[SI-LIST] : TDRs and frequency domain

Jay Diepenbrock ([email protected])
Mon, 15 Dec 1997 08:54:49 -0500

Mark wrote,

>>Another brief question regarding TDR measurements.

>>It occured to me that I'm using a 38 psec rise time TDR to check thhe
impedance
>>of boards where the fundamental frequency is less than a GHz (so harmonics of
>>interest can extend upward towards 10 GHz where I know FR-4 is extrememly
>>lossy). It seems that I should be checking the board impedance at the
>>fundamental, and a few harmonics to really understand the performance of the
>>board. Not knowing very much about how a TDR works, or what the bandwidth of
>>the sampling circuit is, I'm realizing that I'm no longer sure what it is I'm
>>looking at.
A TDR is really a very fast repetitive step generator in the same box with a
high
bandwidth sampling oscilloscope. It stimulates the device under test, displays
the
reflected (or transmitted, in TDT mode) waveform, and does some math to
calculate
the reflection coefficient and impedance by comparing what it sees to what it
would
get driving into a pure 50 Ohm load. H-P's App. Note 62 has a good description
of how a TDR works (I'm sure Tektronix has something similar, but I don't have
it).
The bandwidth of the TDR sampling heads are typically 20 GHz or better,
depending
on the manufacturer and model number. Of course, you'll see impedance
discontinuities
much shorter at that rise time than your circuit may care about, but it depends
on the edge rate (rise/fall times) of your signal. You can approximate the
maximum
bandwidth of a signal using fmax = 1/pi*tr, where tr is the rise/falltime of
your
signal, regardless of the fundamental frequency (some people prefer using
0.35/tr,
but you get the idea). This would yield a max. frequency of approx. 318 MHz at
a
rise time of 1 ns, independent of the fundamental frequency.

>>Can someone give me a brief overview of how a TDR works, and why using a 38
>>psec sampling head is OK (or better) when fundamental frequencies are well
>>below 1 GHz? A network analyzer tells all, but what exactly can I determine
>>about the performance of my boards at various frequencies when using a TDR?
One can't deduce frequency domain behavior directly with a TDR, it's much easier
to just measure it directly with a network analyzer. However, with the use of
an impulse forming network and FFT software, one can calculate it indirectly.
Our Research folks published a couple of papers on this technique in IEEE
journals
back in 1992. Contact me offline if you'd like the references, or I'll post
them
here if there's enough interest.

Jay Diepenbrock

Advisory Engineer
Interconnect Technology & Qualification
IBM Global Procurement, B8UA/061, RTP, NC
Phone: 919-543-8804 Fax: 919-543-3642
Email: [email protected]