Source impedance of mains, a workaround?

[Al (astral highway)]

Power engineers, what's a workaround for source impedance of our mains in a domestic 13A outlet?

I am trying to model the charging behaviour of some very low ESR EVOX-RIFA capacitors (ESR, from data sheet, is just 230 milliOhms at mains frequency).

They will be on a DC bus and will charge to 370VDC.

Because of this low ESR, if the equivalent series resistance of the mains is close to zero, the peak charging current comes up to astronomical figure that is absurd, even if theoretically sound.

There will, however, be very little stray resistance as the capacitors will be on a busbar a few mils thick. The supply will be full wave rectified, so, I'm thinking, the internal series resistance of the rectifier diode assembly is the limiting factor.

I need to do calcs that make sense, and theoretical precision is far less important than practical relevance here.

Thank you.

[GrimJosef]

If I switch on a 2kW fan heater in my workshop, which is fed by a single radial from my consumer unit so may have a relatively high mains impedance, I find the mains voltage drops by about 2V. Roughly speaking 2kW corresponds to 8A at 240V and if drawing 8A drops the voltage by 2V then the mains source impedance is about 0.25ohms. So the mains itself is unlikely to be what limits the charging current for your caps. Whether it's actually the equivalent resistance of the rectifier diodes or of the isolating transformer windings and inter-winding coupling (I'm assuming you're using a transformer) will depend on the relative beefiness of the diodes and the transformer.

Cheers,

GJ

[Al (astral highway)]

Originally Posted by GrimJosef

... (I'm assuming you're using a transformer) will depend on the relative beefiness of the diodes and the transformer.

No transformer in this design. That figure is actually very helpful, as that's probably the average power needed. Thank you, I can now put some meaningful figures in place.

[Hartley118]

It struck me that there's likely some mains wiring inductance and stray capacitance to take into account if you're worried that your series resistance is exceptionally low. So I looked it up and was interested to see that the IEC have a typical value for the figure: viz IEC725:1981 which models the European domestic mains supply as having an impedance of (0.4+j0.25)ohms.

This is quoted in a short note which, as I read it, is responding to audiophool interest in special low impedance mains wiring: http://www.acoustica.org.uk/other/mains_Z.html

Martin

[GrimJosef]

If you rectify the mains into a low ESR capacitor via a reasonably chunky bridge, say one like this http://www.farnell.com/datasheets/23...SAAEgK3x_D_BwE, then you might find that a very high current flows transiently. The bridge is rated for 400A peak and the datasheet indicates that the forward voltage drop across the diodes will then still be less than 2V each. Of course if 400A were to flow even for just one mains half-cycle (10 milliseconds) the total charge transported would be in the low coulombs. I imagine that you don't have several millifarads of capacitance (but then I imagined that you'd have a transformer, so who knows ). If you don't have several millifarads then you will charge whatever capacitance you do have up to 370V on the few millisecond timescale or less. Now the details of the charging current become very dependent on precisely when in the mains cycle you close the switch, or push the mains plug home if you don't have a switch. I fear that the protective devices in your consumer unit (do you have a consumer unit ?) might be the first things to have a sense of humour failure about all this ...

EDIT: j0.25 ohms at 50Hz corresponds to an inductance of a bit less than a millihenry. At peak mains voltage that would allow a rate of rise of current of perhaps 400 amps per millisecond, so for a properly accurate answer the effects of the inductance will need to be included.

Cheers,

GJ

[Al (astral highway)]

Hey Martin,

Thanks for digging that reference out for complex impedance . Seems a plausible idea.

I’m on the move so haven’t read it yet but will do so properly tomorrow.

Cheers

[Al (astral highway)]

Hey GJ, I sense some alarm in your reply. Please don’t worry. It’s a circuit that I will theorise to the nth level before I push any plugs in and it will have overcurrent detection circuitry way faster than my consumer RCD, as well as suitable inrush limiting.

