Do tungsten lamps burn out more quickly when you turn them on and
a lot? One of mine blew out, but only when i was turning it on. It
has been on for about 30 minutes total, but i was turning it on for
only 5 minutes at a time.
Was this just a bad lamp, or should i just leave them on longer?
I was trying to save the useful spectrum life.
Years ago, I ran a test to see how many times a lamp could be turned
on/off before failure. A computer turned the lamp on for one second and
then off for one second, and repeated the sequence until the lamp
failed (determined with a sensor). When the computer was connected to a
mechanical relay, which most closely modeled a regular light switch, the
lamp would fail after a few hundred cycles to a few thousand cycles;
however, when the computer was connected to a SSR (solid state relay)
with zero voltage crossing detection, the lamp lasted longer than the
test (several days).
The test seemed to prove that if you turned a lamp on exactly during the
crossover part of the A.C. sine wave. Remember that in an A.C. circuit,
the voltage starts at zero volts, climbs to about 190 volts positive
(120 volts A.C. is the root mean square of the voltage, not the peak
voltage), drops to zero volts, continues dropping to about 190 volts
negative and then climbs again. In any case, the sine wave crosses the
zero point 120 times per second for those of us who have 60 cycle power.
and 100 times per second for those who have 50 cycle power. Using a
device, such as a solid state relay or dimmer control that has zero
voltage crossover detection should give you the maximum life of the
bulb, all other things being equal.
Do you happen to know the circuit one would use to implement this? I wonder if
this could be done cheaply enough with simple components (e.g., triacs, SCRs,
whatever) to make it worthwhile to build a little box for each bulb whose live
one wishes to extend.
It turns out that, contrary to cutesy lists of absurd laws and email
sigs, it is actually *not* illegal to carry an ice cream cone in one's
There are several low-tech approaches:
1.) Just stick a suitable rheostat or potentiometer in series with the bulb.
2.) Stick a big diode in series with the bulb, and short out the diode
when full brightness is required.
3.) Fancy switch to put the bulbs in series when full output is not
required and in parallel when full output is required.
4.) Bite the bullet and use Quartz-Halogen bulbs. Easier to color
correct for color anyhow, I suspect.
If you want to go the higher tech route, you can drive the bulb from a
constant current source, where the current is about the amount required
to give it normal full power. This protects the bulb from the starting
.~. Jean-David Beyer Registered Linux User 85642.
/V\ Registered Machine 73926.
A low-tech version of this high-tech route is to find a ferroresonant
"constant voltage" transformer whose output rating is equal to or a bit
higher than the set of bulbs you want to power. These are large and
heavy; they weight about 70 pounds per 1000 W of output. They used to
be expensive, but I've found a number of them selling for very little in
surplus places. They were typically used to power expensive server
computers, and those have largely gone the way of the dodo, surplusing
the power conditioning box.
Anyway, constant voltage transformers have a couple of interesting
behaviours. When the input line voltage varies +-15% from its nominal
value, the output of the transformer varies only 2-3%, keeping your
lights at constant brightness and colour temperature. (That's where the
"constant voltage" label comes from.)
They *also* have inherent current limiting. No matter how heavy a load
you connect to their output, including a dead short, the output current
is limited to about twice the nominal rating of the transformer. The
transformer will hum loudly, but it won't burn up or be damaged by a
short of any reasonable duration.
A 500 W transformer is rated to deliver about 4 A at normal voltage,
and it will never put out more than about 8-10 A. This is considerably
less than a 500 W halogen lamp will draw when cold, so if the lamp is
plugged into a 500 W CV transformer, the transformer will provide
automatic current limiting that pampers the lamp.
Now, the CV transformer I'm talking about has no active parts other than
a transformer and one or more capacitors. There are no electronics.
There's another type of voltage-regulating transformer that uses
electronics to switch among several different transformer taps as the
line voltage changes. This protects computers agains brownouts too,
but it's not very useful for powering lamps. It regulates voltage only
in large steps, and it has no inherent current limiting. You don't want
a tap-switching regulator for this.
Seriously OT, but these devices are of considerable conceptual interest.
They are in essence "parametric oscillators" (at least some of them, the
ones I know about are).
