Published 20 Sept 2009.

More CFL experiments

In response to D.Helmore comments

D.Helmore makes some interesting points in his comment to my previous CFL article but I do not agree with some of them. I should add that I have no medical qualifications which would allow me to confirm or refute the veracity of the claims made regarding physiological effects caused by CFL's.

Addressing the last two paragraphs first, I will say that in no way am I "telling you to get used to them". The whole thrust of my articles is to point out both strengths and weaknesses and to conclude that they have to be used properly and that national power savings will be very limited and how would you demonstrate such savings anyway. Secondly, I am not "pushing the introduction of such horrible lights..." My aim is to provide objective information to allow readers to make their own choices. Thirdly, I have not "..considered the effect on young children.." While this might sound callous, it is not, for the reason that I have no evidence of any effect of CFL's on children, giving due regard to my statement above about having no medical qualifications. If I had such evidence, it may be mentioned, but it would be outside the scope of my articles.

I have made a number of further measurements on CFL's to try and confirm some of the assertions made by the writer. These are engineering tests that may or may not assist, but cannot support a conclusion of the presence, or not, of physiological effects to humans by CFL's.

My latest experiments

Aim: To compare CFL's with both linear fluorescent tubes and incandescent bulbs for (a) Light output, (b) Time taken to reach full output, (c) Modulation spectra especially 100Hz components.

Method: Assemble a collection of 9 lamps, including a 60W incandescent and a linear fluorescent 2 x 18W dual tube model with inductive (non-electronic) ballast. The various models and power are shown in each table below. I have used a photodiode detector which responds to visible light and includes an infrared filter. Lamps are placed at a distance from the sensor which results in the same light reading for each. Measurements are made of the spectra over the frequency range 100Hz to 48kHz, using a Liberty Praxis audio measurement system. I have not attempted to measure absolute levels of lux but the results are a good comparison of one lamp with another.

Results Part (a): Light Output

The table below is sorted by lamp giving maximum light to minimum light output. The linear fluorescent gave the most light while a globe-style CFL gave the least. The spiral type CFL's are better at radiating downwards than the ones with longitudinal tubes .

Lamp TypeBrandRated powerDistance to sensor (cm) for equal luxPercent light compared to highestComparitive efficiency: light out for watts in
1Linear fluorescentTritech dual 840mm364710047
9CFL spiraleco20368773
4CFL spiralDEC3A26338354
3CFL spiraleco15297888
2Incandescentno brand60277521
5CFL tubularDEF5B112065100
8CFL tubularOsram15206573
6CFL tubularNoFrills20186151
7CFL globe Radiant13165875

This table illustrates the relative amount of light directed below the lamp, which is aimed directly downwards. All lamps, except the linear fluorescent, are screwed into a standard desk lamp which has a painted white interior. The distance from the lamp to the sensor was measured for the same reading of light output (after warmup). This table shows that the linear fluoro gave the most light, but the 11W tubular CFL had the highest ratio of illumination to electrical watts used. This experiment highlights the fact that CFL's do not radiate light directly underneath very well, assuming they are installed vertically, like in a downlight fitting. For most light below, they should be mounted horizontally. Essentially, the plane where you see the greatest surface area of tube, is the plane where most light will radiate.

Due to heat buildup, it is not a good idea to mount CFL's in an standard lamp fitting facing downwards anyway. Specially designed downlight style fittings are available, which not only let heat escape, they do the best possible job of concentrating all the light emitted from the lamp in a downwards direction.I should add that these figures do not represent the full potential light output of each lamp as would be radiated in all directions. They do go some way to explaining the poor effectiveness of CFL's when used in non-optimised fittings.

So, I do agree with the writer and the quoted reference from the optometrist. CFL's do radiate most light horizontally (when mounted vertically) and do not provide as much light directly underneath as an incandescent of five times the power rating (which is the usual comparison). Eyestrain and resultant headaches from trying to read in poor light is quite likely. Mount them horizontally, and the better CFL's would give the same light. This table does show that some CFL's are better than others. The other option is to ensure lampholders and fittings are designed for CFL's and not just plug them into any old lampholder.

Results Part (b): Time taken to get going

This table is sorted by time it takes each lamp to reach full brightness.

