Probing for Noise

Mike McKinley and Jeff Shuey of Intel recently asked me about noise problems with their high-speed measurements on SDRAM modules. They were using a 5 GS/s digital scope, with an FET probe. The FET probe had a 0.75" ground wire connected to digital GND. With the probe tip touching the VCC supply on the SDRAM module, the scope showed what appeared to be inductive spikes of ±300mV.

By itself, the situation doesn’t sound very unusual. We all have noise on our power bus, but, here’s what really grabbed Mike’s attention. When he probed a spot of digital GND adjacent to the power rail, he saw the same noise.

What causes this effect? How can a probe pick up noise when looking at its own ground? Does this have anything to do with weird ground currents circulating through the probe? If so, how can you ever tell if the noise is really present on GND, VCC, or both?

The key to understanding this situation is Faraday’s Law. Here’s the issue: Any conductive loop you place near a high-speed circuit picks up noise voltages. These noise voltages are caused by the changing magnetic fields which emanate from the board. The loop need not even touch the board to pick them up. We call this kind of effect mutual inductive coupling, or magnetic pickup. Faraday’s Law prescribes the exact amount of coupling you will see in any given situation.

The region encircled by your probe’s ground wire forms a Faraday pickup loop. The ground wire loop picks up magnetic noise just like any other conductive loop. The bigger the area of the ground wire loop, the more noise you see.

Many digital engineers worry that weird currents circulate in the ground system, waiting for a chance to leap into the measurement equipment, from which vantage point they can obfuscate your sensitive measurements. Sometimes, that scenario happens, but rarely. Faraday’s law is a simpler effect, and has nothing to do with your probe ground wire actually touching digital GND.

Here is a simple test that lets you determine whether noise has anything to do with weird ground currents, or just plain old magnetic pickup:

  1. Disconnect the probe (both tip and ground connections) from the board.
  2. Touch the probe tip to its own ground wire and nothing else.
  3. Move the self-grounded probe configuration near the board, but don't let either the probe tip or its own ground come into electrical contact with the board.

The ground wire on the probe now acts as a magnetic-field-pickup loop. It senses any magnetic fields locally present on the board. Because it does not electrically connect to the board, it picks up only the magnetically coupled noise. If you do this experiment, waving the self-grounded probe around near a high-speed design, you will probably see a couple hundred mV worth of noise. This coupling is magnetic in nature, and not due to any mysterious ground currents.

To reduce magnetic noise, try using a probe with a shorter ground wire. With a smaller pickup loop, less noise couples into the scope.

When probing a VCC node that is, presumably, a very low-impedance source, you can use a straight 50 [μm] probe. Forget the fancy FET-input stuff. A three-foot length of RG-174 coax, directly soldered to Vcc and Gnd in the target machine will present a much smaller exposed pickup loop area than the 5-GHz probe he is using. He should take care to use the built-in 50-ohm termination in the scope, and use the AC-coupling mode.

The primary advantages of the RG-174 coax are that its diameter is only 0.100", it is very flexible, and it has plenty of bandwidth for this measurement. Furthermore, since Mike will be probing an extremely low-impedance source, the 50-ohm input impedance of the coax, usually a big disadvantage, won’t matter in this case. A direct coax probe is the perfect solution to Mike’s measurement problem.