3 Phase Frequency On Comms Cable

[Sidewinder]

Ok, let's see whose maths is still there because mine has left the building!

I have, I think, 3ph power being superimposed onto an analogue signal line.

I was expecting to see 50Hz on the scope traces, that is individual interference from each phase.

As they are all synchronised, then I would have expected them to add to the same 50Hz signal and even though phase shifted, the fundamental frequency is the same.

Ive taken a time / voltage trace of the noise, then run it through an FFT to get the frequency components, hence I was expecting 50Hz!

I'm not getting it!

Now I'm still working on the maths & graphs & the possibilities.

However, as we have 3 off 50Hz signals out of phase by 120 deg or 2(Pi)/3 could or would these superimpose a higher fundamental frequency?

If so what would that frequency be?

I don't think it would be 150Hz, because of the phase shift.

My maths head is really gone now!

It's been so long since I've had to look at stuff like this.

Forget about harmonics from loads for a minute, let's just think about the fundamental basic 50 Hz 3ph supply with each phase shifted by 120 deg.

Take it that the load is balanced & the pf is unity, just for now.

Looking for ideas now as I'm not getting what I expected!

 

[ProDave]

I think it would be 150 Hz.

In each full cycle, you will get three +ve peaks, and three -ve peaks, so the superimposed noise will definitely have a 150Hz component.

 

[Sidewinder]

Dave,

I don't think it will be, see attached graph over 2Pi, with 120 Deg that is 2Pi/3 phase shift, done the lazy way mind, I could not think of the right function to use!

You don't get 3 full peaks +/-.

3Ph.pdf

 

[ProDave]

Post the scope trace of the actual interference and the FFT analysis

 

[kerching]

I seem tomremeber it being something to do with phasor additions. Either graphically or via , i think, using negative numbers/j_notation.

This is going back 35 years ago when I had more pressing matters on my mind........20.5p for a pint of Boddingtons at the Kersal nearish to Steps area!,,

So i found this which is probably of no use whatsoever

http://scipp.ucsc.edu/~haber/ph5B/addsine.pdf

Just forgetting

 

[Sidewinder]

Somewhere along those lines Kerch, hence why my maths went down the pub!

It's not so much the amplitude I am interested in, but the fundamental frequency I am looking for, however the fundamental frequency of the 3 waves is identical @ 50Hz.

So, they should show a major frequency component @ 50Hz?...

Oh PD, I'm sorting out the wave & FFT's now, I was doing it when I started the thread, but I have been a little side tracked.

Mrs SW has made some Chinese food for dinner, Hmmm!

 

[Sidewinder]

OK,

I am withholding some information, but this is a deliberate action so that you don't jump to conclusions like wot I did!

The comms system is analogue, think like a block of flats call point system for letting people in, this time though, it is 2 call points in parallel, wired as per manufacturers instruction, to a single point of calling, not multiple flats.

230V a.c. supply to psu's 1 per call point, giving 12V d.c. for the call points.

Receiving device is 230.

230 devices are well apart, and could be on different phases.

Noise is present on call points even if the are powered from my van battery, without the engine running, so totally independent to the supply to them.

They are an am/fm system with a d.c. carrier of around 4 V.

Full volume is circa 0.2mV a.c.

The "Noise" is around 0.02V a.c.

Thus 10% of full volume ish.

I'll get the waves and stuff up shortly.

 

[Sidewinder]

OK pdf of the scope traces & fft's

Screens & waves.

All Wave Forms.pdf

 

[Sidewinder]

No, prawn crackers, a packet or so anyway!

They were left within reach, what do you expect...   :slap

 

[steptoe]

Are you having any bread and butter with that  :innocent   :slap

surely you mean

are you having some Chinese food with your bread & butter,,,,,,,,,,,     :innocent

 

[kerching]

Last time I saw a trace like that I never saw my Gran again!

Just saying...

 

[Bruspark]

Simple Digital equipment can produce all sorts of nonsense results . I suggest you try to isolate the point where the interference gets in by trial and error rather than trying to figure out its exact physics . May be worth checking if you've ended up earthing multiple times and produced an earth loop via the long call wires

 

[Sidewinder]

No earth loops.

The comms kit is analogue.

TBH I don't care where the noise is getting in, I'm only interested in the science at this point.

As I have said I have withheld information, deliberately, else premature conclusions will be drawn.

To find the noise by trial & error will cost in excess of £10k.

Oh, & I didn't install the kit, I just have to find the source of the fault

 

[ProDave]

Well as the dominant frequencies of the interference are NOT multiples, or sub multiples of 50Hz, I would rule out looking for mains interference and look for some other source.

 

[Sidewinder]

I realised that PD, which is what confused me.

I'm just wondering if the partial waveform that would occur in any single time period could throw off the FFT, as per the first 3ph graph I posted, it is "nearly" a full wave form, but not quite, and if the fft algorithm is a little off, I'd hope it is not, but, that could perhaps throw the calculation.

The noise amplitude being circa 10% full scale, the only thing I can think of is boosting the signal at the origin above the noise floor.

Basically I need to improve the SNR somehow.

It could be possible to bypass each section of cable one at a time to try to identify which section is inducing the noise, but that would be a few days work in itself.

Stupid install method chosen by customer!

There is one section of unshielded Cat5E, that could be replaced perhaps, though TBH I would prefer to re-route the whole of the CAT5E run inside the premises, and move the joint to where the cable enters the premises.

 

[kerching]

Going out on a limb here, puts on NOMEX and prepares to be shot down but, how is the shielding on the Cat5 terminated?

One end or both ends?....i have had major problems with network cables being earthed at both ends in the past, BUT this is probably totally irrelevant in your case .........just a thought, straws and clutching

 

[Sidewinder]

Straws & clutching is a good idea ATM!

The shield is floating.

I would normally terminate a shield @ "supply" end.

In this case there is 2 "supply" ends.

Plus there is no shield @ one end as it is converted to UTP part way through.

Just a thought, it is 3 wire comms, I wonder if I double up on the common, and use 2 pairs 1 pr Tx & common, 1 pr Rx & common?

Perhaps this could reduce the noise.

I may just tell the customer it is noise and let them sort it out, when they muck it up again they can call me back to fix it!

I'll let this run for today, to see if anyone has any solutions, or ideas, then perhaps I'll post some more info.

I don't want people jumping to conclusions.

Got a sound recording of the noise, going to try to get that off the phone onto the computer and try an fft on that if I can find software to see if it is the same results as the scope.

I'll post up what I get.

 

[kme]

Having com in each TP "should" alleviate the noise element somewhat - if you have full 8 wire available, I could also be tempted to run a TX/RX pair, purely as an exercise to see what effect that has on the noise.

Erm - whilst typing ( and thinking), you mention a "floating" shield??, which disappears somewhere into a UTP? If the shield is truly "floating", then it could be the route of ingress of the unwanted signal?

 

[ProDave]

Well if there's an unshielded bit in the middle, I would connect the two end shields to earth. You won;t get earth loops with the break in the middle.

 

[kerching]

One make of door intercom I used to install many eons ago, but I cannot remember the make ( terminals were marked. C H R O and. T, if that helps?) HAD to have the voice pair as a twisted pair, if you use burglar alarm cable the system went all moody. AND the pair had to be balanced NOT split..

Yet again probably no use in your case but, you never know

Jus saying