Switching at >1 MHz.

[DaBit]

I took the shortcut on the HID-bulb-on-the-motorbike project by using the existing 220V ballast preceded by a 12VDC -> 260VDC converter. That works now:



But I am still not satisfied. The electronics are still big and bulky. Next try will be a regular hard-switched converter, preferrably pushpull. But to avoid acoustic resonance in the arc discharge, I need to feed the bulb AC with a frequency of <1kHz or >1MHz. Most ballast choose the <1kHz option by doign a DC/DC stepup and a H-bridge inverter. To cut on components and reduce the size, I still find the >1MHz route much more attractive.

Are there special problems that needs to be adressed? Or is it a fairly simple pick the right ferrite, pick the right controller and beef up the gate drivers (and place the power stage very close to the MOSFET's) kind of job?

Is there any chance that a regular power ferrite such as Ferroxcube 3C90 or Epcos N87 material can be used at these frequencies without exceptional core losses? I am not sure I can obtain suitable cores in Ferroxcube 3F4, Epcos N49 or similar ferrite.

[subwo1]

You could consider zero voltage switching which reduces switching losses and drive requirements.

[DaBit]

I'm struggling with that. ZVS control and component choice is not exactly trivial. And the benefits of ZVS are not that high at 12V primary voltage. I'm not interested in extremely high efficiencies; somewhere in the neighborhood of 70% would be nice. But I am interested in a small form factor. Plenty of power to burn, but no space to mount the electronics which burns the power

[subwo1]

Oh, I forgot you were using 12v as the power source. I forget, do you need to regulate the output, or will current limiting be enough?

[Eva]

At 1Mhz you may probably end up consuming more space, particularly due to heatsink size and gate drive complexity.

[DaBit]

Originally posted by subwo1
Oh, I forgot you were using 12v as the power source. I forget, do you need to regulate the output, or will current limiting be enough?

I need to regulate lamp current, but I intend to use a ballasting inductor before the lamp. I need that anyway to keep the HV ignition pulses away from the power source.

Originally posted by Eva
At 1Mhz you may probably end up consuming more space, particularly due to heatsink size and gate drive complexity.

I have two choices:

1) Use a flyback or boost converter, or eventually a pushpull to create a constant current source, and use a H-bridge to create a low-frequency AC square-wave. Control is quite difficult due to the parabolic voltage vs. current characteristic of the lamp.

2) Use a HF converter, and feed the lamp directly, eventually through a ballasting reactor to make the impedance of the lamp appear positive, which simplifies the control loop?

I intend to use the casing as the heatsink. That is also what I did now in the 12V->260V push-pull converter; I mounted the MOSFET's on a copper strip, and bolted that to the aluminium case. The copper distributes heat well, and the aluminium box provides a surface large enough to get rid of the 20-40W.

Do you think option 2 would result in a larger PCB? One or two push-pull gate drivers are smaller than a H-bridge with high Vds MOSFET's and gate drivers. But what about the magnetics? Would those be much smaller? I doubt it; higher frequency also means higher core and AC resistance losses. But comparing a 200kHz and 1MHz case is a bit of homework I can do myself.

[Bakmeel]

You could consider a series resonant tank circuit fed by a half bridge. If you design your inductor such, that the resonant tank will have an appropriate Q, you can control power with frequency modulation.

Use an above-resonant starting circuit to generate the high ignition voltage.

You will have to make careful considerations preventing too much EMI. At these frequencies everything becomes an antenna

[DaBit]

Sounds a little like the output of a 'real' HF transmitter. Maybe not a bad idea, except for the quite high Q required to obtain the voltage gain wanted. Hmm...

[subwo1]

Originally posted by DaBit


I need to regulate lamp current, but I intend to use a ballasting inductor before the lamp. I need that anyway to keep the HV ignition pulses away from the power source.

Maybe you can operate a push-pull, then, near full duty cycle to ease switching problems with the MOSFETs. It can give ZVS characteristics to push-pull with a constant switching frequency. You would have to watch for flux imbalance on the core, though, since it is not AC coupled to its drive waveforms.

[DaBit]

Originally posted by subwo1

Maybe you can operate a push-pull, then, near full duty cycle to ease switching problems with the MOSFETs. It can give ZVS characteristics to push-pull with a constant switching frequency.

Can you explain this a bit more? There is 2x Vprimary between the two end points of the primary winding when one MOSFET is on. Thus worst case I need to switch with 2x Vpri over the switch, best case 1x Vpri (when there is no current flowing through the primary before a switch closes).