2
\$\begingroup\$

I’m currently working on a circuit set up where I’m using high speed serial to send data. However, I took a closer look at the clock signal and it appears to be distorted. I’m trying to determine potential sources of distortion. I’ve narrowed it down to three things: there is an impedance mismatch on the transmission line, the cable is somehow low pass filtering the signal, or the device is simply producing a bad clock signal.

I’ve done some investigation into determining whether or not it’s a transmission line related issue. My setup: A device sourcing the clock signal at 12.5 Mhz with an unknown output impedance. I’m currently waiting to hear back from the manufacturer to determine this output impedance. This signal is then placed on a 10 ft cable (non-coaxial cable - http://multimedia.3m.com/mws/media/343873O/3mtm-shrunk-delta-ribbon-sdr-cable-assembly-ts2120.pdf). It states that the characeteristic impedance of this cable is 100 ohms. The signal is then going into a high speed data acquisition device with input impedance of 50K ohms.

I’ve attempted to impedance match the source, transmission line, and input data acquisition all to 100 ohms. I placed a 100 ohm resistor across the input data acquisition terminals, but I have not done anything related to the clock generating device since I don’t know the output impedance. I have a couple picture of the clock signal running at 6.25 MHz and you can see the distortion.

enter image description here enter image description here

Does this distortion appear to be due to the effects of transmission lines? Or does it look like it is more just due to low pass filtering? However, there are some jagged steps in the signal which don’t seem characteristic of simple low pass filtering of a clock signal.

And I suppose it could also be due a badly generated clock signal to begin with.

Any ideas?

\$\endgroup\$
  • \$\begingroup\$ How does the same signal look before the cable? \$\endgroup\$ – winny Jul 14 '16 at 14:58
  • 1
    \$\begingroup\$ My idea is to do all kinds of measurements yourself. Measure the output impedance as well as the cable impedance. Measure the signal without load, with proper termination, etc. it is pretty impossible to tel if it looks like "if X happened to the signal" when you don't have the before signal to compare to the after one. \$\endgroup\$ – PlasmaHH Jul 14 '16 at 14:58
  • \$\begingroup\$ Have you connected just the 100 resistor across the load end of the line? \$\endgroup\$ – Chu Jul 14 '16 at 15:13
  • 1
    \$\begingroup\$ You're sure your load didn't already have 100 ohm termination before you added your 100 ohms in parallel? Are you sure you used paired wires in your cable correctly? \$\endgroup\$ – The Photon Jul 14 '16 at 16:15
  • 1
    \$\begingroup\$ Disconnect the cable from the driver and what does the driver output look like into an open circuit and into 100 ohns. Get a benchmark! \$\endgroup\$ – Andy aka Jul 14 '16 at 16:48
1
\$\begingroup\$

I've just realized that you are driving a twisted pair from a single ended driver - this would be incorrect. If you are going to do this use coax. The big problem is that twisted pair only sensibly works when it is differentially driven. It appears you are not doing that. You may well compound the problem by having a single ended (not differential) receiver too.

| improve this answer | |
\$\endgroup\$
  • \$\begingroup\$ Excuse my ignorance, but what exactly is the harm in putting a single ended signal on a twisted pair cable? I understand it won't gain the benefits of common noise rejection if it were differntially driven, but what hurt can it cause? \$\endgroup\$ – Izzo Jul 14 '16 at 18:47
  • 1
    \$\begingroup\$ Well, if you are going to do this then you need to terminate the unused ends of the "spare wire" to something because that wire is acting like a secondary winding (a la transformer) and producing a voltage that can cause reflections / add to the misery. So if the sending end has (say) driving impedance of X then you terminate that unused wire to screen in X at the driving end. At the receiving end you should terminate in something but it's all getting a little confusing because this sort of thing isn't a well-recognized scenario. \$\endgroup\$ – Andy aka Jul 14 '16 at 20:12
0
\$\begingroup\$

The plateaus or steps that you see in the waveform are an indicator that you have significant reflections occurring due to impedance mismatch. Make sure you probe the signal at the end of the line so that what you see on the oscilloscope matches what the receiver is seeing as closely as possible. The width of the step is approximately twice the propagation delay of the impedance mismatch.

| improve this answer | |
\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.