What I mean to say is this: Most commercially available units at your

local hardware store use an iron core transformer as the ballast, so

isn't there a readily available piece of junk electronic device (such

as an old tv or something) that I can hack open and grab a transformer

from that will work? How about a flyback transformer.

You ain't gonna make a 60 Hz ballast for an HPS lamp with a piece of

ferrite that small. Easiest hack from a junk TV is if you can find a

vacuum tube one and find the vertical output transformer. A decent iron

core one in a vacuum tube TV set will have a core 2.375 or 2.625 or 3

inches long, 3/4 to 1 inch wide, and about 1.98-2.5 inches tall (plus

perhaps 1/16 inch each way for the mounting fitting whatever they call

that). Those have E-I cores, and are easier to hack than most iron-core

transformers with E-I cores since these are gapped and have all of the E

pieces together and all of the I pieces together, AND these usually do not

have any welds. (Many ballasts have welds.) Most other iron core

transformers have the E pieces and I pieces interlieaved, and at least

sometimes maybe often glued or at least effectively glued by some glop or

another.

Or a step down

transformer from a power supply (like out of a vcr or something) used

in reverse with the output coil as the input coil instead.

None of these will work without serious modifications, many will not

work even with any possible modifications, and most are not that easy to

modify.

I'm not sure what is electronically required to light a 70w or 150w HPS

light bulb, I think it's something like 4kv at low current to sustain

the arc,

That is to start the arc. More typical is 4 KV repeated pulses.

and the ignitor is some sort of bimetal or capacitor used to

create a pulse initially to start the arc.

Capacitor and a triac and a few other bits.

That said, I can get the ignitor, bulb and all other nessecary hardware,

so I don't even really need to know much about that. What I need are

some clues as to exactly what kind of transformer is needed. I'm pretty

sure it can't be too tough. Aren't there some tolerances and forgiveness

involved so as to allow me to make a direct substitution from somewhere

else without causing a serious safety issue? -Evan

Let's suppose you get an E-I core with 1 square inch center leg, overall

dimensions 3 by 2.5 by 1 inches. That may come from a 26 inch color

vacuum tube TV vertical output transformer, or that transformer may have a

core a little smaller.

Start with typical lamp voltage of 55 volts and typical line voltage of

120 volts and a design of simple series inductor ("reactor") ballast.

Voltage across the ballast steady-state is ideally square root of line

voltage squared minus lamp voltage squared, meaning 106.6 volts ideally.

Resistive losses in the ballast screw this up a bit to voltage across the

ballast being less, but I think to about the same extent you need to plan

for high line voltage and low lamp voltage - so plan for 106 volts.

I have heard 12,000 gauss as being a reasonable design for peak magnetic

flux density in steady state operation of an iron core magnetic component

of such size using older type core materials. You can probably get away

with 12,500 maybe 13,000, but I would avoid aggressive operation of a

homebrew device for long where it is not under supervision. 1 square inch

is 6.45 square centimeters. 12,000 times 6.45 times times 60 times 2

times pi is 2.918 abvolts per turn peak. (The abvolt is CGS system

voltage unit where K-prime is 1.) Divide abvolts by 1E-8 to get volts,

and divide by SQR(2) to get RMS volts per turn - that is .206 volts per

turn. For 106 volts, this means you need 514 turns for conservative

operation, 475 turns even for aggressive operation.

The window in a 3 x 2.5 x 1 inch E-I core with 1 inch wide center leg is

1.5 by .5 inches. The largest size magnet wire that will fit that many

turns even theoretically with typical insulation thickness is AWG 21, so

optimistically you could cram in that many turns AWG 22. Kep in mind,

the winding has to generate 4KV pulses, so you need thin insulation

between layers (in addition to the coating that magnet wire has) plus

some decent insulation between the core and the widning.

I will continue the design process assuming that you manage to get 22

AWG wire to fit. I give enough info below to let "someone skilled in the

art" to redesign for a 35 watt HPS lamp without going astray enough to

cause an actual problem should 23 AWG wire be required.

For overall heating concerns, I would design for 12,500 gauss peak at

most (494 turns) and allow at most 5 watts of heat to be generated by

winding resistance when the winding resistance is elevated to what it

would be at 100 degrees C. In fact, I would rather not exceed 4 watts

winding resistance heating in a component of this size. 22 AWG

copper wire resistance at that temperature close enough to 1 ohm per 50

feet. The average turn in a half inch wide window on a 1 inch square

center leg has four segments 1 inch long and four 1/4-circle segments of

1/4 inch radius or .3927 inch ech, for average turn length of 5.571

inches. 494 turns at this rate is 229.3 feet, worth 4.59 ohms. The

roughly 1 amp typical current of a 50 watt HPS lamp means roughly 4.59

watts of heat from winding resistance - borderline good for 50 watt

and no way for 70 watt unless you have some really high temperature magnet

wire and insulation (in which case you may be good with the roughly 1.4

amps of 70-watt).