The core of the device is not actually a transformer, but rather a big
inductance that looks much like and is physically made like a big iron
transformer. This inductance, which would be essentially the output
winding of the transformer if it were a transformer, is paired with an
equally big capacitance to make an LC circuit resonant at 60 Hz (for
U.S. versions anyway).
The input windings on the transformer are arranged such that they induce
no voltage in the output winding, for example by counter-winding two
series-connected input coils on the input leg of the transformer so they
cancel each other. Instead of inducing an output voltage, the input
current merely saturates the iron underneath these coils, thus
time-modulating the L value in the resonant circuit -- and note that
this happens at twice the input frequency, i.e. at 120 rather than 60
Hz, since current in either direction tends to saturate the iron.
So, you have an LC circuit resonant at 60 Hz, with one of its reactive
elements modulated in value at 120 Hz, and that's the essential
condition to have a so-called "degenerate parametric oscillator". If
the modulation is strong enough compared to the losses in the circuit,
it will break into parametric oscillation at 60 Hz, and that's what
drives the load. If you put on too much load -- including a dead short
-- that kills the oscillation, thus giving you complete short-circuit or
This device thus operates on exactly the same basic parametric principle
that governs how a child pumps a swing; radio-frequency "varactor diode"
amplifiers (variable C instead of L) that used to be used as very low
noise RF receivers a few decades back; and optical parametric amplifiers
using nonlinear crystals that split "pump" photons into "signal" and
"idler" photons and are routinely used to generate tunable optical
signals at wavelengths where tunable lasers are not readily available.
It's a neat invention.
A guy named Kent Wanlass (one of two Wanlass brothers) may have invented
it, or at least made and sold these several decades ago.
The device you go on to describe is not what I'm familiar with. The
constant voltage transformers I referred to are made by Sola, Superior
Electric, and Hammond among others.
As I understand it, they are indeed transformers. The simplest version
has fairly ordinary-looking primary and secondary windings wound on the
centre leg of a shell-type transformer core. Unlike a normal
transformer, where the primary is wound over the secondary (or vise
versa), the primary and secondary windings are physically separated.
Magnetic shunts (chunks of transformer steel) are inserted between
the centre and outside legs of the core at a point between the primary
and secondary winding. These magnetic shunts provide a flux path around
the primary that bypasses the secondary winding, producing lots of leakage
inductance. This is what limits the current when the secondary is
Meanwhile, the secondary winding is parallelled by a capacitor, chosen
to make the secondary resonant at 60 Hz. The resonance drives the
portion of the core inside the secondary winding into saturation, which
limits the amplitude of the secondary voltage. Changes in primary
voltage have almost no effect on secondary voltage over the regulating
Now, the above is actually a simplification. In real CV transformers,
the secondary actually has enough turns to step up the voltage by a
factor of several, so the capacitor is operating at several times line
voltage. This allows the capacitor to be lower capacitance for
resonance, which is physically smaller and cheaper than what you'd need
at 115 V. The actual output voltage is obtained from a tap on the
secondary where the voltage is 115 V or so.
Also, the transformer I've described so far outputs a pretty square
waveform. That's great for the input stage of a DC power supply,
but not for some AC loads. The commercial CV transformers I see use a
"harmonic neutralized" design that gives an output closer to a sine
wave. Instead of one secondary winding, there are two, with another
pair of magnetic shunts between the two secondaries. The capacitor is
connected across the two secondary windings in series. The output
voltage is taken from just the "middle" secondary winding. In the Sola
transformer, there's also an air gap in the centre leg of the core, at
the end where the 3rd winding is. I don't understand how the extra
winding and shunt cancel some of the 3rd harmonic output, but they do.
I don't have extensive experience with constant voltage transformers,
but did play quite a bit with one of the "parametric oscillator" type
commercial power supplies a few decades ago out of conceptual interest
and as a classroom demonstration of the basic phenomena.
The experimental stigmata you could look for to see if your devices fall
within my description are:
1) Existence of an oscillation threshold as input pump power is
increased: Put a variac on the input and a medium load on the output,
turn the input voltage up slowly from zero, see if the output voltage
"turns on" or suddenly jumps from low value up to near full value as the
input voltage approaches the specified input value.
