Connecting a grounded single ended source to a differential input amplifier via coaxial cable.

Yes.

Don't see why.

Yes.

Ground loops? From where? The amp supply is isolated.

If you do this, something has to establish the common-mode voltage!

What's the signal bandwidth here?

John

Worst case there is always Mini-Circuits and their little transformers :-)

Since there ain't too much happening in politics right now you could have mentioned split grounds here, then we'd have another hot-blooded discussion ...

Snort

Cheers

PeteS

Hi John, thanks for the reply.

Signal bandwidth is up to 100kHz, but propably a lot less.

I say I would lose the advantage with common mode noise rejection because I connect a coaxial cable, which shields the inner core, to the diff amp, and hence the noise on the 2(1 shield and 1 inner core) conductors are not the same. A twisted pair would have been better- but the customer already has 100s of meter of coax laid out. Or am I wrong here?

The customer wants to connect the shield to the chassis of the measurement instrumentation, and that would create a ground loop, so he will short isolated ground to chassis ground. He claimes that gives better readings. He thinks the noise on the shield needs a good return path, and connecting to the instrument chassis on one end and earth, which is already a very good return path, on the other end is a good idea.

I already have a diff amp, so that's why I am leaning towards it, I also thought it would be a beter solution. The biasing resistors would connect down to isolated ground. What do you mean something must establish the common-mode voltage?

Why do you think the single ended amp would be better? Obviously, then the customer should not short the shield to the chassis. I can make my diff amp single ended by connecting the negative input to (isolated) ground- this is done with a mosfet switch with 1ohm Ron.

Any other suggestions as to how to best measure grounded signal sources with isolated amps?

MJ

Try wrapping the coaxial cable around a ferrite toroid. This turns it into a balun, also known as a transmission line transformer. Googling for "balun 'transmission line transformer' " threw up a number of links. The first one

actually describes more complicated variations on the standard 1:1 balun as does

I had to dig down quite a way to find a link that even talked about the standard 1:1 unbalanced-to-balanced application

My favourite reference is to Matick R E "Transmission-line pulse transformers - theory and applications" from the Proceedings of the IEEE volume 56, pages 47-62 (1968) but you need a university library or a subscription to IEEE Explore to get hold of it.

-- Bill Sloman, Nijmegen

It is not a "ground loop," it is a major line imbalance and uncontrolled reference return...

That is probably right but not for the reason he thinks. The chassis to isolated ground connection eliminates noise current injection into the IA input due to potential differences between the two grounds. This will be going on internal to the chassis.

That used to be called a "coaxer" and it is a very high frequency common mode choke, not a balun...I have used this method and it works well.

The largest interference will most likely arise from external current noise sources magnetically inducing a differential mode voltage into the coax. He has no option except to voltage-to-current convert the sensor output and terminate in a low impedance at the IA in that case, assuming he has to use coax and maintain a wide signal bandwidth.

It's easy to show that common mode rejection absolutely depends on having the impedances for both differential inputs exactly the same. Provided that both inputs are maintained within the common mode voltage range, the diff amp doesn't care, so to speak, whether the signal is single-ended or what some folks call 'true' differential. There is much practical information on this subject available at these websites:

For audio signals, the best results can be obtained using transformers.

Paul Mathews

Thank you all for the replies- the whole problem is still unclear to me though.

Also, this is not an RF application- it is an accelorometer sensor app with a single ended voltage output. The fact that the diff amp runs from an isolated supply make matters worse, sine the customer wants to connect his signal gnd to the diff amp chassis- which is at the same ground as his signal gnd- but the diff amp runs off its own isolated supply.

Would it be wrong to just connect the shield to the one input of the diff amp and the inner core to the + input of the diff amp? And not connecting anything to the chassis. John says a single ended amp would be better. But the single ended amp would also be isolated (all the input channles run from an isolated supply in this measurement system). Most solutions show a 2 wire + shield cable- I have only 2 conductors, the shield and the shielded inner core.

Is it so difficult to measure a ground single ended signal with an isolated diff or single ended input amp?

MJ

Just put a resistor from the 'ground' wire of your signal cable to the local ground. That will keep the common-mode in range if the distant end's 'ground' connection is ever allowed to float. Any other signal-conditioning (filters, clamps, etc.) will be the same regardless of the wiring scheme.

Do be careful, though, that the 'ground' wire of the signal cable doesn't connect IN MULTIPLE PLACES to any kind of building frame or other grounded metal; minor ground-fault currents can be guided through your cable and the resulting I*R voltage drop can swamp a small signal. The resistor-to-ground at the amplifier end should be chosen to be higher value than the wiring resistance so it won't mess measurements up. Much.

I think that if your signal comes in over a cable longer than a meter or two, the few uH of common mode inductance added by this kind of balun is not going to make much of a difference.

Low frequency current flowing on cable shields (common mode current) ends up leaking in because of the cable shield resistance, so shield currents are to be avoided. Baluns don't help against low-frequency common mode current. Transformers and differential amplifiers do. The choice of which is better depends on the signal bandwidth.

Would you hire Bill or me, John? ;-)

Jeroen Belleman

in cases like that, you are losing the balanced use of the input. In my opinion, you should be tying the shield side to common. Just use either input of the amp if polarity isn't a problem, and tie off the other input so that it won't detect any noise. I do think if you tie the unused input to common also, you'll be all set.

quoted text -

My 'common' is isolated and floating with respect to the sensor's common. Tying the shield to common, will short my isolated ground and break the isolation barrier. And a current might flow in the shield, inducing noise in the inner core, because of potential differences in the 2 'common' points. Remember, my input stage is floating.

MJ