Hello,
I would like advice from anyone who has sold / installed a small passive harmonic filter. I am proposing this to be installed to a 100A PDU feed sub-sub-panel. This is the only "noisy" sub-panel for a MGE Epsilon PDU. I've specified and installed AHF filters in front of UPS's and MCC's, but cost and harmonic values rule out bigger units. One device I am looking at is a Mini-Z zero sequence filter made by PQI (Power Quality International). Any experience with these?

Have you looked over here.. http://www.eng-tips.com/
..they are not keen on talking to non-engineer types, but they may be of help

How noisy is the feeder to the PDU? How much current THD and voltage THD? How much load relative to the rating is connected?
Sometimes its not worth the hassle. More new power supplies are provided as time goes on, with less impact on harmonic influence.

The panels I am considering this filter for are sub-fed from a PDU. Harmonic content is about 80% current THD, Voltage THD at 1.5%. Amps are 7.5A / phase, 13A on the neutral (this is on the feeders). These are a pair of feed-thru 100A 42 space panel boards. At the rate PC's are added to these panels, I'm concerned about cooking the neutral and ground.
I have found an RMS value in excess of 35 amps flowing on the ground back to the UPS, on a #4 AWG. There are many frequencies riding on this wire. (up to 10kHz).
I believe I have two issues here; fundemental voltage distortion will increase above 3% soon from additional PC's. And 2nd, stray oscillary transients run along the grounds. If left unchecked, something is going to fail.

Definitely keep your eye on it with a true RMS meter and your THD meter. I wouldn't do anything yet until the panel is a bit more loaded.
You'll find as more load is added, they begin to cancel each other out. Non-linear power supplies draw power when they want. With enough load eventually they start to draw at such "randomness", that the effective THD reduces.
Also be sure that you are using a clamp-on that is somewhat close in size to your load, so as to increase accuracy of the measurement. Some clamp-on's that are designed to measure 1000A, don't measure 14A very accurately.

[This message has been edited by Ron (edited 10-11-2005).]

[This message has been edited by Ron (edited 10-11-2005).]

Thanks Ron. One more final reply and I'll let this rest. These panels are supplying a small army of PC's in the "buisness continuity room" (code speak for disaster recovery department)of a major insurance co. I appreciate your hesitation to add hardware until a larger draw is pulled from the panels. This company buys electronic hardware (PC's, servers, routers ect. in bulk at year's end, ussually devising a plan to connect it all together at once with little engineering). I say this knowing there is a definite probability that 25 more identical workstations will arive this year and will pull similar current patterns from the SMP's. I was introduced to these folks after a 150KVa UPS blew-out this summer. Admittedly I'm trying prevent another business operations failure in guessing worst-case senarios. The collection of PC's has never really been tested under real world conditions.

Assuming 25 new computers, your loads per phase will increase from 7.5A per phase to 17.5A per phase. Still significantly less than the 80A per phase (100A x 80%) that the PDU can tolerate.
Although the load per phase may almost double, if you can keep the loads balanced per phase, I would be interested in how the neutral current would be effected. I would expect the neutral current to remain pretty constant or reduce.

You could spread the harmonic pulses around some more (in terms of phase angle timing) if you operate some of the PCs off of 208V (phase to phase). PC power supplies usually have 120/240 selector switches. What this can do: Say most of the PCs are operating off of 3 phases of 120V. At phase angle 0, 1/3 of the PCs draw a spike of current off phase A, and return it on the neutral. Then at 120 degree phase angle PCs on phase B draw a spike of current, and return it on the neutral. Now if you connect a PC (after setting the power supply to 240V setting) across phase A and B for 208V it will draw, at phase angle 150, a current spike current and return it from phases A and B, and the neutral sees nothing. This is in between times of PCs on phase A and neutral, B and neutral and C and neutral are drawing their current spikes. This off loads the neutral some, and also should make the power company a little happier.

[This message has been edited by wa2ise (edited 10-12-2005).]

Thank you for the insight; abeite un-othordox for I.T. departments to follow. Your concept is electrically valid but would cause a sizable increase and confusion in cost to re-wire and re-label exsiting 120v 5-15 repts to 6-20's, not to mention loading some portion of the panels with 2pole 20's. I can't imagine this working in the long run. (too many crossed human wires). Thanks, I DO follow your reasoning
Eric

Mr. Wise, that's not a bad idea, but the power cords and the receptacles on the circuits feeding them must be changed!

Thanks, but I suppose another downside to my idea is that the IEC ends of all the PC power cords for 120V or the 208V cords are the same, and if in the future someone replaces a computer fed by 208V, unplugs the IEC female from the back of the PC, and doesn't realize that the PC power supply on the new one needs to be set to "220V", and he reuses the old power cord still on 208V, he'll blow up that power supply.... When I've done this 208V or 230V thing I labeled the power cord near the female IEC connector "208V" or "240V", and I know to expect some of the cords to have the higher voltage. But someone else might not expect this....

[This message has been edited by wa2ise (edited 10-13-2005).]

