In article ,
"Michael A. Terrell" wrote:

Telamon wrote:

What are you looking for? Do you want an explanation of why the
connector does not have a big effect on a 100MHz signal and yet has a
great effect on a 1 GHz signal for the same impedance mismatch?


Have you ever used a TDR to look at a transmission line? A 1.5:1
impedance bump can cause problems and the more impedance bumps in the
path the more it degrades the signal.

Yes. Time domain reflectometry.

The techs at Microdyne had both 50 and 75 ohm cables and adapters
available. I had someone tell me the video boards i had just tested and
calibrated were all bad. He was trying to use 50 Ohm cables in a 75 ohm
video test which caused the bandwidth to roll off to -3 dB at 16 Mhz
instead of 20 Mhz. It took all day to convince the old timers that they
had to use the right cables, because they had got away with the wrong
coax on the older, 5 Mhz systems for years.

You need more bandwidth for digital signals because of the faster rise
and fall times. For data to be valid you need a large eye opening in
amplitude and time for margin.

There are two things of significance:
1. The amplitude of the discontinuity.
2. The length of time of the discontinuity.

#1 determines the amplitude of the reflection.
#2 determined the frequency it becomes significant.

The cable is another issue. I know you understand that the attenuation
per foot increases with higher frequency. This can occur more rapidly
with digital signals due to the faster than analog transition times. For
digital you need a path with 3.5 times the bandwidth.

You have to understand another consequence of reflections on a
transmission line and that is the amplitude of the signal varies over
the length of the line due to constructive and destructive interference
of the forward and reverse waves so it is possible to measure a bigger
swing than you set the generator to drive the line with in some places
on the line. I have seen this confound more than one person until I
explain this to them.

To test a path the best thing to do is set the generator to produce a
1/0 pattern, which is the highest frequency of the bit stream and look
at the far end with a scope. Swap the cable or connector and measure
the change in attenuation.

If the impedance of the cable is wrong you will lose signal amplitude
due to reflection along with the loss per foot number.

--
Telamon
Ventura, California