crosstalk

[chorb]

Hey EE guys out there, I am wondering how I should go about setting up a lab measurement of crosstalk between two adjacent wires on a printed circuit board. I have read that the best method is to have 2 wires, the first wire grounded at one end and the other end hooked up to a probe/ ocsilliscope, and the second wire hooked up to function generator and a probe/ocsilliscope. From there run the pulse through the appropriate wire and measure the amount of signal gained on the grounded wire.
Is this a good setup for this sort of thing?
Any other methods of calculating this?

 

[bobsmith1492]

It might be better to hook the receiving wire up to a high-gain, low-noise opamp on the board to amplify whatever it picks up... the scope leads would likely pick up more noise than you could ever hope to realize from a single wire.

Also, how much current are you looking at? If it's low current (like that from a function generator), the output might be so low that it would be impossible to tell - there's only something like a few picofarads between breadboard traces. That capacitance and some inductance (that you could calculate) would likely be the methods of crosstalk signal generation. You could look at those to determine what levels to expect; you can find the typical capacitance of a breadboard online.

I wouldn't expect any noticeable crosstalk (measureably noticeable, that is) unless you ran either high current and/or a high frequency signal through your transmitting line.

 

[Born2bwire]

EDIT: Ahh cursed be the double post.

 

[Born2bwire]

If you have access to a program like ADS Momentum then you could easily run a simulation to see what the crosstalk would be as well. You can lay down the actual trace in the program and find out what the S-parameter measurements are. I would imagine that most microwave labs would have the software on hand. But if you're not thinking about microwave frequencies, more like up to the mid KHz range, then I really wouldn't worry too much about crosstalk. If you really want to, you could try putting a ground trace in between the wires to help act as a shield if that's possible.

 

[bobsmith1492]

D'oh, sorry, I was talking about a breadboard.... you said PCB. Ah, well; the same things apply to a PCB. You can easily get the stats for FR4 (the most common PCB material) online, like here: http://www.e-proto-types.com/resources/information/fr4_laminate.html

From there you can do calculations; google it. A precursory look tells me there is material out there that could help you.

Again, like Born2bwire said, unless you're way up in the megahertz ++ ranges, it's not much of an issue, unless this is for a class or lab, in which case we're not supposed to answer that here at ATHT.

 

[chorb]

No, I am working on this as a project at a internship position, nothing to do with academic studies, although reading through some of these High-Speed Design books makes me feel like it is.

I am going to have to come up with a scheme to test the PCB no matter what (as its what I have been assigned to do), so while theory does help me understand what I am doing, I still have to set-up the simulation and test for crosstalk between wires.

I will be running between 20Hz - 20kHz, up to 180V, and while there might not be much crosstalk, I need to measure it down to the -90dB level.

Google results havent really helped in terms of how to test two adjacent wires, only what I mentioned above; so if anyone has any more ideas, I would be more than willing to hear them.

 

[Born2bwire]

Originally posted by: chorb
No, I am working on this as a project at a internship position, nothing to do with academic studies, although reading through some of these High-Speed Design books makes me feel like it is.

I am going to have to come up with a scheme to test the PCB no matter what (as its what I have been assigned to do), so while theory does help me understand what I am doing, I still have to set-up the simulation and test for crosstalk between wires.

I will be running between 20Hz - 20kHz, up to 180V, and while there might not be much crosstalk, I need to measure it down to the -90dB level.

Google results havent really helped in terms of how to test two adjacent wires, only what I mentioned above; so if anyone has any more ideas, I would be more than willing to hear them.

Do you have access to a network analyzer? While such a device would be more at home in a microwave lab, it would allow you to do a four port measurement on your two test lines. The S21 and S12 parameters would be a measurement of merit for the amount of crosstalk. Take a look at S-parameter measurements. This is the kind of thing that I am familiar with for measuring crosstalk and is probably what they want from you.

Otherwise, I would try to see if I cannot communicate with the function generator and scopes via the computer. Most of them have RS-232 ports that you can communicate with them. You can then write a simple Java program to run an automated test and measurement of your experiment. I've done this before to test voltage monitoring boards. This way, you can automatically sweep your frequencies and voltages quickly and record the results in the computer. Otherwise, I think you will need to do a lot of manual measurements to characterize the crosstalk as a function of frequency and voltage. Also be aware of the noise floor of your scope and measuring devices. -90dB is a common noise floor for devices but if they are old or cheap, they may be above that.

 

[f95toli]

Most network analyzers won't work in that frequency range, they generally work from 10kHz and upwards.

I think you best option would be a dual channel dynamic signal analyzer. This is basically an FFT spectrum analyzer but with the possibiility to measure e.g. S12 (but it is not exactly the same thing and it is usually not called S12 in this case) in a frequency range 1 mHz-100 kHz .

Something like the SR785 should work. I use one from time to time.

http://www.thinksrs.com/products/SR785.htm

I seriously doubt this can be done with any precision using only an oscilloscope.

btw, I am not sure but I guess it should be possible to measure crosstalk using a lock-in amplifier, at least that should give you the S/N you need.

 

[f95toli]

btw,

I found this
http://www.reed-electronics.com/tmworld/article/CA301478.html

How to measure cross-talk at -100 dB using a SB Audigy platinum...

 

[chorb]

thanks for the info, but I dont think my company will spring $13k for their intern, dont even know if they would spring the $200 for the sound card.

Im just gona try the oscilloscope method and if it doesnt work out then I'll but in the request for one of those

 

[f95toli]

Good luck.

However, is this a measurment they really need? If so, they need to realize that some measurements simply can't be done properly without the right equipment.
As I wrote above, I seriously doubt this can be done using an oscilloscope.
And the soundcard is a very neat -and cheap- solution but I don't think I would trust a measurement done that way UNLESS I was able to somehow test it using a known sample to verify the numbers.

One option would be to rent a signal analyzer. Renting equipment is often a good solution if you only need it for a week or so.