Re: GIs - the saga continues.

Pete C wrote:
On Thu, 06 Jul 2006 17:21:05 +0100, Gibbo <gibbo@xxxxxxxxxxxxxxxx>
wrote:

For dealing with this I wouldn't use a capacitor in parallel across
the GI to bypass the high freqs, but a choke in series before the GI
to block them.
Aye, now someone else has mentioned it!

Was going to mention it last night and had spent some time looking at
common mode/differential chokes.


Ok.

It's pretty basic stuff, even a wifebeater-drinking fule like Marvo
mentioned it ;)


Oi! I drink wifebeater!

In your earlier reply you said a capacitor. You even gave details of the proposed cap. i.e. 1000uF, non polarised - 35 pence

Yes, to bypass up to 30mA of AC leakage current across the diodes at
50Hz.


Which isn't the problem.

A choke in series with AC will do *nothing* to attenuate the RF until you try to draw some current from it. It doesn't matter what value you use.

Anyway let's run through this with some figures (as you seem to like them).

Let's go back to my 40 amp charger which puts (fundamentally) 70kHz down the earth lead. With the ground disconnected let's assume it was trying to put 20 volts of this on the earth lead (it's actually somewhat more). This was clamped at +/- 1.3 volts by the GI thus putting it into conduction.

So on the load (boat) side of the choke we have 20 volts at 70kHz.

At 70kHz the 0.7mH choke has an impedance of 308 ohms.

Until the GI conducts there is *no* current. Zero current through 308 ohms = zero voltage drop. So until the GI conducts the choke will do *nothing*. All it will do is limit the current once the GI starts to conduct. But that is too late, it's already conducting, so might as well not be there.

Yes, that's why I added to include a cap, for best attenuation.


I think you added the cap *after* you said one wasn't needed and you wouldn't use one, then realised that was wrong :-)

So a cap will have to be added from the GI (boat side) to the GI (outside world side).

We need to get this 20 volts at 70kHz down to below about 0.5 volts. So we need to attenuate it by a factor of 40.

So the cap has to have an impedance at 70kHz of (roughly) 1/40th (actually 1/39th) the inductor. i.e. 308/40 = 7.7 ohms. Call it 8.

8 ohms impedance at 70kHz = 280nF. Call it 270nF (nearest preferred value).

I'd look at one of these, cost around 33p + VAT.

<http://www.niccomp.com/Catalog/npx.pdf>

No mention of ESR but gives a dV/dT of 100v/us.


Irrelevant.

Could even use a 1uF one for 71p.


Could also put some baked beans in the oven. That won't cure the problem either but it's as good as your suggestion.

Now even assuming we can find one with an ESR of zero this means we need a cap of value 270nF. Dead simple. That gets the 70kHz down well below the problem voltage across the GI.

Problem sorted.

Good!!!

Except...........

We now have a 270nF cap and a 700uH choke forming a series resonant circuit at roughly 12kHz (from the head - I haven't calculated it).

So any interference coming back down the earthlead from other equipment at 12kHz will get *amplified* not attentuated. Any other RFI coming down at subharmonics of 12kHz will kick it into resonance and thus force the GI into conduction (this isn't wild theory - this is reality).

Were talking SMPS here. Correct me if I'm wrong, but the RFI will be
at the switching frequency and above only, and not below 70kHz.


Well for my particular charger the fundamental switching frequency is 70kHz. 1) That doesn't mean everyone else's is. 2) There is a very common type of switch mode PSU that suffers quite horrible subharmonic instabilities and consequent subharmonics. But they continue to be used because they have so many other advantages (like they don't blow up for one). My charger is of this type, I just chose to concentrate on the 70kHz fundamental.

Line frequency triac controlled chargers (still very common, even though they fail the EMC directive) are extremely common. These produce RFI (if you can call it that at such low frequencies) right down to 100Hz. And *huge* amounts of it. Enough to actually distort the incoming voltage waveform. Thay also have these lovely offending caps in order to try to at least get close to the EMC directive.

Quote:

Spectrum analyser showed peak at 70kHz (the fundamental switching
frequency of this particular charger - different for each item) plus
harmonics as expected up to several MHz.

