: Type Unit Price Lifetime Number for 8K Hours Cost for 8K Hours CFL $7.23 8000 Hours 1 $7.23 Incandescent $0.70 750 Hours 10.67 $7.47 Capital cost is essentially a wash. I'm ignoring the time cost of money in this simplified analysis, but that won't make too much of an error. (If you can find a 100 watt equivalent CFL lamps for $5.00 each, the capital cost shifts by $4.00 or so towards the CFL.) To look for the main savings, therefore, we must look at power costs, where the CFL consumes 77 watts less power. Over an 8,000 hour operating life, therefore, the CFL consumes 616 fewer kilowatt hours. Northern Virginia Electrical Coop (NOVEC) serves Clifton and has prices considerably below many other utilities--the consumption-based portion of our power bill in Clifton is 7.81 cents/kwh according to this month's bill. Hence, our power savings would be $48.10 over the CFL's lifetime, assuming no price increases and again neglecting the time value of money. Electrical power in other areas can be considerably higher. http://www.eia.doe.gov/bookshelf/brochures/rep/index.html provides state-by-state averages (2006 data), and shows Virginia at 8.49 cents/kwh. New York, in contrast, is nearly 17 cents/kwh. (Good for me as a small shareholder of ConEd, but not so good for New York residents) California averages 14.3 cents/kwh. The lowest priced state seems to be West Virginia at 6.35 cents/kwh, and New York is the most expensive. The 2006 national average is 10.4 cents/kwh. State 2006 Price (cents per kwh) CCFL Savings West Virginia (lowest) 6.35 $39.11 NOVEC (Virginia) 7.81 $48.11 Michigan 9.77 $60.18 National Average 10.4 $64.06 California 14.33 $88.27 New York (highest) 16.89 $104.04 The claimed $61 savings are therefore quite believable. It is worth noting that the packaging notes that the CFL life may be reduced (and light output is also reduced) when the lamp is used in base-down position. It also notes that lamp life is reduced when used in enclosed or recessed fixtures. It is also the case that CFLs experience a marked reduction in lifetime when cycled with short on times. In that regard, I've found that CFL flood lamp replacements used in ceiling fixtures have a very short life. However, the operating cost savings are so considerable that even if the CFL lifetime is 3000 or 4000 hours, it still is cost effective. I also found McMaster-Carr carries a line of CFL lamps with 10,000-15,000 hour life ratings, although to get the price down to $10 each, 12 units or more have to be purchased.
What's Inside a CFL? A CFL contains a DC power supply and an AC inverter typically operating in the 40-50 KHz range. The voltage inverter causes a current-limited current to flow through the gases contained within the CFL envelope. The ultraviolet light produced by the gases is converted to visible light by fluorescent phosphors deposited on the glass tube's inside surface. The inverter is designed to limit the current through the tube to its rated value, and usually employs a resonant tank-type output. Historically, conventional fluorescent lamps, powered by the 120V line, limit lamp current through a series inductor, called a "ballast." For that reason, the CFL's electronics inverter module is often called an "electronic ballast." I had not intended to disassemble the lamp,  but that plan was changed when I dropped it onto the concrete floor in my basement shop. After cleaning up the remains (and being suitably careful to avoid potential mercury vapor exposure from the broken lamp), I disassembled the electronics module.
CFL lamp partially disassembled. The four wires (one is broken and no longer connected to the printed circuit board) are the fluorescent tube connections.
Component side of PCB.
Current and Power As the CFL module has a DC power supply, we may expect that it will behave like any DC power supply, i.e., it will draw current from the mains power only when the instantaneous AC voltage exceeds the DC voltage on the filter capacitor, plus the rectifier diode drop. To see whether this is the case, I powered the CFL from the AC mains, and monitored the mains  voltage and supplied current. (This must be done carefully, using an isolation transformer. Don't do this at home unless you thoroughly understand the safety of life issues and have the necessary equipment and knowledge to use it correctly.) I used a Tektronix TDS430 digital oscilloscope, with the line voltage on Channel 1 and the current, sensed through a TCP202 Hall-effect current probe on Channel 2. I also used the TDS430's mathematical option package to compute and display the instantaneous power, i.e., the product of the voltage (CH 1) and current (CH 2) traces. This  trace is displayed as M2. In addition, I used the display measurement feature to compute the average of the instantaneous power, shown at the right margin as "M2 Mean." The display identifies current is VΩ and average power as VV. In the case of CH 2, the scale should be understood as 1.00 A/div and 200 watts/division for M2.
