I got a gear tray from a 2x58W T8 lantern to repair (well, i already returned it, but i can still ask to get another look at it if we find a solution)

The lantern had a cheepee HF ballast that failed. This was what needed repairing. I put in a VS HF ballast in its place and it works with the new ballast

The lantern also have an EM module - Charger/inverter module with relay, that switches one of the lamps between the normal ballast (for normal operation) or it's own output (in emergency). And this module causes a problem with the lamp detection by the ballast :

If the light switch is on when power returns (that means, 230V is applied at the same time to the HF ballast and EM module), the EM module clicks the relay to connect the lamp to the ballast, but at this point the ballast allready detected that there is no lamp and shut down, so it does not light up (both lamps stay off). Cycling the light switch for a second (that is, switching the HF ballast off, and back on while the lamps are already connected to it) will get the lamps to start normally



The wiring scheme on the HF ballast :




The wiring scheme on the EM module :


It can be figured out, and i confirmed by opening the module and following the PCB traces, that this is what's inside :




The original luminaire wiring :


(the pinout of the original ballast was different, but i drew the schematic here with the VS pinout)

Assuming that the lamp detection is on the "cold" end, this wiring scheme can be expected to have the lamp detection problem. So i didn't test it, i modified it right away. Since the original ballast was essentially a "self ballasted CFL" sort of ballast, i guess that it would actually work with the wiring above just fine



Anyway, i rewired it to :


And it does exactly the problem i tried to avoid....



I took the lamp sockets off the gear tray and plugged them on the lamps freely. Then i connected power to ballast and EM pack at once, to recreate the "not working" condition. And then, i took off and connected back one socket at a time. to see which socket disconnect/connect triggers the ballast's inserted lamp detection. Result : The ballast detects the lamps on the cold end (terminals 1/2/3)

So, when the EM pack reconnects the lamp (a split second after power applied) the lamp is not detected. If then, i take off and put back on the socket on the same lamp end (same as where the EM relay is), then it does detect the lamp

Also, it does detect the lamp and start if the cold end of the lamp is inserted by hand when the ballast is powered on (few seconds after power applied, unlike few 10's mSec ? with the EM relay)



I tried to exchange the lamp roles on the ballast (swap terminals 2-3, 4-6, 5-7) to see if the other lamp's circuit does not have this glitch, but it have it as well



I guess that it is a glitch of the ballast's lamp detection logic, which can't be resolved except by playing with timings (add transistor with RC delay in series with the relay coil in the EM module ? or increase the value of some "+ main electrolitic cap to chip Vcc" initial startup resistor in the ballast ?)

Would like to read your input on the matter

Maybe the original ballast should try to repair.

The described problem could happen with ballasts using voltage mode heating.
How these ballasts are usually internally connected:
There the heater is connected to an auxiliary winding of the ballasting choke via few 100's nF coupling capacitor.
The lamp detection usually works by pulling the not grounded pin of the cold end filament up to the internal ~12V supply via a high ohmic resistor (high ohmic to not load it that much, there is only small current available when the main inverter chip is off.
That means after power up, it takes some time to charge the coupling capacitor to shut down the chip and reset the protection, if the lamp is missing.
So even with the lamp missing, the ballast attempts to start and open filament causes it to shut down quickly.
Normally with really a missing lamp, the delayed lamp detection signal shut it down and resets theprotection register, so after the lamp is reconnected, it starts normally.
Here the few 10s of ms of the EM relay delay is enough to trigger theprotection, but does not allow the capacitor to be charged enough to reset it.

One solution could be the delay, othe that came to my mind is to provide an external bias (charging it to about 5V should be enough, the detection input of the ballast IC uses standard 5V logic thresholds to allow easy intfacing to standard microcontrollers used in smart ballasts) from the EM, which will keep the filament coupling capacitor charged when the EM is active, so once power reapplied, the chip immediately sees high state and stays in reset till the relay reconnect the lamp back.