Calculate induced/coupled voltage in a wire

[ianholmes]

I am trying to work out the filtering requirements for some inputs to an Instrumentation Amp design to stop nasty RF getting up it. I would like to be able to calculate the magnitude of the induced voltage and current based on the impedance of the amp and the length of the cable connecting it to my sensor. My amp is designed to operate with (-138dBV) small input so I'm concerned about common and diff mode injection.
I'd rather not "put a capacitor on it" and see what happens.

I did some maths but I wouldn't like to bet my beer that it has any solid background!

Can anyone point me in the right direction for a technique?
I have lots of clean paper and a sharp pencil!

Ta

 

[Warpspeed]

That is a toughie, nasty RF is that at 100Khz or 50Ghz ? And is the coupling mechanism inductive or capacitive ? The biggest problem is working out the actual RF field strength because it can be highly influenced by nearby objects.

About the only way you are going to come up with a figure is place the equipment in an RF anechoic chamber, and sweep it with a known field strength from several directions with different polarization. An accredited EMI lab can do that, but it will not be cheap.

I doubt if you can do a theoretical analysis because there are too many unknowns. Best test it and find out at what frequencies it is most susceptible and try to work out the coupling mode that is creating the problem. Just design it using best practice and then test it.

Shielding, filtering, unwanted internal resonances, grounding, and impedance levels all can have major effects on the results.

Try and find a copy of "Noise Reduction in Electronic Systems" by Henry W. Ott ISBN 0-471-85068-3

It is a pretty good introduction to the subject.

 

[ianholmes]

Thanks Warpspeed.

I'm looking at 100kHz through 1GHz. You're right about the costs of extenal testing! We do emissions on site but don't have amplifier for immunity, we usually hire a bunch or handhelds and wave them about. I was hoping to do some cigarette packet maths assuming a far field incident wave with a uniform strength. My musings on it suggested that a lower impedance on the end of the cable would be a good thing as the voltage would be smaller (induced), but I'm not sure it would stand up in court! Many years ago I had a degree in Physics but my brains gone all woolly after years in electronics! I was hoping someone might be able to spark it back to life. Thanks for the hint on the book, I'll have a look for it.

I'm still up for the maths if anyone knows where to find it.

 

[Warpspeed]

Your best bet might be to select an impedance suitable for maximum energy transfer within the system, and then try to keep the nasties out with good design practice.

Try and get a copy of Ott's book it is very informative about stray coupling modes, grounding strategies, screening, and good practice generally. It is not highly mathematical, more aimed at the practical working engineer.

There is quite a bit on the theory of electromagnetic propagation, coaxial cable, twisted pairs, screening and grounding. Lots of good stuff in there for good EMC design.

Ott started out teaching EMC fundamentals within AT&T Bell Laboratories way back in the dim past. He then went on conducting training courses on these subjects to a wider range of engineers. His book is probably still the best introduction to practical EMC design you will find anywhere.

 

[logbook]

It is a very worthy concept trying to calculate what effect nasty RF will have on your system. It is however a largely pointless exercise if you can just build the system and see what you get. Building it gives you a 1:1 real time simulation!

You say your amplifier is sensitive to -138dBV. If my maths is correct your amplifier is sensitive to approximately 0.1µV. Now such an amplifier will not have much bandwidth anyway but it may well do nasty non-linear things if you inject too much nasty RF into it.

Let’s try to get real here. You are not gong to be able to calculate the pickup into the cable from every frequency from 100kHz to 1GHz and sum the resulting interference. The task is immense. Just use a screened twisted pair and if that is not good enough stick that in a copper pipe.

The biggest ambient field will undoubtedly be the mains frequency which will be all pervasive at the sensor frequencies indicated by your sensitivity ( a few tens of hertz at most). You still need a capacitor between the inputs of your amplifier to reduce RF injection. By far the best way of getting rid of the mains induced noise is to integrate over an integral number of mains cycles. This is what is done in 5½ digit DVMs, and above, to get nice quite readings.

 

[ianholmes]

Thanks Logbook. I guess I'll build it then! The application is a loadcell amplifier that needs to operate with a wide range of sampling periods from the sub hertz to the 10 hz region (i.e. this is the time available to get the data from the loadcell). The problems here are long term stability (temp, offset drift) and the ususal suspects. The layout is critical to the performance (I have on board signal processing too) so I want to protoype on a real pcb not a breadboard which is more costly. I just wondered (hoped) that there might be a way to check out my "best practice" design with a calculator.
I haven't had any problems in the past with the approach you suggest, I'm just a bit anxious to get this right. WHen I've finished the design will be tested by Weights and Measures which a) costs loads and b) they don't take prisoners!

Perhaps I should just get on with it. Thanks for everyone's input.