We can learn quite a bit from studying this oscilloscope capture. Looking at the current, for example, we see that, as expected, current is drawn over a relatively small fraction of the total AC cycle. An expanded view of the current trace is shown below. Although the average power consumed by the CFL is 21.2 watts, the peak current is 1.17 amperes. And, current is drawn only for 2.6 milliseconds for each half-cycle. At 60 Hz, a half-cycle is 8.33 ms, so the lamp's full required power is consumed over about 31% of the AC cycle. Looking at the M2, or instantaneous power waveform, in the oscilloscope trace above we see a similar effect—the instantaneous power peak is a  bit over 200 watts.
The oscilloscope capture below shows a 40 watt incandescent lamp in the same test setup. It looks quite different  than the CFL. Although the instantaneous power still varies with time—as it should since the lamp looks more or less like a resistor and the instantaneous current varies proportionally with the instantaneous voltage—the peak effect is much less pronounced and the power is more uniformly spread over the full cycle.
So what, you're probably thinking. I have many electronic power supplies running everything from  televisions to computers to amateur radio equipment and all have this same peak cycle current draw. Quite true. I can't find a definitive number in a quick Internet check, but the figures thrown around say that 15-20 percent of electrical power consumption in the US results from lighting and the great majority of this represents incandescent lamps. If we switch to CFL lamps, the total lighting-related power consumption will drop, a good thing generally. However, having a significant proportion of the total electrical power load represented by AC waveform peak devices such as the Sylvania CFL I looked at is definitely not a good thing. It increases the harmonics in the power line and also presents a highly non-linear load to the power utilities. These things complicate both electrical generation and the distribution network, including power transformers. The European Union, in fact, has adopted a standard, EN61000-3-2, placing limits on the relative amplitude of the current pulse harmonics, with the thought of avoiding or at least reducing problems caused to the electrical power grid by peak-charging loads such as  the CFL. This is sometimes said to be a "power factor" issue. Normally the power factor relates to the phase between the voltage and current waveforms. Power factor is the cosine of the phase difference between the voltage and current, sometimes expressed as a percentage and sometimes as a ratio over the  range -1 to +1. A resistive load has a power factor of 1.0, as the voltage and current are in phase. Motors are inductive and have a power factor that varies with motor size and construction, but is usually in the 0.8 to 0.9 range. Overall, a power utility's load is normally inductive due to customer's motor loads, and the power factor can be brought closer to 1.00 with capacitors, if necessary. You will sometimes see a capacitor package installed on pole tops by some utilities for power factor correction. (At the risk of being inaccurate due to brevity, the power generation and distribution network must be sized for apparent power, i.e., VARs or the product of voltage times current, called volt-amperes. The actual load that residential and small business customers pay for, however, is voltage times current times power factor or watts. Hence, it's most efficient for the utility if power factor = 1.00. Large scale industrial and commercial users, however, often are billed for both VARs and kwh.) The issue with a CFL is not so much the power factor, as it is that the entire power is drawn over a relatively small segment of  the 60 Hz waveform. This spike-type current waveform, if analyzed in a Fourier series, or looked at with a spectrum analyzer, will show that it causes harmonics in the power system. It also causes a problem similar to the power factor issue, in that the power network must be designed for the instantaneous voltage/current peaks
