3 phase Power filtering with unbalanced loads?

[max_torque]

I have a device that takes a 3phase 400VAC 16A input, and drives a range of "load" devices. The big loads primarily resistive (heating elements) but a few are inductive as they are small (250W) AC induction motors.   Those loads are not all on at any given moment, and are controlled by logic, so any given load can be on at any given moment and viseversa.  The Loads are between each line and neutral, and have been allocated to be as evenly spread as possible, ie the 3 big heaters are across L1->N, L2->N and L3->N.  If all loads were on, then there would only be a small neutral current due to that balancing, but if the loads are unbalanced, then the neutral current will be around 16A.

Ok, so that's the setup, the question resides around what sort of AC power filter architecture i need to provide, primarily to prevent external noise coming into this unit and messing things up (loads are Thyristor controlled and therefore to some degree potentially noise sensitive). The unit is going to be used in a high noise environment with regard to the AC supply  (there can be up to 2MW of high performance prototype inverters / motor controllers running in the same room....)

I think:

1) The loads are unbalanced so that rules out a common mode type filter arrangement for the 3 phases
2) Neutral must be considered a power carrying conductor in this application


Schaffner sell what looks to be a suitable power filter for 3Ph + N asymmetric loads

http://www.farnell.com/datasheets/2188999.pdf

Options include simply buying that (expensive) filter device and including it inside our unit, or desiging our own power filter (we can reverse engineer the Schaffner part......  ) to include on the power switching pcb that our unit already requires


Comments anyone?   

[MagicSmoker]

Most 3-phase filter chokes (aka "reactors") are wound on a single E-core so are coupled by default. This gives very good performance both for common mode noise current, as well as *balanced* normal mode load currents, but it will not work very well, and might even saturate, if asked to supply unbalanced load currents returned to neutral. So, you'll likely have to use individual chokes for your filters, and because of the phase-angle control of the heating elements these will need to be Tee filters, rather than Pi or LC.

EDIT - pretty much exactly like the Schaffner filter which I checked *after* writing the above...

[David Hess]

That Schaffner filter only implements common and differential mode filtering and does nothing for load balancing.

Wouldn't two Scott-T transformers connected back to back and then in parallel serve to balance a 3 phase load?

[NiHaoMike]

Wouldn't it be easier to add snubbers to the thyristors in order to increase immunity to spikes? Put in provision for chokes on the board but try putting in wire links at first.

[T3sl4co1l]

How are the loads unbalanced, are you sinking them to neutral or to ground?

Even if you were, that just means using a 4 or 5 line CMC.

If you need diff mode filtering as well, yeah, just use single inductors, of whatever size and rating you can get.  Be sure to get saturation current higher than peak current draw -- which is 1.4 * Irms for sinusoidal loads only, and can be significantly higher for pulsed loads (rectifier + cap input filter).

Nice thing about 3ph is you can save a lot of ground leakage current by canceling it out.  Typically, 'X' type caps are wired in a wye configuration, and the "neutral" node can be 'Y'-capped to ground with fairly aggressive values (10nF+).  The ground current goes up considerably if one phase drops, of course.

Tim

[max_torque]

The loads are unbalanced because they are controlled by the process and so dependant upon exact circumstance.

I already have snubbed Triacs, and a small upstream choke for each output (outputs are Line to Neutral)

My concern is the horribly noisy supply, because of the other (non certified and certainly non standard) equipment driven off the same supply, hence i want to robustly block as much of that noise as possible.  Most of the noise coming down the 3ph supply is common mode, and will have complex and changing harmonics depending on the operating state of various equipment within the facility.

Luckily most of the noise is also significantlyhigher frequency than the fundamental (50Hz) supply, so it ought to be possible to reject it reasonably easily?

[MagicSmoker]

The loads are unbalanced because they are controlled by the process and so dependant upon exact circumstance.

If you use a (relatively uncommon) 4-way common mode choke - so you can put inductance in series with each phase and neutral - then it won't matter that the phase loads are unbalanced because the neutral current will exactly cancel out the imbalance.

I already have snubbed Triacs, and a small upstream choke for each output (outputs are Line to Neutral)

My concern is the horribly noisy supply...

Note that the primary purpose of triac/thyristor snubbing is to limit dV/dt during turn-off, not to reduce the emission of, or susceptibility to, noise.

That said, phase angle control of loads is itself a prodigious producer of noise, and presents a terrible power factor as well, so a filter is in order even if your concern about susceptibility is a bit misplaced.


EDIT - grammar

[T3sl4co1l]

The complement to the parallel R+C snubber (for dV/dt) is the series L||R snubber (for dI/dt).  With both, you can save a lot of noise in the first place (i.e., from the TRIAC), and clean up what's left with whatever additional filtering.

Tim

[NiHaoMike]

That said, phase angle control of loads is itself a prodigious producer of noise, and presents a terrible power factor as well, so a filter is in order even if your concern about susceptibility is a bit misplaced.

Hence why zero cross switching generally makes more sense than phase angle control for heating devices. It's very unusual for the thermal time constant to be small enough for that to make a difference in thermal regulation.

[max_torque]

yup, we are not running phase angle control, just ZCS with a sort of integrated power kind of control scheme.

The device runs fine on the bench, and the traics and drivers all work and are snubbed correctly, my concern is as mentioned over what is going to come down the AC supply line in the actual test facility where the devices will be used. In that facility prototype high perfmormance power electronics and motor controls are developed and tested, often for motorsport use (including F1 / formulaE etc) and those DUTs are generally extremely noisy, because electrical noise is several orders of magnitude less important than other factors and the DUTs are not intended to be certified or type approvedin any way. As a result, all the switching devices run at high dv/dt and di/dt (to get the physical size and mass down, and to get the most robustness) and that causes a lot of noise to be passed back up the chain.  Various filters are in place, but the overall environment is still horendously noisy (we have to use fibreoptic CAN and ethernet to get datacoms around the test facility, but that of course also gives us a nice safe galvanic isolation for operators etc).

I've scoped the AC supply lines and it's not uncommon to see 50 to 100V spikes at hgih frequency on top of the normal AC waveform!

As a result i want to include a very robust supply filter to these units!