2) Sudden cessation of oscillation if output is overloaded. Gradually
increase the output current drawn from the secondary with fixed input
voltage, e.g, by adding more and more small light bulbs in parallel
across the output. As you add more load the output voltage may
gradually droop, but if at some point a small increase in load causes
the output to suddenly drop to essentially zero, that argues for the
parametric type behavior.
Of the CV transformers I have around to play with, the Sola does seem to
behave like an oscillator. As you slowly increase the input voltage,
the output voltage also slowly increases for a while (no CV action).
Then, suddenly, the transformer starts its characteristic humming, and
the output voltage jumps up near its standard output value. After that
point, increasing the input voltage further produces small changes in
output voltage. In addition, if you take the input voltage back down,
the transformer drops out of oscillation at a lower voltage than that
needed to start it. So there is some hysteresis in the oscillation
All of the above describes what happens with no load on the output.
I think that under load, you get the same behaviour but it takes a
higher input voltage to produce oscillation.
The other two (Superior Electric and Hammond) do not suddenly start
oscillating in this way. As you raise the input voltage from 0 to about
70 V, the output voltage is just 1.5 times the input voltage. Then the
voltage out curve begins to flatten out, and by about 90 or 95 V in the
curve is quite flat but still positive slope. As input voltage goes
from 100 to 120 V, output voltage rises by only 2-3 V. There is no
point at which the transformer suddenly starts humming, and no
hysteresis in the voltage in/out curve.
I haven't tried that. The transformers are currently buried in the
basement, so it's not easy to make new measurements.
On the theory of ferroresonant transformers:
Several years ago I was filing a patent
(04356433) on a controllable HID power supply.
The attorney mentioned he had represented
Sola in a patent infringement suit against
At the conclusion of the trial he was convinced
that no one at either firm had the foggiest
idea of how the device really worked.
The effect was first discovered by accident and
the transformer variant was developed by
mucking around on the bench.
Search the patent literature for further confusion.
BTW: Be aware that a ferro draws a hefty amount
of power even with no load: be sure to turn
the transformer off at the conclusion of the
darkroom session or the electric bill will
rise by $100/year -- A 100 Watt load left on
for a month costs $7.30 at 10 cents/KW/Hour.
Also a ferro has quite an overshoot when turned
on -- don't turn the thing on with the lamp
connected to the output, i.e. don't control
the lamp by connecting the timer to the input
to the ferro.
Nicholas O. Lindan, Cleveland, Ohio firstname.lastname@example.org
Consulting Engineer: Electronics; Informatics; Photonics.
in article RbVkb.6081$ email@example.com, Nicholas O.
Lindan at firstname.lastname@example.org wrote on 10/20/03 10:46 AM:
At one time, I was working at a company that used Wabash transformers for
charging capacitor banks. The idea was that you could run them into a short
circuit (discharged capacitors) and then not over charge when you hit the
constant output voltage. They ran HOT! The power factor was not all that
great either. It probably averaged about 40% over the charging cycle. I
looked at the waveform of current in an external capacitor supplied with the
transformer. It was not a pretty sight. It looked like, IIRC, an
asymmetrical sawtooth. Although I read up on ferroresonance, it never made
One solution I tried was to get a current limiting (high leakage inductance)
transformer. A capacitor was put in parallel with it. That caused the power
factor to change from leading to lagging during the charging cycle. This
gave a reasonable average power factor.
Another interesting item is that you would get rectified current flowing
into the capacitors only while the instantaneous ac voltage exceeded the
bank voltage. This meant that the charge rate slowed down as the bank
charged. That is, you only charged near the peak of the sinusoidal inpaut.
To get over that, I used a much higher open circuit voltage than what was
otherwise necessary. To preven accidental overvoltage, I used two voltage
measuring circuits to detect when proper voltage was obtained.
To get around this problem, I devised the monocyclic transformer described
here in an earlier post.
Polytechforum.com is a website by engineers for engineers. It is not affiliated with any of manufacturers or vendors discussed here.
All logos and trade names are the property of their respective owners.