The caps are in fact 330uF.

I resolutely don’t want this thread to head off into the realm of safety. Some circuits need a low impedance power supply , just as an electric heater does. I’m not selling it, giving anyone else acces to it or taking any risks.

Cheers

[GrimJosef]

Actually I'm not concerned much about personal safety from a supply impedance point of view. Supplies with surprisingly high impedances are more than capable of killing us. Your caps are storing enough energy to be getting into defibrillator territory on their own, never mind the supply that's charging them. I'm mostly bothered by the possibility of hundreds of amps of peak current causing the kit to be 'hurt' rather than the people, and also by the fact that so many things become important at this sort of power level that it's very difficult to be sure that any theorising really has accounted for every possible factor. I used to build big exotic electrical pulsers for a living. We theorised to the best of our ability but eventually the day would come when we just had to turn the thing on. It went bang sufficiently often at this stage that my department head took to accusing us of never being happy unless we'd just managed to destroy something expensive .

Cheers,

GJ

[Al (astral highway)]

Originally Posted by GrimJosef

Your caps are storing enough energy to be getting into defibrillator territory on their own, never mind the supply that's charging them.

Indeed! And they need to be.

I'm mostly bothered by the possibility of hundreds of amps of peak current causing the kit to be 'hurt' rather than the people...

That's very thoughtful of you. And that's the reason behind my question.

When I'm working with power switching, I'll theorise, take a step back, use my intuition, do a thought experiment, do another, make notes, repeat... before even getting a soldering iron out. I can be quite obsessive. In this case, I'll make a mock-up with a tenth and then a half of the power of the final design, and I record everything.

And yes, sometimes, I'll plug in some numbers into differential equations. But only once I know that I'm aware of the big picture, the properties of the exact components I'm using, and the numbers that are actually relevant.

All of those things!

...and also by the fact that so many things become important at this sort of power level that it's very difficult to be sure that any theorising really has accounted for every possible factor.

I appreciate that you have a long perspective as a professional power engineer, and I can imagine you have many relevant examples of just this phenomenon.

I'm aware of what I know and what I don't know that I need to know. I have a deeply inquiring disposition and I'll turn every stone until I know what my curiosity tells me I need to. I'll also trial things and retrial them until I'm happy, especially with waveforms etc.

For example, the current transformers I'm using are home-made and I am onto my third iteration. The test period for the model I last wound will be another week or so. (Not full time, but including refining my test circuit, ordering things and mulling over.)

The 'plug it in and stand back' approach isn't for me.

And... I'm using some quite pricey IGBT's and the protection circuitry that I'm putting in for them has to trigger in around 5uS, or two RF periods of my design.

[broadgage]

Actual measured values of mains supply impedance typically range from 0.1 ohm up to 1 ohm. Values as low as 0.01ohm do exist but are uncommon.
This is at the "point of supply" which is generally taken to be the consumers side of the electricity meter.

A value as low as 0.01 ohm implies a short circuit current of 24,000 amps, just about possible if you are next door to a very large transformer but almost unknown domestically.

0.1 ohms implies a short circuit of about 2,400 amps, a very typical figure.
1 ohm implies a short circuit current of only 240 amps, a very low figure, which also implies significant voltage drop under load. A load of 25 amps on such a supply might drop the voltage from 250 volts off load to 225 volts at full load.
A 1 ohm supply impedance at the point of supply would only be expected on a restricted supply with a 25 amp fuse in the suppliers cut out.

A 1 ohm mains supply impedance is however much more likely at the point of use. A shed at the far end of a large garden, with an undersized cable from the house, could give such a figure.

[GrimJosef]

Originally Posted by astral highway

I'm aware of what I know and what I don't know that I need to know. I have a deeply inquiring disposition and I'll turn every stone until I know what my curiosity tells me I need to. I'll also trial things and retrial them until I'm happy, especially with waveforms etc.