An SMPS won't operate at a frequency that's audible, for obvious
reasons.


Apart form the ones that do. Like the average CRT TV for a start (either 15.625kHz or occassionally 7.8kHz)

The only reason manufacturers moved to higher frequencies was to get the component size down. This was a struggle due to swtching losses. New techniques (such as zero volt/current switching) have allowed it. That doesn't mean there aren't millions of switchmode PSUs out there operating down to 15Khz and even lower in some cases.

Have you seen any RFI from equipment at 12kHz and below, that's likely
to be used on a boat?


Well if we list stuff in general it's almost endless. If we stick to stuff likely to be used on a boat then:-

The average CRT TV, most triac controlled chargers such as: Heart Interface Freedom range, Heart Fleetpower series, Heart Interface EMS series, Heart load manager (not a charger), Mastervolt badged Traces (and the earlier Victron badged Traces), Trace truck series, Trace DR series, early range Victron Pheonix series, Trace M series, early Matservolt Dakars, Tecsup Tecpro range, Earlier Sterling Marine combis and chargers (when they actually made them instead of importing them), Vanner combis, older Studer combis and chargers, how long does this list have to be? There is one *hell* of a lot of this equipment around.

Even a bog standard line frwquency transformer/rectifier will preduce interference at 100, 150, 200, 400, 800 Hz etc up to several Khz. But now they have these lovely caps in to shunt it all down to ground.

If you could try one out and let me know how it goes I'd be most
grateful...

Not much point!

Why not? You give up far too easily!

I might agree with you there.

It's just getting interesting!
(not for everyone tho!!! :) )


I think you will probably mess around with this for a few months then say to yourself "oh he was right". And remain quiet. If you come up with a solution then I will salute you. I'll even put your name on my website with the solution.

Remember I have nothing to gain from this. I'm just warning people of a problem.

You're recommending people ditch their galvanic isolator and buy an

Show me where I said that.

expensive isolation transformer, and you haven't properly investigated
the alternatives.


I have investigated it enough to know that a test is necessary to see if the GI is being forced into conduction by RFI. Which is what I said in the original post, and is what I say on the website.

Quote:

No matter how I look at this, no matter what calculations I do, no matter what experiments I do, I cannot think of a way out of it. Other than using an isolation transformer.

Even if a tiny amount of current gets through the GI, it's well worth
measuring this (with the aid of a small series resistor) to determine
whether it's really a problem or not.


No, the test is to put a 'scope across the GI. If the RFI is anywhere near even a half the conduction voltage of the GI then something needs to be done.

It's no good someone sitting there thinking "I've got a GI so I'm not fizzing my hull".

It's no good someone measuring the DC voltage across it and drawing the same conclusion.

It's no good someone measuring the DC current through it. A) what is acceptable? and B) the measurement could be miles out due to the RFI confusing the meter.

A proper test needs to be done. And if it shows a problem then something needs to be done to cure it. Be that removing all the offending equipment or changing over to an isolatiobn transformer.

I know which I am doing.

Obviously some installations will have no problem whatsoever, in which case their GI is fine. But many would find a problem.

--
Gibbo

Emails to this email address are deleted at the server. I don't even get to see them. If you need to email me you can do so via my website.

--
Posted via a free Usenet account from http://www.teranews.com

.

Relevant Pages

  • Re: Choosing a ferrite bead
    ... it is exactly a choke. ... plane then it is a capacitor, ... the 70's, who was an rf type, telling me that ferrite beads were predominantly lossy, so acted more to increase trace resistance than like chokes. ... That's why he could get about as much effect from tacking one on top of the trace as he could cutting the trace and running it through the bead, as I wanted to. ...
    (sci.electronics.design)
  • Re: Choke on Yamaha 650
    ... Well apparently my whole choke issue boils down to the fact that I had ... having a hell of a time cold starting it and then it stayed off ... it wouldn't hold idle, with choke off it would stall under 3,500 rpms. ... while with the choke ON before switching it off? ...
    (rec.motorcycles)