[This message has been edited by wa2ise (edited 10-13-2005).]

You could always put a small DY transformer next to the PDU supplying some of the circuits and achieve the same result - with the advantage that all the external wiring stays as is without any dangers of PSU failure.

M.

Marc,
Other than isolating the harmonic problem to the secondary side of the transformer, how does the DY xfmr reduce overloading of the neutral in the PDU?
BTW, I still think that after additional loading, the problem will not be severe enough to warrant the neutral to be in excess of the phase conductors. Now that I think of it, I assumed that the PDU feeder had a full sized neutral or better, right?

Let's as pure example assume each workstation takes 2A and there are to be 24 workstations.

It is always advisable to assume the PCs are non-PF corrected and have very sharp conduction angles. This means the Neutral can carry up to 3 x phase current.

So, in no transformer scenario:

24 PCs over 3 phases = 8 PCs per phase = 16A per phase and possible 48A on Neutral.

If we put in DY xfos and split the load between straight feed and DY feed, then:

9 PCs over 3 phases = 3 PCs per phase = 6A per phase and possible 18A on Neutral.

Next 15 PCs are on DY. Secondary currents are 10A per phase and possible 30A on local Neutral. However, primary side (Delta) has 10A per phase but no Neutral current therefore total current on each supplied phase is 16A per phase (back to our original).

The trick is the current curve presented to each phase is a combination of 2-pulse and 6-pulse resulting in a much reduced harmonic content, and the whole system is capable of running on a 3P+1N cable.

As for the Ground/Earth currents, I suspect circulating currents and would need to do tests to establish where they were coming from (because there is just no way on 7.5A per phase he can combine to be in excess of 22.5A - and as is 13A is already on the Neutral, that Ground current must be coming from another system).

M.


[This message has been edited by marcspages (edited 10-13-2005).]

I'm not understanding how the transformer is connected into the distribution to result in the 9PC's over 3 phases then 15PC over 3 phases. Are you suggesting a isolation transformer serving a subpanel for the 15 PC's?

I haven't done the FFT on the input to a PC's switchmode power supply, but I believe that the entire draw is not at 180 hz relative to the nominal 60hz. Even if it was only drawing 180hz (and other triplens), I thought when you add the triplens, you can theoretically get a max of 1.73 times the phase current on the neutral. I will try to draw three phases of 180hz, 120 degrees out of phase from each other and see what I get.

[This message has been edited by Ron (edited 10-13-2005).]

Ron,

The DY is effectively an isolation transformer; The primary is connected to the incoming three phases, the secondary feeding a portion of the PDU that has been split from the original supply - except the Neutral is connected to the original Neutral so as to maintain the original wiring fed from the PDU.

Believe me, there is a huge amount of badly written material stating the maximum on the Neutral can be 1.73 (the primary mistake being they are relating this to the way voltages are calculated in a 3-phase supply). We're dealing with current, a completely different kettle of fish - I've even shown the Neutral taking 3 times Phase in practice and made many an engineer shudder!

What I omitted to add in the previous post is a DY xfos is much cheaper than a filter! (but you do require the load to be relatively constant - but most office situations are).

M.

PC power supplies can be thought of as being a bridge rectifier (at 230V setting) or a voltage doubler rectifier (at the 120V setting), feeding a big capacitor that has a more or less resistive load across it. The rectifier diodes only conduct at the very peak of the input AC voltage to "top off" the capacitor to replace what the resistance drained off between voltage peaks. This makes for current spikes happening at the input AC voltage peaks, and no current draw at all between peaks. These spikes last for about 5% of the cycle time of the 60Hz waveform cycle. Thus there is *NO* overlap from phase to phase. PCs on phase A draw their current spikes at the peak of A from A and the neutral. Then likewise for phase B, and then C. The neutral thus sees 3 times as many spikes as any one phase wire does, and over time will get as hot as the same guage of wire carrying 3 time the current of any one of the phases. You just need to think in terms of the time domain to see this.

Heard it said that the European equivalent of UL and/or FCC will soon require that switching power supplies draw current off the lines like a purely resistive load would. This technology exists; it just costs more....

Exactly! And you can even go to a maximum of 60 degrees conduction angle per half-cycle and still get a whole 3 times phase current on the Neutral (assuming all three phases are equally loaded with the same hi-tech load).

It's the thinking in real time that is a problem for a lot of folk, hence the introduction of Fourier analysis - but this has led to so much incorrect teaching. But lest you think I'm complaining, I'm not. It is this that keeps me in business sorting out issues others can't (tee-hee!). An example is "the 3rd circulating in the delta winding" - absolute rot! Sadly, I'm not prepared to publish my findings on this because there are institutes in the PQ fraternity with a little more clout than me busy plagiarizing my work left, right, and centre and calling it their own!

G5/4 is the primary driving force here in the UK to get power supplies to incorporate both PF and harmonic correction. The extra costs pays for itself in a very short space of time through electrical systems not failing through high harmonic content.

M.