I've noticed a few statements at LG about today's CFLs they're mostly resonantly started. Wikipedia speaks about a semi-resonant start which seems to be something different. The following two statements from the article suggest it's indeed another way.

As the electrodes heat, the lamp slowly, over three to five seconds, reaches full brightness.
Semi-resonant start circuits are mainly restricted to use in commercial installations because of the higher initial cost of circuit components.

Also the included circuit drawing doesn't much look like what I can see inside a common CFL. But the long article about fluorescent lamps does not write about another -resonant start. So how are those CFLs* that don't preheat for a short time started?

*here's an example of such CFL

I have three feit cfls that are kind of semi resonant. When the garage is dark, you can see the cathodes heat and then the arc strikes.

The semi-resonant means "like resonant", with the ballasts it means the circuit looks LIKE a series resonant circuit, but the way how it function, it isn't.
It describes the arrangements of mainly T12 magnetic ballasts where the ballasting coil has the winding split to two (parallel) sections.
The wiki article is actually wrong in stating the resonance is doubling the voltage or so, it is not the case (2x240V=480VAC, so 678Vpeak, would ignite the T8 without any problems, but clearly the T8's do not work on SRS, so I guess that is quite strong proof).
The reason is, once the tube is not lit, the ballast sections are connected effectively in series in such way, with any current there the magnetic field of one section cancels out the field from the other section. As the magnetic field cancels out, there is no inductance in the circuit (except small leakage inductance, but that is really low).
So in the circuit remains just the PFC capacitor just in series with the heater windings. With normal PFC cap design, that means heating current of about 80% of the lamp rating, so with that after few seconds the heaters will reach emission temperature.
As the magnetic field is not excited in the ballast, the mains voltage (240V) is present directly across the lamp. And for T12this is sufficent for an ignition with hot electrodes, so after the electrodes warm up, the lamp ignites.
Once the lamp ignites, the ballast starts to limit teh current in an usual way as in normal preheat circuit. The only difference is, the PFC connection and mains input go through separate wires. Because the polarity connection of the windings, that does not influence how the circuit behaves in an overall way.
There is just one difference towards normal preheat connection: As the arc is being fed from both ends (the reactive current part come from the capacitor side, the real power part from the mains), so the maximum filament load is about 20% lowerthan the regular preheat. (80% of the arc current from the caoacitor, about 50% of the arc current from the mains side; because phase shifted by 90deg, the vector sum is 100%)


Now the "Resonant start" in the CFL's works differently:
There the ballasting inductor is really connected in series with the capacitor, so the circuit does exhibit the resonance. And this resonance is really used to generate the high heating current for fast warmup together with high voltage for the ignition. And the circuit really operates at the resonance frequency (the selfoscillating ballasts) or the inverter justsweeps through the resonant frequency (the intelligent controller types) for the ignition.
After the lamp ignites, the arc resistance is usually so low, the Q of the resonant circuit becomes very low, so the resonance disappears after the lamp starts.
Because of the resonance effect, it is quite easy to generate really high voltages for ignition even from rather low AC voltage, so a simple two transistor half-bridge, generating just half of it's supply as the AC output (150V) is sufficient to power a lamp with 100V arc. Or with some ballast designs even higher (then the Q does not fall that low and the resinance is used to boost the voltage for the lamp).

@Medved: Thanks a lot for explaining this! Very interesting!

So in the circuit remains just the PFC capacitor just in series with the heater windings. With normal PFC cap design, that means heating current of about 80% of the lamp rating, so with that after few seconds the heaters will reach emission temperature.
As the magnetic field is not excited in the ballast, the mains voltage (240V) is present directly across the lamp. And for T12this is sufficent for an ignition with hot electrodes, so after the electrodes warm up, the lamp ignites.

I'm still struggling to understand why "the lamp slowly, over three to five seconds, reaches full brightness". The principle you've just explained doesn't suggest such a behaviour. I can imagine the ignition would look like that by an electronic fluorescent starter, wouldn't it? No reason for a fade-in. Actually, this 11-year-old Opple 2D lamp lits gradually up within about 1 second but there's most probably another cause for it. I'm afraid I've never seen that three to five second fade-in as described in Wikipedia!

Btw. do you remember first fluorescent lighting in our classical trams (used since early 1980's?)? There was a switch on the dashboard with a warning label that drivers should switch the lighting every time they stopped in final station thus preventing its damage. I suspect it had something in common with 600V DC used in this kind of traction. The polarity had to be altered? I have no idea how these fluorescents were ballasted though.

@Medved: Thanks a lot for explaining this! Very interesting!
I'm still struggling to understand why "the lamp slowly, over three to five seconds, reaches full brightness". The principle you've just explained doesn't suggest such a behaviour. I can imagine the ignition would look like that by an electronic fluorescent starter, wouldn't it? No reason for a fade-in. Actually, this 11-year-old Opple 2D lamp lits gradually up within about 1 second but there's most probably another cause for it. I'm afraid I've never seen that three to five second fade-in as described in Wikipedia!

