Page 1 of 2

Fluorescent Lamp Inverter

Posted: Sun Nov 06, 2005 12:42 am
by Bosium
Hi All

My folks are big into camping and 4x4'ing, as there is a lot to do where I live. My dad has one of these little 12V powered 8W fluorescent tubes with battery clips on the one end that he uses a lot.

I recently opened it up to replace the (crappy) wires it had come with from the factory. I was amazed to see so few components on the PCB, literally 2 resistors or so, a cap or two, 1 transistor and a little transformer.

I have read quite a lot about FL tube driving, and so I thought I'd give designing a little inverter a go. I have built several push-pull SMPS's for car audio use so I know something about switchmodes.

I also thought that 8W is pitiful, and I'd rather try to design for a 15W 1.5ft tube.

So I designed a simple flyback converter based on the TL494 as I have some experience with this IC and it allows for frequency and duty-cycle adjustments quite easily.

It draws almost nothing at all when the FL tube is unattached, but when the tube is attached it draws anywhere between 100mA and current limit on my PSU, ie 2.5A depending on what duty cycle I select. It strikes instantly and dims well although it is not very bright and my FETs cook! I thought they would stay cool and it would be brighter.

I had better luck when I first wound it for 1:15 turns ratio, although it would not strike first time unless the duty cycle was set reasonably high. At least it was bright. I think maybe the secondary resistance is too high at 9ohms, too many turns with ultra-thin (0.2mm 32AWG) wire?

I also experimanted with using a cap as a current limiter in series with the tube but with little success.

Can anyone shed some light or perhaps point me in the right direction? I really was hoping to get the lamp to run bright without too much current drawn and without my FETs getting so hot, ideally I would have liked to use just one FET - they are quite sturdy devices.

Thanks in advance!

PS: I cant seem to post a PDF of the circuit diagram in this forum, whats up with that?

Re: Fluorescent Lamp Inverter

Posted: Sun Nov 06, 2005 2:05 am
by Michael J
Oatley Electronics here in Australia have done
a few of these circuits in the past few years.

They have made a few that will run 240V CFL
tubes,(folded compact fluoroecent) they don't
work with the 110V USA types as far as I know
because internal diode issues etc.

What voltage is common in your area,,?

Re: Fluorescent Lamp Inverter

Posted: Sun Nov 06, 2005 8:42 am
by Chris Smith
Shorter pulses, more current in each pulse. Over all current is reduced.

Fly back transformers are not the type you need, those little transformers that are cube type work the best. You can drive them with the low voltage high current, short on times, and get your voltage boost.

Your fly back I suspect has slower on times from Zero, more saturation, and the impedance builds up around saturation levels? Just guessing on that one with out seeing your set up?

If you can shorten your pulse and increase your current in the pulse and run this into a cube type transformer similar to the one that was already there, the tube will light up bright, draw less current over all, and not need the heater elements found in some tube ends to start the tube.

Mine was 12 volt, 50khz, .4 amp current, 21 watt tube, and It hurt to stare into it. You could read a book at 100 yards.

If you going to use a flyback for voltage boosting, rectify it back into DC, and then run your Fets at high speed HV, low on times, and use the Dc HV to drive the tube. 300 to 600 volts is plenty.

The Best examples for you to follow are in the HE-NE laser gas tube designs running on DC low voltages that boost upwards of 1000vdc, and then have a starter that boosts upwards of 15kv. Similar, but not the same. You dont need the starter.

Re: Fluorescent Lamp Inverter

Posted: Sun Nov 06, 2005 11:04 pm
by Bosium
Hi Michael J, we run on 220VAC here. However, I want an inverter to drive a tube directly, not one to drive a light fitting.

Hi Chris, how do I get higher current with smaller pulses? It makes no sense. Less duty cycle, less current surely. The load is not changing. If you decrease pulse-width then you decrease the output voltage and therefore the current drawn.

I am not modifying an old inverter, I am making one from scratch. My transformer is a EI-core transformer - ie one of those "cube" transformers. It is ferrite cored and I rewound it for 5 turns on the primary and 250 on the secondary to get +-600V for the tube. It is definately not saturating. At +-40KHz I could get several hundred watts through it with enough copper on the core, so 15W really should not be a problem.

When I had wound it for 180V output, the transformer worked much better but the circuit had trouble firing the tube. I had some success using a RC-delay to force the TL494 to run at full duty cycle for a second or so at startup to fire the tube and then switch to lower duty cycle to limit current. It still drew 1.6A at 12V when running at the lower duty cycle, that equates to about 15W assuming a 80% efficiency and the tube was about as bright as it was on a big AC inductive ballast.

