Does characterization of Transmission lines need buffers

[prasu443]

Hello every one,

For my project work, I need to design high speed buffer structures on a test chip. These buffers are used as on-chip data reference sources for characterization of transmission lines. My question is, why characterization of transmission lines need on chip data reference source and how buffers help in characterizing transmission lines ???

Somebody plz help me out,,,,


Regards
Prashanth

 

[jgruszynski]

If the characterization is digital BER, then it makes sense. It's not how you'd do it for analog applications or bleeding edge digital (which is actually analog).

 

[prasu443]

Hello jgruszynski,

Thank you for your reply. But still I have some confusion. The high speed buffers are the only structures on the test chip along with some SOLT calibration structures.

Why buffers are used for the characterization of transmission lines? Even if it is digital BER test. Can't we input the signal directly to the transmission lines?

Waiting for your valuable reply

Regards
Prashanth

 

[jgruszynski]

BER is a purely digital test that doesn't really pay attention to analog subtleties like impedance match or reflection coefficients like an analog microwave application would worry about. The "quality" of the transmission line would be determined by the BER rather than some physically object measure like mismatch loss.

It's an idea that drives analog folks crazy in frustration but I've seen that kind of thing for "transmission line" interboard connectors. Impedance control in the digital world tends to be a 0th order approximation of the concept usually taken to 1st or 2nd order by analog folks.

Having buffers that resemble product line drivers or receivers makes BER a more meaningful way of determining go/no-go as functional test and could conceivable (but needn't) give a meaningful parametric metric. Whatever bizarro source or termination impedances would at least be somewhat controlled for in that you have then in both cases. In digital circuits, thresholding give you lots of slop in manufacturing parameter SPC. Analog isn't so forgiving. Still, it could bite you in some cases.

Normally doing characterization of a transmission line for analog applications would not include buffers - you'd just touch down on the ends and do s-parameters. Buffers would be bad for s-parameter measurement of a transmission line because they invariably have bizarro impedance with frequency dependencies that is certainly not at the transmission line characteristic impedance.

There's also a possibility that the buffers really are not part of transmission line testing but rather are for characteristic the buffers themselves rather than transmission lines. Transmission lines might be present (especially if you find them alone also) for calibration and de-embedding for buffer testing. Just a thought.