That approach of combining theory and trial (experiment) is always best. Sometimes the experiments can tell us things that we couldn't or didn't anticipate - indeed we wouldn't have to do them if we were 100% certain of the answer before we started. What we learn can then be incorporated into the theory of course, so it makes better predictions the next time around.

On a point of detail I'm not sure how relevant a datasheet ESR value might be when it comes to operating an electrolytic capacitor in a high-current regime. I guess it goes without saying that they're not usually charged like that. The datasheet from Kemet (who now own Evox-Rifa I think) for their ELG series caps shows that the ESR varies by a factor of two or so between test frequencies of 100Hz and 1kHz (graph near the bottom of p18 here https://content.kemet.com/datasheets/KEM_A4021_ELG.pdf). There's also some data on the subsequent pages which might be useful as far as modelling an electrolytic capacitor's effective impedance across the frequency spectrum goes. Unfortunately there doesn't seem to be nearly as much information available in the time domain as in the frequency domain. The good Monsieur Fourier can help us get from one to the other of course. But he's not great when it comes to transients, especially if there might be nonlinearities involved.

Cheers,

GJ

[Ed_Dinning]

Hi Gents, from memory the MCB's etc used in domestic DB's are rated at about 7KA max, so better keep below that !!!

Ed

[broadgage]

I believe that you are correct, but there is some wiggle room in the relevant regulations and approvals.
Whilst individual MCBs are indeed rated in the low single figures of KA breaking capacity, a complete assembly of such MCBs in a consumer unit has a much higher "conditional short circuit rating" provided that it is installed in domestic or similar installations with backup protection by a BS 88 fuse or equivalent not exceeding 100 amps.
This fuse is normally provided by the electricity company in their sealed cut out.
This means that in some truly exceptional cases, a fault in the consumers installation will trip the MCB and also blow the suppliers fuse.

This sounds a bad design but is fine in practice.
MOST domestic and similar installations have an available short circuit current well within the breaking capacity of the MCB.
In the small number of installations with an unusually high short circuit current, MOST faults will be at least 2 meters away from the consumer unit, and the impedance of the 2 meters of small sub circuit cable will limit the fault current to within what the MCB can interrupt.

In a handful of cases with an exceptionally high available short circuit current, AND a short circuit very close to the consumer unit, then the suppliers cut out fuse may operate. The MCB may also be destroyed and need replacing after the fault.

[Herald1360]

If you know where this kit is going to be plugged in, why not just plug in a 3kW kettle there and measure the voltage drop?

No need to overthink that.....

[Argus25]

Al,

One thing that makes it difficult to model is the behavior of the rectifier impedance under massive overload currents which will be likely higher than the mains source impedance. Then there is the possibility of internal conductors inside the rectifiers or the capacitors fusing or the rectifier junctions melting or tripping off the dwelling's circuit breaker.

I think the best way to handle this is to set up a situation where you know exactly what the peak current will be because you have placed a low value current limiting resistor in series, it assumes the bulk of the series resistance and then you can easily calculate the peak current, make it a suitably safe value and a short time later after the known & controlled turn on surge, relay contacts short out the resistor.

Again coincidentally that article I mentioned on the switch mode power supply, which has the current transformer/current detector in it, which I will post very soon, uses exactly this technique to avoid surge currents in the charging of the filter capacitors, charged from the mains with a bridge rectifier (they use a 4 Ohm resistor in a 200W supply, with a 3A slow fuse). Also it has a clever reset mechanism (using a programmable UJT or PUT) where if you toggle the main power switch on & off, the relay's delay system is quickly reset so it can't be "fooled" to allow a turn on surge.

[Argus25]

Al, I just posted about that power supply with the surge limiter and current transformer system in the vintage computer section.

[GrimJosef]

Originally Posted by Argus25

One thing that makes it difficult to model is the behavior of the rectifier impedance under massive overload currents which will be likely higher than the mains source impedance.

