Measuring Crosstalk 1

[bpelec]

I am trying to measure crosstalk between channels in a slip ring. Due to the assembly of the slip ring, there are 2-3m cables on each side of it, and my test signals have to go through these cables.

I am using a HP 8714B Network Analyser, in transmission mode, to measure the crosstalk.

I have connected the input of one of the slip ring channels to the RFout and have terminated the output of this slip ring channel (on the other side of the slip ring) with a 50 ohm load. I have then connected another slip ring channel to the RFin of the analyser, and have also terminated this channel with a 50 ohm load.

The input and output of the analyser are on the same side of the slip ring - the terminating loads are on the other.

Ok.. Here are my questions:

1) What are the effects of background noise on the crosstalk measurements. The tests are being performed on the factory floor, and there are several machines in operation. It is likely that the ground connection in the test area has various harmonics on it from this machinery, as well as there just being EMI in the area.. Would this be a problem, or would it couple to both channels of the analyser and hence not make a difference?

2) Related to 1, if I ground the case of the slip ring, might noise on the ground connection affect the crosstalk measurement?

3) The cables that connect to each side of the slip ring include several coaxial cables, and a few unshielded. I am testing between shielded channels... How much of an effect are the cables going to have on my crosstalk compared to the slip ring itself?

4) I think the worst of the crosstalk is being casued by a ground loop that connects the shields of the two channels I am testing... Is there anything I can do about this?

Sorry for such a long post and thank you very much for taking the time to read it. I would greatly appreciate any help and advice that you have to offer.

Best regards,

BPELEC.

 

[VEBill]

1) Try it and see. Narrow band noise would be less of a problem than broad band noise.

Q: What frequency range is the slip ring req'd to operate over, and over what range are you testing?

2) Try it and see. To start, use the connections most representative of the final application, and then if req'd make changes if perssible and if req'd.

3) Possible considerable. But if the cables are separate, then separate them (distance). Can you add shielding over the conductors if red'd?

4) Improve the grounding or isolate the grounds.

 

[bpelec]

VE1BLL, thank you for helping me again.

The slip ring has to meet a crosstalk specification that ranges from 300 kHz to 35 Mhz.

I have tried a number of tests on the unit and have varied which parts are grounded... So many things make a difference it is difficult to get a handle on the fundamental causes...

With regards to your first point, why would narrow band noise be less of a problem than broadband? Do you just mean that it would only cause a small error at one point, rather than across the whole band? My feeling is that as long as both channels get the same background noise, it shouldn't affect the crosstalk measurement - do you think that is correct?

With regard to your last point above, how would I improve the grounding, and/or how would I isolate the grounds/shields?

Best regards,

BPELEC.

 

[VEBill]

When you do a sweep, a narrow band noise might show up as a spike in the curve at a certain frequency. You try the sweep again and it might be gone. You disconnect the source and run the sweep again and the the spike is still there. Pay attention to the exact frequency to see if it is one you recognize (like an ISM frequency, maybe 13.56MHz, from nearby equipment). All this means that you can basically look at it and say, 'Noise, ignore it.'

It would be difficult and probably not required to set up a differential measurement to try to cancel noise. The setup you describes included only one input channel, so I'm not sure what you're thinking.

If two signal share a poor ground and this cause problems, then you can either separate the grounds (joining them later at a single point gound), or improve the quality of the ground as best you can (less resistance and inductance).

 

[bpelec]

Thank you for your help. I am going to go back to do some more tests...

You're right, I only have one input channel, but there are two connections to the analyser (RFout and RFin). I thought that if there was background noise from nearby equipment, then as long as both the RFout and RFin channels had the same noise on them, then the noise would not be an issue in the crosstalk measurement... Thinking about it again, I think I am wrong - the analyser will compare what it generates with what it receives, and as it is not generating the noise itself, it will not take into account any noise being added to the Rfout channel. Therefore, the analyser will detect noise on the RFin channel and will include it as crosstalk.

 

[VEBill]

I've sometimes thought that such an analyzer could, instead of simply emitting plain old CW, emit some complicated and changing waveform that it could distinguish from any random noise and it would thereby be more or less immune to external noise.

If the noise is constant (often a bad assumption) then you can sometimes calibrate it out to some degree.

Or just come back and do the sweep at 3am when things are quieter.