[TowerTalk] TDR

[Jim Lux]

On 9/4/12 1:37 PM, Jim Brown wrote:
> On 9/4/2012 1:04 PM, Roger D Johnson wrote:
>> I'm used to the old Tek pulse type TDRs. What would be the effect on
>> the trace of a swept frequency TDR
>> with FFT when the line is terminated with an antenna with several
>> resonant points,
>
> I don't have a good answer to that.  I do know that both my Ten Tec TAPR
> VNA and the DG8SAQ VNA can provide a good TDR that sees a lot of detail
> on the line with an antenna attached, and I think the AIM units can
> too.  I've also seen what I think are reflections from antenna elements
> and from the earth, but might be "wraparound" in the FFT.
>

It shouldn't make any difference.  That is, the time domain and 
frequency domain are interconvertible, in the "continuous" transform. 
However with discrete (sampled) transforms, you can run into problems.

Where you WILL get into trouble is if the phase difference between two 
successive frequencies is > 180 degrees.  Then, the unwrapping doesn't 
always work right.  (for a long cable, with high frequencies, this is 
easy to do, because you have many cycles in the cable)

You need to have enough points to make sure this doesn't happen.  It's a 
particular problem if you have a sharp resonance, because the phase 
changes very rapidly as you go through resonance.

If your 3dB bandwidth is, say, 100kHz, that implies that you're going to 
get a phase change of 90 degrees in Z in about that much distance. You 
get most of 180 degrees in double that (200kHz).  So, if your analyzer 
is sampling every 100 kHz, (e.g. 5000 points from 0 to 500 MHz), you'll 
probably get into trouble.

A smart analyzer will allow you to tell it something about the system so 
that when it does the transform it can constrain things.  Also, the 
"easy" transform is the one with evenly spaced points, but you can also 
do transforms with unevenly spaced points.  You can sample more 
frequently in areas where the Z is changing rapidly, for instance.