Lamp TypeBrandRated power% light at coldtime to 70% output (secs)time to 100% output (secs)
2Incandescentno brand60100-0
7CFL globeRadiant1380-40
9CFL spiraleco20441448
4CFL spiralDEC3A26502050
6CFL tubularNoFrills20481850
3CFL spiraleco15441360
5CFL tubularDEF5B113534120
8CFL tubularOsram153090260
1Linear fluorescentTritech dual 840mm3672-300

Of course the incandescent globe reaches full output straight away. The linear fluorescent took longest although it started at a relatively high 72% brightness. Included is a column showing the starting brightness. Some of the lamps are abysmal. There is also a column showing the time to reach 70% brightness which I imagine is an acceptable percentage of full output, where you would be able to avoid stumbling over objects in the gloom. The Osram CFL is one of the worst offenders. Starting at 30% output, it takes a minute and a half to even reach 70% and over 4 minutes to get to full output.

Results Part (c): Spectra.

Table of certain recorded modulation spectra from 100Hz to 48kHz. This table is sorted by the amount of 100Hz in the spectra, all relative to the lamp with the highest level of 100Hz flicker, which is the linear fluorescent tube.
Refer to the 5th column, headed '100Hz component rel lamp 1 (%)'.

TypeBrandRated power (W)100Hz component rel to lamp 1 (dB)100Hz component rel lamp 1 (%)200Hz component wrt 100Hz (dB)300Hz component wrt 100Hz (dB)400Hz component wrt 100Hz (dB)500Hz component wrt 100Hz (dB)HF component wrt 100Hz (dB)HF component freq band (Hz)
Linear fluorescentTritech dual 840mm360100-16-22-28-33niln/a
Incandescentno brand60-1225.1-24-30-41-45niln/a
CFL spiraleco20-1517.8-10-19-19-3436 to >48kHz
CFL spiralDEC3A26-218.9-12-17-20-24-3236 to 40kHz
CFL tubularDEF5B11-246.3-11-18-20-24-2933 to 36kHz
CFL tubularNoFrills20-246.3-11-17-21-24-2940 to 42kHz
CFL tubularOsram15-255.6-12-19-22-27-3332 to 34kHz
CFL spiraleco15-2651-6-24-25-2939 to >48kHz
CFL globeRadiant13-293.5-12-19-21-19-2433 to 34kHz

This table records the modulation spectra of each lamp at the distance where each gave the same amount of lux on the photodiode. Of most interest is the 100Hz component. The linear fluorescent produced the most 100Hz and the other lamps are shown relative to it. The incandescent lamp gave just 1/4 of the 100Hz flicker of the linear tube. All the CFL's were less than 20% with most less than 10% of the 100Hz flicker of the linear tube.

This suggests that if professed physiological effects were not apparent with incandescent lighting, then neither would CFL's cause problems, assuming it was the 100Hz flicker that was responsible. Other harmonics of 100Hz are also measured relative to the 100Hz component. CFL's also produce a supersonic component in the light spectra. The frequency range of supersonic light output for the lamps in this test was from 32kHz to over 48kHz, which is where my measurements stopped. It is quite possible that those CFL's with components in the 36kHz to 38kHz range will prevent infrared remotes working normally.

This test demonstrates for me that CFL's produce much less 100Hz flicker than either the long tubular fluorescents or standard incandescent lamps. So any attributed physiological issues from CFL's must have a different explanation. It might be suggested that the supersonic component in the range above 30kHz is responsible, since this is exclusive to CFL's, however the output light component of supersonic flicker is very low relative to the 100Hz output of each lamp. I do have it in mind to check if there is a supersonic acoustic output from CFL's but that will have to wait for another rainy day.

So, while I accept that 100Hz flicker from the older style long tubular fluorescents might cause a problem, I have shown that CFL's ought to be much less of a problem in this regard. Less of a problem than even incandescent lighting. Any person suffering from such flicker phenomena would not be able to work a computer since PC screens refresh at generally lower than 100Hz rates. Finally, I will have to add here that I have not dismissed the claim that there are physiological effects from CFL's. I simply don't see it as being caused by 100Hz flicker.

Axino-tech September 2009