Sesekali ndongeng teori baku, walo bukan keterikatan pasti..
Optimised just now
https://www.electroschematics.com/fluorescent-light-driver/
View original
Advertisement
Home » Lights » Fluorescent Light Driver
Fluorescent Light Driver
JORIBO
Share this:
Tweet
More
Fluorescent light driver, my project No. 10 (versions 1 to 9 will be reported when I have my own homepage ready)
A 12 V fluorescent lamp driver is one of the most critical circuits which you can make. The correct function relies on special effects! There are for example resonances by winding capacities, by coincidence of a tuned circuit with a switching frequency, influence of the lamp capacity, of the leakage flux and saturation properties of the ferrite core etc.
I have made these circuits as accessory for solar power projects since 1981, but most of them worked with bad efficiency or problems. Now, after 32 years, I made some systematic measurements and wrote down the key points to eliminate problems.
The standard schematic, which you find in many versions everywhere, looks like this:
This schematic is also used as lamp driver for scanners or some UV lighters, it is very common.
The key points
1. Sine-wave oscillator but pure switching
On the primary side the device is a saturation-controlled sine wave oscillator. There is a feedback winding to drive the transistors. These operate in pure switching mode, just on/off. Radio amateurs call this a C-mode operation (A is proportional like an audio amplifier, B is mixed and C is pure switching). The pure on/off results in very small losses in the transistors. The transistors have to be fast switching ones, high-frequency compliant or especially made for switch operation. A 2N3055 will develop high losses at 40kHz and is unsuitable.
2. The function of the two capacitors
We have two very important capacitors, the primary side capacitor, making a resonant circuit with the transformer winding, and the secondary one, making ALSO a resonant circuit but as a series-resonant circuit with the lamp inside. Both have to be selected to achieve correct function.
At first the secondary C is selected. It is stressed with a very high frequency, voltage and voltage slope! It is advisable to buy impulse-resistant FKP types rated at 2000V. These will have a long life and will not become warm. Also MKP is ok, if rated at over 1000V. The absolute limit is a 400V rating. Ceramic capacitors are excluded due to their high losses. Mica is overdone, leave these for the millionaires
If you cannot get foil-C’s with over 400V rating, you may use double the capacity and connect two in series. A good FKP capacitor costs about 1 to 3 Euro here in Finland.
The secondary C determines largely the frequency of oscillation. It needs to be selected to get the whole circuit into the correct frequency, for which it had been designed. I had dimensioned the transformer for 40kHz and 0.2 T (Tesla= Vsec/m2) of flux density. Using still wrong size capacitors, it worked first at 100 kHz, then at 15, but finally with the correct value at 37kHz which was close enough to my aim.
A ceramic capacitor was taken, as I had no high-voltage high-quality foil-C’s at hand. It was replaced later (the photos still show the ceramic one) because of too high losses, became hot.
Now please look at the hand sketches below the hand-sketched schematic. In the middle is the primary voltage without any primary capacitor, measured over one transistor. It is a rectangle with an induction peak. This is not optimum, so I added a primary capacitor (still too small) and the result was like the right-hand sketch: there is a superposition of the oscillation of the primary and secondary side.
The primary side C needs to be adjusted to equal resonance frequency as the secondary side, which needed (after some testing) 180nF. Minimum voltage rating is here 100V, better to be 400V.
All values refer to my transformer, you will need different values but you can use the same principles of selection.
3. The secondary C is needed anyhow
One might ask why there is a C on the secondary side in series with the lamp. This is because a fluorescent lamp can become easily asymmetric by wear, and then acts like a diode and takes a DC part of the current. This would drive the small ferrite transformer into asymmetry; one side of the push-pull arrangement would magnetize it more into saturation than the other side. This results in high losses and can bring the whole function to a stop. This problem can happen after a certain time, when the lamp wears down. The secondary C prevents this harmful DC content.
4. The heating filaments
The fluorescent lamps have heating filaments, which need 6 to 10V for start. For steady-state operation the voltage can be reduced to about half, or switched off completely. The filaments are heated by the gas discharge also, but on an inverter I would recommend a steady small heating voltage, as the lamp operates more stable, has less tendency to flicker. The 7 turns for the filaments were optimum for starting but too much for operation. So I added resistors to reduce the filament heating voltage. Next time I would probably use only 5 turns instead of 7.
Using very rough voltage containing spikes all time, or using additionally a separate ionization wire on the lamp makes the filament heating unnecessary. But you pay this advantage with a shorter lamp life or HF disturbances.
5. The transformer core
This is the key component. You need to match your switching frequency and winding data to the ferrite material. Using an old flyback transformer from a TV or monitor limits your frequency to about 20 kHz. They are designed for ca 16 kHz. Typical ferrite cores can handle 40 kHz well with 0.2 to 0.3 T and some go up to 100kHz but then only 0.1T. The higher the frequency, the lower is the permitted flux density.
You need to get info’s from the manufacturer, as the ferrites look all alike. In my case, the transformer was very probably from REINHÖFER, marketed by CONRAD, order number 516678, Material Manifer 196 from TRIDELTA, and after browsing the internet, I found the necessary data sheets here:
http://www.tridelta.de/viomatrix/imgs/download/manifer_196.pdf
Looking at the specific loss numbers for f=40 kHz and B= +-0.2 T, we come to about P= 250mW/cm3 of core volume.
The core volume is 3.86cm3 (dimensions measured). So the expected iron loss is roughly around 1W. These figures must not be taken very serious. The diagram does not tell us if it is for rectangular or sinusoidal voltage, also not if it is based on a push-pull magnetization (both directions) or with a simple forward-converter (one direction). But you can see that if we would double the frequency and cut the flux density in half (same power) we would have less losses in the core (more in the transistors!). The red curves are for a hot core and blue for a cold one, surprisingly the losses are less with 100C. But the copper losses increase with rising temperature and especially the insulation material is stressed far more. Running the transformer at 100C core temperature would need silicone/glass insulation, class H, or the transformer will fail soon. As I prefer “normal” laquer, epoxy, paper, the transformer must not become hot, here it goes to ca 40C in 20C room temperature.
The air gap is selected to make the transformer insensitive to a certain amount of voltage asymmetry, and also to separate the primary and secondary winding magnetically. A certain amount of leakage flux is desirable because else the secondary voltage could not rise up to resonance peaks, it would be “hold tight” by the primary winding. As a starter value, I recommend 0.2 to 0.5% of the average magnetic path length as air gap. Here I used laquered paper of 0,04mm thickness all over both cores, making 0.08mm distance between the two E-cores. And as the air gap is 2 times in one magnetic path, it comes to 0,16mm.
You might ask, where is the magnetic field strength H and why did I not use the magnetic length and the B/H curve anywhere. In fact, for finding a reasonable number of turns, it is not necessary. You could predict the magnetization current, or make a more accurate simulation model of the little transformer. For example, to put it into P-SPICE. That I did not do, students of electronics may do so.
6. The windings and number of turns
We assume that we have both transistors conductive over each half of the period.
The frequency is assumed to be f=40 kHz. This means a period time of T=1/f = 25usec.
In half of this time (the conducting period of a transistor) we magnetize the core from -B to +B, so over a range of 2*B. The time is t=T/2= 12.5usec.
The maximum flux density B= (U*t) / (2*n*A) with n= number of turns and U= applied voltage and A the iron cross-section. (The mentioned factor of 2from the 2*B is now in the denominator).
The primary winding for one direction is assum