RF Noise As radio amateurs, we are also concerned with radio noise generated and radiated by electronic devices such as the CFLs. Measuring emissions requires a calibrated test range and specialized equipment, such as a line impedance stabilization network or LISN,  which I do not have. However, I did look at a conducted differential line noise, i.e., radio frequency noise measured as current over one side of the power line. At the lower radio frequencies, the power wiring looks like a transmission line and crud induced on it will not necessarily be completely radiated or coupled into your antenna or receiving equipment. To measure the differential mode induced noise current, I used a Tektronix P6022 current probe. To increase the probe sensitivity, I wrapped 10 turns of small diameter wire through the probe jaws and connected the CFL to the wire. A more complete investigation would also look at common mode noise, i.e., noise excited on both the hot and neutral conductors against ground, as this signal has a greater potential to radiate. However, these measurements are not going to mean much without a defined environment with respect to the common mode line impedance. For the range up to 100 KHz, I connected the P6022 current probe to an HP3562A dynamic signal analyzer. The capture below shows the noise for a 40 watt incandescent lamp. The P6022 rolls off and is not accurate below 10 KHz, so you should take the data below 10 KHz with some caution.
Compare the CFL data below with the incandescent data above. We see  the inverter fundamental at around 47-48 KHz and its second harmonic. In addition, there's a tremendous amount of broadband hash, being 40 to 50 dB above the level seen from the 40 watt incandescent lamp.
I  traced the CFL's input circuit to see what the designers did for RFI suppression. The circuit fragment below shows that the RFI filter consists of a single bypass capacitor and an RF choke. Incidentally, the component ratings seem marginal for reliability. R1 is a resistor, but its primary purpose seems to be as a fuse. It's a 1/4 watt, 0.47 ohm resistor which will quickly open if excessive current is drawn. The filter capacitor is rated at only 200 V, even though it will normally see 180 volts applied. That's a thin margin indeed. Likewise, C1's 250 volt rating is much lower than would normally be used for a line bypass. The RF choke is unmarked, but looking at the core material, it is likely in the few millihenry range. Earl, N8ERO, has written to say he has disassembled several CLF ballast assemblies and found the chokes identical at 1.3 mH, and with a Q of 50 at 60 KHz. He also reports the self-resonant frequency is around 300 KHz.
I also looked at the differential noise over a wider frequency range with the same test setup, but employing an Advantest R3466 spectrum analyzer. Since the spectrum analyzer has a 50 ohm input impedance, I added a 10 db gain Z10000-U broadband amplifier. (The P6022 current probe is designed to operate into a 1 MΩ instrument, such as an oscilloscope or the HP3562A DSA.) As a calibration point, I connected the output of an HP8657A signal generator at 1 MHz and -40 dBm output. This results in a current of approximately 89 microamperes through the current sensing circuit.
The image below shows the test setup with a CFL installed, but the power supply switch turned off. One strong local AM broadcast station is seen around 1440 KHz, and a marine navigation beacon at just over 200 KHz.
Engaging the AC line power starts the CFL and the noise floor increases substantially, particularly below 1 MHz. With the CFL powered up, the stray broadcast pickup is greatly diminished, as can be seen in the 1440 KHz AM signal, which is not visible when the CFL is powered.
The remaining question is how much of the CFL trash induced into the power line is radiated. I can't provide a good answer to this question at the moment. Casual listening with my antenna system shows no noise, but my antennas are 100 feet or more from the CFL test position.   http://www.cliftonlaboratories.com/compact_fl.htm

[lima]

If you actually read this, here is the thought of where I am going...

The ambient RF noise floor is going to creep up with dozens or hundreds of CFLs in the neighborhood...  All neighborhoods...

 

The effect of one or two may not be noticeable, but the incremental increase multiplied by the total number has got to have an impact.

[SydTheSkeptic]

OMG

You lost me early on, but COOL visuals of the components.  I love stuff like that.

K, I skimmed it, and so the point is that you think the more we use these, the more ambient noise it'll create?

[lima]

yes and how will this effect us?