Reducing transformer inductance

[ion]

I am trying to power a small mercury discharge tube, but since I can't get my hands on a proper ballast I decided to use an old transformer that has sufficient inductance on one of the windings.

I measured around 10.5 Henrys on the primary, which on 240V 50Hz mains should limit the current to around 100mA, but I'm only getting 20mA - not enough to light the tube.
I don't know enough about transformers and inductors to figure out why this is the case, and ideally I need something in the region of 150mA - which I thought would be about 7H for the inductor.

The transformer is similar to this design, but without the centre tap on the secondary:
https://www.electronicscomp.com/image/cache/catalog/6-0-6-3amp-step-down-transformer-800x800.jpg

Would removing the secondary winding reduce the inductance of the primary?

[T3sl4co1l]

Two problems:

1. What signal level did you measure inductance with?  Guessing it was small, a few volts, not the full 240?

This matters because the high-permeability laminated iron core is nonlinear.  At low field, mu is relatively low (initial permeability).  At higher levels (say, 10s of V at 50Hz) it begins to rise up to its nominal value (which may be a tenfold or greater change).  At still higher levels (>240V at 50Hz), it falls again, rather sharply this time, as the field "saturates".

2. There's not much you can do on the secondary, well, besides putting an inductor in parallel.  But if you had an inductor of the right size, you'd just use a slightly different one instead of the primary altogether.

But there is still something you can do: take the frame off, and with the help of a utility blade, flat screwdriver, hammer, and heat if necessary: remove the laminations from the bobbin.  Restack them so all the 'E's are on one side and the 'I's on the other.  Put the frame back on, cinch it up, and measure again.

What this does, is remove the overlap between laminations -- instead, field has to flow entirely through each piece, then jump the small airgap that is inevitably present between the E and I stacks.  You can control this by adding spacer material (preferably paper or PET film) between them.

Air gap acts to reduce the effective permeability of the core.  It also acts to stabilize it.  While the core material itself might vary over a factor of 10, if the air gap is ten times more significant, the core variation can only contribute a 10% error, i.e. going from small to large signal only increases the inductance by 10%.

For small air gaps (gap length much less than the width/height of the core face), the air gap is inversely proportional to the inductance.  So you shouldn't need too many tries to home in on the right value.

And the final part: the core still saturates above 240V.  Air gap has little to no effect on saturation voltage and frequency.  This solution is not suitable for a resonant ignition method.  If you have a one of those bimetallic starters, or an electronic one, or the tube just does it by itself, no worry.

Oh, a final final part as well... is your transformer rated for the desired 100mA primary current in the first place?  In other words is it 24VA -- roughly the size of a fist?  If smaller, it'll run rather hot, or burn up completely.  (The pictured unit, with apparently (2 * 6V) * 3A = 36VA, would do fine.  I guess you've picked that picture out of a catalog, check its dimensions for comparison. )

Tim

[mag_therm]

There are 2 types of mercury vapor lamps:
1) the original type which has an auxiliary start electrode.
2) The newer pulse start type which needs a costly starter/ballast dedicated to the part number.

As I found out the hard way with 2), it is difficult to make a home brew circuit that applies 3000 V (or so) pulses that are on the lamp terminals and also across the ballast winding.
I had flashovers both on the lamp base and on the inductor while the lamp stayed cold and dark!.

So I gave up and now use the legacy 1) type which starts at low impedance and runs on current controlled pwm DC.

[ion]

Two problems:

1. What signal level did you measure inductance with?  Guessing it was small, a few volts, not the full 240?

This matters because the high-permeability laminated iron core is nonlinear.  At low field, mu is relatively low (initial permeability).  At higher levels (say, 10s of V at 50Hz) it begins to rise up to its nominal value (which may be a tenfold or greater change).  At still higher levels (>240V at 50Hz), it falls again, rather sharply this time, as the field "saturates".

I used my DE-5000, so that explains the lower than expected current.

2. There's not much you can do on the secondary, well, besides putting an inductor in parallel.  But if you had an inductor of the right size, you'd just use a slightly different one instead of the primary altogether.

But there is still something you can do: take the frame off, and with the help of a utility blade, flat screwdriver, hammer, and heat if necessary: remove the laminations from the bobbin.  Restack them so all the 'E's are on one side and the 'I's on the other.  Put the frame back on, cinch it up, and measure again.

....

The 'E's were a nightmare to remove - a bit of corrosion didn't help.  Quite a few got slightly bent, it certainly wasn't possible to reinsert all of them around the coil.  It's probably not good that there are even fewer 'I's on the other side.  I removed the secondary as it wasn't attached to the primary anyway and it made reassembly a bit easier without it.

