Am I missing any potential pitfals with this soft start / power sequence design?

[Metatronic_Mods]

I'm just wondering if there's any glaring problems with this design which I have overlooked. The basic idea is to use a relatively large resistance to limit the inrush current to the transformer and reservoir caps. Once the caps charge to a specified voltage, relay RL1 closes, bypassing the resistor array. Then at some slightly higher voltage RL2 closes, connecting the load to the caps.

RL2 is necessary in order to ensure the supply voltages come up with a fast edge and reach their specified levels at roughly the same time, otherwise one of the micros can be pretty temperamental. I know without RL2 (ie. load always connected) the res caps would never be able to reach a sufficient voltage to close RL1 (not with such a large ballast resistance at least), and from my understanding this is one of the reasons inrush current limiting for the transformer and res caps is advised to be handled separately.

The only other issue I can see (vs. existing valid designs) is overload protection in the event RL1 fails to close; which is overcome easily enough with the addition of F2.

Is there anything else I should be worried about? (Never mind about particular component values, or the relay coil charge/hold circuitry - it may prove over simplistic as drawn, but I plan to cross that bridge after I select which relays to use).

[Gyro]

The only real issue [EDIT: issues] I can see (assuming all your RC time constants work with the relay coil resistances), is that the relays will close relatively gently as their coil currents increase. Slow closure could possible lead to excessive burning of the contacts (or maybe not), probably more of an issue on the mains contact one.

The whole thing would be a lot more predictable if you used a couple of comparators to monitor the capacitor bank voltage and actively switch the relays at the required voltage thresholds.

Relay switching the supply to your "temperamental" micro's regulator could also cause problems due to contact bounce - A MOSFET might be better.

[Metatronic_Mods]

The whole thing would be a lot more predictable if you used a couple of comparators to monitor the capacitor bank voltage and actively switch the relays at the required voltage thresholds.

100% agree. There's a few solutions I'm considering, dependent on PCB area budget. The RC networks were really just placeholders.

Relay switching the supply to your "temperamental" micro's regulator could also cause problems due to contact bounce - A MOSFET might be better.

My only concern would be the extra heat dissipation.  But I don't expect bounce to be a problem. The problem with the micro is due to a Vref and Vdd connected to separate rails, so long as they stabilize at nominal levels within ~10ms of one another there's no issue. Still, something to thing about that I hadn't yet, thanks!

[Gyro]

My only concern would be the extra heat dissipation.

I really wouldn't worry about that. It's easy to get low voltage MOSFETs with on resistances in the milliohms region, no additional heatsinking required.

[jmelson]

The only real issue [EDIT: issues] I can see (assuming all your RC time constants work with the relay coil resistances), is that the relays will close relatively gently as their coil currents increase.

Not much of a problem.  As the relay armature starts to move, the gap closes, and the magnetic field becomes stronger.  The relay will close just about as sharply as if voltage is applied with a switch.  Now, this circuit may have a slow OPENING of the contacts, but presumably that doesn't happen under load.

One issue is if the relay ever fails to close, the resistors burn up.  So, use flame-proof resistors there.  Also, as the transformer needs magnetizing current to develop flux, if the resistors are too high in value, the filter caps may never charge.  You might have to decrease the resistor value quite a bit.

Jon

[Tomorokoshi]

When RL2 closes and engages the regulators connected to the load, if the rectifier network drops low enough RL1 could disengage, leading to an oscillation. Check at your lowest AC voltage input and derate from there.

If this goes to EMC testing how will the circuitry react to burst / dip / EFT / etc.

[Metatronic_Mods]

One issue is if the relay ever fails to close, the resistors burn up.  So, use flame-proof resistors there.  Also, as the transformer needs magnetizing current to develop flux, if the resistors are too high in value, the filter caps may never charge.  You might have to decrease the resistor value quite a bit.

Jon

Do they burn up though? If the resistance were much lower (the more common approach I think?) then obviously you're right. But for an equivalent resistance on the order of ~10k the power dissipated will be ~1-2W right? Using 4x 3W resistors for the array should be plenty to ensure they don't get too hot. Or am I forgetting something to do with the phase relationship when the transformer is unloaded?

[Metatronic_Mods]

When RL2 closes and engages the regulators connected to the load, if the rectifier network drops low enough RL1 could disengage, leading to an oscillation.

Good call. I only have a general idea of what a given load connected will look like, lots of unknowns. I'll have to run some tests to figure out what the usage limitations will be.

[jmelson]

One issue is if the relay ever fails to close, the resistors burn up.  So, use flame-proof resistors there.  Also, as the transformer needs magnetizing current to develop flux, if the resistors are too high in value, the filter caps may never charge.  You might have to decrease the resistor value quite a bit.

Jon

Do they burn up though? If the resistance were much lower (the more common approach I think?) then obviously you're right. But for an equivalent resistance on the order of ~10k the power dissipated will be ~1-2W right? Using 4x 3W resistors for the array should be plenty to ensure they don't get too hot. Or am I forgetting something to do with the phase relationship when the transformer is unloaded?

The problem is the transformer can require a pretty high magnetizing current to develop full magnetic field.  If the magnetizing current causes the voltage across the primary to sag below the mains voltage, then the rectifier-filter will not charge up to the expected voltage.  Remember, also, that an unloaded power transformer is a huge inductance.  I'm quite sure your circuit as designed will NOT charge the caps to more than a few hundred mV, and will just sit there, and the relays will never pull in.

Jon

[duak]

There is a logical hazard when powering the circuit off.  The main reservoir caps and then the delay caps for the relays have to discharge enough for the relays to drop out.  This could be significant and would prevent the circuit from cleanly starting up again if the time off is just right.  Also, one relay may drop out and the other won't.

Note that the transformer primary will not saturate if there is enough secondary current.  In this case, charging the reservoir caps is often enough.

If a supply voltage sequence is required, the later voltage is derived from or gated by the earlier one.

If a particular voltage should not be applied until another one is present, then a clamp diode and current limiting resistor may work very well.  I'm thinking of an A/D that was paricularly sensitive to an analog input applied without power.  An additional clamp diode from the errant input to the power supply limited the voltage while a series resistor limited the current.

Cheers,

[NiHaoMike]

How about use a NTC for inrush limiting?

[Metatronic_Mods]

How about use a NTC for inrush limiting?

Efficiency. I was never able to find an NTC that could adequately limit the inrush and also drop to a negligible resistance at normal operating loads. Plus I don't necessarily know what the load will even be. Seems to me NTC's become problematic when you're after a wide operating range.