I would be very interested to see the circuit diagram of your 21W 0.4A tube, sounds almost too good to be true?

I have the tube directly connected to the secondary winding of the transformer. I am not using the heater elements or any capacitive or inductive ballast.

I would really like this circuit to work a little better. Im sure the FETs shouldnt cook so badly.
I think I will rewind the tranny for +-300V output, that seems like a good compromise and should fire the tube dont you think? It's only 1.5ft long.

Re: Fluorescent Lamp Inverter

Posted: Sun Nov 06, 2005 11:24 pm
by Bosium
Oh, I forgot to mention that I did NOT gap the core. Perhaps an air gap would help? It is a EI-33 core, a standard type and quite big actually. From an old PC power supply. I have used this exact core wound for push-pull operation before in a car-amp stepup converter. It is my 'test' core. I will use a smaller one ultimately, like an EI16 or EI20 core.


Re: Fluorescent Lamp Inverter

Posted: Mon Nov 07, 2005 12:58 am
by rshayes
This page ( has a fair amount of information about the lamps and balasts.

You do need some kind of ballast for these lamps. This can be either capacitive or inductive. Striking a lamp requires several hundred volts. Operating voltage is usually under 100 volts.

Heating the cathodes reduces the striking voltage needed. Cold cathode lamps require a kilovolt or so to strike.

The recommended waveform is a sine wave rather than short pulses. High peak currents overstress the cathodes and shorten bulb life.

Frequency does not appear to be a critical factor. Since high voltages are involved, the inverter efficiency will probably be poor with switching frequencies above the 100 KHz range. Stray capacitance starts to become very significant.

Re: Fluorescent Lamp Inverter

Posted: Mon Nov 07, 2005 11:05 am
by Chris Smith
The trick to get more current in the pulse is by reducing the on resistance, increasing the voltage available, while shortening the pulse duration. You do this by multiple factors, way past simply doubling.

The math is based on ohms law at 100% duty cycle, but your duty cycle is only 5 % to 10 %, which means that you can run 10 to 20 times the amps by increasing the voltage appropriately, and reducing the circuit resistance as well when ever possible and still equal the same over all current draw.

One amp for one second is equal to 2 amps for ½ second.

1000 amps for 1/1000 of a second is equal to one amp for one second, but the pulse is far more effective.

In order to not use any ballast at all you need a sufficient voltage and current to over come the natural resistance in the gas medium, and this easily over come as you know in living in a country with 250 volt mains.

Many experiment using fluro tubes with 250 volts, don’t require any ballast at all, or even heater elements. So when you bump the running voltage up around 300 plus volts, AND have the current to stand behind it, [short high current pulses are fine] Then the over all current is at 5% to 10% of your original figure, yet each pulse has a much higher current in it to conduct, and the speed of the pulses are fast enough to sustain the gas in a “on” configuration.

My Fluro from the 80s is long gone and although I have a shelf full of old projects I don’t have that original blue print to work off of. But the knowledge of how to accomplish this has been used by me over the years to build many power supplies for many types of tubes including the HeNe laser.

One project I did back then was to simply match a better final transistor to one of those over the counter fluro circuits, but I matched it with a better spec transistor, [one with a TO-220 package instead of the small current TO-92/3 that came with it],... removed or lowered the resistor feeding the source because the package can handle more volts and current, and then changed the oscillator cap using my O-Scope to increase the speed, and this alone dropped the current by 2/3rds while making the tube run much brighter. These over the counter assemblies are all about money, as little as possible, so modifying them can be done with much success. And this is a crude analog setup with inefficient pulses.

If I was you I would use your flyback to charge a HV cap to around 400 plus volts. Then use a low “on” resistance HV FET to drive the pulses, or even increase the voltage to 600 volts and use a HV transistor with a higher “on” resistance.

Then I would use some thing like the LS123 series Retriggerable Monostable Multivibrator or better, remove the timing cap and replace it with a 914 signal diode to increase the transition speed to its max, and use this to drive the gate or base of your circuit. This is the basics only, and you can work out the sundry items needed to accomplish the rest of the circuit. You will find that this will be running in the range of about 200 NS on time, but with a quad or dual package in the LS123, you can slow down the “on” time, while shortening or lengthening the transition time between pulses to get you current where you want it. Also a variable timing circuit in the lab will let you know what works the best, and you can even leave one of the timing variables in to dim the tube if you so chose.