Lamp amps is normally a bit more than ratio of lamp watts to lamp volts

since the lamp (bulb) has power factor a bit less than 1 due to the arc

characteristics causing distortion of voltage and current waveforms.

So, let's go for the 50 watt lamp. Lamp voltage is nominally 55 volts

steady-state (varies widely with age and condition for that matter). I

guesstimate lamp power factor of .95 - 50 divided by 55 and divided by .95

is .957 amp.

Now for typical ballast voltage: I guesstimate 80 degree phase angle

between lamp voltage drop and ballast voltage drop (ideally this is 90

degrees). The ballast has voltage drop leading current by a few degrees

less than 90 and the HPS lamp's arc has voltage drop leading current by a

couple to maybe a few degrees.

Cosine law is that line voltage squared is ballast voltage

squared plus lamp voltage squared plus twice the cosine of phase angle

times product of lamp and ballast voltages. Using 120V line voltage, 55

volt lamp voltage, and my guesstimate of 80 degrees, the voltage drop

across the ballast to the nearest volt is 99 volts. Divide by .957 amp

and the ballast requires impedance of about 103.5 ohms.

Assume the ballast has VA (volt-amps) of 99 times .957 or about 95, and

6 watts actual power dissipation. This is getting to be an educated guess

of mine, but I consider it reasonably accurate. Arc-cosine of 6/95

is 86.379 degrees, sine of which is .998 so voltage across the inductive

reactance is close enough to 99.8% of the total voltage across the ballast

so basically unchanged from the above 99 volts to whatever extent that

above 99 volts is a good figure for voltage across the ballast. That

means inductive reactance is down only .2% from the 103.5 ohm ballast

impedance that I got above (or from whatever it actually should be - close

enough for an HPS lamp), so go for 103.3 ohms.

103.3 ohms inductive reactance divided by 2, pi and 60 means inductance

of .274 henry.

Now, designing a gap to make this thing achieve an inductance of .274

henry.

Inductance of a gapped inductor in abhenries (CGS inductance unit

with K-prime being unity, which is conveniently same as nanohenries) is 4

times pi times magnetic path cross section area times square of number of

turns divided by effective gap thickness.

.274 henry is 274,000,000 abhenries. Divide by square of above 494

turns to get inductance of a 1-turn winding on same core with same gap, or

1122.785 abhenries. Divide by core cross section area of 6.45 square

centimeters, and that is 174.089 abhenries if gap is unchanged but

magnetic cross section is reduced to area of 1 square centimeter. Divide

by 4 and pi to get reciprocal of effective gap in centimeters being

13.8536 or effective gap thickness in centimeters of .0721836.

Next step - consider effective gap thickness of the core material. Go

along the centerline of a path through half of this core and that is 4

inches or 10 centimeters, a bit less considering rounding of flux paths

turning through corners, so I would say 9.5 centimeters. Divide that by

the permeability of the core material to get effective gapping of the core

material itself - I am falling short of citable numbers here but I pull

out of a hat 20,000. This means that the core accounts for .000475

centimeter of gap, subtract that from .0721836 to get .0717 centimeter of

required actual gap.

Since a gap between an E-stack and an I-stack is crossed by magnetic

flux twice, you need your gap between the E-stack and the I-stack to be

half this, or .03585 centimeter or .3585 millimeter. It is fair to round

this to .36 millimeter since an HPS lamp can easily tolerate power input

being 10% off in either direction.

4 sheets of 20-pound copier paper appears to me to be "in-range" here.

There is another step: The winding needs a tap at roughly the 4% point

(20 turns from one end of the winding I think) for the ignitor to make

this ballast generate a 3-4 KV starting pulse. The tap should be close to

whichever endpoint of the ballast has skimpiest insulation from the core,

since the high voltage with respect to everything else will be at the

other end of the winding. Given convenience of putting the tap in an

outer layer of the winding, you need the inner layers of the winding

adequately insulated from the core for the full 4KV.

**************************************

So, if you pull this off, what do you get? A homebrew ballast! It is

not certified by UL or whatever other organization that does safety

testing that could cost $$$$$! This means extra liability should a fire

worth an insurance claim start at this thing, even if the fire started for

a reason other than flaws in design or construction.

MY WARRANTY: If anything goes wrongo, blammo or blooie or if any fire

problems arise or if anyone gets electrocuted or so much as shocked into

breaking someting (including body parts) or falling off a ladder or

anything else even somewhat along these lines goes wrong from someone

using info that I posted, I will refund what I got paid for posting this

and no more. Info here is worth education only, and putting it into

application has (maybe at best) risks typically adequately mitigated by

having devices (whether or not made using info like that above) passing

actual certification testing by a recognized safety certification body

such as UL or CSA.

If you want a ballast that does not require constant supervision by

qualified technical personnel on hand to quickly disconnect power and

quickly hit it with a sufficiently convenient and suitable fire

extinguisher with both A and B ratings and to do adequately quickly

whatever else ends up being necessary should the ballast go blooie, then

don't homebrew it but buy something approved by UL, CSA or whatever.

- Don Klipstein (

[email protected])