Fluoro driver

[commanda]

fluoro.pdf




This is currently breadboarded and working nicely. The transformer is wound on a 2 inch  length of ferrite rod, similar to what bthumble has described here.

R1 and R2 monitors the battery voltage, and shuts down on low battery.

R12 monitors the current, adjustable via P1.

The 7812 was added for 24 volt applications. The transformer primary would require more turns for this.

The heater windings (I4 & I5) probably require less turns.

The reference voltage from the TL494 was divided down by R5, R6, P1, so that R3 could provide hysteresis for the low voltage shutdown. This hasn't been tested with a real battery yet, just a bench power supply.

The fets I'm using are IRF540; mainly because I had them on hand. But they're generally readily available and cheap.

I've had it running with a 2 foot 18 watt tube. Currently using a 15 watt 18 inch tube.

The main criteria for this design was reliability. It is intended for the upcoming Fiji trip.

Amanda

[maker of toys]

Cool!

any feeling for the amount of radiated EMI from that ferrite rod?

-Dan

[dinges]

Darn. Now I need to get me more PC PSUs to build fluoro inverters.

For anyone interested: the TL494 is the same as the KA7500; they can be found in old PC power supplies, free for the taking. Along with a few FETs. So, if you've got the an old PSU, you can build this thing without going to the electronics store.

Ferrite rods are in MW radios. Also free for the taking. Design criterium was apparently not just reliability, but also availability...

Peter thinks Commanda has Scottish roots. Or Dutch.

I assume the circuit also works for smaller FLs, and CFLs ? What's the maximum power you think the design can handle? I assume the ferrite to be the limiting factor.

What's the advantage over BTHumble's design?

[dinges]

I forgot: how do you adjust P1? What's the exact goal/function...

[commanda]

Maximum power I don't know; my bench supply limits at 3 amps.

BTHumble's doesn't auto-off on flat battery. He's had several failures at least that he knows about. He thinks this has something to do with it.

The single-ended flyback type circuits tend to destroy tubes fairly quickly, due to the pulsed dc sucking all the mercury up to one end.

And this push-pull design, without the filament heaters, tends to not fire 2 foot long tubes.

P1 adjusts the amount of current the circuit draws. Without it, it runs pretty much 100% pwm.

Before you get too excited, just remember the circuit hasn't necessarily been optimised or fully (reliable under all circumstances) tested.

Amanda

[Flux]

Looks promising. You are right to use the heaters even at the expense of a bit of efficiency.

You are probably right about the need for low volt cut off, low volts kill most circuits but normally I would use a low volt disconnect, but if the lights are the critical factors there is no harm in including it in every driver.

I have toyed with this cycle by cycle current limit but never actually got round to it. I am not sure how much leakage reactance of your ferrite rod transformer plays in the operation, it may be important.

I have used 3525s in a similar design, but I used a closed ferrite transformer and resonant L C ballast with tube heaters in the series circuit.

The oscillator RC components being tuned to the resonant frequency of the ballast.

Ideally my circuit should be started above resonant frequency to pre-heat the heaters and then the frequency dropped to resonance. This can be done with a fet or transistor to alter the chip oscillator frequency across R9 in your circuit.

For CFL s I have found it satisfactory to omit the frequency shift and run up in soft start mode ( available on 3525, not sure about your chip)

If you are aiming for Fiji and availability of components is a big issue then your circuit avoids the ferrite closed core and the fairly stressed and possibly difficult to find resonant capacitor.

Good luck, I think you are on a winner.

Flux

[dinges]

If your supply limits at 3A, use the 12V from a PC power supply. About 10A. That should be able to let the magic smoke out of your circuit.

'P1 adjusts the current the circuit draws'. Ok, then I'll adjust it for minimum current (10mA or so)... That way I save on the power  Not nitpicking, but, again, how do you adjust it? You measure current, and adjust P1 until...? P1 influences light output too?

As far as the circuit not being optimally tested; I have blind faith in you. Still working on the dumploadcontroller (waiting for parts) and so far, it works as advertised. As far as the testing goes: being in Australia, I understand you can't test in rain, snow and ice. No worries, I can do that for you over here. Once, when I was young, I remember having seen the sun. I still remember it was very bright.

[commanda]

'P1 adjusts the current the circuit draws'. Ok, then I'll adjust it for minimum current (10mA or so)... That way I save on the power  Not nitpicking, but, again, how do you adjust it? You measure current, and adjust P1 until...?

18 watt tube. 12 volts. Current = 1.5 amps. Simple.

Maybe not.

Was just reading something about electronic ballasts being more efficient (20KHz vs 50Hz) and only needing about 70% rated power input.

P1 influences light output too?

Without a proper luxmeter, very hard to judge.

This may be Sydney, Australia. It may be the first week of spring. We just had the wettest September day on record for one hundred and twenty something years. The wind not far from here was hitting 140 Km/Hr. And it snowed on the central tablelands. Again, not far from here.