The 5s is about the time required for the filaments to reach the temperature.
The lamp may ionize even before the filament reaches the full temperature, but because the discharge needs higher voltage, it either does not light or it lights only very dimly (just capacitive glow discharge).
But as electrodes warms up, their emission is able to support stronger and stronger discharge current, so that manifest itself by a slow brightness ramp up.
This effect is most pronounced on lamps with cathode guards: Extra electrodes taking over the ion bombardment. The ion bombardment (as a result from the starting discharge) heats up the place where the ions land, so when that place is the emission coated filament, it heats it up rather quickly, so you see the lamp starting in a flash upon ignition. But if the lamp has some form of cathode guards, the ion bombardment is redirected to these (or it could be just the filament support wires), so there is no heat to the cathode from it, so the cathode continues to warm up, so reach the full emission slowly, hence the supported current and so brightness rises slowly as well. Once the emission is able to take majority of the current, it redirect the discharge root (so the ion target, as well as the electron emission point) to itself and then the discharge becomes capable to maintain it.
Such slow brightness rampup is quite common for all the real rapid start ballasts (the SRS belongs to this category) with certain lamps.

With the electronic ballast (most likely the Opple "2D") the reason for the slow ramp up could have two mechanisms behind, one very similar to the rapid start described above (but different circuit) and the other for quite different reason:
The first one:
One of the programmed start methods uses PTC shunting the arc for some time after warmup before ignition. Just the PTC has one drawback: After the lamp ignites, the PTC is kept hot to remain OFF, so once the lamp is shut down for short time, the filaments cool down, but the PTC does not, so remains open, so the lamp cold starts.
In some designs this PTC uses to be in series with a VDR (or someone had experimented with a SIDAC with very similar results), so the combo is conductive only for some voltage. The voltage uses to be selected such, the ballast exceeds it dureng the resonance when the lamp is not lit yet, but after the lamp starts, the voltage is below the VDR threshold, so there is no current, so the PTC can cool down and become ready for the next start even when the lamp is still lit. Now what happens before the PTS heates and switches off is, the voltage becomes just limited by the VDR. And with some lamps that could be high enough to support the hot electrode discharge, so the startup could be exactly the same as with the SRS (although maybe on higher voltages or so).
The other cause of the slow startup comes with one frequency sweep method:
The controller IC's have the ability to steer the frequency, so they may operate the circuit high enough above the resonance, so the voltage across the lamp is below the ignition voltage, but there is high enough current to the filaments. This concept then replaces the PTC (as it uses to be quite unreliable component, due to the fact it has to be repeatedly heated to rather high tempetratures). Some ballasts use fixed "preheat" frequency for some time and then do rather fast sweep across the resonance to ignite the lamp, other do just plain slow sweep from something way above resonance, through the resonance towards the operating frequency. The later concept has the advantage of spending longer time in the high current/voltage mode, so the lamp filaments get's more heating boost till really the lamp ignites and seconds, the controller gets simpler (just a VCO controlled from a voltage on a slowly charging capacitor, so no need for comparators switching over modes and an additional capacitor for the ignition sweep).
Now the first method (with the fixed preheating frequency) could have the component tolerance "mixed" so, the voltage is really above the threshold, when the lamp hot ignites, so we have again the case of "rapid starting". However this behavior is not intended and should not be there (the cathodes are not yet fully warm), but I have one F15T8 fixture doing exactly that.
The slow sweep type controllers have the gradual brightening coming from the main functionality, so it is always present there:
You probably know, the fluorescents are dimmed by increasing the frequency. Well, that is, what exactly happened, just the frequency went from above down to the normal, so at the moment the lamp ignites, it correspond to some reduced power "dimming", while as the sweep continues to the normal frequency, the power, so brightness increases.

Btw. do you remember first fluorescent lighting in our classical trams (used since early 1980's?)? There was a switch on the dashboard with a warning label that drivers should switch the lighting every time they stopped in final station thus preventing its damage. I suspect it had something in common with 600V DC used in this kind of traction. The polarity had to be altered? I have no idea how these fluorescents were ballasted though.

You mean the T3 or (KT3) before the modernizations in the top right corner of the dashboard? That switch is indeed a manual polarity switch for the 600V supplied fluorescents (in two 90cm? 25W tubes in series per one ballast). In the western world were frequently used automatic polarity swap relays (every turn OFF and ON it swapped the polarity, so it didn't relied on the driver), but the manual switch is way simpler, so less likely to stop working.
After the modernization the lighting was replaced by PLL or T5 fluorescents supplied from the 24V battery (using the DC powered programmed start electronic ballasts). The main reason for that is, the regulations do require external power independent lighting, so with more extensive rebuild requiring new certification (replacement of the old resistive PSCC-style accelerators with DC series motors by the more efficient electronic inverter fed induction motor and the complete control system), the lighting had to be changed as well.