<small>[ November 07, 2005, 02:58 PM: Message edited by: Chris Smith ]</small>

Re: Fluorescent Lamp Inverter

Posted: Mon Nov 07, 2005 11:09 am
by Chris Smith
Some timing examples..........

100 amps = 100 volts at one ohm

200 nano seconds on time = one in 5 million/sec

100 amp pulses, times 1000 cycles at 200ns on time is .02 amps over all. [per second]

100 amp pulses, times 50,000 cycles at 200ns on time is 1 amp over all. [per second]

<small>[ November 07, 2005, 07:06 PM: Message edited by: Chris Smith ]</small>

Re: Fluorescent Lamp Inverter

Posted: Mon Nov 07, 2005 11:10 am
by Chris Smith
<small>[ November 07, 2005, 02:56 PM: Message edited by: Chris Smith ]</small>

Re: Fluorescent Lamp Inverter

Posted: Mon Nov 07, 2005 11:24 am
by Chris Smith

<small>[ November 07, 2005, 11:52 AM: Message edited by: Chris Smith ]</small>

Re: Fluorescent Lamp Inverter

Posted: Sat Nov 12, 2005 11:39 pm
by Bosium
Hi Guys

I appreciate your input. I have experimented quite a lot and I have favoured square-wave excitement as it allows for a lot of flexibility in design.

My design uses the TL494 switchmode controller IC. I have used this before many times for other projects as it is a good solid device and is extremely common. I actually pulled it right out of an old PC power supply.

I am using an EI-33 core, also from a PW power supply with 22T+22T turns on the primary bifilar wound, 4 AWG32 (0.2mm) wires in parallel. Secondary is 500T of single-strand 0.2mm. With about a 1mm air-gap in the core the inductance is somewhere around 47mH on the secondary.

Runing the oscillator at 50KHz, push-pull configuration and about a 25% duty cycle or so, the tube runs at full-brightness and draws some 1.2A off 12V. This I think is pretty good. The FETs get a little hot still, but I think they need to be turned on harder so I will put a little bipolar driver stage before them to improve efficiency.

I used a capacitor and resistor-divider combo on the dead-time in put to make it run at full duty cycle on startup to strike the tube and then drop off to +- 25% DC. This could be changed to allow for differnent brightness levels and / or bigger tubes. I used a F15-T12.

I did NOT use a capacitor in series with the tube. I am relying on the inductance of the secondary to limit the current. This seems to work very well.

If I could post a schematic in this stupid forum I would.

Thanks and regards

Re: Fluorescent Lamp Inverter

Posted: Sun Nov 13, 2005 9:28 am
by Newz2000
Use to host the image. It will give you a link that you can paste right into the forum. It's free and doesn't even require you to register or give your e-mail address.

Re: Fluorescent Lamp Inverter

Posted: Sun Nov 13, 2005 9:15 pm
by Chris Smith
You can completely control your current draw by upping the over all current [voltage stored in a cap for discharge], while decreasing the on time and duty cycle.

Most of the parasitic circuits can be abolished, and start up wont need any buck or boost from any parallel, passive, or parasitic circuitry.

If the circuit fails to start the tube, simply turn it off and then back on.

This is all you will need to ignite the tube.

Work on a higher speed control of your square wave to reduce and control your current draw.

Re: Fluorescent Lamp Inverter

Posted: Mon Nov 14, 2005 11:12 pm
by VIS
I wonder if there is a way to run those tubes forever without them turning black, like with Cold Cathodes, or Neon. I think that the tubes will light up near a CB antenna when you talk on it without any connection. Flyback transformers and Tesla Coils definitely can do that too.

Re: Fluorescent Lamp Inverter

Posted: Wed Nov 16, 2005 6:39 am
by Bosium
Chris, I noticed the tube draws more current and runs brighter the shorter the pulses are, but I cant make the pulses too short without it drawing tons of current. Do you use a ballast capacitor or inductor? Or do you rely on the inductance in the secondary?

I have a 18W tube running at full brightness at about 1.4A, or two 15W tubes in series running at full bright also at about 1.4A. Makes me think that the two 15'ers are more efficient connected directly than the 18 is, I guess cos there will be a greater volt drop across the two 15's so it allows me to run shorter pulses, but perhaps with a ballast cap on the 18 it would help?

What a fiddly bloody circuit.

VIS - There is a thing called electrode-less discharge or something like that. Nikola Tesla used it in the 19th century or so iv read. It requires a very strong RF field, but I stand to be corrected.