Mike McKinley of Intel (Mike_McKinley@ccm2.hf.intel.com) writes:
Here is the question: a co-worker and I were using a FET probe with a .75" to 1" ground wire (both were tried) in a 66Mhz environment. Using a 5 GS/s digital scope, we tried to measure the stability of VCC on an SDRAM memory module (a DIMM). The scope showed what appeared to be an inductive spike of +/-300mV or greater. When my co-worker measured at the same Place that he referenced ground, he still got a +/-250mV "inductive" reaction. What we are curious about is whether there is a delay in the Scope or from the wire that would cause two separate spikes--one on VCC and one on VSS--to appear on the scope as an inductive spike. (This is in a 3.3V environment.) If so, what would be a way to accurately measure the stability of VCC in reference to VSS so that they are aligned in time?
Mike, you have run smack into a classic measurement problem.
Let me see if I can restate your problem:
- Using an FET probe
- wire a 1" ground wire
- with the ground wire connected to digital ground at point A,
- you touch the probe to the same digital ground at point A, and
- you see 250 mV noise spikes.
HERE'S THE ISSUE
Any conductive loop near your circuit will pick up noise voltages. These voltages are related to magnetic fields which are circulating on the board, near the point of measurement. The loop doesn't even have to be touching the board to pick them up. We call this kind of noise *magnetic pickup*.
The relevant loop in this case starts at point A, runs up the ground wire to your probe, passes through a 1 pF effective capacitance (that's the ground-to-tip capacitance of your probe), and then comes back down the probe to its tip at point A.
Your question indicates that you believe you are seeing the effects of some kind of weird current emanating from the ground connection. You probably aren't. Fortunately, we won't have to argue back and forth on this issue for long, because there is a *very simple* test you can use to prove to yourself that the noise has nothing to do with weird ground voltages.
(a) Disconnect the probe from the board.
(b) Disconnect the probe ground wire from the board
(c) Connect the probe ground wire to its own tip (this makes a little free-space magnetic field sensor loop)
(d) Move the probe, with its ground connected to its tip, near the troublesome spot on your board. Don't let the probe or its ground touch the board.
(e) You will see the magnetic pickup, probably about 250 mV worth.
OK, so let's say you are getting some of this kind of magnetic pickup, (please write back if you aren't). What can you do about it?
YOU HAVE THREE CHOICES
(1) Learn to live with it. If it's small enough, this is a practical solution. In your case, since you are looking for small power supply fluctuations, this probably won't work.
(2) Probe the ground point, as you have been doing, and store the noise on your scope. Now look at the signal (keeping the ground point at A fixed), and digitally subtract the known interference. This will somewhat improve your clarity of view, but is far from a perfect solution for two reasons. First, the magnetic noise pickup is a function of the exact placement of the probing loop. Move the loop, and the noise changes. Second, unless you have a good trigger, and the noise is synchronous to your other signals, you will never be able to accomplish the subtraction.
(3) Use a probe with an even smaller pickup loop. The magnetic pickup will be proportional to the pickup loop size. Since you are probing a power supply point (presumably a very low impedance source), you can just use a straight 50-ohm probe. Forget the fancy FET-input stuff. Just use a piece of RG-174 coax, with a BNC crimped onto one end (for the scope). At the other end of the coax, solder the coax signal and ground connections onto your board, using as small an exposed loop area between signal and ground as possible. Of course, before you look at a signal you will want to solder both coax signal and coax ground to point A, so you can measure the effective noise floor with this new probe, and then you can hunting for the Vcc-Vss connections that you want to see.
You are not alone. I just worked with another engineer on Wednesday of this week who had precisely the same problem. Any ground wire affixed to your probe, even a short one-inch connection, degrades measurement accuracy and also picks up noise.
Dr. Howard Johnson