OK, I think I can understand SRS lamp fade in now.

One of the programmed start methods uses PTC shunting the arc for some time after warmup before ignition. Just the PTC has one drawback: After the lamp ignites, the PTC is kept hot to remain OFF, so once the lamp is shut down for short time, the filaments cool down, but the PTC does not, so remains open, so the lamp cold starts.

That's exactly the behaviour of older Osram Duluxstars. I posted a video some time ago - the correct preheating start only. When it is turned off and on after a few seconds, it lights up without preheating. Despite of this it's a superior endurance lamp which - although switched often for many years - has got its filaments depleted only very slightly.

In some designs this PTC uses to be in series with a VDR (or someone had experimented with a SIDAC with very similar results), so the combo is conductive only for some voltage. The voltage uses to be selected such, the ballast exceeds it dureng the resonance when the lamp is not lit yet, but after the lamp starts, the voltage is below the VDR threshold, so there is no current, so the PTC can cool down and become ready for the next start even when the lamp is still lit.

Provided the environment inside a fitting isn't too hot to heat up the PTC anyways...?

I don't know much about DC motors legislation but forcing EC motors because of Energy Efficiency Directive is just the same stupid as with lighting.

EDIT: I mixed up forcing higer IEC classes (1-4) with EC motors. These are still optional, I believe. But I don't like the idea much because of unnecessary (and potentially unreliable) electronics inside.

It sounds like with the electronic ballasts, most seem to be of the resonant-start type. While most start the lamp instantly (or like an instant-start lamp, not preheating the cathodes), many have a PTC and some are true programmed-start which prevents the lamp from ever starting before the cathodes are preheated. It sounds like these are all really like the rapid start magnetic ballast, only that the control of the cathode heating is different.

One of the CFL manufacturers refers to the PTC-type starting as "modified rapid start" so maybe that's what it should be called.

I've had magnetic rapid start & preheat fixtures start a lamp instantly at times. Also I've seen instant start electronic and magnetic ballasts, and even preheat fixtures (before the starter could heat up) where the lamp starts with a cold-cathode glow discharge for a second and then lights fully, so it "looks like" rapid start.

@arcblue: the CFLs you speak of that are PTC are like teh feit ones I installed in the garage. I had to use those since they are the only ones that will start and continue to glow bright. I can see the cap discharge when I flick the switch, wait 3secs, and then turn it off.

It sounds like with the electronic ballasts, most seem to be of the resonant-start type.

I may say it differently: The only ballast circuit not being resonant start are the simple ELV blocking and Royer oscillators. Although the Royer uses the resonance to operate the inverter, but not for any voltage boosting.

All the mains powered HF ballasts in use do use resonance for the voltage boost.



And even when there is no means to preheat the electrodes before ignition, the ballasts with the filaments connected in series with the resonator (most CFL's and cheepeese electronic ballasts) do wear the lamps way less than the "single wire per lamp end" "true" instant start types, just because they provide quite large heating current to the filaments during the cold cathode mode, so the lamp burns in that phase really short time. The cathode wear is really comparable to the RS ballasts (there the wear is smaller due to the lower voltage, but it takes quite long time - the whole gradual "brightening" the cathode fall is way greater than the normal 10..15V).

And every time the bluish glow happens, it means really quite severe sputtering.
With IS it glows for a second or so, a time needed for electrode warm up.
With RS it may be the result of insufficient heating of that end, or just that end approaches an EOL. The thing is, mainly the HPF RS ballasts are able to feed the cold electrode mode, when it happens only on one side, because the series capacitor may act with the rectifying lamp as a voltage doubler. So if it does that with just old lamp, it is nothing to worry about (normal lamp EOL approaching). But whe it happens with new lamp, it is time to check the filament connections...

Medved, is possible make an srs T5 magnetic ballast?
If i remember correctly, T5 have argon.

Despite T5 have argon, their narrow diameter and their long tube length, cause them to have a high starting voltage as well, so preheat/switchstart is the only option for a magnetic ballast for them. Also, certain types of T5, have high lamp voltage, so their magnetic ballast would need to be autotransformer ballasts and not simple series choke ballasts.

The older "miniature" types (up to 13W) will work, but the modern HE/HO won't.
The HE/HO have frequently problems even with glowbottle starters, even when the arc voltage is below 120V. The reason behind is, these modern T5's were designed primarily for the resonant start HF electronic ballasts, so ballasts with way sufficient voltage reserve for the ignition and mainly operating at high frequency, so the lamp does not need to be able to reignite after the rather lengthy (compare to the HF operation) current zero crossing on the magnetic. The results are rather large reignition overshoots, causing weak discharge in, or sometimes false triggering, the glowbottle starters.

ok, thank you! 

I'm wondering, do they make semi resonant start fixtures for use on 120v, or is SRS just restricted to 240v land?

I know about magnetic SRS only in the UK and here in Israel (Eltam SR40 Rapid Start, which was an SRS ballast. Discontinued).