Is Transmission Line Reflection Frequency Dependent?

Question

I'm currently sending high speed serial over a coax cable. I've got an oscilloscope hooked up (on the load end) trying to watch the clock pulse signal. I've got a 10 ft coax cable with a 12.5 Mhz clock signal. I'm unsure of the coax impedance and the load impedance.

The clock signal appears to be deformed. I'm curious if this deformation is due to the cable acting like a transmission line and specifically due to transmission line reflection.. However, I'm unsure on some of the properties of the reflection.

Is the reflection of a signal frequency dependent? Thus the superposition of all of the frequencies in the square wave would be phase adjusted and report a bad clock signal?

Answer 1

^{simulate this circuit – Schematic created using CircuitLab}

If \$Zc=\sqrt{\dfrac{L}{C}}=Z_{source}=Z_{load}\$, then there is no frequency dependance, the signal is just delayed but not distorted. If any part of above equation isn't equal, then you get standing waves.

Answer 2

Try reducing the reflections (if there are any).

By adding an attenuator. Something like 3 dB, 6 dB attenuation? Attach this in between the coax and the oscilloscope. Did the deformation go away? The rx signal will be much smaller, but the attenuator will make the load appear matched thus reducing reflections.

[Edit] If you don't have an attenuator, add more cable.

You can also try to see if you can change the oscilloscopes load impedance to 50Ω and see if the distortion goes away.

Answer 3

Assuming your clock source is producing (approximate) square waves with a fundamental frequency of 12.5MHz, an expected rise time might be about 4ns.

Assuming the signal propagates at the speed of light (in reality, it's less than this, about half or a third), then to travel 10ft it takes 10ns, which indicates that transmission line effects might be important.

I wrote this online transmission line simulator so you can visualize what is happening to the signal inside of the transmission line. The model assumes a constant characteristic impedance (which is pretty good in practice for most things).

Parameters tested:

• tRise = 4ns
• tFall = 4ns
• tOn = 36ns
• tOff = 36ns
• tDelay = 10ns
• Zeq = 50 ohms
• R1 = 0 ohms (source terminator)
• R2 = 1 megohms (sink terminator)
• other parameters at their default values

Here are a few snapshots of the results (green is ideally terminated, red is poorly terminated, and blue is the source voltage):

First reflection

Second reflection

During the first reflection, the poorly terminated signal was 40% higher than the desired line signal, and during the second reflection the signal was about 10% lower than the desired line signal.

Repeating by simulating a shorter cable (approx. 1ft cable, i.e. tDelay = 1ns):

First reflection

Second reflection

Notice that the overshoot is much smaller (about 10% for first reflection). This happens even if Zeq is constant!

tl;dr: You should consider properly terminating your cable when using a 10ft long cable and/or getting a short cable. Also, the measured signal is a product of wave interference, even if the cable's equivalent impedance is near constant. In all cases the properly terminated signal always propagates to the correct level without any distortion, and is only delayed because of the finite signal speed. Only with a properly terminated/short cable can you tell if your clock signal is really bad, or if you're just measuring reflections.

Answer 4

You're absolutely right. For every frequency source you'll have a different behaviour in your transmission line since its impedance varies with frequency. You didn't say but as you mentioned 12.5MHz clock I suppose this is a square wave, right? If so, you can expect different reflection magnitude for each frequency component of your clock, hence leading to a highly distortion signal in the end.