I'm getting just under 5H on the DE-5000, so I had a go at powering it up with a resistive load.  But now it's making a scary buzzing noise, no doubt because of the terrible job I did reassembling the core.  I ran it for a couple of seconds, just enough to get a current reading, and it seems I hit right around the 150mA I wanted.

I'm a bit worried about the buzzing though - is it just annoying or will the transformer melt/catch fire/explode if I ran it for a couple of minutes (just enough to see if the tube lights)?

And the final part: the core still saturates above 240V.  Air gap has little to no effect on saturation voltage and frequency.  This solution is not suitable for a resonant ignition method.  If you have a one of those bimetallic starters, or an electronic one, or the tube just does it by itself, no worry.

It's one of those small G23 socket tubes with the built in starter so I'm just guessing, but I think it's a bimetallic one.

Oh, a final final part as well... is your transformer rated for the desired 100mA primary current in the first place?  In other words is it 24VA -- roughly the size of a fist?  If smaller, it'll run rather hot, or burn up completely.  (The pictured unit, with apparently (2 * 6V) * 3A = 36VA, would do fine.  I guess you've picked that picture out of a catalog, check its dimensions for comparison. )

Tim

I really have no idea what the ratings are, I don't even remember where the transformer came from.  The picture was a random one I googled that looks about the same size and construction.  I'll just have to keep an eye on the temperature till I'm sure it can handle the power.

[ion]

There are 2 types of mercury vapor lamps:
1) the original type which has an auxiliary start electrode.
2) The newer pulse start type which needs a costly starter/ballast dedicated to the part number.

As I found out the hard way with 2), it is difficult to make a home brew circuit that applies 3000 V (or so) pulses that are on the lamp terminals and also across the ballast winding.
I had flashovers both on the lamp base and on the inductor while the lamp stayed cold and dark!.

So I gave up and now use the legacy 1) type which starts at low impedance and runs on current controlled pwm DC.

It's definately the simpler type - it looked like it was about to light on just the mains voltage with a capacitive ballast ( I hadn't discovered the long forgotten transformer yet).  It just needed a bit more voltage.

[bob91343]

The buzzing is magnetostrictive vibration and causes no harm other than the noise and sometimes a slight reduction in Q.

[james_s]

There are 2 types of mercury vapor lamps:
1) the original type which has an auxiliary start electrode.
2) The newer pulse start type which needs a costly starter/ballast dedicated to the part number.

As I found out the hard way with 2), it is difficult to make a home brew circuit that applies 3000 V (or so) pulses that are on the lamp terminals and also across the ballast winding.
I had flashovers both on the lamp base and on the inductor while the lamp stayed cold and dark!.

So I gave up and now use the legacy 1) type which starts at low impedance and runs on current controlled pwm DC.

There is an easy solution to that. In Europe/UK you can get 3 wire igniters meant to work with a simple choke ballast which is the standard over in 240V land. These are self contained and are wired up between the choke and lamp socket and superimpose high voltage spikes across the lamp without utilizing the choke to develop the spike. They are easier on the ballast and will work well in remote ballast installations where the length of wire will tend to attenuate the pulses.

My recollection is that ballast chokes usually use gapped cores. It's possible to wire chokes in parallel to increase lamp current. I've used multiple small fluorescent lamp chokes in parallel to run larger HID lamps, and in a few cases multiple HID chokes to run high wattage lamps.

[Zero999]

A capacitor will work, but tends to shorten the lamp life because there will be a current surge every time the tube ignites, as the capacitor has a very low impedance to sudden changes in voltage.

Why not simply use a fluorescent ballast? An 8W ballast will probably be fine, as it'll give roughly 150mA.

[T3sl4co1l]

Ha yes, you'll have a hard time assembling it without spare pieces; a few missing is okay.  The buzzing is metal slapping itself.  It shouldn't be bad, but besides being annoying, it's probably a slight wear issue (the slapping).

Try to clamp it down better (crimp the frame tighter?), or better yet, soak it in suitable glue or varnish to bond everything together again.

Suitable glues I think would be, most that are thin and solvent-based, which dry leaving a firm deposit, and which bond reasonably well to metal and existing varnished surfaces.  Varnish, your average non-stained oil based (alkyd) varnish or polyurethane should do.

Won't have as high a temperature rating as a proper electrical varnish, so monitor it and don't let it get too toasty.  Once the surface has dried, bake it to cure the inside as well.  Use a modest temperature (maybe 100C, or whatever maximum temperature is allowed for your product) and enough hours until it stops smelling.  Use a small (toaster?) oven outdoors if possible.

Also, ideally you'd pull a vacuum to draw goo into every crevice, but if you don't have one of those available, just letting it soak for a few days is the best you can do.

Tim