If your supply limits at 3A, use the 12V from a PC power supply. About 10A. That should be able to let the magic smoke out of your circuit.

I could just drag it across the room and connect it straight to a 12 volt, 100 AmpHr battery. Unfortunately, being bread-boarded with jumper leads everywhere, it's a bit fragile at the moment.

Just don't confuse your blind faith with the blind leading the blind. When you're blind, it can sometimes be difficult to tell the difference.

Amanda

[oztules]

Flux,

Yes the soft start is available with this chip.  (tl494 ka7500)

A cap (10uf or so) in series with pin 13/14 to pin 4, and a resistor (100k etc) from pin 4 to ground will do it. It simply pulls the pin to +vref at start and inhibits pulse width to zero, the cap discharges through the 100k to ground, and the chip wakes up as the voltage heads toward zero on pin 4 (pulse width control to maximum at 0 volts above ground.). At this stage the max pulse width will be controlled by the two op amps on board.... voltage shut off and current regulation.

This is a voltage mode controller (tl494). It is not using true pulse by pulse current limit.

The 3525 you mentioned I think is a current mode device and can limit based on pulse by pulse cut off.

..........oztules

[RP]

One source for closed ferrites would be just about anything with a CRT in it.  The flyback transformers used for the anode voltage use a ferrite core at ~15Khz.  Many of these can be dissasembled to remove the flyback coil leaving a little rectangle made of round ferrite sticks.

If they have broken TVs and computer monitors on Fiji you mght be able to score some.

[BT Humble]

BTHumble's doesn't auto-off on flat battery. He's had several failures at least that he knows about. He thinks this has something to do with it.

Yep, the old flyback design seems to work pretty well if people are disciplined about switching them off when the battery gets low.  The tube usage isn't such a big deal, the 1 person we've got who keeps their battery full has been using one of the first circuits every night for 18 months now, still on the original tube with no problems.

Mind you, Amanda's circuit has much nicer bells and whistles.  I like it.

BTH

Off to collect dead PSU's again...

[commanda]

An updated circuit has just been uploaded.

Changed R9 to get the frequency nearer where it should be.

Removed the capacitor off the current sense circuit.

Added a capacitor in series with the secondary. If you mess up winding the transformer and miss a turn on the primary the secondary positive & negative swings will be unequal leading to premature blackening of the tube. This capacitor is entirely optional. it will work without it.

Amanda

[commanda]

I've done some more work on this circuit.

fluoro1.pdf




I deleted the original fluoro.pdf referred to in the first post.

I'm now using the dead time control (pin 4) to control the pwm.

R7 compensates for changing supply voltage. The voltage generated across R12 as a functon of the current is fed back via R13 to the bottom end of P1.

Although I've shown a 7812 on the circuit, the values of R1 and R7 are set for a 12 volt supply.

I also added a gate driver chip, because I wasn't happy with the waveforms and switching characteristics of the fets when driven directly from the TL494. They also ran too warm for my liking. You can, of course, use 2 pairs of TO92 transistors in place of the dip8 package.

At first glance, you might think that C5 is placed incorrectly. To cleanly fire the tube, you want a good kick start. Using soft start is not a good idea.

I've had it running all day every day for a couple of weeks now, with no visible blackening of the tube ends at all. This is a 2 foot tube.

Probably a few more refinements to come yet, but it's really close.

Amanda

[dinges]

It's slowly getting bigger & bigger... More parts Even exotic parts (never heard of a TC4427)... the original goal was to keep it simple & buildable with local parts, IIRC <evil grin>. I doubt there are many TC4427s around on Fiji. I know there are none in my parts bin...

I haven't built it yet, but when I do I will use the transistors to drive the FETs. BTW, wasn't the TL494 able to drive about 100mA or so? And this is NOT (fast) enough for the FETs? Amazing.

In a PC PSU, aren't the FETs driven directly by the TL494 too?

I think the Chinese are much better at designing than you are. I've seen fluoro inverters made by them with just 5 or 6 parts...

[commanda]

Peter,

The TL494 does not have a totem-pole output stage. It can turn the fet on, but can't actively turn it off.

A PC psu has driver transistors and a transformer between the TL494 and the fets.

I can also build an inverter with 5 or 6 parts. The single transistor flyback design has certain inherent problems, which I outlined earlier. The second type is the quasi-resonant circuit, which Flux alluded to earlier. The magnetics in these is very difficult, and doesn't lend itself to one-off scratch built versions.

Constructive criticism is always welcome. Now that I've picked myself up off the floor, I'll press on.

Amanda

[commanda]

I've done yet more work on this, following Peter's very constructive criticism.

I've chucked out the TL494, and replaced it with an SG3525. This chip has a totem-pole output stage, and works much better.

I've only had it running for half a day, but am pretty happy with it as it stands.

fluoro2.pdf

I've deleted the previous circuit.

As always, constructive criticism welcome.

Amanda