Does anyone know why a fluorescent lamp oscillates when run on DC? I’ve noticed that almost all linear lamps 4W to 20W and 22 Watt Circline lamps will oscillate from 1500-2800 Hz when powered by DC using a resistor ballast. There is a sine wave a few Volts p-p across the lamp when viewed on a DC blocked scope. If you run a preheat choke ballast type fixture on DC by adding a series resistor to limit current, the oscillation will be audible as a whistle that is heard coming from the choke. If you run 2 lamps in series, each one will oscillate at its own frequency  creating a beat note as well.
Actually a preheat choke type fixture will oscillate on AC as well. I have heard a high frequency whistle mixed in with a 120Hz AC power buzz emanating from small 8 Watt preheat fixture choke ballast.  I asked this once before and no one knew why... I thought it might be worth a shot to post it again....

This is conjecture, but could it be do to waveform distortion? I know fluorescent lamps only have a power factor of around 0.9 when run on a resistive ballast, due to the waveform distortion.

The only other thing I can think of would be that you accidentally formed a negative resistance oscillator, but I don’t know if the capacitance between the wires would be enough to cause oscillations, considering how small it is.

Have you tried looking at the waveform of the oscillations? If it is a sawtooth wave, you might have a negative resistance oscillator.

Thanks for your quick reply. I haven’t looked at the waveform recently but vaguely remember it being more of a sine wave than sawtooth... but is certainly worth looking at again. As you I believed the oscillation was caused by negative resistance. The frequency got lower when I bridged the F4T5 “oscillator “ with a 0.22 uF cap.  Back in the 70’s as a kid I heard a strange whistle like sound mixed in with 60 Hz coming out of a 2 wire choke ballasted F15T8 and F8T5 fixtures. It was not until the late 80’s when I was in college I heard that strange sound coming out of my classic brown 2 lamp F15T8 articulating arm desk lamp with 2 choke ballasts in the base. This peaked my curiosity enough to build a simple “suicide” DC supply out of a bridge rectifier and big electrolytic capacitors and got HV DC from the 120V mains. I current limited my desk lamp with a 75 Watt incandescent light bulb in series. The DC voltage eliminated the 60 Hz component of the noise and only the 1500-2000 Hz whistle emanated from the choke ballasts. Since that time I always wondered what caused that oscillation. My recent DC lamp was an F4T5 run from 48 V DC (4 twelve volt gel cells) through 120 Ohm 5 W resistor no choke. Once the lamp starts you can clearly hear a 2KHz tone by placing a “hound” inductive amplifier that’s used to trace telephone lines near the lamp. Clearly it oscillates without a choke...

Discharges exhibit a bit of negative resistance (higher current leads to higher cathode spot temperature, so lower cathode drop),
but also the dynamics of the ionization may cause the system to become unstable and oscillate: In an equilibrium state the free charge avalanche multiplication is suppossed to generate equal amount of free electron/ion pairs, as they recombine.
But both effects have some time delay, so cause a phase lag (when expressed as how they respond to a small superimposed sinewave like changes). So we have a system with two phase lags in the loop (one is voltage so multiplication rate -> amount of charges generated means an integration, so 90deg phase lag, other is gradual recombination, forming a kind of exponential decay, so could be described as a low pass filter with some time constant, so again up to 90deg phase lag), which could lead to phase lag approaching 180deg.
That means in the loop (voltage -> charge multiplication + charge decay -> amount of charges -> conductivity -> current -> voltage) which is has normally 180deg phase shift (it is a negative feedback, at least per static behavior), the extra 180deg phase shift means the originally negative feedback becomes positive (see Bode stability  criteria) when dynamic behavior taken into account. The remaining few degrees lag to the full 360deg may come from either the electrical response (e.g. a capacitance adding extra lag current -> voltage) or even within the tube when taking the finite speed of sound waves within the plasma (how fast the electrons and ions distribute, so causing extra delay), you may loose all of the remaining phase margin so get an oscillator.


And for the power factor:
The PF=~0.9 comes just from the fact at mains frequency AC supply, the current is nearly a sinewave, but the voltage is nearly a square wave, so they have quite a mismatch. And if you calculate just an ideal sinewave vs an ideal square wave, you get PF = (2/pi) / sqrt(1/2) = 0.900316...