That was my first thought Hugo. But then I checked the datasheet for a chunky bridge (see above) and to my surprise it really is rated to pass 400A with less than 2V across it. It's far from Ohmic, of course, but its effective resistance at that point is just 5 milliohms, compared with perhaps a few hundred milliohms for the mains.

Then there is the possibility of internal conductors inside the rectifiers or the capacitors fusing ...

Yes. Capacitors which are designed for huge pulses often come with a maximum current rating. In a repetitively pulse environment the current can cause the internal foils, particularly the tabs leading out of the spiralled main cap, to twitch and, eventually, to snap through metal fatigue.

Cheers,

GJ

[trobbins]

Are you loading the caps so that the diodes will conduct at the waveform peak, and trying to simulate that, with flat topping the mains waveform as the outcome?

That may be a little different to using a resistive load to determine cable impedance, as the crest factor would be quite high.

[Argus25]

Originally Posted by GrimJosef

That was my first thought Hugo. But then I checked the datasheet for a chunky bridge (see above) and to my surprise it really is rated to pass 400A with less than 2V across it.

That really is a "Chunky Bridge" !

The problem remains though, that if you load the mains with what is close to a transient near dead short the tens of milli-Ohms zone, the current is only limited by the less than a few hundred milli Ohms of the mains itself, the peak currents are nasty and will likely damage components, switches etc. It really is better not to do it and limit the current with at least an additional ohm or two initially in series, it is easy to switch it out after the surge has gone and the capacitors are charged.

[Al (astral highway)]

Originally Posted by Argus25

Al, I just posted about that power supply with the surge limiter and current transformer system in the vintage computer section.

Thanks, Hugo, I'll look that up!

Everyone: here is some data to give you the more detailed picture. I hope this will be useful in clarifying the context for my original question.

My (intended) circuit is a very conventional IGBT half bridge, with full wave rectification using D22-20 stud rectifier diodes.

(These are very modest looking, in fact tiny compared to some vintage ones I have from the 1980's.)

However, I chose them because they can pass an average forward current of 20A each, and a surge current of 275A each.

They can do this for 2mS, and then they can pass 200A for the next 2mS.

Ahah, you might say, a period at 50Hz is 20mS, so they're smoke!

However, they are backed up by 4 x 1.5KE 220CA bi-directional transient voltage suppressors. Each can dissipate 1.5KW and can tolerate a surge current of 200A for 8.3mS. So together, they can dissipate 6kW instantaneously.

Icing on the cake, there is a repetitive pulse rated 1200V, 1.5uF, capacitor across the DC- and DC+ rails.

The electrolytics are the EVOX-RIFA PEH169 series. These have a tiny inductance of 16nH, which is irrelevant at the frequency I'm concerned with.
But relevant to this thread, they have a ripple tolerance of 15A at 10KHz. But as I say, the bi-directional TVS will take the heat before they can.

My original frustration came when I was working with the well-known capacitor discharging and charging equations, e.g,

dQ/dt, equal to the reciprocal (negative) of the time constant times the charge on the capacitor

which resolves to Q(t)=Qo * e ^-t/RC

I basically had too many unknowns, especially R.

But I also had the bigger picture...

I=V/R, simplifying, by dropping, for the time being, the effect of e * ^-t/RC.

So I could see that dividing the volts on the capacitors after 20mS=370V, by the tiny, few hundred milliOhm ESR of the capacitors, I'd get a literally astronomical current.

It was just a thought experiment - the usual, '...what happens if we divide something quite large by something tiny?' so no calculator needed.

This wouldn't pass the common sense test - so I became focussed on the complex impedance of the mains at the outlet.

I'm not in a shed, but I am deliberately going to be using a small, not coiled up, extension lead, 2 metres long. My common sense, bearing in mind all that I now know, says that the peak current will be somewhere in the hundreds of A for a couple of periods.

Does that help with the wider context?