50 ohm coax to 100 ohm balanced line (for forward/return/next/fext testing)

[forrestc]

I do quite a few Ethernet power injection designs.   So far, I've validated my designs with a cable validation tool, but I'd like to do better validation as time goes on.

I'm looking seriously at buying a spectrum analyzer with tracking generator (or similar instrument such as a network analyzer) which I can use for more detailed testing such as forward and return loss, along with NEXT.  One of the challenges I'm running into is not the test instrument itself - but instead understanding and finding the right cables/adapters to convert the 50 ohm unbalanced coax output to the 100 ohm balanced line which is used in Ethernet and CAT5.   I realize I could just electrically hook it up, but that seems wrong for some reason.

Anyone done this before and can point me in the right direction?   Even knowing what to search for in google would be helpful. 

[nctnico]

You can buy these kind of baluns off the shelve. Farnell/Element14 has (had?) them.

[T3sl4co1l]

You may also want a 150 ohm common mode driver, for EMC testing.  This isn't too hard to make:
http://seventransistorlabs.com/Precompliance/index.html

The balun I made for testing wasn't very good and needs some work, but yeah, they aren't hard to find, either.

Tim

[uncle_bob]

Hi

The bigger question:

What do you want to test?

Often cable plant issues are caused by things other than simple frequency response of a properly driven / terminated cable. With balanced lines, you have lines that ... aren't balanced. With multiple pair cables, you get crosstalk. When people pull a cable tight around a corner you get impedance bumps.

Each thing you decide to test takes it's own setup or even dedicated tester.

Bob

[forrestc]

What do you want to test?

This is a midspan PoE Injector application.  I'm not testing cables at all - which is sort of why I want to move away from the high end CAT5 validator and into something a bit more 'test equipment like'. 

I'm hoping to be able to come up with a real number for Insertion Loss, Return Loss and Crosstalk, to ensure they don't exceed the 802.3 limits.  I also am looking for some validation that the impedance through these devices are relatively close to 100 ohms.  I realize these are all interdependent.

Some of the challenge with the baluns I've looked at is the frequency range.  Ethernet seems to need testing from 0.1Mhz through 125Mhz, which seems to be pretty wide for a balun - which probably means I'll need more than one.

[Marco]

Can't you just use 10 GBASE-T ethernet transformer and impedance match with a resistive L-pad?

Isn't this more commonly done with TDR/TDT rather than spectrum analyzers and tracking generators? Interchangeable, but still.

[uncle_bob]

What do you want to test?

This is a midspan PoE Injector application.  I'm not testing cables at all - which is sort of why I want to move away from the high end CAT5 validator and into something a bit more 'test equipment like'. 

I'm hoping to be able to come up with a real number for Insertion Loss, Return Loss and Crosstalk, to ensure they don't exceed the 802.3 limits.  I also am looking for some validation that the impedance through these devices are relatively close to 100 ohms.  I realize these are all interdependent.

Some of the challenge with the baluns I've looked at is the frequency range.  Ethernet seems to need testing from 0.1Mhz through 125Mhz, which seems to be pretty wide for a balun - which probably means I'll need more than one.

Hi

A very practical way to to the job is with a 4 port rather than a 2 port network analyzer. There is no particular magic to matching everything up in a test like you are talking about. Calibrate everything carefully and do the math to extract the data. Been there / done that many times (with 4 port analyzers). The application wasn't ethernet, but it was something similar.

Bob

[chris_leyson]

MiniCircuits make some nice baluns and they even specify differential phase error for some of them. Off the shelf ethernet transformers may do the job but I don't think the specs are as good, internally the placement of the ferrite ring cores looks a bit random. I'm looking to measure crosstalk down 0.5mm flat flex wired as G S+ S- G S+ S- G etc I might just pull some of the ring cores out of an off the shelf ethernet transformer and use those.