Address: P. O. Box 1457, Los Gatos,
CA 95031
TEL:
800-323-3956/408-356-4186
FAX:
408-358-3799
Mobile: 408-858-4528
URL:
www.dsmith.org
Email: doug@dsmith.org
Figure 1. A Wire
Well, maybe... The wire in Figure 1 has some interesting properties that bear on signal integrity and system reliability. The voltage drop across a short wire (less than 1/10 wavelength) is given by:
e = L·di/dt + Ri
where:
For reasonable size signal and return loops where the wire is more than
a a few cm from its return path (a ground plane for instance), the inductance
can be on the order of 10 nH/cm. For 2 cm of wire (about an inch) that
is about 20 nH. Table 1 lists a few signal and noise sources, a possible
di/dt value, and the resulting voltage drop across 2 cm of wire.
|
|
|
Logic signal |
|
|
Switching power supplies |
|
|
Electrostatic Discharge (ESD) |
|
|
Table 1. Voltage Drops Along a Wire Due to Various Sources of Current
It is possible for a simple HC240, a relatively low speed buffer, to generate 2 volts across about 10 cm of wire! In my experience, the first time I became aware of the magnitude of this effect was years ago at Bell Labs. The system was one of the first stored program controlled business telephone systems, and it had a clock in the backplane. The software folks wanted to know what the effect of shorting the clock in the backplane would be, such as might happen if a defective board was plugged in, in hopes of having the software recognize the pattern of resulting system problems. This ability would allow the software to print the proper error code. An engineer shorted the clock to signal ground with a 6 inch test clip, but the system continued to work perfectly! He asked me why we had this unnecessary clock in the backplane.
It turned out that the L·di/dt drop across the 6 inch lead was enough to let the system work. The clock wave form was a square wave and the L·di/dt drop was a series of alternating positive and negative spikes (the differential of the square wave). The positive spikes passed through the logic threshold and were of sufficient width to be recognized as a valid edge. So the system worked! When the clock was shorted to the adjacent ground pin with a screwdriver blade, the desired result was achieved.
As speeds have increased, L·di/dt voltage drops have become ever more important. I have observed well over one volt drop across a bonding wire in a chip package. Click here to see my paper "A Method for Troubleshooting Noise Internal to an IC" for more information on noise associated with chip packages and how to measure this form of it.
With today's sub-nanosecond risetimes, even the few nanohenries of good bypass capacitors and small board features must be taken into account.
The numbers for switching supplies in Table 1 apply to external leads connecting to the supply, the input and outputs. I have seen many switching power supplies that caused a peak drop of 2 volts/cm on connected leads. In one case, signal integrity problems were caused a meter away in a part of the system that was not even powered from the noisy supply. A typical symptom is an intermittent problem that occurs a few times per hour. This kind of problem can be a real time burner for engineers working on a prototype system.
Switching power supplies and electrostatic discharge can be potent noise sources that can wreak havoc in a system, even remotely, from a distance. Switching supplies can be especially problematical in systems with optical interfaces, telephone connections, magnetic storage devices, or other low level signals. These topics will be covered in future Technical Tidbits.