I decided to set up a standard preheat circuit using my Robertson L48 ballast, but instead of a starter, I instead used, in what was perhaps a dangerous endeavor, the 280VAC, 6 uF capacitor from my mercury vapor fixture. I got some pretty interesting results.

In trying a 4 watt lamp, the lamp started instantly upon applying power. It was just very slightly dimmer until I removed the capacitor from the circuit. With an 8 watt lamp, the lamp started in a small fraction of a second, and I could see the cathodes glow before it did.

Next I tried an F15T12, and things got a little more crazy. Obviously the wrong type of ballast, but I knew it would be possible - just more difficult - to strike. Upon applying power, the lamp instantly struck a rather violent cold cathode discharge. The lamp was bright, and the cathodes were bright blue. After a fraction of a second, things calmed down and the lamp behaved normally.

What kind of monster did I create here?   I'm assuming I created a resonant circuit which multiplied the OCV? Is there risk of damage to the ballast or the capacitor by doing this? With the right value of capacitor, could you turn a regular preheat circuit into a nifty resonant starting circuit like a CFL?

Have you posted anywhere or can you post a pic or a diagram of exactly what you did? TIA.

I just made a standard preheat circuit, but with the capacitor in place of the starter. I whipped up a quick diagram for you, anyhow. 

I think that the british SRS ballasts and most electronic ballasts operates in a similar manner. SRS ballast uses a capacitor to allow the lamp to hot start directly from plain 240V.

I think that the british SRS ballasts and most electronic ballasts operates in a similar manner. SRS ballast uses a capacitor to allow the lamp to hot start directly from plain 240V.

Not really. The winding sections are polarised so, when there is no arc, the magnetic field cancels each other, so there is no effective inductance in the capacitor circuit. The capacitor current is then utilized to provide cathode heating and across the lamp is practically an unchanged mains voltage (so 240V). That is sufficient to start the F40T12, but not the F36T8.
When the lamp ignites, the arc current splits so, the reactive component (90deg lagging behind the mains voltage) flows from the capacitor branch and the resistive component (no phase shift) flows from the mains. That mean the arc is effectively fed from both pins, so the total arc current means lower load for the filament.
The circuit is called "semi-resonant" and not "resonant", because the presence of the inductor and capacitor may suggest a resonance, but the inductive coupling effectively kills the resonance for the starting completely.


With electronic the inductor does not have it's section in series with the capacitor, so the inductance stays unchanged in the circuit, so the circuit does exhibit the resonance and this resonance is utilized to boost the voltage for ignition. Therefore this circuit does not have problems to ignite a lamp requiring about 1kV peak for a reliable ignition from an inverter with just 160Vrms output (NPF F36T8).
Principally this could work on any frequency, but to prevent the resonant capacitor ffrom forming a huge current spikes through the arc, it requires the plasma ionization time constant to be way longer than the operating frequency. And as no one can change that with given lamps, it is the mercury discharge properties, what restrict this resonant concept for HF operation only (above 20kHz).

So this is actually how an SRS system works? Can what I've done here actually work in the long run, with no harm done to the ballast? If so, I'll totally make a demo-test fixture out of this - that would be way cool!   Now, the trick would be finding the proper capacitor... I assume one much smaller than 6 uF since that seems to cause rather extreme results.  

What you have created is not a SRS, but a really resonant circuit.

The ballast would be OK there, what will suffer a lot is the lamp.
I don't know the spec of your ballast, but 6uF would be way too much even for a 0.43A F36T8 ballast.
The capacitor reactance has to be slightly lower than the ballast inductive reactance. The "slightly" mean the circuit will operate a bit off the resonance peak, so the voltage and currents would be rather limited even when the lamp won't start.
And definitely the voltage rating would have to be at least 750VDC/500VAC as a minimum.
So for a 0.43A F36T8 ballast the capacitance would be about 5.5uF

What would be the proper capacitor for an F14T12 on the proper choke, 3 or 4 uF? Also, would a film capacitor be okay for the job? Since they're cheaper than oil-filled ones...

No capacitor OK for the resonant circuit at the mains frequency. The resonance would require too high capacitances, so the lamp would get severely stressed by the peak current when the capacitor discharges at the reignition after the current zero cross. So although the circuit would be able to generate sufficient electrode heating current and voltage for ignition, it is the discharge itself, what does not tolerate the capacitors.

For the resonance to work, the 0.16A ballast will generate sufficient current and voltage with 2uF, so again way too much for a fluorescent.

The lamps could tolerate maximum ~10..33nF, but not much more. And that capacitance can not do more than prevent the arc buildup across starter contacts.

Operation at high frequency suffice with just few nF for the resonant circuit to work, what is way safe for the lamps, so inside a HF ballast the resonance concept could be used. But by far not at 50 or 60Hz.

Ah, okay. Thanks for the info. It was a cool concept, anyhow.

With this type of setup you are shifting the phase of the lamp in relation to the supply. The voltage changes across the ballast would be large and sharp. The voltage may touch 240v across the ballast then abruptly switch the other way as the mains cycle reverses causing a massive collapse of the magnetic field of the ballast and a huge surge of voltage and current being sent straight into the lamp. No wonder it glowed blue!
I tried something similar too here in 240v land. I used an 8uf capacitor on a 15w t8 and it absolutely obliterated the electrodes! (this was the same tube that would get melted down later)
I suggest you place a second lighting ballast in series with the capacitor (about 36w rated for a 20w t12) to help cancel out the phase offset caused by the capacitor, it would significantly reduce the damage to the electrodes and still maintain the heating during startup. I have used this system successfully for a while now.

It is not the phase shift on the main (50Hz) harmonic, that should be capacitive for the resonant concept to work.
But it is the capacitors low impedance at higher frequency components coming from the lamp (the abrupt voltage drop when it reignites after a current zero cross). The resulting effect would be the same as driving the lamp via a capacitive ballast .

Adding an inductor in series to eliminate this effect would theoretically work (but the capacitance would have to be made lower, so the combination of the inductor and capacitor would yield the required reactance). The inductor would have to be chosen so, the resonance would be at about 120Hz (for a 50Hz system; for 60Hz the resonance would have to be at about 150Hz) But that would mean the capacitor would have to handle about 50% higher voltage than required for the lamp. As that inductor will impose extra losses, the complete circuit would become rather inefficient and due to the required capacitor voltage rating (a 2kV cap for e.g. an F36T8).

I'm wanting to experiment a bit more with this when I get the time. Perhaps try and hack up a semi-resonant start system, or maybe an LOA-style system that could run an F40T12 from 120V. But I need some capacitors. If I wanted, say, a 1 uF capacitor, would one like this by okay?

http://www.ebay.com/itm/450V-AC-1uF-50-60Hz-Polypropylene-Film-Motor-Fan-Start-Run-Capacitor-CBB61-/331293865892?pt=LH_DefaultDomain_0&hash=item4d22a737a4

I have no idea how much current capacitors like these are meant to handle.

1uF won't be enough.
The capacitor with all the inductances has to be in (or close to) resonance at 50/60Hz.


And although "semiresonant" ballast has the word "resonant" in it's name, there is no resonance effect taking place during starting. When there is no discharge in the lamp, the ballast windings are polarised so, all the inductance seen in the circuit (phase-1'st winding-1'st filament-2'nd winding-capacitor-2'nd filament-Neutral)is cancelled out, so there is just the capacitor current providing the filament heating and the mains voltage is directly transferred across the tube.

Don't worry about the capacitance. My question is, would that particular capacitor be safe to use for these particular experiments? Because I have no knowledge of what the different types of capacitors are able to handle.