Quick & dirty impedance, noise, and jitter measurements

-December 22, 2014

In a previous post I discussed an impedance measurement technique for coaxial cable. The same trick can be used for twisted pair, with the addition of a small pulse or RF transformer.

Generally for twisted pair, the impedance is around 100 to 120 ohms, so you will want a turns ratio of 1.4:1 to 1.5:1, with the lower turns count towards the scope, but it's not critical; 1:1 will do if that’s all you have handy. The transformer is essential to convert the unbalanced BNC and coaxial into a balanced signal where both conductors are differential to each other and both see the same capacitance to ground. Attempting to measure without the transformer will give erroneous results.

So what’s all this good for? Well, I once wondered what the impedance of AWG 30 Kynar insulated wire (the standard wire wrap stuff) is when twisted into a pair with an electric drill. The stuff measured 102 ohms with the trimpot test. A few years later, with access to better equipment (a real Tektronix differential reflectometer plug-in), I tried the same thing. I predicted the impedance to a junior engineer, and, sure enough, the instrument read 102 ohms. He was impressed. I should have bet him a bottle of Scotch…

Now, you know how to find a defective installed cable. Terminate the far end in a resistor equal to the characteristic impedance: You should not see much of a reflection. If you do there is a cable fault such as a kink or broken conductor at the location you calculate by the measured round-trip delay.

Jitter measurement
You can measure sine wave or deterministic jitter amplitude on a scope, and if there is only one jitter frequency you can get a rough idea of what that frequency is. Gaussian jitter is a bit more difficult on an analog scope due to fuzzy edges on the zero crossings, but if you wait long enough, a digital scope does a pretty good job.

Assume sine wave jitter for this example. The trick is to use the delayed sweep to look at the signal a half-period of the jitter later in time. If viewing a square wave, use the scope bandwidth limit to reduce the risetimes and make the transitions easier to see. The transitions will spread out in time; adjust the main timebase and delay times for the maximum widening of the transitions, and the delayed sweep timebase to display one transition for maximum resolution. If this spreading effect occurs over a wide range of delay times, you probably have more than one jitter frequency.

What is happening is that the scope triggers randomly on any transition, always on the same spot far to the left of the delayed sweep display. When you look at the signal later in time, the jitter has raised or lowered the signal frequency and you see this as an accumulated change in period over many cycles, or edge spreading plus and minus the center frequency. The maximum that can occur corresponds to the positive and negative peaks of the sine wave, one half the sine period. The time between the maximum spread extents is the jitter amplitude -- convert to a percentage of the bit time to calculate the unit interval.

White noise measurement
Here is a challenge for all you math and lab whizzes. I forget exactly how this worked, but it would be interesting to try to calculate and compare to what those with decent equipment measure.

Apply the same Gaussian noise signal to two vertical channels (both on the same voltage range) on an analog scope (I don’t think this will work on a digital scope -- maybe on a digital phosphor scope -- another interesting experiment). Use the vertical position controls to place the traces in the top and bottom halves of the screen, then slowly merge the traces together until the dimmer space between them combines to the same brightness as the center portions of each trace. The noise peaks of each trace are filling each other in, for equal intensity across the center of the screen. For best results use a photometer and a large core plastic optical fiber run along the screen; probably more accurate than the human eye.

Now remove the input noise signal and measure the apparent voltage on the graticule between the two quiet traces. This reading should correspond to the noise amplitude in the bandwidth of the scope. I forget what the correction factor is to convert this to RMS.

That’s what the challenge is all about: What is the correction factor to convert the trace separation to RMS noise voltage? Measure or math, your choice…

Also see:

Loading comments...

Write a Comment

To comment please Log In