PWM Dimming of a Low-Voltage DC Incandescent Filament (Thermal Shock?)

Question

I plan on driving some Numitron tube filaments from an IV-16 tube (datasheet here). For ease of access I'll give the key specs here:

• Nominal filament voltage: 3.15V
• Max filament voltage (steady-state): 4.5V
• PWM frequency (to avoid mechanical resonance): <105Hz or >1000Hz

I have two main questions:

1. Assuming I was to 'fade in' a segment using PWM, over a relatively long period of time, such as in the design by the legendary DiodeGoneWild here, would I not exacerbate the effects of inrush current/thermal shock on the filament? Essentially wouldn't the filament experience more 'harsh transitions' under cold-resistance conditions than just switching it on once?
2. What actually causes the failure of an incandescent filament? Is it the inrush current at cold? Is it the thermal cycling? The mechanical stresses caused by repeated expansion and contraction from the thermal cycling?

Combining those two questions, is it actually a case of: Yes more high-current pulses than just turning on and leaving on BUT less stress on filament because thermal inertia of filament averages out the heat produced by those pulses into a gradual ramp up of tmperature and that prevents the underlying failure mode of the filament?

For the purposes of this question, assume driving specs of 5V, >20kHz square-wave PWM for the filaments, no additional series resistance or other current limiting.

And as a final bonus question: are there better (for filament longevity) drive methods: Linear? Other digital techniques (Pulse Density Modulation, for example)?

Thanks in advance for any input.

Answer 1

It takes time for a filament to heat up. If your PWM frequency is much faster than the thermal time constant of of the lamp, you effectively won't be rapidly turning the lamp on and off, but rather limiting the amount of current going to it. As such, the thermal shock won't be any worse than turning it on once--and will actually be better, because you're slowing down the temperature change.

Answer 2

Before high output LEDs became available, I made a softstart + voltage regulator for incandescent bulbs to go from 1000+ hours lifetime down to single digit numbers to gain ~50 % efficiency from baseline for flashlight applications. At 1000 hours lifetime, inrush isn’t a big problem for the bulb itself, but your MOSFET will suffer from the ~12x inrush current into a cold filament, so you may want to dim up slowly for that reason. Operating very close to the melt point of tungsten, you have to soft start it. I sacrificed a lot of 12 V halogen bulbs in the process.

Answer 3

PWM to the filament works OK as long as 100% duty cycle drives the bulb at about rated current/voltage. Thermomechanical resonances have to be avoided as you stated as well.

Some people try PWM-ing a low voltage bulb - say 12V - with say 60V pulses. They forget that at a suitably high voltage across the filament, a light bulb is a directly-heated diode tube with anode connected to filament. The higher voltage pulses of correct polarity will bypass the filament and flow as they would in a diode, via free electron current in vacuum.

When using PWM with voltages much above the rated bulb volrage, polarity matters and you need to connect the bulb such that the free-electron diode is reverse-biased (!).

Answer 4

"Thermal shock" effects incandescent lamps are harmful not because of expansion and contraction, but because parts of the filament that are cold will convert far less power to heat than parts that are already hot. If current is passed through a cold filament, any parts that are thinner will have higher resistance per unit length, and thus drop more voltage and dissipate more heat than parts which are thicker, and will have less thermal mass to limit the rate of temperature increase. Then, as those parts get hotter, their resistance will increase which will in turn result in them dissipating an even larger fraction of the overall power.

What prevents thermal runaway are the facts that thermal radiation and thermal conduction both increase with temperature; the parts of the filament that are hotter will release more heat through radiation and through conduction to the surrounding gas and supporting materials than the parts that are colder, and that as the entire filament gets hotter, the relative differences in electrical resistance between the hottest and coldest parts will diminish.

If one drives a filament with e.g. 25% PWM at a rate which is much faster than the rate it which it can heat up or cool down, then during the brief period the filament is on, the thinner parts will heat up more than the rest, but then there will be an interval during which the heat can spread from the hotter parts to the cooler parts without more heat being added to the hotter parts. Then there will be another burst during which the thinner parts are heated more than the thicker parts, but because the temperatures will have had some chance to equalize the difference won't be as great as it would have been without that idle time.

Using PWM to drive a bulb at a relatively low current may result in an equilibrium resistance ratio that's higher than the equilibrium ratio at full brightness, but under those conditions no part of the filament would be getting very hot, and thus even sustained operation under such conditions would have no detrimental effects. What causes accelerated wear are conditions where parts of the bulb reach temperatures above those intended for sustained operations. If nothing in the bulb gets that hot, there won't be any accelerated wear.